Video Book Ch1 3+5

Wireless Video Communications:
Second to Third Generation Systems and

Beyond
by
c L. Hanzo, P.J. Cherriman, J. Streit
Department of Electronics and Computer Science,

University of Southampton, UK
Contents
Preface
xxiii
Acknowledgments
xxix
Contributors
xxxi
I Transmission Issues
1 Information Theory
1.1 Issues in Information Theory . . . . . . . . . . . . . . . . . . . . . . .

1.2 Additive White Gaussian Noise Channel . . . . . . . . . . . . . . . . .
1.2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.2 Practical Gaussian Channels . . . . . . . . . . . . . . . . . . .
1.2.3 Gaussian Noise . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3 Information of a Source . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4 Average Information of Discrete Memoryless Sources . . . . . . . . . .
1.4.1 Maximum Entropy of a Binary Source . . . . . . . . . . . . . .
1.4.2 Maximum Entropy of a q-ary Source . . . . . . . . . . . . . . .
1.5 Source Coding for a Discrete Memoryless Source . . . . . . . . . . . .
1.5.1 Shannon-Fano Coding . . . . . . . . . . . . . . . . . . . . . . .
1.5.2 Human Coding . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6 Average Information of Discrete Sources Exhibiting Memory . . . . .
1.6.1 Two-State Markov Model for Discrete Sources Exhibiting Memory
1.6.2 N -State Markov Model for Discrete Sources Exhibiting Memory
1.7 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.1 Two-State Markov Model Example . . . . . . . . . . . . . . . .
1.7.2 Four-State Markov Model for a 2-Bit Quantizer . . . . . . . . .
1.8 Generating Model Sources . . . . . . . . . . . . . . . . . . . . . . . . .
1.8.1 Autoregressive Model . . . . . . . . . . . . . . . . . . . . . . .
1.8.2 AR Model Properties . . . . . . . . . . . . . . . . . . . . . . .
vii
3
7
7
8
8
11
12
13
15
15
16
18
22
22
24
25
25
27
28
28
29
viii
CONTENTS
1.8.3 First-Order Markov Model . . . . . . . . . . . . . . . . . . . .

1.9 Run-Length Coding for Discrete Sources Exhibiting Memory . . . . .
1.9.1 Run-Length Coding Principle . . . . . . . . . . . . . . . . . . .
1.9.2 Run-Length Coding Compression Ratio . . . . . . . . . . . . .
1.10 Information Transmission via Discrete Channels . . . . . . . . . . . . .
1.10.1 Binary Symmetric Channel Example . . . . . . . . . . . . . . .
1.10.2 Bayes' Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.10.3 Mutual Information . . . . . . . . . . . . . . . . . . . . . . . .
1.10.4 Mutual Information Example . . . . . . . . . . . . . . . . . . .
1.10.5 Information Loss via Imperfect Channels . . . . . . . . . . . .
1.10.6 Error Entropy via Imperfect Channels . . . . . . . . . . . . . .
1.11 Capacity of Discrete Channels . . . . . . . . . . . . . . . . . . . . . . .
1.12 Shannon's Channel Coding Theorem . . . . . . . . . . . . . . . . . . .
1.13 Capacity of Continuous Channels . . . . . . . . . . . . . . . . . . . . .
1.13.1 Practical Evaluation of the Shannon-Hartley Law . . . . . . . .
1.13.2 Shannon's Ideal Communications System for Gaussian Channels
1.14 Shannon's Message and Its Implications for Wireless Channels . . . .
1.15 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . .
2 The Propagation Environment
2.1 The Cellular Concept . . . . . . . . . . . . . . . . . . . . . .

2.2 Radio Wave Propagation . . . . . . . . . . . . . . . . . . . . .
2.2.1 Background . . . . . . . . . . . . . . . . . . . . . . . .
2.2.2 Narrowband Fading Channels . . . . . . . . . . . . . .
2.2.3 Propagation Path-loss Law . . . . . . . . . . . . . . .
2.2.4 Slow-Fading Statistics . . . . . . . . . . . . . . . . . .
2.2.5 Fast-Fading Statistics . . . . . . . . . . . . . . . . . .
2.2.6 Doppler Spectrum . . . . . . . . . . . . . . . . . . . .
2.2.7 Simulation of Narrowband Fading Channels . . . . . .
2.2.7.1 Frequency-Domain Fading Simulation . . . .
2.2.7.2 Time-Domain Fading Simulation . . . . . . .
2.2.7.3 Box-Muller Algorithm of AWGN Generation
2.2.8 Wideband Channels . . . . . . . . . . . . . . . . . . .
2.2.8.1 Modeling of Wideband Channels . . . . . . .
2.3 Summary and Conclusions . . . . . . . . . . . . . . . . . . . .
3 Convolutional Channel Coding

3.1
3.2
3.3
3.4
Brief Channel Coding History . . . . . . . . . . . .

Convolutional Encoding . . . . . . . . . . . . . . .
State and Trellis Transitions . . . . . . . . . . . . .
The Viterbi Algorithm . . . . . . . . . . . . . . . .
3.4.1 Error-Free Hard-Decision Viterbi Decoding
3.4.2 Erroneous Hard-Decision Viterbi Decoding
3.4.3 Error-Free Soft-Decision Viterbi Decoding .
3.5 Summary and Conclusions . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
30
31
31
32
34
34
38
39
40
42
43
49
53
55
58
62
62
65
67
67
71
71
73
73
76
77
83
85
86
86
87
87
87
92
93
93
94
96
98
98
101
104
106
CONTENTS
ix
4 Block-Based Channel Coding
107
5 Modulation and Transmission Techniques
161
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

4.2 Finite Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.2.1 Denitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.2.2 Galois Field Construction . . . . . . . . . . . . . . . . . . . . . 111
4.2.3 Galois Field Arithmetic . . . . . . . . . . . . . . . . . . . . . . 113
4.3 Reed-Solomon and Bose-Chaudhuri-Hocquenghem Block Codes . . . . 114
4.3.1 Denitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.3.2 RS Encoding . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4.3.3 RS Encoding Example . . . . . . . . . . . . . . . . . . . . . . . 118
4.3.4 Linear Shift-Register Circuits for Cyclic Encoders . . . . . . . 122
4.3.4.1 Polynomial Multiplication . . . . . . . . . . . . . . . . 122
4.3.4.2 Systematic Cyclic Shift-Register Encoding Example . 123
4.3.5 RS Decoding . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.3.5.1 Formulation of the Key Equations [26, 61, 63, 66{70, 73] 126
4.3.5.2 Peterson-Gorenstein-Zierler Decoder . . . . . . . . . . 130
4.3.5.3 PGZ Decoding Example . . . . . . . . . . . . . . . . . 133
4.3.5.4 Berlekamp-Massey Algorithm [26, 61, 63, 66{70, 73] . . 138
4.3.5.5 Berlekamp-Massey Decoding Example . . . . . . . . . 144
4.3.5.6 Computation of the Error Magnitudes by the Forney
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . 148
4.3.5.7 Forney Algorithm Example . . . . . . . . . . . . . . . 151
4.3.5.8 Error Evaluator Polynomial Computation . . . . . . . 153
4.4 RS and BCH Codec Performance . . . . . . . . . . . . . . . . . . . . . 156
4.5 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 158
5.1 Modulation Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
5.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
5.1.2 Quadrature Amplitude Modulation [84] . . . . . . . . . . . . . 164
5.1.2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . 164
5.1.2.2 Modem Schematic . . . . . . . . . . . . . . . . . . . . 164
5.1.2.2.1 Gray Mapping and Phasor Constellation . . 167
5.1.2.2.2 Nyquist Filtering . . . . . . . . . . . . . . . . 168
5.1.2.2.3 Modulation and Demodulation . . . . . . . . 170
5.1.2.2.4 Data Recovery . . . . . . . . . . . . . . . . . 171
5.1.2.3 QAM Constellations . . . . . . . . . . . . . . . . . . . 172
5.1.2.4 16QAM BER versus SNR Performance over AWGN
Channels . . . . . . . . . . . . . . . . . . . . . . . . . 175
5.1.2.4.1 Decision Theory . . . . . . . . . . . . . . . . 175
5.1.2.4.2 QAM Modulation and Transmission . . . . . 177
5.1.2.4.3 16QAM Demodulation in AWGN . . . . . . 178
5.1.2.5 Reference-Assisted Coherent QAM for Fading Channels181
5.1.2.5.1 PSAM System Description . . . . . . . . . . 181
5.1.2.5.2 Channel Gain Estimation in PSAM . . . . . 183
5.1.2.5.3 PSAM Performance [128] . . . . . . . . . . . 185
CONTENTS
5.2
5.3
5.4
5.5
5.1.2.6 Dierentially Detected QAM [84] . . . . . . . . . . . .

5.1.3 Adaptive Modulation . . . . . . . . . . . . . . . . . . . . . . .
5.1.3.1 Background to Adaptive Modulation . . . . . . . . .
5.1.3.2 Bit Error Rate and Channel Capacity Optimization of
Adaptive Modems . . . . . . . . . . . . . . . . . . . .
5.1.3.3 Bit Error Rate and Channel Capacity of Adaptive
Modulation . . . . . . . . . . . . . . . . . . . . . . . .
5.1.3.4 Equalization Techniques . . . . . . . . . . . . . . . . .
Orthogonal Frequency Division Multiplexing . . . . . . . . . . . . . .
Packet Reservation Multiple Access . . . . . . . . . . . . . . . . . . . .
Flexible Transceiver Architecture . . . . . . . . . . . . . . . . . . . . .
Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . .
6 Video Trac Modeling and Multiple Access
6.1 Video Trac Modeling . . . . . . . . . . . . . . . . . . . .

6.1.1 Motivation and Background . . . . . . . . . . . . .
6.1.2 Markov Modeling of Video Sources . . . . . . . . .
6.1.3 Reduced-Length Poisson Cycles . . . . . . . . . . .
6.1.4 Video Model Matching . . . . . . . . . . . . . . . .
6.2 Multiple Access . . . . . . . . . . . . . . . . . . . . . . . .
6.2.1 Background . . . . . . . . . . . . . . . . . . . . . .
6.2.2 Classication of Multiple Access Techniques . . . .
6.2.3 Multiframe Packet Reservation Multiple Access . .
6.2.3.1 Performance of MF-PRMA . . . . . . . .
6.2.4 Statistical Packet Assignment Multiple Access . .
6.2.4.1 Statistical Packet Assignment Principles .
6.2.4.2 Performance of the SPAMA Protocol . .
6.3 Summary and Conclusions . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2 Factors Controlling Co-Channel Interference . . . . . . . . . . .
7.2.1 Eect of Fading on Co-Channel Interference . . . . . . .
7.2.2 Cell Shapes . . . . . . . . . . . . . . . . . . . . . . . . .
7.2.3 Position of Users and Interferers . . . . . . . . . . . . .
7.3 Theoretical Signal-to-Interference Ratio . . . . . . . . . . . . .
7.4 Simulation Parameters . . . . . . . . . . . . . . . . . . . . . . .
7.5 Results for Multiple Interferers . . . . . . . . . . . . . . . . . .
7.5.1 SIR Prole of a Cell . . . . . . . . . . . . . . . . . . . .
7.5.2 Signal-to-Noise-Plus-Interference Ratio (SINR) . . . . .
7.5.3 Channel Capacity . . . . . . . . . . . . . . . . . . . . .
7.6 Results for a Single Interferer . . . . . . . . . . . . . . . . . . .
7.6.1 Simple Model for SINR in a Single Interferer Situation .
7.6.2 Eect of SIR and SNR on Error Rates . . . . . . . . . .
7.6.3 Time-Varying Eects of SIR and SINR . . . . . . . . . .
7.6.4 Eect of Interference on the H.263 Videophone System .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7 Co-Channel Interference
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
186
190
190
193
195
197
197
201
202
204
205
205
205
207
210
215
223
223
225
227
228
237
237
242
243
247
247
248
248
249
250
252
255
258
258
262
263
269
270
272
275
280
CONTENTS
xi
8 Channel Allocation
287
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287

8.2 Overview of Channel Allocation . . . . . . . . . . . . . . . . . . . . . . 288
8.2.1 Fixed Channel Allocation . . . . . . . . . . . . . . . . . . . . . 289
8.2.1.1 Channel Borrowing . . . . . . . . . . . . . . . . . . . 291
8.2.1.2 Flexible Channel Allocation . . . . . . . . . . . . . . 292
8.2.2 Dynamic Channel Allocation . . . . . . . . . . . . . . . . . . . 292
8.2.2.1 Centrally Controlled DCA Algorithms . . . . . . . . . 294
8.2.2.2 Distributed DCA Algorithms . . . . . . . . . . . . . . 295
8.2.2.3 Locally distributed DCA algorithms . . . . . . . . . . 296
8.2.3 Hybrid Channel Allocation . . . . . . . . . . . . . . . . . . . . 297
8.2.4 The Eect of Handovers . . . . . . . . . . . . . . . . . . . . . . 298
8.2.5 The Eect of Transmission Power Control . . . . . . . . . . . . 299
8.3 Simulation of the Channel Allocation Algorithms . . . . . . . . . . . . 299
8.3.1 The Mobile Radio Network Simulator, \Netsim" . . . . . . . . 299
8.3.1.1 Physical Layer Model . . . . . . . . . . . . . . . . . . 302
8.3.1.2 Shadow Fading Model . . . . . . . . . . . . . . . . . . 302
8.3.2 Channel Allocation Algorithms Investigated . . . . . . . . . . . 304
8.3.2.1 Fixed Channel Allocation Algorithm . . . . . . . . . . 304
8.3.2.2 Distributed Dynamic Channel Allocation Algorithms 304
8.3.2.3 Locally Distributed Dynamic Channel Allocation
Algorithms . . . . . . . . . . . . . . . . . . . . . . . . 306
8.3.3 Performance Metrics . . . . . . . . . . . . . . . . . . . . . . . . 306
8.3.4 Nonuniform Trac Model . . . . . . . . . . . . . . . . . . . . . 309
8.4 Performance Comparisons . . . . . . . . . . . . . . . . . . . . . . . . . 310
8.4.1 System Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 310
8.4.2 Carried Trac with Quality Constraints . . . . . . . . . . . . . 311
8.4.3 Comparing the LOLIA with FCA . . . . . . . . . . . . . . . . . 312
8.4.4 Eect of the \Reuse Distance" Constraint on the LOLIA and
LOMIA DCA Algorithms . . . . . . . . . . . . . . . . . . . . . 314
8.4.5 Comparison of the LOLIA and LOMIA with the LIA . . . . . 317
8.4.6 Interference Threshold-Based Distributed DCA Algorithms . . 318
8.4.7 Performance Comparison of Fixed and Dynamic Channel
Allocation Algorithms Using Nonuniform Trac Distributions 321
8.4.8 Eect of Shadow Fading on the FCA, LOLIA, and LOMIA . . 324
8.4.9 Eect of Shadow Fading Frequency and Standard Deviation on
the LOLIA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
8.4.10 Eect of Shadow Fading Standard Deviation on FCA and LOLIA327
8.4.11 SINR Prole across Cell Area . . . . . . . . . . . . . . . . . . . 329
8.4.12 Overview of Results . . . . . . . . . . . . . . . . . . . . . . . . 332
xii
CONTENTS
9 Second-Generation Mobile Systems
9.1 The Wireless Communications Scene . . . . . . . . .

9.2 Global System for Mobile Communications | GSM
9.2.1 Introduction to GSM . . . . . . . . . . . . . .
9.2.2 Overview of GSM . . . . . . . . . . . . . . .
9.2.3 Logical and Physical Channels in GSM . . .
9.2.4 Speech and Data Transmission in GSM . . .
9.2.5 Transmission of Control Signals in GSM . . .
9.2.6 Synchronization Issues in GSM . . . . . . . .
9.2.7 Gaussian Minimum Shift Keying in GSM . .
9.2.8 Wideband Channel Models in GSM . . . . .
9.2.9 Adaptive Link Control in GSM . . . . . . . .
9.2.10 Discontinuous Transmission in GSM . . . . .
9.2.11 Summary and Conclusions . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.2 Basic CDMA System . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.2.1 Spread Spectrum Fundamentals . . . . . . . . . . . . . . . . .
10.2.1.1 Frequency Hopping . . . . . . . . . . . . . . . . . .
10.2.1.2 Direct Sequence . . . . . . . . . . . . . . . . . . . .
10.2.2 The Eect of Multipath Channels . . . . . . . . . . . . . . .
10.2.3 RAKE Receiver . . . . . . . . . . . . . . . . . . . . . . . . . .
10.2.4 Multiple Access . . . . . . . . . . . . . . . . . . . . . . . . . .
10.2.4.1 Down-Link Interference . . . . . . . . . . . . . . . .
10.2.4.2 Up-Link Interference . . . . . . . . . . . . . . . . . .
10.2.4.3 Gaussian Approximation . . . . . . . . . . . . . . .
10.2.5 Spreading Codes . . . . . . . . . . . . . . . . . . . . . . . . .
10.2.5.1 m-sequences . . . . . . . . . . . . . . . . . . . . . .
10.2.5.2 Gold Sequences . . . . . . . . . . . . . . . . . . . .
10.2.5.3 Extended m-sequences . . . . . . . . . . . . . . . .
10.2.6 Channel Estimation . . . . . . . . . . . . . . . . . . . . . . .
10.2.6.1 Down-Link Pilot-Assisted Channel Estimation . . .
10.2.6.2 Up-Link Pilot-Symbol Assisted Channel Estimation
10.2.6.3 Pilot-Symbol Assisted Decision-Directed Channel
Estimation . . . . . . . . . . . . . . . . . . . . . . .
10.2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.3 Third-Generation Wireless Mobile Communication Systems . . . . .
10.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.3.2 UMTS Terrestrial Radio Access (UTRA) . . . . . . . . . . .
10.3.2.1 Characteristics of UTRA . . . . . . . . . . . . . . .
10.3.2.2 Transport Channels . . . . . . . . . . . . . . . . . .
10.3.2.3 Physical Channels . . . . . . . . . . . . . . . . . . .
10.3.2.3.1 Dedicated Physical Channels . . . . . . . .
10.3.2.3.2 Common Physical Channels . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
10 CDMA Systems: Third-Generation and Beyond
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
339
339
342
342
345
346
347
351
357
358
359
360
363
363
365
365
366
366
367
368
371
374
378
379
380
383
385
385
386
387
387
388
389
390
392
392
392
395
395
399
400
401
404
CONTENTS
xiii
10.3.2.3.2.1 Common Physical Channels of the FDD

Mode
404
10.3.2.3.2.2 Common Physical Channels of the TDD
Mode
408
10.3.2.4 Service Multiplexing and Channel Coding in UTRA . 410
10.3.2.4.1 CRC Attachment . . . . . . . . . . . . . . . 411
10.3.2.4.2 Transport Block Concatenation . . . . . . . 411
10.3.2.4.3 Channel-Coding . . . . . . . . . . . . . . . . 411
10.3.2.4.4 Radio Frame Padding . . . . . . . . . . . . . 414
10.3.2.4.5 First Interleaving . . . . . . . . . . . . . . . 414
10.3.2.4.6 Radio Frame Segmentation . . . . . . . . . . 414
10.3.2.4.7 Rate Matching . . . . . . . . . . . . . . . . . 414
10.3.2.4.8 Discontinuous Transmission Indication . . . 415
10.3.2.4.9 Transport Channel Multiplexing . . . . . . . 415
10.3.2.4.10 Physical Channel Segmentation . . . . . . . 415
10.3.2.4.11 Second Interleaving . . . . . . . . . . . . . . 415
10.3.2.4.12 Physical Channel Mapping . . . . . . . . . . 415
10.3.2.4.13 Mapping Several Multirate Services to the
UL Dedicated Physical Channels in FDD Mode416
10.3.2.4.14 Mapping of a 4.1 Kbps Data Service to the
DL DPDCH in FDD Mode . . . . . . . . . . 417
10.3.2.4.15 Mapping Several Multirate Services to the
UL Dedicated Physical Channels in TDD Mode420
10.3.2.5 Variable-Rate and Multicode Transmission in UTRA 420
10.3.2.6 Spreading and Modulation . . . . . . . . . . . . . . . 422
10.3.2.6.1 Orthogonal Variable Spreading Factor Codes 423
10.3.2.6.2 Up-Link Scrambling Codes . . . . . . . . . . 426
10.3.2.6.3 Down-Link Scrambling Codes . . . . . . . . 426
10.3.2.6.4 Up-Link Spreading and Modulation . . . . . 426
10.3.2.6.5 Down-Link Spreading and Modulation . . . . 428
10.3.2.7 Random Access . . . . . . . . . . . . . . . . . . . . . 429
10.3.2.7.1 Mobile-Initiated Physical Random Access
Procedures . . . . . . . . . . . . . . . . . . . 429
10.3.2.7.2 Common Packet Channel Access Procedures 430
10.3.2.8 Power Control . . . . . . . . . . . . . . . . . . . . . . 430
10.3.2.8.1 Closed-Loop Power Control in UTRA . . . . 430
10.3.2.8.2 Open-Loop Power Control in TDD Mode . . 431
10.3.2.9 Cell Identication . . . . . . . . . . . . . . . . . . . . 432
10.3.2.9.1 Cell Identication in the FDD Mode . . . . . 432
10.3.2.9.2 Cell Identication in the TDD Mode . . . . . 434
10.3.2.10 Handover . . . . . . . . . . . . . . . . . . . . . . . . . 436
10.3.2.10.1 Intra-Frequency Handover or Soft Handover 436
xiv
CONTENTS
10.3.2.10.2 Inter-Frequency Handover or Hard Handover 436

10.3.2.11 Intercell Time Synchronization in the UTRA TDD Mode438
10.3.3 The cdma2000 Terrestrial Radio Access . . . . . . . . . . . . . 439
10.3.3.1 Characteristics of cdma2000 . . . . . . . . . . . . . . 439
10.3.3.2 Physical Channels in cdma2000 . . . . . . . . . . . . 441
10.3.3.3 Service Multiplexing and Channel Coding . . . . . . . 443
10.3.3.4 Spreading and Modulation . . . . . . . . . . . . . . . 445
10.3.3.4.1 Down-Link Spreading and Modulation . . . . 446
10.3.3.4.2 Up-link Spreading and Modulation . . . . . . 447
10.3.3.5 Random Access . . . . . . . . . . . . . . . . . . . . . 447
10.3.3.6 Handover . . . . . . . . . . . . . . . . . . . . . . . . . 450
10.3.4 Performance-Enhancement Features . . . . . . . . . . . . . . . 452
10.3.4.1 Down-Link Transmit Diversity Techniques . . . . . . 452
10.3.4.1.1 Space Time Block Coding-Based Transmit
Diversity . . . . . . . . . . . . . . . . . . . . 452
10.3.4.1.2 Time-Switched Transmit Diversity . . . . . . 452
10.3.4.1.3 Closed-Loop Transmit Diversity . . . . . . . 452
10.3.4.2 Adaptive Antennas . . . . . . . . . . . . . . . . . . . 453
10.3.4.3 Multi-User Detection/Interference Cancellation . . . . 453
10.3.5 Summary of 3G Systems . . . . . . . . . . . . . . . . . . . . . . 454
II Video Systems Based on Proprietary Video Codecs

11 Fractal Image Codecs
11.1 Fractal Principles . . . . . . . . . .

11.2 One-Dimensional Fractal Coding .
11.2.1 Fractal Codec Design . . .
11.2.2 Fractal Codec Performance
11.3 Error Sensitivity and Complexity .
11.4 Summary and Conclusions . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12.1 Video Codec Outline . . . . . . . . . . . . . . . . . . . . . . .

12.2 The Principle of Motion Compensation . . . . . . . . . . . . .
12.2.1 Distance Measures . . . . . . . . . . . . . . . . . . . .
12.2.2 Motion Search Algorithms . . . . . . . . . . . . . . . .
12.2.2.1 Full or Exhaustive Motion Search . . . . . .
12.2.2.2 Gradient-Based Motion Estimation . . . . .
12.2.2.3 Hierarchical or Tree Search . . . . . . . . . .
12.2.2.4 Subsampling Search . . . . . . . . . . . . . .
12.2.2.5 Post-Processing of Motion Vectors . . . . . .
12.2.2.6 Gain-Cost-Controlled Motion Compensation
12.2.3 Other Motion Estimation Techniques . . . . . . . . . .
12.2.3.1 Pel-Recursive Displacement Estimation . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12 Very Low Bit-Rate DCT Codecs
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
457
459
459
462
465
467
471
473
475
475
477
481
482
483
484
485
486
487
487
489
490
CONTENTS
12.2.3.2 Grid Interpolation Techniques . . . . . . . .

12.2.3.3 MC Using Higher Order Transformations . .
12.2.3.4 MC in the Transform Domain . . . . . . . .
12.2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . .
12.3 Transform Coding . . . . . . . . . . . . . . . . . . . . . . . .
12.3.1 One-Dimensional Transform Coding . . . . . . . . . .
12.3.2 Two-Dimensional Transform Coding . . . . . . . . . .
12.3.3 Quantizer Training for Single-Class DCT . . . . . . .
12.3.4 Quantizer Training for Multiclass DCT . . . . . . . .
12.4 The Codec Outline . . . . . . . . . . . . . . . . . . . . . . . .
12.5 Initial Intra-Frame Coding . . . . . . . . . . . . . . . . . . . .
12.6 Gain-Controlled Motion Compensation . . . . . . . . . . . . .
12.7 The MCER Active/Passive Concept . . . . . . . . . . . . . .
12.8 Partial Forced Update of the Reconstructed Frame Buers . .
12.9 The Gain/Cost-Controlled Inter-Frame Codec . . . . . . . . .
12.9.1 Complexity Considerations and Reduction Techniques
12.10The Bit-Allocation Strategy . . . . . . . . . . . . . . . . . . .
12.11Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12.12DCT Codec Performance under Erroneous Conditions . . . .
12.12.1Bit Sensitivity . . . . . . . . . . . . . . . . . . . . . .
12.12.2Bit Sensitivity of Codec I and II . . . . . . . . . . . .
12.13DCT-Based Low-Rate Video Transceivers . . . . . . . . . . .
12.13.1Choice of Modem . . . . . . . . . . . . . . . . . . . . .
12.13.2Source-Matched Transceiver . . . . . . . . . . . . . .
12.13.2.1 System 1 . . . . . . . . . . . . . . . . . . . .
12.13.2.1.1 System Concept . . . . . . . . . . .
12.13.2.1.2 Sensitivity-Matched Modulation . .
12.13.2.1.3 Source Sensitivity . . . . . . . . . .
12.13.2.1.4 Forward Error Correction . . . . . .
12.13.2.1.5 Transmission Format . . . . . . . .
12.13.2.2 System 2 . . . . . . . . . . . . . . . . . . . .
12.13.2.2.1 Automatic Repeat Request . . . . .
12.13.2.3 Systems 3{5 . . . . . . . . . . . . . . . . . .
12.14System Performance . . . . . . . . . . . . . . . . . . . . . . .
12.14.1Performance of System 1 . . . . . . . . . . . . . . . .
12.14.2Performance of System 2 . . . . . . . . . . . . . . . .
12.14.2.1 FER Performance . . . . . . . . . . . . . . .
12.14.2.2 Slot Occupancy Performance . . . . . . . . .
12.14.2.3 PSNR Performance . . . . . . . . . . . . . .
12.14.3Performance of Systems 3{5 . . . . . . . . . . . . . . .
12.15Summary and Conclusions . . . . . . . . . . . . . . . . . . . .
xv
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
490
490
491
491
492
492
493
496
497
499
502
502
503
504
506
508
509
510
512
513
515
516
516
517
517
517
518
518
518
519
520
522
523
524
524
527
527
529
530
531
535
xvi
CONTENTS
13 VQ Codecs and Multimode Video Transceivers
13.1 Introduction . . . . . . . . . . . . . . . . . . . . . .
13.2 The Codebook Design . . . . . . . . . . . . . . . .
13.3 The Vector Quantizer Design . . . . . . . . . . . .
13.3.1 Mean and Shape Gain Vector Quantization
13.3.2 Adaptive Vector Quantization . . . . . . . .
13.3.3 Classied Vector Quantization . . . . . . .
13.3.4 Algorithmic Complexity . . . . . . . . . . .
13.4 Performance under Erroneous Conditions . . . . .
13.4.1 Bit-Allocation Strategy . . . . . . . . . . .
13.4.2 Bit Sensitivity . . . . . . . . . . . . . . . .
13.5 VQ-Based Low-Rate Video Transceivers . . . . . .
13.5.1 Choice of Modulation . . . . . . . . . . . .
13.5.2 Forward Error Correction . . . . . . . . . .
13.5.3 Architecture of System 1 . . . . . . . . . .
13.5.4 Architecture of System 2 . . . . . . . . . .
13.5.5 Architecture of Systems 3{6 . . . . . . . . .
13.6 System Performance . . . . . . . . . . . . . . . . .
13.6.1 Simulation Environment . . . . . . . . . . .
13.6.2 Performance of Systems 1 and 3 . . . . . .
13.7 Summary and Conclusions . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
14 Low Bit-Rate Parametric Quad-Tree-Based Codecs and

Multimode Videophone Transceivers
14.1 Introduction . . . . . . . . . . . . . . . . . .
14.2 Quad-Tree Decomposition . . . . . . . . . .
14.3 Quad-Tree Intensity Match . . . . . . . . .
14.3.1 Zero-Order Intensity Match . . . . .
14.3.2 First-Order Intensity Match . . . . .
14.3.3 Decomposition Algorithmic Issues .
14.4 Model-Based Parametric Enhancement . . .
14.4.1 Eye and Mouth Detection . . . . . .
14.4.2 Parametric Codebook Training . . .
14.4.3 Parametric Encoding . . . . . . . . .
14.5 The Enhanced QT Codec . . . . . . . . . .
14.6 Performance under Erroneous Conditions .
14.6.1 Bit Allocation . . . . . . . . . . . . .
14.6.2 Bit Sensitivity . . . . . . . . . . . .
14.7 QT-Codec-Based Video Transceivers . . . .
14.7.1 Channel Coding and Modulation . .
14.7.2 QT-Based Transceiver Architectures
14.8 QT-Based Video-Transceiver Performance .
14.9 Summary of QT-Based Video Transceivers .
14.10Summary of Low-Rate Codecs/Transceivers
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
537
537
537
541
544
546
548
549
550
550
552
554
554
554
555
557
558
558
558
560
561
563
564
567
567
568
571
571
573
573
576
577
580
581
582
583
584
586
586
586
588
591
595
595
CONTENTS
xvii
III High-Resolution Image Coding
601
15 Low-Complexity Techniques
15.1 Introduction and Video Formats . . . . . . . . . . . . . . . .

15.2 Dierential Pulse Code Modulation . . . . . . . . . . . . . . .
15.2.1 Basic Dierential Pulse Code Modulation . . . . . . .
15.2.2 Intra/Inter-Frame Dierential Pulse Code Modulation
15.2.3 Adaptive Dierential Pulse Code Modulation . . . . .
15.3 Block Truncation Coding . . . . . . . . . . . . . . . . . . . .
15.3.1 The Block Truncation Algorithm . . . . . . . . . . . .
15.3.2 Block Truncation Codec Implementations . . . . . . .
15.3.3 Intra-Frame Block Truncation Coding . . . . . . . . .
15.3.4 Inter-Frame Block Truncation Coding . . . . . . . . .
15.4 Subband Coding . . . . . . . . . . . . . . . . . . . . . . . . .
15.4.1 Perfect Reconstruction Quadrature Mirror Filtering .
15.4.1.1 Analysis Filtering . . . . . . . . . . . . . . .
15.4.1.2 Synthesis Filtering . . . . . . . . . . . . . . .
15.4.1.3 Practical QMF Design Constraints . . . . . .
15.4.2 Practical Quadrature Mirror Filters . . . . . . . . . .
15.5 Run-Length-Based Intra-Frame Subband Coding . . . . . . .
15.5.1 Max-Lloyd-Based Subband Coding . . . . . . . . . . .
15.6 Summary and Conclusions . . . . . . . . . . . . . . . . . . . .
16 High-Resolution DCT Coding

16.1
16.2
16.3
16.4
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . .
Intra-Frame Quantizer Training . . . . . . . . . . . . . . . .
Motion Compensation for High-Quality Images . . . . . . .
Inter-Frame DCT Coding . . . . . . . . . . . . . . . . . . .
16.4.1 Properties of the DCT transformed MCER . . . . .
16.4.2 Joint Motion Compensation and Residual Encoding
16.5 The Proposed Codec . . . . . . . . . . . . . . . . . . . . . .
16.5.1 Motion Compensation . . . . . . . . . . . . . . . . .
16.5.2 The Inter/Intra-DCT Codec . . . . . . . . . . . . . .
16.5.3 Frame Alignment . . . . . . . . . . . . . . . . . . . .
16.5.4 Bit-Allocation . . . . . . . . . . . . . . . . . . . . . .
16.5.5 The Codec Performance . . . . . . . . . . . . . . . .
16.5.6 Error Sensitivity and Complexity . . . . . . . . . . .
16.6 Summary and Conclusions . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
603
603
608
608
610
611
613
613
614
615
617
618
620
620
623
624
627
630
633
637
639
639
639
644
650
650
657
658
660
661
662
665
666
667
669
IV Video Systems Based on Standard Video Codecs
673
17 An ARQ-Assisted H.261-Based Recongurable Multilevel

Videophone System
675
17.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 675

17.2 The H.261 Video Coding Standard . . . . . . . . . . . . . . . . . . . . 675
17.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 675
xviii
17.3
17.4
17.5
17.6
CONTENTS
17.2.2 Source Encoder . . . . . . . . . . . . . . . . . . . . . . . . . . .

17.2.3 Coding Control . . . . . . . . . . . . . . . . . . . . . . . . . . .
17.2.4 Video Multiplex Coder . . . . . . . . . . . . . . . . . . . . . . .
17.2.4.1 Picture Layer . . . . . . . . . . . . . . . . . . . . . . .
17.2.4.2 Group of Blocks Layer . . . . . . . . . . . . . . . . . .
17.2.4.3 Macroblock Layer . . . . . . . . . . . . . . . . . . . .
17.2.4.4 Block Layer . . . . . . . . . . . . . . . . . . . . . . . .
17.2.5 Simulated Coding Statistics . . . . . . . . . . . . . . . . . . . .
17.2.5.1 Fixed-Quantizer Coding . . . . . . . . . . . . . . . . .
17.2.5.2 Variable Quantizer Coding . . . . . . . . . . . . . . .
Eect of Transmission Errors on the H.261 Codec . . . . . . . . . . . .
17.3.1 Error Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . .
17.3.2 Error Control Mechanisms . . . . . . . . . . . . . . . . . . . . .
17.3.2.1 Background . . . . . . . . . . . . . . . . . . . . . . . .
17.3.2.2 Intra-Frame Coding . . . . . . . . . . . . . . . . . . .
17.3.2.3 Automatic Repeat Request . . . . . . . . . . . . . . .
17.3.2.4 Recongurable Modulations Schemes . . . . . . . . .
17.3.2.5 Combined Source/Channel Coding . . . . . . . . . . .
17.3.3 Error Recovery . . . . . . . . . . . . . . . . . . . . . . . . . . .
17.3.4 Eects of Errors . . . . . . . . . . . . . . . . . . . . . . . . . .
17.3.4.1 Qualitative Eect of Errors on H.261 Parameters . . .
17.3.4.2 Quantitative Eect of Errors on a H.261 Data Stream
17.3.4.2.1 Errors in an Intra-Coded Frame . . . . . . .
17.3.4.2.2 Errors in an Inter-Coded Frame . . . . . . .
17.3.4.2.3 Errors in Quantizer Indices . . . . . . . . . .
17.3.4.2.4 Errors in an Inter-Coded Frame with Motion
Vectors . . . . . . . . . . . . . . . . . . . . .
17.3.4.2.5 Errors in an Inter-Coded Frame at Low Rate
A Wireless Recongurable Videophone System . . . . . . . . . . . . .
17.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17.4.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17.4.3 Bit-Rate Reduction of the H.261 Codec . . . . . . . . . . . . .
17.4.4 Investigation of Macroblock Size . . . . . . . . . . . . . . . . .
17.4.5 Error Correction Coding . . . . . . . . . . . . . . . . . . . . . .
17.4.6 Packetization Algorithm . . . . . . . . . . . . . . . . . . . . . .
17.4.6.1 Encoding History List . . . . . . . . . . . . . . . . . .
17.4.6.2 Macroblock Compounding . . . . . . . . . . . . . . .
17.4.6.3 End of Frame Eect . . . . . . . . . . . . . . . . . . .
17.4.6.4 Packet Transmission Feedback . . . . . . . . . . . . .
17.4.6.5 Packet Truncation and Compounding Algorithms . .
H.261-Based Wireless Videophone System Performance . . . . . . . . .
17.5.1 System Architecture . . . . . . . . . . . . . . . . . . . . . . . .
17.5.2 System Performance . . . . . . . . . . . . . . . . . . . . . . . .
Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . .
676
679
680
680
681
683
684
686
687
689
692
692
693
693
693
693
694
694
695
696
696
699
700
702
705
705
708
710
710
710
711
711
715
715
716
717
719
720
720
721
721
725
731
CONTENTS
18 Comparison of the H.261 and H.263 Codecs
18.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18.2 The H.263 Coding Algorithms . . . . . . . . . . . . . . . . . . .
18.2.1 Source Encoder . . . . . . . . . . . . . . . . . . . . . . .
18.2.1.1 Prediction . . . . . . . . . . . . . . . . . . . .
18.2.1.2 Motion Compensation and Transform Coding .
18.2.1.3 Quantization . . . . . . . . . . . . . . . . . . .
18.2.2 Video Multiplex Coder . . . . . . . . . . . . . . . . . . .
18.2.2.1 Picture Layer . . . . . . . . . . . . . . . . . . .
18.2.2.2 Group of Blocks Layer . . . . . . . . . . . . . .
18.2.2.3 H.261 Macroblock Layer . . . . . . . . . . . .
18.2.2.4 H.263 Macroblock Layer . . . . . . . . . . . .
18.2.2.5 Block Layer . . . . . . . . . . . . . . . . . . . .
18.2.3 Motion Compensation . . . . . . . . . . . . . . . . . . .
18.2.3.1 H.263 Motion Vector Predictor . . . . . . . . .
18.2.3.2 H.263 Subpixel Interpolation . . . . . . . . . .
18.2.4 H.263 Negotiable Options . . . . . . . . . . . . . . . . .
18.2.4.1 Unrestricted Motion Vector Mode . . . . . . .
18.2.4.2 Syntax-Based Arithmetic Coding Mode . . . .
18.2.4.2.1 Arithmetic coding [562] . . . . . . . .
18.2.4.3 Advanced Prediction Mode . . . . . . . . . . .
18.2.4.3.1 Four Motion Vectors per Macroblock
18.2.4.3.2 Overlapped Motion Compensation for
Luminance . . . . . . . . . . . . . . .
18.2.4.4 P-B Frames Mode . . . . . . . . . . . . . . . .
18.3 Performance Results . . . . . . . . . . . . . . . . . . . . . . . .
18.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . .
18.3.2 H.261 Performance . . . . . . . . . . . . . . . . . . . . .
18.3.3 H.261/H.263 Performance Comparison . . . . . . . . . .
18.3.4 H.263 Codec Performance . . . . . . . . . . . . . . . . .
18.3.4.1 Gray-Scale versus Color Comparison . . . . . .
18.3.4.2 Comparison of QCIF Resolution Color Video .
18.3.4.3 Coding Performance at Various Resolutions . .
18.4 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . .
19 A H.263 Videophone System for Use over Mobile Channels
xix
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
19.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19.2 H.263 in a Mobile Environment . . . . . . . . . . . . . . . . . . . . . .
19.2.1 Problems of Using H.263 in a Mobile Environment . . . . . . .
19.2.2 Possible Solutions for Using H.263 in a Mobile Environment . .
19.2.2.1 Coding Video Sequences Using Exclusively Intra-Coded
Frames . . . . . . . . . . . . . . . . . . . . . . . . . .
19.2.2.2 Automatic Repeat Requests . . . . . . . . . . . . . .
19.2.2.3 Multimode Modulation Schemes . . . . . . . . . . . .
19.2.2.4 Combined Source/Channel Coding . . . . . . . . . . .
19.3 Design of an Error-Resilient Recongurable Videophone System . . . .
733
733
735
735
735
735
736
736
738
738
739
740
744
745
745
746
747
747
749
749
751
752
753
754
757
757
758
761
764
765
767
768
776
777
777
777
777
778
779
779
779
780
781
xx
CONTENTS
19.3.1
19.3.2
19.3.3
19.3.4
19.3.5
19.3.6
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Controling the Bit Rate . . . . . . . . . . . . . . . . . . . . . .
Employing FEC Codes in the Videophone System . . . . . . .
Transmission Packet Structure . . . . . . . . . . . . . . . . . .
Coding Parameter History List . . . . . . . . . . . . . . . . . .
The Packetization Algorithm . . . . . . . . . . . . . . . . . . .
19.3.6.1 Operational Scenarios of the Packetizing Algorithm .
19.4 H.263-Based Video System Performance . . . . . . . . . . . . . . . . .
19.4.1 System Environment . . . . . . . . . . . . . . . . . . . . . . . .
19.4.2 Performance Results . . . . . . . . . . . . . . . . . . . . . . . .
19.4.2.1 Error-Free Transmission Results . . . . . . . . . . . .
19.4.2.2 Eect of Packet Dropping on Image Quality . . . . .
19.4.2.3 Image Quality versus Channel Quality without ARQ .
19.4.2.4 Image Quality versus Channel Quality with ARQ . .
19.4.3 Comparison of H.263 and H.261-Based Systems . . . . . . . . .
19.4.3.1 Performance with Antenna Diversity . . . . . . . . . .
19.4.3.2 Performance over DECT Channels . . . . . . . . . . .
19.5 Transmission Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . .
19.5.1 ARQ Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19.5.2 Implementation of Transmission Feedback . . . . . . . . . . . .
19.5.2.1 Majority Logic Coding . . . . . . . . . . . . . . . . .
20 Error Rate Based Power Control
20.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20.2 Power Control Algorithm . . . . . . . . . . . . . . . . . . . . . .
20.3 Performance of the Power Control . . . . . . . . . . . . . . . . .
20.3.1 Frame Error Rate Performance . . . . . . . . . . . . . . .
20.3.2 Signal-to-Interference Ratio Performance . . . . . . . . .
20.3.3 Signal-to-Interference-Plus-Noise Ratio Performance . . .
20.4 Multimode Performance . . . . . . . . . . . . . . . . . . . . . . .
20.5 Average Transmission Power . . . . . . . . . . . . . . . . . . . .
20.6 Optimization of Power Control Parameters . . . . . . . . . . . .
20.6.1 Joint Optimization of IPC and DPC Parameters . . . . .
20.6.2 Joint Optimization of NEF and NFE . . . . . . . . . . . .
20.6.3 Joint Optimization of IPSS and DPSS . . . . . . . . . . .
20.6.4 Conclusions from Optimizing the Power Control
Algorithm Parameters . . . . . . . . . . . . . . . . . . . .
20.7 Power Control Performance at Various Speeds . . . . . . . . . . .
20.7.1 Power Control Results for Pedestrians . . . . . . . . . . .
20.7.2 Channel Fading . . . . . . . . . . . . . . . . . . . . . . . .
20.7.3 Tracking of Slow Fading . . . . . . . . . . . . . . . . . . .
20.7.4 Power Control Error . . . . . . . . . . . . . . . . . . . . .
20.8 Multiple Interferers . . . . . . . . . . . . . . . . . . . . . . . . . .
20.8.1 Frame Error Rate Performance . . . . . . . . . . . . . . .
20.8.2 Further Eects of Power Control on System Performance
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
781
782
784
785
786
787
788
790
790
792
792
793
795
796
798
800
802
806
811
811
812
816
819
819
819
824
824
828
828
832
834
838
839
841
843
844
845
845
845
848
851
855
855
857
CONTENTS
xxi
21 Adaptive Single-Carrier, Multicarrier, and CDMA-based Video

Systems
21.1 Turbo-equalised H.263-based videophony for GSM/GPRS . . . . . . .

21.1.1 Motivation and Background . . . . . . . . . . . . . . . . . . . .
21.1.2 System Parameters . . . . . . . . . . . . . . . . . . . . . . . . .
21.1.3 Turbo Equalization . . . . . . . . . . . . . . . . . . . . . . . . .
21.1.4 Turbo-equalization Performance . . . . . . . . . . . . . . . . .
21.1.4.1 Video Performance . . . . . . . . . . . . . . . . . . . .
21.1.4.2 Bit Error Statistics . . . . . . . . . . . . . . . . . . .
21.1.5 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . .
21.2 Adaptive QAM-based Wireless Videophony . . . . . . . . . . . . . . .
21.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21.2.2 Adaptive Video Transceiver . . . . . . . . . . . . . . . . . . . .
21.2.3 Burst-by-Burst Adaptive Videophone Performance . . . . . . .
21.2.4 Switching Thresholds . . . . . . . . . . . . . . . . . . . . . . .
21.2.5 Turbo-coded Video Performance . . . . . . . . . . . . . . . . .
21.3 UMTS-like Burst-by-burst Adaptive CDMA Videophony . . . . . . . .
21.3.1 Motivation and Video Transceiver Overview . . . . . . . . . . .
21.3.2 Multimode Video System Performance . . . . . . . . . . . . . .
21.3.3 Burst-by-Burst Adaptive Videophone System . . . . . . . . . .
21.4 H.263/OFDM-Based Video Systems for Frequency-Selective Wireless
Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21.4.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21.4.2 System Overview . . . . . . . . . . . . . . . . . . . . . . . . . .
21.4.2.1 The WATM System . . . . . . . . . . . . . . . . . . .
21.4.2.2 The UMTS-type Framework . . . . . . . . . . . . . .
21.4.3 The Channel Model . . . . . . . . . . . . . . . . . . . . . . . .
21.4.4 Video-Related System Aspects . . . . . . . . . . . . . . . . . .
21.4.4.1 Video Parameters of the WATM System . . . . . . .
21.4.4.2 Video parameters of the UMTS scheme . . . . . . . .
21.4.5 System Performance . . . . . . . . . . . . . . . . . . . . . . . .
21.5 Adaptive Turbo-coded OFDM-Based Videotelephony . . . . . . . . . .
21.5.1 Motivation and Background . . . . . . . . . . . . . . . . . . . .
21.5.2 AOFDM Modem Mode Adaptation and Signaling . . . . . . .
21.5.3 AOFDM Subband BER Estimation . . . . . . . . . . . . . . .
21.5.4 Video Compression and Transmission Aspects . . . . . . . . . .
21.5.5 Comparison of Subband-Adaptive OFDM and Fixed Mode
OFDM Transceivers . . . . . . . . . . . . . . . . . . . . . . . .
21.5.6 Subband-Adaptive OFDM Transceivers Having Dierent
Target Bit Rates . . . . . . . . . . . . . . . . . . . . . . . . . .
21.5.7 Time-Variant Target Bit Rate OFDM Transceivers . . . . . . .
861
861
861
862
865
868
869
873
875
875
875
877
879
886
890
892
894
894
899
902
908
908
908
909
913
916
917
918
918
921
922
926
927
927
929
929
930
930
936
941
xxii
CONTENTS

21.6 Digital Terrestrial Video Broadcasting for Mobile Receivers . . . . . .
21.6.1 Background and Motivation . . . . . . . . . . . . . . . . . . . .
21.6.2 MPEG-2 Bit Error Sensitivity . . . . . . . . . . . . . . . . . . .
21.6.3 DVB Terrestrial Scheme . . . . . . . . . . . . . . . . . . . . . .
21.6.4 Terrestrial Broadcast Channel Model . . . . . . . . . . . . . . .
21.6.5 Data Partitioning Scheme . . . . . . . . . . . . . . . . . . . . .
21.6.6 Performance of the Data Partitioning Scheme . . . . . . . . . .
21.6.7 Nonhierarchical OFDM DVBP Performance . . . . . . . . . . .
21.6.8 Hierarchical OFDM DVB Performance . . . . . . . . . . . . . .
21.7 Satellite-Based Video Broadcasting . . . . . . . . . . . . . . . . . . . .
21.7.1 Background and Motivation . . . . . . . . . . . . . . . . . . . .
21.7.2 DVB Satellite Scheme . . . . . . . . . . . . . . . . . . . . . . .
21.7.3 Satellite Channel Model . . . . . . . . . . . . . . . . . . . . . .
21.7.4 The Blind Equalizers . . . . . . . . . . . . . . . . . . . . . . . .
21.7.5 Performance of the DVB Satellite Scheme . . . . . . . . . . . .
21.7.5.1 Transmission over the Symbol-Spaced Two-Path
Channel . . . . . . . . . . . . . . . . . . . . . . . . . .
21.7.5.2 Transmission over the Two-Symbol Delay Two-Path
Channel . . . . . . . . . . . . . . . . . . . . . . . . . .
21.7.5.3 Performance Summary of the DVB-S System . . . . .
21.9 Wireless Video System Design Principles . . . . . . . . . . . . . . . . .
950
950
950
951
962
965
966
972
983
990
993
996
996
997
999
999
1003
1003
1010
1010
1017
1018
1020
Glossary
1023
Bibliography
1033
Subject Index
1065
Author Index
1065
About the Authors
1065
Preface
The Wireless Multimedia Communications Scene
Against the backdrop of the emerging third-generation wireless personal communications standards and broadband access network standard proposals, this book is
dedicated to a range of topical wireless video communications aspects. The transmission of multimedia information over wireline based links can now be considered a
mature area, where a range of interactive and distributive services are oered by various providers across the globe, such as Integrated Services Digital Network (ISDN)
based on H.261/H.263 assisted videotelephony, video on demand services using the
Motion Pictures Expert Group (MPEG) video compression standards, multimedia
electronic mail, cable television and radio programs, and so on. A range of interactive
mobile multimedia communications services are also realistic in technical terms at the
time of writing, and their variety, quality, as well as market penetration are expected
to exceed those of the wireline-oriented services during the next few years.
The wireless multimedia era is expected to witness a tremendous growth with the
emergence of the third-generation (3G) personal communications networks (PCN) and
wireless asynchronous transfer mode (WATM) systems, which constitute a wireless
extension of the existing ATM networks. All three global 3G PCN standard proposals,
which originated in the United States, Europe, and Japan, are based on Code Division Multiple Access (CDMA) and are capable of transmitting at bit-rates in excess
of 2 Mbps. Furthermore, the European proposal was also designed to support multiple simultaneous calls and services. The WATM solutions often favor Orthogonal
frequency Division Multiple Access (OFDM) as their modulation technique. Indeed,
the Broadband Access Network (BRAN) standard also advocates OFDM. A range
of WATM video aspects and mobile digital video broadcast (DVB) issues are also
reviewed in Part IV of the book.
Research is also well under way toward dening a whole host of new modulation and signal processing techniques, and a further trend is likely to dominate this
new era: namely, the merger of wireless multimedia communications, multimedia consumer electronics, and multimedia computer technologies. These trends are likely to
hallmark the community's future research in the forthcoming years. This book is
naturally limited in terms of its coverage of these aspects, simply because of space
limitations. We endeavored, however, to provide the reader with a broad range of
applications examples, which are pertinent to scenarios, such as transmitting lowlatency interactive video as well as distributive or broadcast video signals over the
existing second-generation (2G) wireless systems, 3G arrangements, and the forthcoming fourth-generation systems. We also characterized the video performance of
xxiii
xxiv
PREFACE
Video
quality
Complexity
Broadcast or
interactive
video
solution
ETC.
Bit rate
Figure 1:
Error
resilience
Delay
Contradictory system design requirements of various video communications systems.
a range of high bit-rate Local Area Network (LAN) type systems as well as various
video broadcast systems, transmitting broadcast-quality video signals to mobile receivers, both within the home and farther aeld, to demanding business customers on
the move.
These enabling technologies facilitate a whole host of wireless services, such as
videotelephony, electronic commerce, city guide, Internet access for games, electronic
mail, and web browsing. Further attractive applications can be found in wireless inhome networks, DVB reception in busses, trains, cars, on board ships, and elsewhere
| for example, using multimedia laptop PCs. Again, the book does not delve into
the area of specic applications; rather it oers a range of technical solutions, which
are applicable to various propagation and application environments.
We hope that the book oers you a range of interesting topics, sampling | and
hopefully without gross aliasing errors, the current state of the art in the associated
enabling technologies. In simple terms, nding a specic solution to a distributive or
interactive video communications problem has to be based on a compromise in terms
of the inherently contradictory constraints of video-quality, bit-rate, delay, robustness
against channel errors, and the associated implementational complexity, as suggested
by Figure 1. Analyzing these trade-os and proposing a range of attractive solutions
to various video communications problems are the basic aims of this book.
Video over Wireless Systems
Over the past decade second-generation (2G) wireless systems have been installed
throughout the world and in some countries about half of the population possesses
a mobile telephone. These systems typically exhibit a higher spectral eciency than
their analog counterparts and oer a signicantly wider range of services, such as data,
PREFACE
Video
Format
xxv
Uncompressed
Bit-Rate (Mbit/s)
10 frame/s
30 frame/s
Gray
Color
Gray
Color
SQCIF
128 x 96
12 288
0.983
1.47
2.95
4.42
QCIF
176 x 144
25 344
2.03
3.04
6.09
9.12
CIF
352 x 288
101 376
8.1
12.2
24.3
36.5
4CIF
704 x 576
405 504
32.4
48.7
97.3
146.0
16CIF
1408 x 1152 1 622 016
129.8
194.6
389.3
583.9
CCIR 601
720 x 480
345 600
27.65
41.472 82.944 124.416
HDTV 1440 1440 x 960 1 382 400
110.592 165.888 331.776 497.664
HDTV
1920 x 1080 2 073 600
165.9
248.832 497.664 746.496
SQCIF: Sub-Quarter Common Intermediate Format
QCIF: Quarter Common Intermediate Format
CIF: Common Intermediate Format
HDTV: High Denition Television
Table 1:
Luminance
Dimensions
No. of Pels
per Frame
Various video formats and their uncompressed bit-rate. Upon using compression
10-100 times lower average bit-rates are realistic.
fax, e-mail, short messages, and high-speed circuit switched data. However, because of
their relatively low bit-rates, the provision of interactive wireless videotelephony has
been hindered. Potentially there are two dierent options for transmitting video over
the 2G systems, namely, over their data channel, or | provided that the standards
can be amended accordingly | by allocating an additional speech channel for video
transmissions. Considering the second option rst, the low-rate speech channel of the
2G systems constrains the achievable bit-rate to such low values that the spatial video
resolution supported is limited to the 174 144-pixel Quarter Common Intermediate
Format (QCIF) or to the 128 96-pixel Sub-QCIF (SQCIF) at a video-frame scanning
rate of 5 to 10 frames/s.
The range of standard video formats is summarized in Table 1, along with their
uncompressed bit-rates at frame scanning rates of both at 10 and 30 frames/sec for
both gray and color video signals. This table indicates the extremely wide range of
potential bit-rate requirements. Clearly, the higher resolution formats can be realistically used only, for example, in the context of high-rate WATM systems.
The Cordless Telephone (CT) schemes of the second generation typically have a 32
kbit/s speech rate, which is more readily amenable to interactive videotelephony. For
the sake of supporting a larger video-frame size, such as the 352 x 288-pixel Common
Intermediate Format (QCIF), higher bit-rates must be supported, which is possible
over the Digital European Cordless Telecommunications (DECT) system upon linking
a number of slots at a rate in excess of 500 kbps.
By contrast, the data channel of the 2G systems can often oer a higher data rate
than that of the speech channel, for example, by linking a number of time-slots, as was
proposed in the DECT scheme or in the high-speed circuit switched data (HSCSD)
mode of the Global System of Mobile communications known as GSM. CT schemes
typically refrain from invoking channel coding, since they typically operate over benign
xxvi
PREFACE
channels. Hence, they do not employ channel interleavers, which is advantageous

in video delay terms but disadvantageous in terms of error resilience. The data
transmission mode of cellular systems, however, typically exhibits a high interleaving
delay, which helps increase the system's robustness against channel errors. This is
advantageous in terms of reducing the channel-induced video impairments but may
result in \lip-synchronization" problems between the speech and video output signals.
Both the speech and data channels of the 2G systems tend to support a xed
constant bit-rate. However, the existing standard video codecs, such as the H.263
and MPEG2 codecs, generate a time-variant bit-rate. This is because they endeavor
to reduce the bit-rate to near the lowest possible bit-rate constituted by the entropy of
the source signal. Since this is achieved by invoking high-compression variable-length
coding schemes, their time-variant bitstream becomes very sensitive to transmission
errors. In fact, a single transmission error may render the video-quality of an entire
video-frame unacceptable. Hence, the existing standard-based video codecs, such as
the H.263 and MPEG2 schemes, require ecient system-level transport solutions in
order to address the above-mentioned deciencies. (This issue will be discussed in
more depth in Part IV of the book). An alternative solution is invoking constant-rate
proprietary video codecs, which | to a degree | sacrice compression eciency for
the sake of an increased robustness against channel errors. This philosophy is pursued
in Part II of the book, which relies on much of the compression and communications
theory, as well as on the various error correction coding and transmission solutions
presented in Part I.
At the time of writing, the standardization of 3G systems is approaching completion in Europe, the United States, and Japan. These systems, which are characterized
in Part I of the book, along with their 2G counterparts, were designed to further enrich the range of services supported. They are more amenable to interactive wireless
videotelephony, for example, than their 2G conterparts. This book aims to propose
a range of video system solutions that bridge the evolutionary avenue between the
second- and third-generation wireless systems.
Part I of the book provides an overview of the whole range of associated transmission aspects of the various video systems proposed and investigated. Specically,
Chapter 1 summarizes the necessary background on information, compression, and
communications theory. This is followed by the characterization of wireless channels
in Chapter 2. The impairments produced by these channels can be counteracted by
the channel codecs of Chapters 3 and 4. Various modulation and transmission schemes
are the topic of Chapter 5. We then provide a discussion of video trac modeling
and evaluate the proposed model's performance in the context of various statistical
multiplexing and multiple access schemes in Chapter 6. The eects of co-channel
interferences | which represent the most dominant performance-limiting factor of
multiple access-based cellular systems | are described in Chapter 7. Dynamic channel allocation schemes | which rely on knowledge of co-channel interference and the
multiple access scheme employed | are the topic of Chapter 8. The video transmission capabilities of 2G wireless systems are discussed in Chapter 9. These elaborations
are followed by an in-depth discussion of various CDMA schemes in Chapter 10, including a variety of novel residual number system-based CDMA schemes and on the
global 3G CDMA proposals, which concludes Part I of the book.
PREFACE
xxvii
Part II is dedicated to a host of xed but arbitrarily programmable rate video

codecs based on fractal coding, on the discrete cosine transform (DCT), on vector
quantized (VQ) codecs, and quad-tree-based codecs. These video codecs and their
associated quadrature amplitude modulated (QAM) video systems are described in
Chapters 11{14. Part III book focusses on high-resolution video coding, encompassing
Chapters 15 and 16.
Part IV presents Chapters 17{21, which characterize the H.261 and H.263 video
codecs, constituting one of the most important representatives of the family of stateof-the-art hybrid DCT codecs. Hence, the associated ndings of these chapters can
be readily applied in the context of other hybrid DCT codecs, such as the MPEG
family, including the MPEG2 and MPEG4 codecs. Chapters 17{21 also portray the
interactions of these hybrid DCT video codecs with recongurable multimode QAM
transceivers. The book concludes in Chapter 21 by oering a range of system design studies related to wideband burst-by-burst adaptive TDMA/TDD, OFDM, and
CDMA interactive as well as distributive mobile video systems and their performance
characterization over highly dispersive transmission media.
Motivation
The rationale of this book has already been outlined from a technical perspective.
Another important motivation of the book is to bring together two seemingly independent research communities, namely, the video compression and the wireless
communications communities, by bridging the philosophical dierence between them.
These philosophical dierences are based partially on the contradictory requirements
portrayed and discussed in the context of Figure 1. Specically, while a range of
exciting developments have taken place in both the image compression and wireless
communications communities, most of the video compression research has been cast
in the context of wireline-based communications systems, such as ISDN and ATM
links. These communications systems typically exhibit a low bit error rate (BER)
and a low packet- or cell-loss rate. For example, ATM systems aim for a cell-loss rate
of 10 9. Thus, the error-resilience requirements of the video codecs were extremely
relaxed.
In the increasingly pervasive wireless era, however, such extreme transmission
integrity requirements are simply unrealistic, because they impose unreasonable constraints on the design of wireless systems, such as, for example, WATM systems. For
example, the ATM cell-loss rate of 10 9 could only be maintained over wireless links
at a high implementational cost, potentially invoking Automatic Repeat Requests
(ARQs). ARQs, however, would increase the system delay, potentially precluding
real-time interactive video communications, unless innovative design principles are
invoked. Again, all these trade-os are the subject of this book.
Part I of the book seeks to provide background for readers who require an overview
of wireless communications, potentially, for example, video compression experts. Part II
assumes a sound knowledge of the issues treated in Part I of the book, while oering
an eortless introduction to the associated video compression aspects. Wireless experts may therefore skip Part I and begin reading Part II of the book. Part III focuses
xxviii
PREFACE
exclusively on video compression. Hopefully, readers from both the video compression
and wireless communications communities will nd Part IV informative and fun to
read, since it integrates the knowledge base of both elds, aiming to design improved
video systems.
It is our hope that the book underlines the range of contradictory system design
trade-os in an unbiased fashion and that you will be able to glean information from it
in order to solve your own particular wireless video communications problem. Most of
all however we hope that you will nd it an enjoyable and relatively eortless reading,
providing you with intellectual stimulation.
Lajos Hanzo, Peter J. Cherriman, and Jurgen Streit
Department of Electronics and Computer Science
University of Southampton
Acknowledgments
We are indebted to our many colleagues who have enhanced our understanding of the
subject, in particular to Prof. Emeritus Raymond Steele. These colleagues and valued
friends, too numerous to be mentioned, have in uenced our views concerning various
aspects of wireless multimedia communications. We thank them for the enlightenment
gained from our collaborations on various projects, papers, and books. We are grateful
to Jan Brecht, Jon Blogh, Marco Breiling, Marco del Buono, Clare Sommerville,
Stanley Chia, Byoung Jo Choi, Joseph Cheung, Peter Fortune, Sheyam Lal Dhomeja,
Lim Dongmin, Dirk Didascalou, Stephan Ernst, Eddie Green, David Greenwood, Hee
Thong How, Thomas Keller, W. H. Lam, C. C. Lee, M. A. Nofal, Xiao Lin, Chee
Siong Lee, Tong-Hooi Liew, Matthias Muenster, Vincent Roger-Marchart, Redwan
Salami, David Stewart, Je Torrance, Spyros Vlahoyiannatos, William Webb, John
Williams, Jason Woodard, Choong Hin Wong, Henry Wong, James Wong, Lie-Liang
Yang, Bee-Leong Yeap, Mong-Suan Yee, Kai Yen, Andy Yuen, and many others with
whom we enjoyed an association.
We also acknowledge our valuable associations with the Virtual Centre of Excellence in Mobile Communications, in particular with its chief executives, Dr. Tony
Warwick and Dr. Walter Tuttlebee, and other members of its Executive Committee,
namely Dr. Keith Baughan, Prof. Hamid Aghvami, Prof. Ed Candy, Prof. John
Dunlop, Prof. Barry Evans, Dr. Mike Barnard, Prof. Joseph McGeehan, Prof. Peter Ramsdale and many other valued colleagues. Our sincere thanks are also due
to the EPSRC, UK for supporting our research. We would also like to thank Dr.
Joao Da Silva, Dr Jorge Pereira, Bartholome Arroyo, Bernard Barani, Demosthenes
Ikonomou, and other valued colleagues from the Commission of the European Communities, Brussels, Belgium, as well as Andy Aftelak, Mike Philips, Andy Wilton,
Luis Lopes, and Paul Crichton from Motorola ECID, Swindon, UK, for sponsoring
some of our recent research. Further thanks are due to Tim Wilkinson at HP in
Bristol for funding some of our research eorts.
We feel particularly indebted to Chee Siong Lee for his invaluable help with proofreading the manuscript and to Rita Hanzo as well as Denise Harvey for their skillful
assistance in typesetting the manuscript in Latex. Similarly, our sincere thanks are
due to Ken Moore, Cathy Faduska, Surendra Bhimani, Linda Matarazzo and a number of other sta members of IEEE Press for their kind assistance throughout the
preparation of the camera-ready manuscript. The valuable support of Betty Pessagno and Suzanne Ingrao in copy editing and proof reading the manuscript is also
gratefully acknowledged. We are also indebted to Professor John B. Anderson, the
series editor, who incorporated the book in his prestigious IEEE communications series. Finally, our sincere gratitude is due to the numerous authors listed in the Author
xxix
xxx
ACKNOWLEDGMENTS
Index | as well as to those, whose work was not cited due to space limitations | for
their contributions to the state of the art, without whom this book would not have
materialized.
Lajos Hanzo, Peter J. Cherriman, and Jurgen Streit
Department of Electronics and Computer Science
University of Southampton
Contributors
Video Trac Modelling and Multiple Access;

J. Brecht, M. del Buono, and L. Hanzo
Chapter 10: CDMA Systems: Third-Generation and Beyond;
K. Yen and L. Hanzo
Chapter 21: Adaptive Single-Carrier, Multicarrier, and CDMA-Based Video Systems;
P. Cherriman, L. Hanzo, B.L. Yeap, C.H. Wong, E.L. Kuan, T. Keller,
C.S. Lee, S. Vlahoyiannatos
Chapter 6:
Chapter 21 is partially based on our research sponsored by the Virtual Center of

Excellence in Mobile and Personal Communications (MVCE), UK.
xxxi
Part I
Transmission Issues
Chapter
Information Theory
1.1
Issues in Information Theory
The ultimate aim of telecommunications is to communicate information between two

geographically separated locations via a communications channel with adequate quality. The theoretical foundations of information theory accrue from Shannon's pioneering work [1{4], and hence most tutorial interpretations of his work over the past
fty years rely fundamentally on [1{4]. This chapter is no exception in this respect.
Throughout this chapter we make frequent references to Shannon's seminal papers and
to the work of various authors oering further insights into Shannonian information
theory. Since this monograph aims to provide an all-encompassing coverage of video
compression and communications, we begin by addressing the underlying theoretical
principles using a light-hearted approach, often relying on worked examples.
Early forms of human telecommunications were based on smoke, drum or light
signals, bonres, semaphores, and the like. Practical information sources can be
classied as analog and digital. The output of an analog source is a continuous
function of time, such as, for example, the air pressure variation at the membrane of
a microphone due to someone talking. The roots of Nyquist's sampling theorem are
based on his observation of the maximum achievable telegraph transmission rate over
bandlimited channels [5]. In order to be able to satisfy Nyquist's sampling theorem
the analogue source signal has to be bandlimited before sampling. The analog source
signal has to be transformed into a digital representation with the aid of time- and
amplitude-discretization using sampling and quantization.
The output of a digital source is one of a nite set of ordered, discrete symbols
often referred to as an alphabet. Digital sources are usually described by a range of
characteristics, such as the source alphabet, the symbol rate, the symbol probabilities,
and the probabilistic interdependence of symbols. For example, the probability of u
following q in the English language is p = 1, as in the word \equation." Similarly,
the joint probability of all pairs of consecutive symbols can be evaluated.
In recent years, electronic telecommunications have become prevalent, although
3
CHAPTER 1. INFORMATION THEORY
most information sources provide information in other forms. For electronic telecommunications, the source information must be converted to electronic signals by a
transducer. For example, a microphone converts the air pressure waveform p(t) into
voltage variation v(t), where
v(t) = c p(t );
(1.1)
and the constant c represents a scaling factor, while is a delay parameter. Similarly,
a video camera scans the natural three-dimensional scene using optics and converts
it into electronic waveforms for transmission.
The electronic signal is then transmitted over the communications channel and
converted back to the required form, which may be carried out, for example, by a
loudspeaker. It is important to ensure that the channel conveys the transmitted signal with adequate quality to the receiver in order to enable information recovery.
Communications channels can be classied according to their ability to support analog or digital transmission of the source signals in a simplex, duplex, or half-duplex
fashion over xed or mobile physical channels constituted by pairs of wires, Time Division Multiple Access (TDMA) time-slots, or a Frequency Division Multiple Access
(FDMA) frequency slot.
The channel impairments may include superimposed, unwanted random signals,
such as thermal noise, crosstalk via multiplex systems from other users, man-made
interference from car ignition, uorescent lighting, and other natural sources such as
lightning. Just as the natural sound pressure wave between two conversing persons
will be impaired by the acoustic background noise at a busy railway station, similarly
the reception quality of electronic signals will be aected by the above unwanted
electronic signals. In contrast, distortion manifests itself dierently from additive
noise sources, since no impairment is explicitly added. Distortion is more akin to
the phenomenon of reverberating loudspeaker announcements in a large, vacant hall,
where no noise sources are present.
Some of the channel impairments can be mitigated or counteracted; others cannot.
For example, the eects of unpredictable additive random noise cannot be removed or
\subtracted" at the receiver. Its eects can be mitigated by increasing the transmitted
signal's power, but the transmitted power cannot be increased without penalties, since
the system's nonlinear distortion rapidly becomes dominant at higher signal levels.
This process is similar to the phenomenon of increasing the music volume in a car
parked near a busy road to a level where the amplier's distortion becomes annoyingly
dominant.
In practical systems, the Signal-to-Noise Ratio (SNR) quantifying the wanted and
unwanted signal powers at the channel's output is a prime channel parameter. Other
important channel parameters are its amplitude and phase response, determining its
usable bandwidth (B ), over which the signal can be transmitted without excessive
distortion. Among the most frequently used statistical noise properties are the probability density function (PDF), cumulative density function (CDF), and power spectral
density (PSD).
The fundamental communications system design considerations are whether a
high-delity (HI-FI) or just acceptable video or speech quality is required from a system, which predetermines, among other factors, its cost, bandwidth requirements, as
1.1. ISSUES IN INFORMATION THEORY
Figure 1.1:
Basic transmission model of information theory.
well as the number of channels available, and has implementational complexity ramications. Equally important are the issues of robustness against channel impairments,
system delay, and so on. The required transmission range and worldwide roaming
capabilities, the maximum available transmission speed in terms of symbols/sec, information condentiality, reception reliability, convenient lightweight, solar-charged
design, are similarly salient characteristics of a communications system.
Information theory deals with a variety of problems associated with the performance limits of the information transmission system, such as that depicted in Figure 1.1. The components of this system constitute the subject of this monograph;
hence they will be treated in greater depth later in this volume. Suce it to say at
this stage that the transmitter seen in Figure 1.1 incorporates a source encoder, a
channel encoder, an interleaver, and a modulator and their inverse functions at the
receiver. The ideal source encoder endeavors to remove as much redundancy as possible from the information source signal without aecting its source representation
delity (i.e., distortionlessly), and it remains oblivious of such practical constraints as
a nite delay and limited signal processing complexity. In contrast, a practical source
encoder will have to retain a limited signal processing complexity and delay while
attempting to reduce the source representation bit rate to as low a value as possible.
This operation seeks to achieve better transmission eciency, which can be expressed
in terms of bit-rate economy or bandwidth conservation.
The channel encoder re-inserts redundancy or parity information but in a controlled manner in order to allow error correction at the receiver. Since this component is designed to ensure the best exploitation of the re-inserted redundancy, it is
expected to minimize the error probability over the most common channel, namely,
the so-called Additive White Gaussian Noise (AWGN) channel, which is characterized by a memoryless, random distribution of channel errors. However, over wireless
channels, which have recently become prevalent, the errors tend to occur in bursts
due to the presence of deep received signal fades induced by the distructively superimposed multipath phenomena. This is why our schematic of Figure 1.1 contains an
interleaver block, which is included in order to randomize the bursty channel errors.
Finally, the modulator is designed to ensure the most bandwidth-ecient transmission
of the source- and channel encoded, interleaved information stream, while maintaining
the lowest possible bit error probability. The receiver simply carries out the corresponding inverse functions of the transmitter. Observe in the gure that besides the
direct interconnection of the adjacent system components there are a number of additional links in the schematic, which will require further study before their role can be
highlighted. Thus, at the end of this chapter we will return to this gure and guide
the reader through its further intricate details.
Some fundamental problems transpiring from the schematic of Figure 1.1, which
were addressed in depth by a range of references due to Shannon [1{4], Nyquist [5],
Hartley [6], Abramson [7], Carlson [8], Raemer [9], and Ferenczy [10] and others are
as follows:
What is the true information generation rate of our information sources? If

we know the answer, the eciency of coding and transmission schemes can be
evaluated by comparing the actual transmission rate used with the source's
information emission rate. The actual transmission rate used in practice is
typically much higher than the average information delivered by the source,
and the closer these rates are, the better is the coding eciency.
Given a noisy communications channel, what is the maximum reliable information transmission rate? The thermal noise induced by the random motion of
electrons is present in all electronic devices, and if its power is high, it can seriously aect the quality of signal transmission, allowing information transmission
only at low-rates.
Is the information emission rate the only important characteristic of a source,

or are other message features, such as the probability of occurrence of a message
and the joint probability of occurrence for various messages, also important?
In a wider context, the topic of this whole monograph is related to the blocks of
Figure 1.1 and to their interactions, but in this chapter we lay the theoretical
foundations of source and channel coding as well as transmission issues and
dene the characteristics of an ideal Shannonian communications scheme.
Although numerous excellent treatises are available on these topics, which treat
the same subjects with a dierent avor [8, 10, 11], our approach is similar to that of
1.2. ADDITIVE WHITE GAUSSIAN NOISE CHANNEL
the above classic sources; since the roots are in Shannon's work, references [1{4,12,13]
are the most pertinent and authoritative sources.
In this chapter we consider mainly discrete sources, in which each source message
is associated with a certain probability of occurrence, which might or might not be
dependent on previous source messages. Let us now give a rudimentary introduction
to the characteristics of the AWGN channel, which is the predominant channel model
in information theory due to its simplicity. The analytically less tractable wireless
channels of Chapter 2 will be modeled mainly by simulations in this monograph,
although in Chapter 10 some analytical results are also provided in the context of
Code Division Multiple Access (CDMA) systems.
1.2
Additive White Gaussian Noise Channel
1.2.1 Background
In this section, we consider the communications channel, which exists between the
transmitter and the receiver, as shown in Figure 1.1. Accurate characterization of this
channel is essential if we are to remove the impairments imposed by the channel using
signal processing at the receiver. Here we initially consider only xed communications
links whereby both terminals are stationary, although mobile radio communications
channels, which change signicantly with time, are becoming more prevalent.
We dene xed communications channels to be those between a xed transmitter
and a xed receiver. These channels are exemplied by twisted pairs, cables, wave
guides, optical ber and point-to-point microwave radio channels. Whatever the nature of the channel, its output signal diers from the input signal. The dierence
might be deterministic or random, but it is typically unknown to the receiver. Examples of channel impairments are dispersion, nonlinear distortions, delay, and random
noise.
Fixed communications channels can often be modeled by a linear transfer function, which describes the channel dispersion. The ubiquitous additive Gaussian noise
(AWGN) is a fundamental limiting factor in communications via linear time-invariant
(LTI) channels. Although the channel characteristics might change due to factors such
as aging, temperature changes, and channel switching, these variations will not be apparent over the course of a typical communication session. It is this inherent time
invariance that characterizes xed channels.
An ideal, distortion-free communications channel would have a at frequency response and linear phase response over the frequency range of 1 : : : + 1, although in
practice it is sucient to satisfy this condition over the bandwidth (B ) of the signals
to be transmitted, as seen in Figure 1.2. In this gure, A(!) represents the magnitude
of the channel response at frequency w, and (w) = wT represents the phase shift
at frequency w due to the circuit delay T .
Practical channels always have some linear distortions due to their bandlimited,
non at frequency response and nonlinear phase response. In addition, the group-delay
response of the channel, which is the derivative of the phase response, is often given.

()
()
B
Figure 1.2:
Ideal, distortion-free channel model having a linear phase and a at magnitude

response.
1.2.2 Practical Gaussian Channels

Conventional telephony uses twisted copper wire pairs to connect subscribers to the
local exchange. The bandwidth is approximately 3.4 kHz, and the waveform distortions are relatively benign.
For applications requiring a higher bandwidth, coaxial cables can be used. Their
attenuation increases approximately with the square root of the frequency. Hence,
for wideband, long-distance operation, they require channel equalization. Typically,
coaxial cables can provide a bandwidth of about 50 MHz, and the transmission rate
they can support is limited by the so-called skin eect.
Point-to-point microwave radio channels typically utilize high-gain directional
transmit and receive antennas in a line-of-sight scenario, where free-space propagation
conditions may be applicable.
1.2.3 Gaussian Noise

Regardless of the communications channel used, random noise is always present. Noise
can be broadly classied as natural or man-made. Examples of man-made noise are
those due to electrical appliances, and uorescent lighting, and the eects of these
sources can usually be mitigated at the source. Natural noise sources aecting radio
transmissions include galactic star radiations and atmospheric noise. There exists
a low-noise frequency window in the range of 1{10 GHz, where the eects of these
sources are minimized.
Natural thermal noise is ubiquitous. This is due to the random motion of electrons,
and it can be reduced by reducing the temperature. Since thermal noise contains
practically all frequency components up to some 1013 Hz with equal power, it is often
referred to as white noise (WN) in an analogy to white light containing all colors with
1.2. ADDITIVE WHITE GAUSSIAN NOISE CHANNEL
Figure 1.3:
Power spectral density and autocorrelation of WN.
equal intensity. This WN process can be characterized by its uniform power spectral
density (PSD) N (!) = N0 =2 shown together with its autocorrelation function (ACF)
in Figure 1.3.
The power spectral density of any signal can be conveniently measured by the help
of a selective narrowband power meter tuned across the bandwidth of the signal. The
power measured at any frequency is then plotted against the measurement frequency.
The autocorrelation function R( ) of the signal x(t) gives an average indication of
how predictable the signal x(t) is after a period of seconds from its present value.
Accordingly, it is dened as follows:
Z
1 1
R( ) = Tlim
x(t)x(t + )dt:
(1.2)
!1 T 1
For a periodic signal x(t), it is sucient to evaluate the above equation for a single
period T0, yielding:
Z
1 T0 =2
R( ) =
x(t)x(t + )dt:
(1.3)
T0 T0 =2
The basic properties of the ACF are:

The ACF is symmetric: R( ) = R( ).

The ACF is monotonously decreasing: R( ) R(0).
For = 0 we have R(0) = x2 (t), which is the signal's power.
The ACF and the PSD form a Fourier transform pair, which is formally stated
as the Wiener-Khintchine theorem, as follows:
1
R( ) =
2
1
1
N (!)ej! d!
10

NB(f)
R( )=NoB.sinc(2 B )
No
2
1
2B
-B
Figure 1.4:
Power spectral density and autocorrelation of bandlimited WN.

Z
1 1 N0 ej!
d!
2 1 2
Z
N
1 N0 1 j!
e d! = 0 ( );
=
2 2 1
2
(1.4)
where ( ) is the Dirac delta function. Clearly, for any timed-domain shift > 0, the
noise is uncorrelated.
Bandlimited communications systems bandlimit not only the signal but the noise
as well, and this ltering limits the rate of change of the time-domain noise signal,
introducing some correlation over the interval of 1=2B . The stylized PSD and ACF
of bandlimited white noise are displayed in Figure 1.4.
After bandlimiting, the autocorrelation function becomes:
Z
1 B N0 j!
N B j2f
e d! = 0
e
df
2 B 2
2 B

N0 ej2f B
=
2 j 2 B
1
[cos 2B + j sin 2B cos 2B + j sin 2B ]
=
j 2
sin(2B )
= N0 B
;
(1.5)
2B
which is the well-known sinc-function seen in Figure 1.4.
In the time-domain, the amplitude distribution of the white thermal noise has a
normal or Gaussian distribution, and since it is inevitably added to the received signal,
it is usually referred to as additive white Gaussian noise (AWGN). Note that AWGN is
therefore the noise generated in the receiver. The probability density function (PDF)
is the well-known bell-shaped curve of the Gaussian distribution, given by
2
1
(1.6)
p(x) = p e (x m)=2 ;
2
R( ) =
1.3. INFORMATION OF A SOURCE
11
where m is the mean and 2 is the variance. The eects of AWGN can be mitigated
by increasing the transmitted signal power and thereby reducing the relative eects of
noise. The signal-to-noise ratio (SNR) at the receiver's input provides a good measure
of the received signal quality. This SNR is often referred to as the channel SNR.
1.3
Information of a Source
Based on Shannon's work [1{4,12,13], let us introduce the basic terms and denitions
of information theory by considering a few simple examples. Assume that a simple
8-bit analog-to-digital (ADC) converter emits a sequence of mutually independent
source symbols that can take the values i = 1; 2; : : : 256 with equal probability.
One may wonder, how much information can be inferred upon receiving one of these
samples? It is intuitively clear that this inferred information is denitely proportional
to the \uncertainty" resolved by the reception of one such symbol, which in turn
implies that the information conveyed is related to the number of levels in the ADC.
More explicitly, the higher the number of legitimate quantization levels, the lower the
relative frequency or probability of receiving any one of them and hence the more
\surprising," when any one of them is received. Therefore, less probable quantized
samples carry more information than their more frequent, more likely counterparts.
Not suprisingly, one could resolve this uncertainty by simply asking a maximum
of 256 questions, such as \Is the level 1?" \Is it 2? : : :" \Is it 256?" Following
Hartley's approach [6], a more ecient strategy would be to ask eight questions, such
as: \Is the level larger than 128?" No. \Is it larger than 64?" No. : : : \Is it larger
than 2?" No. \Is it larger than 1?" No. Clearly, the source symbol emitted was of
magnitude one, provided that the zero level was not used. We could therefore infer
that log2 256 = 8 \Yes/No" binary answers were needed to resolve any uncertainty
as regards the source symbol's level.
In more general terms, the information carried by any one symbol of a q-level
source, where all the levels are equiprobable with probabilities of pi = 1=q, i = 1 : : : q,
is dened as
I = log2 q:
(1.7)
Rewriting Equation 1.7 using the message probabilities pi = 1q yields a more convenient form:
1
(1.8)
I = log2 = log2 pi ;
pi
which now is also applicable in case of arbitrary, unequal message probabilities pi ,
again, implying the plausible fact that the lower the probability of a certain source
symbol, the higher the information conveyed by its occurrence. Observe, however,
that for unquantized analog sources, where as regards to the number of possible
source symbols we have q ! 1 and hence the probability of any analog sample
becomes innitesimally low, these denitions become meaningless.
Let us now consider a sequence of N consecutive q-ary symbols. This sequence
can take qN number of dierent values, delivering qN dierent messages. Therefore,
the information carried by one such sequence is:
IN = log2 (qN ) = N log2 q;
(1.9)
12
which is in perfect harmony with our expectation, delivering N times the information
of a single symbol, which was quantied by Equation 1.7. Doubling the sequence
length to 2N carries twice the information, as suggested by:
I2N = log2 (q2N ) = 2N log2 q:
(1.10)
Before we proceed, let us brie y summarize the basic properties of information

following Shannon's work [1{4, 12, 13]:
If for the probability of occurrences of the symbols j and k we have pj < pk ,
then as regards the information carried by them we have: I (k) < I (j ).

1.4
If in the limit we have pk ! 1, then for the information carried by the symbol
k we have I (k) ! 0, implying that symbols, whose probability of occurrence
tends to unity, carry no information.
If the symbol probability is in the range of 0
information carried by it we have I (k) 0.
pk 1, then as regards the
For independent messages k and j , their joint information is given by the sum
of their information: I (k; j ) = I (k) + I (j ). For example, the information
carried by the statement \My son is 14 years old and my daughter is 12" is
equivalent to that of the sum of these statements: \My son is 14 years old" and
\My daughter is 12 years old."
In harmony with our expectation, if we have two equiprobable messages 0
and 1 with probabilities, p1 = p2 = 12 , then from Equation 1.8 we have
I (0) = I (1) = 1 bit.
Average Information of Discrete
Memoryless Sources
Following Shannon's approach [1{4,12,13], let us now consider a source emitting one
of q possible symbols from the alphabet s = s1 ; s2 ; : : : si : : : sq having symbol probabilities of pi , i = 1; 2; : : : q. Suppose that a long message of N symbols constituted
by symbols from the alphabet s = s1 ; s2 ; : : : sq having symbol probabilities of pi
is to be transmitted. Then the symbol si appears in every N -symbol message on
the average pi N number of times, provided the message length is suciently long.
The information carried by symbol si is log2 1=pi and its pi N occurrences yield an
information contribution of
1
(1.11)
I (i) = pi N log2 :
pi
Upon summing the contributions of all the q symbols, we acquire the total information
carried by the N -symbol sequence:
I=
q
X
i=1
pi N log2
1
[bits]:
pi
(1.12)
1.4. AVERAGE INFORMATION OF DISCRETE MEMORYLESS SOURCES
13
Averaging this over the N symbols of the sequence yields the average information per
symbol, which is referred to as the source's entropy [2] :
q
I X
1
H = = pi log2 =
N i=1
pi
q
X
i=1
pi log2 pi [bit/symbol]:
(1.13)
Then the average source information rate can be dened as the product of the information carried by a source symbol, given by the entropy H and the source emission
rate Rs :
R = Rs H [bits/sec].
(1.14)
Observe that Equation 1.13 is analogous to the discrete form of the rst moment or in
other words the mean of a random process with a probability density function (PDF)
of p(x), as in
Z 1
x=
x p(x)dx;
(1.15)
1
where the averaging corresponds to the integration, and the instantaneous value of
the random variable x represents the information log2 pi carried by message i, which
is weighted by its probability of occurrence pi quantied for a continuous variable x
by p(x).
1.4.1 Maximum Entropy of a Binary Source

Let us assume that a binary source, for which q = 2, emits two symbols with probabilities p1 = p and p2 = (1 p), where the sum of the symbol probabilities must be
unity. In order to quantify the maximum average information of a symbol from this
source as a function of the symbol probabilities, we note from Equation 1.13 that the
entropy is given by:
H (p) = p log2 p (1 p) log2 (1 p):
(1.16)
As in any maximization problem, we set @H (p)=@p = 0, and upon using the dierentiation chain rule of (u v)0 = u0 v + u v0 as well as exploiting that (loga x)0 = x1 loga e
we arrive at:
p
(1 p)
@H (p)
= log2 p

log2 e + log2 (1 p) +
log e = 0
@p
p
(1 p) 2
log2 p = log2 (1 p)
p = (1 p)
p = 0:5:
This result suggests that the entropy is maximum for equiprobable binary messages.
Plotting Equation 1.16 for arbitrary p values yields Figure 1.5, in which Shannon
suggested that the average information carried by a symbol of a binary source is low,
if one of the symbols has a high probability, while the other a low probability.
14

1.0
H(p)
0.8
0.6
0.4
0.2
0.0
0.0
Figure 1.5:
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Entropy versus message probability p for a binary source. c Shannon [2], BSTJ,
1948.
Example: Let us compute the entropy of the binary source having message
probabilities of p1 = 81 , p2 =
The entropy is expressed as:
7
8
1
1
log2
8
8
Exploiting the following equivalence:
H=
7
7
log2 :
8
8
log2 (x) = log10 (x) log2 (10) 3:322 log10 (x)
(1.17)
we have:
3 7
7

3:322 log10 0:54 [bit/symbol],
8 8
8
again implying that if the symbol probabilities are rather dierent, the entropy
becomes signicantly lower than the achievable 1 bit/symbol. This is because
the probability of encountering the more likely symbol is so close to unity that
it carries hardly any information, which cannot be compensated by the more
\informative" symbol's reception. For the even more unbalanced situation of
p1 = 0:1 and p2 = 0:9 we have:
H =

0:1 log2 0:1 0:9 log2 0:9

(0:3322 log10 0:1 + 0:9 3:322 log10 0:9)
0:3322 + 0:1368
0:47 [bit/symbol].
In the extreme case of p1 = 0 or p2 = 1 we have H = 0. As stated before,

the average source information rate is dened as the product of the information
1.5. SOURCE CODING FOR A DISCRETE MEMORYLESS SOURCE
15
carried by a source symbol, given by the entropy H and the source emission
rate Rs , yielding R = Rs H [bits/sec]. Transmitting the source symbols via
a perfect noiseless channel yields the same received sequence without loss of
information.
1.4.2 Maximum Entropy of a q -ary Source

For a q-ary source the entropy is given by:
H=
P
q
X
i=1
pi log2 pi ;
(1.18)
where, again, the constraint pi = 1 must be satised. When determining the
extreme
value of the above expression for the entropy H under the constraint of
P
pi = 1, the following term has to be maximized:
D =
q
X
i=1
"
pi log2 pi + 1
q
X
i=1
pi ;
(1.19)
where is the so-called Lagrange multiplier. Following the standard procedure of

maximizing an expression, we set:
@D
pi
= log2 pi
log2 e = 0
@pi
pi
leading to
log2 pi = (log2 e + ) = Constant for i = 1 : : : q;
which implies that the maximum entropy of a q-ary source is maintained, if all message
probabilities are identical, although at this stage the value of this constant probability
is not explicit. Note, however, that the message probabilites must sum to unity, and
hence:
q
X
pi = 1 = q a;
(1.20)
i=1
where a is a constant, leading to a = 1=q = pi , implying that the entropy of

any q-ary source is maximum for equiprobable messages. Furthermore, H is always
bounded according to:
0 H log2 q:
1.5
(1.21)
Source Coding for a Discrete

Memoryless Source
Interpreting Shannon's work [1{4, 12, 13] further, we see that source coding is the
process by which the output of a q-ary information source is converted to a binary
16
sequence for transmission via binary channels, as seen in Figure 1.1. When a discrete
memoryless source generates q-ary equiprobable symbols with an average information
rate of R = Rs log2 q, all symbols convey the same amount of information, and
ecient signaling takes the form of binary transmissions at a rate of R bps. When the
symbol probabilities are unequal, the minimum required source rate for distortionless
transmission is reduced to
R = Rs H < Rs log2 q:
(1.22)
Then the transmission of a highly probable symbol carries little information and hence
assigning log2 q number of bits to it does not use the channel eciently. What can be
done to improve transmission eciency? Shannon's source coding theorem suggests
that by using a source encoder before transmission the eciency of the system with
equiprobable source symbols can be arbitrarily approached.
Coding eciency can be dened as the ratio of the source information rate and
the average output bit rate of the source encoder. If this ratio approaches unity,
implying that the source encoder's output rate is close to the source information rate,
the source encoder is highly ecient. There are many source encoding algorithms,
but the most powerful approach suggested was Shannon's method [1], which is best
illustrated by means of the following example, portrayed in Table 1.1, Algorithm 1,
and Figure 1.6.
1.5.1 Shannon-Fano Coding

The Shannon-Fano coding algorithm is based on the simple concept of encoding frequent messages using short codewords and infrequent ones by long codewords, while
reducing the average message length. This algorithm is part of virtually all treatises
dealing with information theory, such as, for example, Carlson's work [8]. The formal
coding steps listed in Algorithm 1 and in the owchart of Figure 1.6 can be readily
followed in the context of a simple example in Table 1.1. The average codeword length
is given by weighting the length of any codeword by its probability, yielding:
(0:27 + 0:2) 2 + (0:17 + 0:16) 3 + 2 0:06 4 + 2 0:04 4 2:73 [bit].
The entropy of the source is:
H =
=

X
i
pi log2 pi
(log2 10)
X
i
(1.23)
pi log10 pi
3:322 [0:27 log10 0:27 + 0:2 log10 0:2

+0:17 log10 0:17 + 0:16 log10 0:16
+2 0:06 log10 0:06 + 2 0:04 log10 0:04]
2:691 [bit/symbol].
17
Algorithm 1 (Shannon-Fano Coding) This algorithm summarizes the ShannonFano coding steps. (See also Figure 1.6 and Table 1.1.)
1. The source symbols S0 : : : S7 are first sorted in descending order

of probability of occurrence.
2. Then the symbols are divided into two subgroups so that the
subgroup probabilities are as close to each other as possible.
This is symbolized by the horizontal divisions in Table 1.1.
3. When allocating codewords to represent the source symbols, we
assign a logical zero to the top subgroup and logical one to the
bottom subgroup in the appropriate column under ``coding steps.''
4. If there is more than one symbol in the subgroup, this method is
continued until no further divisions are possible.
5. Finally, the variable-length codewords are output to the channel.
Symb. Prob. Coding Steps Codeword

1 2 3 4
S0
0.27 0 0
00
S1
0.20 0 1
01
S2
0.17 1 0 0
100
S3
0.16 1 0 1
101
S4
0.06 1 1 0 0
1100
S5
0.06 1 1 0 1
1101
S6
0.04 1 1 1 0
1110
S7
0.04 1 1 1 1
1111
Table 1.1:
Shannon-Fano Coding Example Based on Algorithm 1 and Figure 1.6
Since the average codeword length of 2.73 bit/symbol is very close to the entropy of
2.691 bit/symbol, a high coding eciency is predicted, which can be computed as:
2:691
98:6 %:
E
2:73
The straightforward 3 bit/symbol binary coded decimal (BCD) assignment gives an
eciency of:
2:691
E
89:69 %:
3
In summary, Shannon-Fano coding allowed us to create a set of uniquely invertible
mappings to a set of codewords, which facilitate a more ecient transmission of the
source symbols, than straightforward BCD representations would. This was possible
with no coding impairment (i.e., losslessly). Having highlighted the philosophy of the
Shannon-Fano noiseless or distortionless coding technique, let us now concentrate on
the closely related Human coding principle.
18
Sort source symbols in

descending order of probabilities
Derive subgroups of
near-equal probabilities
Assign zero & one to top and

bottom branches, respectively
Yes
More than
one symbols are in
subgroups ?
No
Stop, output encoded symbols
Figure 1.6:
Shannon-Fano Coding Algorithm (see also Table 1.1 and Algorithm 1).
1.5.2 Human Coding

The Human Coding (HC) algorithm is best understood by referring to the owchart
of Figure 1.7 and to the formal coding description of Algorithm 2, while a simple practical example is portrayed in Table 1.2, which leads to the Human codes summarized
in Table 1.3. Note that we used the same symbol probabilities as in our Shannon-Fano
coding example, but the Human algorithm leads to a dierent codeword assignment.
Nonetheless, the code's eciency is identical to that of the Shannon-Fano algorithm.
The symbol-merging procedure can also be conveniently viewed using the example
of Figure 1.8, where the Human codewords are derived by reading the associated 1
and 0 symbols from the end of the tree backward, that is, toward the source symbols
S0 : : : S7 . Again, these codewords are summarized in Table 1.3.
In order to arrive at a xed average channel bit rate, which is convenient in many
communications systems, a long buer might be needed, causing storage and delay
problems. Observe from Table 1.3 that the Human coding algorithm gives code-
19
Arrange source symbols in

descending order of probabilities
Merge two of the lowest prob.

symbols into one subgroup
Assign zero & one to top and

bottom branches, respectively
Yes
Is there
more than one unmerged node ?
No
Stop, read transition bits on the branches
from top to bottom to generate codewords
Figure 1.7:
Human coding algorithm (see also Algorithm 2 and Table 1.2).
Algorithm 2 (Human Coding) This algorithm summarizes the Human coding
steps.
1. Arrange the symbol probabilities pi in decreasing order and

consider them as ``leaf-nodes,'' as suggested by Table 1.2.
2. While there is more than one node, merge the two nodes having the
lowest probability and assign 0/1 to the upper/lower branches,
respectively.
3. Read the assigned ``transition bits'' on the branches from top to
bottom in order to derive the codewords.
20
Symb. Prob.
S0
S1
S2
S3
S4
S5
S6
S7
Step 1 & 2
Code Prob.
0.27
0.20
0.17
0.16
0.06
0.06
0.04
0.04
0
1
0
1
Symb. Prob.
S23
S0
S1
S4567
Table 1.2:
0.33
0.27
0.20
0.20
0.12
0.08
Step 5 & 6
Code Prob.
0
0.6
1
0
0.4
1
Step 3 & 4
Code Prob.
0
1
0
0
1
1
Group Code
0.33
S0
S1
S23
0.20
S4567
Step 7
Code Prob.
0
1.0
1
0
1
00
01
10
11
Codeword
00
01
10
11
Human Coding Example Based on Algorithm 2 and Figure 1.7 (for nal code
assignment see Table 1.3)
Symbol Probability BCD Human Code

S0
0.27
000 01
S1
0.20
001 10
S2
0.17
010 000
S3
0.16
011 001
S4
0.06
100 1100
S5
0.06
101 1101
S6
0.04
110 1110
S7
0.04
111 1111
Table 1.3:
Human Coding Example Summary of Table 1.2
0.27
0.20
S1=0.20
0.20
0.17
S2=0.17
0.06
S5=0.06
0.06
S6=0.04
0.08
0.16
0.20
0.27
0
1
0
1
0
1
0
1
0
1
S7=0.04
Figure 1.8:
0.60
0.33
0.20
0.17
0.12
0.40
0.27
0.20
0.16
0.08
S4=0.06
0.33
0.20
0.17
0.16
S3=0.16
0.27
0.40
0
1
0
1
0.27
S0=0.27
Tree-based Human coding example.

21
22
words that can be uniquely decoded, which is a crucial prerequisite for its practical
employment. This is because no codeword can be a prex of any longer one. For
example, for the following sequence of codewords . . . , 00, 10, 010, 110, 1111, . . . the
source sequence of : : : S0 ; S1 ; S2 ; S3 ; S8 : : : can be uniquely inferred from Table 1.3.
In our discussions so far, we have assumed that the source symbols were completely
independent of each other. Such a source is usually referred to as a memoryless source.
By contrast, sources where the probability of a certain symbol also depends on what
the previous symbol was are often termed sources exhibiting memory. These sources
are typically bandlimited sample sequences, such as, for example, a set of correlated
or \similar-magnitude" speech samples or adjacent video pixels. Let us now consider
sources that exhibit memory.
1.6
Average Information of Discrete

Sources Exhibiting Memory
Let us invoke Shannon's approach [1{4, 12, 13] in order to illustrate sources with and
without memory. Let us therefore consider an uncorrelated random white Gaussian
noise (WGN) process, which was passed through a low-pass lter. The corresponding
autocorrelation functions (ACF) and power spectral density (PSD) functions were
portrayed in Figures 1.3 and 1.4. Observe in the gures that through low-pass ltering a WGN process introduces correlation by limiting the rate at which amplitude
changes are possible, smoothing the amplitude of abrupt noise peaks. This example
suggests that all bandlimited signals are correlated over a nite interval. Most analog
source signals, such as speech and video, are inherently correlated, owing to physical
restrictions imposed on the analog source. Hence all practical analog sources possess
some grade of memory, a property that is also retained after sampling and quantization. An important feature of sources with memory is that they are predictable to a
certain extent, hence, they can usually be more eciently encoded than unpredictable
sources having no memory.
1.6.1 Two-State Markov Model for Discrete

Let us now introduce a simple analytically tractable model for treating sources that
exhibit memory. Predictable sources that have memory can be conveniently modeled
by Markov processes. A source having a memory of one symbol interval directly \remembers" only the previously emitted source symbol and depending on this previous
symbol it emits one of its legitimate symbols with a certain probability, which depends
explicitly on the state associated with this previous symbol. A one-symbol-memory
model is often referred to as a rst-order model. For example, if in a rst-order model
the previous symbol can take only two dierent values, we have two dierent states,
and this simple two-state rst-order Markov model is characterized by the state transition diagram of Figure 1.9. Previously, in the context of Shannon-Fano and Human
coding of memoryless information sources, we used the notation of Si , i = 0; 1; : : :
for the various symbols to be encoded. In this section, we are dealing with sources
1.6. ENTROPY OF SOURCES EXHIBITING MEMORY
23
exhibiting memory and hence for the sake of distinction we use the symbol notation of
Xi , i = 1; 2; : : :. If, for the sake of illustration, the previous emitted symbol was X1 ,
the state machine of Figure 1.9 is in state X1 , and in the current signaling interval
it can generate one of two symbols, namely, X1 and X2 , whose probability depends
explicitly on the previous state X1 . However, not all two-state Markov models are
as simple as that of Figure 1.9, since the transitions from state X1 to X2 are not
necessarily associated with emitting the same symbol as the transitions from state
X2 to X1 . Thus more elaborate example will be considered later in this chapter.
Observe in Figure 1.9 that the corresponding transition probabilities from state X1
are given by the conditional probabilities p12 = P (X2 =X1) and p11 = P (X1 =X1 ) = 1
P (X2 =X1 ). Similar ndings can be observed as regards state X2 . These dependencies
can also be stated from a dierent point of view as follows. The probability of occurrence of a particular symbol depends not only on the symbol itself, but also on the
previous symbol emitted. Thus, the symbol entropy for state X1 and X2 will now be
characterized by means of the conditional probabilities associated with the transitions
merging in these states. Explicitly, the symbol entropy for state Xi ; i = 1; 2 is given
by:
Hi =
2
X
j =1
pij log2
1
i = 1; 2
pij
1
1
+ p log2 ;
(1.24)
pi1 i2
pi2
yielding the symbol entropies, that is, the average information carried by the symbols
emitted in states X1 and X2 , respectively, as:
1
1
H1 = p11 log2 + p12 log2
p11
p12
1
1
H2 = p21 log2 + p22 log2 :
(1.25)
p21
p22
Both symbol entropies, H1 and H2 , are characteristic of the average information
conveyed by a symbol emitted in state X1 and X2 , respectively. In order to compute
the overall entropy H of this source, they must be weighted by the probability of
occurrence, P1 and P2 , of these states:
= pi1 log2
H =
=
2
X
i=1
2
X
i=1
Pi Hi
Pi
2
X
j =1
pij log2
1
:
pij
(1.26)
Assuming a highly predictable source having high adjacent sample correlation, it

is plausible that once the source is in a given state, it is more likely to remain in
that state than to traverse into the other state. For example, assuming that the state
machine of Figure 1.9 is in state X1 and the source is a highly correlated, predictable
source, we are likely to observe long runs of X1 . Conversely, once in state X2 , long
strings of X2 symbols will typically follow.
24
'

-
6
&
$
?

%
p12 = P (X2 =X1 )
P (X1 )
P (X2)
X1
p11
X2
p21 = P (X1 =X2 )
P (X1=X1 ) = 1 ; P (X2 =X1)

p22
Figure 1.9:
1.6.2
=1
; P (X1=X2 ) = P (X2=X2 )
Two-state rst-order Markov model.
N -State Markov Model for Discrete Sources

Exhibiting Memory
In general, assuming N legitimate states, (i.e., N possible source symbols) and following similar arguments, Markov models are characterised by their state probabilities
P (Xi ); i = 1 : : : N , where N is the number of states, as well as by the transition probabilities pij = P (Xi =Xj ), where pij explicitly indicates the probability of
traversing from state Xj to state Xi . Their further basic feature is that they emit a
source symbol at every state transition, as will be shown in the context of an example
presented in Section 1.7. Similarly to the two-state model, we dene the entropy
of a source having memory as the weighted average of the entropy of the individual
symbols emitted from each state, where weighting is carried out taking into account
the probability of occurrence of the individual states, namely Pi . In analytical terms
, the symbol entropy for state Xi ; i = 1 : : : N is given by:
Hi =
N
X
j =1
pij log2
1
i = 1 : : : N:
pij
(1.27)
The averaged, weighted symbol entropies give the source entropy:
H =
=
N
X
i=1
N
X
i=1
Pi Hi
Pi
N
X
j =1
pij log2
1
:
pij
(1.28)
Finally, assuming a source symbol rate of vs , the average information emission rate
R of the source is given by:
R = vs H [bps].
(1.29)
1.7. EXAMPLES
25
'

-
6
&
$
?

%
p12 =0.1
X1
X2
0.9
0.1
p21 =0.9
P1 = 0:8
P2 = 0:2
Figure 1.10:
1.7
Two-state Markov model example.
Examples
1.7.1 Two-State Markov Model Example

As mentioned in the previous section, we now consider a slightly more sophisticated
Markov model, where the symbols emitted upon traversing from state X1 to X2 are
dierent from those when traversing from state X2 to X1 . More explicitly:
Consider a discrete source that was described by the two-state Markov model
of Figure 1.9, where the transition probabilities are
p11 = P (X1 =X1 ) = 0:9 p22 = P (X2 =X2 ) = 0:1

p12 = P (X1 =X2 ) = 0:1 p21 = P (X2 =X1) = 0:9;
while the state probabilities are
P (X1 ) = 0:8 and P (X2 ) = 0:2:
(1.30)
The source emits one of four symbols, namely, a, b, c, and d, upon every state
transition, as seen in Figure 1.10. Let us nd
(a) the source entropy and
(b) the average information content per symbol in messages of
one, two, and three symbols.
Message Probabilities
Let us consider two sample sequences acb and aab. As shown in Figure 1.10, the
transitions leading to acb are (1 1), (1 2), and (2 2). The probability of
encountering this sequence is 0:8 0:9 0:1 0:1 = 0:0072. The sequence aab has a
26
Message Probabilities
Pa = 0:9 0:8 = 0:72
Pb = 0:1 0:2 = 0:02
Pc = 0:1 0:8 = 0:08
Pd = 0:9 0:2 = 0:18
Paa = 0:72 0:9 = 0:648
Pac = 0:72 0:1 = 0:072
Pcb = 0:08 0:1 = 0:008
Pcd = 0:08 0:9 = 0:072
Pbb = 0:02 0:1 = 0:002
Pbd = 0:02 0:9 = 0:018
Pda = 0:18 0:9 = 0:162
Pdc = 0:18 0:1 = 0:018
Table 1.4:
Information conveyed
(bit/message)
Ia = 0:474
Ib = 5:644
Ic = 3:644
Id = 2:474
Iaa = 0:626
Iac = 3:796
Icb = 6:966
Icd = 3:796
Ibb = 8:966
Ibd = 5:796
Ida = 2:626
Idc = 5:796
Message Probabilities of Example
probability of zero because the transition from a to b is illegal. Further path (i.e.,
message) probabilities are tabulated in Table 1.4 along with the information of
I = log2 P of all the legitimate messages.
Source Entropy
{ According to Equation 1.27, the entropy of symbols X1 and X2 is computed

as follows:
p11 log2 p11

1
= 0:1 log2 10 + 0:9 log2
0:9
0:469 bit/symbol
= p21 log2 p21 p22 log2 p22
0:469 bit/symbol
H1 =
H2
p12 log2 p12
(1.31)
(1.32)
{ Then their weighted average is calculated using the probability of occur-
rence of each state in order to derive the average information per message
for this source:
H 0:8 0:469 + 0:2 0:469 0:469 bit/symbol:
{ The average information per symbol I2 in two-symbol messages is computed from the entropy h2 of the two-symbol messages as follows:
h2 =
8
X
Psymbol Isymbol
= Paa Iaa + Pac Iac + : : : + Pdc Idc

1:66 bits/2 symbols,
(1.33)
1.7. EXAMPLES
27
giving I2 = h2 =2 0:83 bits/symbol information on average upon receiving

a two-symbol message.
{ There are eight two-symbol messages; hence, the maximum possible information conveyed is log2 8 = 3 bits/2 symbols, or 1:5 bits/symbol. However,
since the symbol probabilities of P1 = 0:8 and P2 = 0:2 are fairly dierent, this scheme has a signicantly lower conveyed information per symbol,
namely, I2 0:83 bits/symbol.
{ Similarly, one can nd the average information content per symbol for
arbitrarily long messages of concatenated source symbols. For one-symbol
messages we have:
I1 = h1 =
4
X
Psymbol Isymbol
= Pa Ia + : : : + Pd Id
0:72 0:474 + : : : + 0:18 2:474
0:341 + 0:113 + 0:292 + 0:445
1:191 bit/symbol.
(1.34)
We note that the maximum possible information carried by one-symbol

messages is h1max = log2 4 = 2 bit/symbol, since there are four one-symbol
messages in Table 1.4.
Observe the important tendency, in which, when sending longer messages of
dependent sources, the average information content per symbol is reduced. This
is due to the source's memory, since consecutive symbol emissions are dependent
on previous ones and hence do not carry as much information as independent
source symbols. This becomes explicit by comparing I1 1:191 and I2 0:83
bits/symbol.
Therefore, expanding the message length to be encoded yields more ecient coding schemes, requiring a lower number of bits, if the source has a memory. This
is the essence of Shannon's source coding theorem.
1.7.2 Four-State Markov Model for a 2-Bit Quantizer

Let us now augment the previously introduced two-state Markov-model concepts with
the aid of a four-state example. Let us assume that we have a discrete source constituted by a 2-bit quantizer, which is characterized by Figure 1.11. Assume further that
due to bandlimitation only transitions to adjacent quantization intervals are possible,
since the bandlimitation restricts the input signal's rate of change. The probability
of the signal samples residing in intervals 1{4 is given by:
P (1) = P (4) = 0:1; P (2) = P (3) = 0:4:

The associated state transition probabilities are shown in Figure 1.11, along with
the quantized samples a; b; c; and d; which are transmitted when a state transition
28
P(2)=0.4
p12=0.8
p =0.2
11
p =0.4
22
P(1)=0.1
p =0.2
21
p 32=0.4
p23=0.4
p =0.2
44
p43=0.8
p =0.4
33
p(4)=0.1
Figure 1.11:
p =0.2
34
P(3)=0.4
Four-state Markov model for a 2-bit quantizer.
takes place, that is, when taking a new sample from the analog source signal at the
sampling-rate fs .
Although we have stipulated a number of simplifying assumptions, this example
attempts to illustrate the construction of Markov models in the context of a simple
practical problem. Next we construct a simpler example for augmenting the underlying concepts and set aside the above four-state Markov-model example as a potential
exercise for the reader.
1.8
Generating Model Sources
1.8.1 Autoregressive Model

In evaluating the performance of information processing systems, such as encoders and
predictors, it is necessary to have \standardized" or easily described model sources.
Although a set of semistandardized speech and images test sequences is widely used by
researchers in codec performance testing, in contrast to analytical model sources, real
speech or image sources cannot be used in analytical studies. A widely used analytical
model source is the Autoregressive (AR) model. A zero mean random sequence y(n)
1.8. GENERATING MODEL SOURCES
29
is called an AR process of order p, if it is generated as follows:
y(n) =
p
X
k=1
ak y(n k) + "(n); 8n;
(1.35)
where "(n) is an uncorrelated zero-mean, random input sequence with variance 2 ;

that is,
E f"(n)g = 0
E f"2(n)g = 2
E f"(n) y(m)g = 0:
(1.36)
From Equation 1.35 we surmise that an AR system recursively generates the present
output from p number of previous output samples given by y(n k) and the present
random input sample "(n).
1.8.2 AR Model Properties

AR models are very useful in studying information processing systems, such as speech
and image codecs, predictors, and quantizers. They have the following basic properties:
1. The rst term of Equation 1.35, which is repeated here for convenience,
y^(n) =
p
X
k=1
ak y(n k)
denes a predictor, giving an estimate y^(n) of y(n), which is associated with

the minimum mean squared error between the two quantities.
2. Although y^(n) and y(n) depend explicitly only on the past p number of samples
of y(n), through the recursive relationship of Equation 1.35 this entails the
entire past of y(n). This is because each of the previous p samples depends on
their predecessors.
3. Then Equation 1.35 can be written in the form of:
y(n) = y^(n) + "(n);
(1.37)
where "(n) is the prediction error and y^(n) is the minimum variance prediction
estimate of y(n).
4. Without proof, we state that for a random Gaussian distributed prediction error
sequence "(n) these properties are characteristic of a pth order Markov process
portrayed in Figure 1.12. When this model is simplied for the case of p = 1,
we arrive at the schematic diagram shown in Figure 1.13.
30
"(n)
y(n)
6
y(n p)
y(n 1)
ap

?
"(n)
Markov model of order p.
y(n)
a2
a
? ?
Figure 1.12:
?a
p
Figure 1.13:
First-order Markov model.
5. The power spectral density (PSD) of the prediction error sequence "(n) is that
of a random \white-noise" sequence, containing all possible frequency components with the same energy. Hence, its autocorrelation function (ACF) is the
Kronecker delta function, given by the Wiener-Khintchine theorem:
E f"(n) "(m)g = 2 (n m):
(1.38)
1.8.3 First-Order Markov Model

A variety of practical information sources are adequately modeled by the analytically
tractable rst-order Markov model depicted in Figure 1.13, where the prediction order
is p = 1. With the aid of Equation 1.35 we have
y(n) = "(n) + ay(n 1);
1.9. RUN-LENGTH CODING
31
where a is the adjacent sample

recursion:
y(n 1)
..
.
correlation of the process y(n). Using the following
"(n 1) + a1 y(n 2)
..
.
y(n k) = "(n k) + a1 y(n k 1)
(1.39)
y(n) = "(n) + a1 ["(n 1) + ay(n 2)]

= "(n) + a1 "(n 1) + a2 y(n 2);
which can be generalized to:
1
X
y(n) =
aj "(n j ):
(1.40)
=
..
.
we arrive at:
j =0
Clearly, Equation 1.40 describes the rst-order Markov process by the help of the
adjacent sample correlation a1 and the uncorrelated zero-mean random Gaussian
process "(n).
1.9
Run-Length Coding for Discrete

1.9.1 Run-Length Coding Principle [8]

For discrete sources having memory, (i.e., possessing intersample correlation), the
coding eciency can be signicantly improved by predictive coding, allowing the
required transmission rate and hence the channel bandwidth to be reduced. Particularly amenable to run-length coding are binary sources with inherent memory, such
as black and white documents, where the predominance of white pixels suggests that
a Run-Length-Coding (RLC) scheme, which encodes the length of zero runs, rather
than repeating long strings of zeros, provides high coding eciency.
Following Carlson's interpretation [8], a predictive RLC scheme can be constructed
according to Figure 1.14. The q-ary source messages are rst converted to binary bit
format. For example, if an 8-bit analog-digital converter (ADC) is used, the 8-bit digital samples are converted to binary format. This bit-stream, x(i), is then compared
with the output signal of the predictor, x^(i), which is fed with the prediction error
signal e(i). The comparator is a simple mod-2 gate, outputting a logical 1, whenever
the prediction fails; that is, the predictor's output is dierent from the incoming bit
x(i). If, however, x(i) = x^(i), the comparator indicates this by outputting a logical
0. For highly correlated signals from sources with signicant memory the predictions
are usually correct, and hence long strings of 0 runs are emitted, interspersed with
an occasional 1. Thus, the prediction error signal e(i) can be eciently run-length
encoded by noting and transmitting the length of zero runs.
The corresponding binary run-length coding principle becomes explicit from Table 1.5 and from our forthcoming coding eciency analysis.
32
Q-ary
Source
x^(i)
Q-ary to x(i) ?
Binary - +
e(i)
Conv.
Predictor
Predictor
Q-ary
Output

-
x^(i)
Binary
?
to Q-ary +
e(i)
Conv. x(i)
Figure 1.14:
RL
Encoder
RL
Decoder
Predictive run-length codec scheme. c Carlson [8].
Length of 0-run Encoder Output Decoder Output

l
(n-bit codeword)
0
00 000
1
1
00 001
01
2
00 010
001
3
00 011
0001
..
..
..
.
.
.
N 1
11 110
00 01
N = 2n 1
11 111
00 00
Table 1.5:
Run-length Coding Table c Carlson, 1975 [8]
1.9.2 Run-Length Coding Compression Ratio [14]

Following Jain's interpretation [14], let us now investigate the RLC eciency by
assuming that a run of r successive logical 0s is followed by a 1. Instead of directly
transmitting these strings, we represent such a string as an n-bit word giving the
length of the 0-run between successive logical ones. When a 0-run longer than N =
2n 1 bits occurs, this is signaled as the all 1 codeword, informing the decoder
to wait for the next RLC codeword before releasing the decoded sequence. Again,
the scheme's operation is characterized by Table 1.5. Clearly, data compression is
achieved if the average number of 0 data bits per run d is higher than the number of
bits, n, required to encode the 0-run length. Let us therefore compute the average
number of bits per run without RLC. If a run of r logical zeros are followed by a 1,
the run-length is (r + 1). The expected or mean value of (r + 1), namely, d = (r + 1),
is calculated by weighting each specic (r + 1) with its probability of occurrence that
1.9. RUN-LENGTH CODING
33
0.9
$RCSfile: gofl.gle,v $
p=0.1
p=0.9
CDF of g( )
g( ) = p (1-p)
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Figure 1.15:
10
12
14
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
$RCSfile: cdfgofl.gle,v $
p=0.1
p=0.9
10
12
14
CDF and PDF of the geometric distribution of run-length l.
is, with its discrete PDF c(r) and then averaging the weighted components, in:
d = (r + 1) =
N
X1
r=0
(r + 1) c(r) + Nc(N ):
(1.41)
The PDF of a run of r zeros followed by a 1 is given by:
c(r) =
pr (1 p) 0 r N
pN
r = N;
(1.42)
since the probability of N consecutive zeros is pN if r = N , while for shorter runs the
joint probability of r zeros followed by a 1 is given by pr (1 p). The PDF and CDF of
this distribution are shown in Figure 1.15 for p = 0:9 and p = 0:1, where p represents
the probability of a logical zero bit. Substituting Equation 1.42 in Equation 1.41
gives:
d = N pN +
N
X1
r=0
(r + 1) pr (1 p)
= N pN + 1 p0 (1 p) + 2 p (1 p) + : : : + N pN 1 (1 p)
= N pN + 1 + 2p + 3p2 + : : : + N pN 1 p 2p2 : : : N pN
= 1 + p + p2 + pN 1:
(1.43)
Equation 1.43 is a simple geometric progression, given in closed form as:
1 pN
d=
:
(1.44)
1 p
RLC Example: Using a run-length coding memory of M = 31 and a zero
symbol probability of p = 0:95, characterize the RLC eciency.
Substituting N and p into Equation 1.44 for the average run-length we have:
1 0:9531 1 0:204
d=
0:05 15:92:
(1.45)
1 0:95
34
The compression ratio C achieved by RLC is given by:
C=
1 pN
d
=
n n(1 p)
155:92 3:18:
(1.46)
The achieved average bit rate is

n
B = 0:314 bit/pixel,
d
and the coding eciency is computed as the ratio of the entropy (i.e., the lowest
possible bit rate and the actual bit rate). The source entropy is given by:
0:95 3:322 log 0:95

0:286 bit/symbol;
10
0:05 3:322 log10 0:05
(1.47)
giving a coding eciency of:
E = H=B 0:286=0:314 91%:

This concludes our RLC example.
1.10
Information Transmission via

Discrete Channels
Let us now return to Shannon's classic references [1{4, 12, 13] and assume that both
the channel and the source are discrete, and let us evaluate the amount of information
transmitted via the channel. We dene the channel capacity characterizing the channel and show that according to Shannon nearly error-free information transmission is
possible at rates below the channel capacity via the binary symmetric channel (BSC).
Let us begin our discourse with a simple introductory example.
1.10.1 Binary Symmetric Channel Example

Let us assume that a binary source is emitting a logical 1 with a probability of
P (1) = 0:7 and a logical 0 with a probability of P (0) = 0:3. The channel's error
probability is pe = 0:02. This scenario is characterized by the binary symmetric
channel (BSC) model of Figure 1.16. The probability of error-free reception is given by
that of receiving 1, when a logical 1 is transmitted plus the probability of receiving a 0
when 0 is transmitted, which is also plausible from Figure 1.16. For example, the rst
of these two component probabilities can be computed with the aid of Figure 1.16 as
the product of the probability P (1) of a logical 1 being transmitted and the conditional
probability P (1=1) of receiving a 1, given the condition that a 1 was transmitted:
P (Y1 ; X1 ) = P (X1 ) P (Y1 =X1)

P (1; 1) = P (1) P (1=1) = 0:7 0:98 = 0:686:
(1.48)
1.10. INFORMATION TRANSMISSION VIA DISCRETE CHANNELS
1 pe =0.98
P (1) = 0:7
P (0) = 0:3
pe =0.02
TX
pe =0.02
1 pe =0.98
0
Figure 1.16:
-3
35
RX
s-
The binary symmetric channel. c Shannon [3], BSTJ, 1948.
Similarly, the probability of the error-free reception of a logical 0 is given by:
P (Y0 ; X0 ) = P (X0 ) P (Y0 =X0)

P (0; 0) = P (0) P (0=0) = 0:3 0:98 = 0:294;
giving the total probability of error-free reception as:
Pcorrect = P (1; 1) + P (0; 0) = 0:98:

Following similar arguments, the probability of erroneous reception is also given by
two components. For example, using Figure 1.16, the probability of receiving a 1 when
a 0 was transmitted is computed by multiplying the probability P (0) of a logical 0
being transmitted by the conditional probability P (1=0) of receiving a logical 1, given
the fact that a 0 is known to have been transmitted:
P (Y1 ; X0 ) = P (X0 ) P (Y1 =X0)

P (1; 0) = P (0) P (1=0) = 0:3 0:02 = 0:006:
Conversely,
P (Y0 ; X1 ) = P (X1 ) P (Y0 =X1)

P (0; 1) = P (1) P (0=1) = 0:7 0:02 = 0:014;
yielding a total error probability of:
Perror = P (1; 0) + P (0; 1) = 0:02;

which is constituted by the above two possible error events.
Viewing events from a dierent angle, we observe that the total probability of receiving 1 is that of receiving a transmitted 1 correctly plus a transmitted 0 incorrectly:
P1 = P (1) (1 pe ) + P (0) pe
= 0:7 0:98 + 0:3 0:02 = 0:686 + 0:006 = 0:692:
(1.49)
36

1.0
0.75
0.75
0.5
P(0,0)
P(1,1)
$RCSfile: bsc-performance.gle,v $
1.0
pe=0
pe=0.125
pe=0.25
pe=0.375
pe=0.5
0.25
0.0
0.0
0.25
pe=0
pe=0.125
pe=0.25
pe=0.375
pe=0.5
0.5
0.25
0.5
0.75
0.0
0.0
1.0
0.25
1.0
1.0
0.75
0.75
0.5
pe=0.5
pe=0.375
pe=0.25
pe=0.125
0.25
0.0
0.0
0.25
0.5
Figure 1.17:
0.75
1.0
0.75
1.0
0.5
pe=0.5
pe=0.375
pe=0.25
pe=0.125
0.25
P(1)
0.5
P(0)
P(1,0)
P(0,1)
P(1)
0.75
1.0
0.0
0.0
0.25
0.5
P(0)
BSC performance for pe = 0, 0.125, 0.25, 0.375, and 0.5.
On the same note, the probability of receiving 0 is that of receiving a transmitted 0

correctly plus a transmitted 1 incorrectly:
P0 = P (0) (1 pe ) + P (1) pe
= 0:3 0:98 + 0:7 0:02 = 0:294 + 0:014 = 0:308:
(1.50)
In the next example, we further study the performance of the BSC for a range of
dierent parameters in order to gain a deeper insight into its behavior.
Example: Repeat the above calculations for P (1) = 1; 0:9; 0:5, and pe =
0; 0:1; 0:2; 0:5 using the BSC model of Figure 1.16. Compute and tabulate the
probabilities P (1; 1), P (0; 0), P (1; 0), P (0; 1), Pcorrect; Perror ; P1 , and P0 for
these parameter combinations, including also their values for the previous example, namely, for P (1) = 0.7, P (0) = 0.3 and pe = 0.02. Here we neglected
the details of the calculations and summarized the results in Table 1.6. Some of
the above quantities are plotted for further study in Figure 1.17, which reveals
the interdependency of the various probabilities for the interested reader.
Having studied the performance of the BSC, the next question that arises is, how
much information can be inferred upon reception of a 1 and a 0 over an imperfect
(1 pe )
pe
P (1) P (0) = P (1=1) P (1; 1) P (0; 0) = P (1=0) P (1; 0) P (0; 1)
= P (0=0)
= P (0=1)
1
0
1
1
0
0
0
0
0.9
0.1
1
0.9
0.1
0
0
0
0
0.7
0.3
1
0.7
0.3
0
0
0
0.5
0.5
1
0.5
0.5
0
0
0
1
0
0.98
0.98
0
0.02
0
0.02
0.1
0.98
0.882
0.098
0.02
0.002
0.018
0.02 0.9
0.7
0.3
0.98
0.686
0.294
0.02
0.006
0.014
0.5
0.5
0.98
0.49
0.49
0.02
0.01
0.01
1
0
0.9
0.9
0
0.1
0
0.1
0.9
0.1
0.9
0.81
0.09
0.1
0.01
0.09
0.1
0.7
0.3
0.9
0.63
0.27
0.1
0.03
0.07
0.5
0.5
0.9
0.45
0.45
0.1
0.05
0.05
1
0
0.8
0.8
0
0.2
0
0.2
0.9
0.1
0.8
0.72
0.08
0.2
0.02
0.18
0.2
0.7
0.3
0.8
0.56
0.24
0.2
0.06
0.14
0.5
0.5
0.8
0.40
0.40
0.2
0.1
0.1
1
0
0.5
0.5
0
0.5
0
0.5
0.9
0.1
0.5
0.45
0.05
0.5
0.05
0.45
0.5
0.7
0.3
0.5
0.35
0.15
0.5
0.15
0.35
0.5
0.5
0.5
0.25
0.25
0.5
0.25
0.25
pe
P0
0
0.1
0.3
0.5
0.02
0.116
0.308
0.509
0.1
0.189
0.277
0.545
0.2
0.278
0.434
0.5
0.5
0.5
0.5
0.5
P1
1
0.9
0.7
0.5
0.98
0.884
0.692
0.491
0.9
0.811
0.723
0.455
0.8
0.722
0.566
0.5
0.5
0.5
0.5
0.5
Table 1.6:
BSC Performance Table
37
38
X0
X1
Figure 1.18:
P (Y0 =X0 )
P (Y1 =X0 )
P (Y0 =X1 )
P (Y1 =X1 )
-3
Y0
s-
Y1
Forward transition probabilities of the nonideal binary symmetric channel.
(i.e., error-prone) channel. In order to answer this question, let us rst generalize the
above intuitive ndings in the form of Bayes' rule.
1.10.2 Bayes' Rule

Let Yj represent the received symbols and Xi the transmitted symbols having probabilities of P (Yj ) and P (Xi ), respectively. Let us also characterize the forward transition probabilities of the binary symmetric channel as suggested by Figure 1.18.
Then in general, following from the previous introductory example, the joint probability P (Yj ; Xi ) of receiving Yj , when the transmitted source symbol was Xi , is
computed as the probability P (Xi ) of transmitting Xi , multiplied by the conditional
probability P (Yj =Xi ) of receiving Yj , when Xi is known to have been transmitted:
P (Yj ; Xi ) = P (Xi ) P (Yj =Xi );
(1.51)
a result that we have already intuitively exploited in the previous example. Since for
the joint probabilities P (Yj ; Xi ) = P (Xi ; Yj ) holds, we have:
P (Xi ; Yj ) = P (Yj ) P (Xi =Yj )

= P (Xi ) P (Yj =Xi ):
(1.52)
Equation 1.52 is often presented in the form:
P (Xi ; Yj )
P (Yj )
P (Yj ) P (Xi =Yj )
;
=
P (Yj )
P (Xi =Yj ) =
(1.53)
which is referred to as Bayes' rule.

Logically, the probability of receiving a particular Yj = Yj0 is the sum of all joint
probabilities P (Xi ; Yj0 ) over the range of Xi . This corresponds to the probability of
39
receiving the transmitted Xi correctly, giving rise to the channel output Yj0 plus the
sum of the probabilities of all other possible transmitted symbols giving rise to Yj0 :
P (Yj ) =
Similarly:
P (Xi ) =
X
X
X
Y
P (Xi ; Yj ) =
P (Xi ; Yj ) =
X
X
X
Y
P (Xi )P (Yj =Xi ):
(1.54)
P (Yj )P (Xi =Yj ):
(1.55)
1.10.3 Mutual Information

In this section, we elaborate further on the ramications of Shannon's information
theory [1{4, 12, 13]. Over nonideal channels impairments are introduced, and the
received information might be dierent from the transmitted information. In this
section, we quantify the amount of information that can be inferred from the received
symbols over noisy channels. In the spirit of Shannon's fundamental work [1] and
Carlson's classic reference [8], let us continue our discourse with the denition of
mutual information. We have already used the notation P (Xi ) to denote the probability that the source symbol Xi was transmitted and P (Yi ) to denote the probability
that the symbol Yj was received. The joint probability that Xi was transmitted
and Yj was received had been quantied by P (Xi ; Yj ), and P (Xi =Yj ) indicated the
conditional probability that Xi was transmitted, given that Yj was received, while
P (Yj =Xi ) was used for the conditional probability that Yj was received given that Xi
was transmitted.
In case of i = j , the conditional probabilities P (Yj =Xj )j = 1 q represent the
error-free transmission probabilities of the source symbols j = 1 q. For example,
in Figure 1.18 the probabilities P (Y0 =X0 ) and P (Y1 =X1 ) are the probabilities of the
error-free reception of a transmitted X0 and X1 source symbol, respectively. The
probabilities P (Yj =Xi )j 6= i, on the other hand, give the individual error probabilities,
which are characteristic of error events that corrupted a transmitted symbol Xi to
a received symbol of Yj . The corresponding error probabilities in Figure 1.18 are
P (Y0 =X1) and P (Y1 =X0 ).
Let us dene the mutual information of Xi and Yj as:
P (Xi ; Yj )
P (Yj =Xi )
P (Xi =Yj )
= log2
= log2
bits; (1.56)
P (Xi )
P (Xi ) P (Yj )
P (Yj )
which quanties the amount of information conveyed, when Xi is transmitted and
Yj is received. Over a perfect, noiseless channel, each received symbol Yj uniquely
identies a transmitted symbol Xi with a probability of P (Xi =Yj ) = 1. Substituting
this probability in Equation 1.56 yields a mutual information of:
1
I (Xi ; Yj ) = log2
;
(1.57)
P ( Xi )
which is identical to the self-information of Xi and hence no information is lost over
the channel. If the channel is very noisy and the error probability becomes 0.5, then
I (Xi ; Yj ) = log2
40
the received symbol Yj becomes unrelated to the transmitted symbol Xi , since for a
binary system upon its reception there is a probability of 0.5 that X0 was transmitted
and the probability of X1 is also 0.5. Then formally Xi and Yj are independent and
hence
P (Xi ; Yj ) P (Xi ) P (Yj )
P (Xi =Yj ) =
=
= P (Xi );
(1.58)
P (Yj )
P (Yj )
giving a mutual information of:
I (Xi ; Yj ) = log2
P (Xi )
= log2 1 = 0;
P (Xi )
(1.59)
implying that no information is conveyed via the channel. Practical communications

channels perform between these extreme values and are usually characterized by the
average mutual information dened as:
I (X; Y ) =
=
X
x;y
X
x;y
P (Xi ; Yj ) I (Xi ; Yj )
P (Xi ; Yj ) log2
P (Xi =Yj )
[bit/symbol].
P (Xi )
(1.60)
Clearly, the average mutual information in Equation 1.60 is computed by weighting

each component I (Xi ; Yj ) by its probability of occurrence P (Xi ; Yj ) and summing
these contributions for all combinations of Xi and Yj . The average mutual information
I (X; Y ) dened above gives the average amount of source information acquired per
received symbol, as distinguished from that per source symbol, which was given by
the entropy H (X ). Let us now consolidate these denitions by working through the
following numerical example.
1.10.4 Mutual Information Example

Using the same numeric values as in our introductory example as regards to the
binary symmetric channel in Section 1.10.1, and exploiting that from Bayes' rule in
Equation 1.53, we have:
P (Xi ; Yj )
P (Xi =Yj ) =
:
P (Yj )
The following probabilities can be derived, which will be used at a later stage, in
order to determine the mutual information:
P (1; 1) 0:686
=
0:9913
P (X1 =Y1 ) = P (1=1) =
P1
0:692
and
P (0; 0) 0:294
P (X0 =Y0) = P (0=0) =
=
0:9545;
P0
0:3080
where P1 = 0:692 and P0 = 0:3080 represent the total probability of receiving 1 and
0, respectively, which is the union of the respective events of error-free and erroneous
41
receptions yielding the specic logical value concerned. The mutual information from
Equation 1.56 is computed as:
P (X1 =Y1 )
P (X1 )
0:9913
(1.61)
log2 0:7 0:502 bit
0:9545
I (X0 ; Y0 ) log2
1:67 bit:
(1.62)
0:3
These gures must be contrasted with the amount of source information conveyed by
the source symbols X0 ; X1 :
I (X1 ; Y1 ) = log2
I (0) = log2
and
1
0:3
log 3:33 1:737

2
bit/symbol
(1.63)
1
log2 1:43 0:5146 bit/symbol:
(1.64)
0:7
The amount of information \lost" in the noisy channel is given by the dierence
between the amount of information carried by the source symbols and the mutual
information gained upon inferring a particular symbol at the noisy channel's output.
Hence, the lost information can be computed from Equations 1.61, 1.62, 1.63, and 1.64,
yielding (1.737 - 1.67) 0.067 bit and (0.5146 - 0.502) 0.013 bit, respectively. These
values may not seem catastrophic, but in relative terms they are quite substantial and
their values rapidly escalate, as the channel error probability is increased. For the sake
of completeness and for future use, let us compute the remaining mutual information
terms, namely, I (X0 ; Y1 ) and I (X1 ; Y0 ), which necessitate the computation of:
I (1) = log2
P (X0 =Y1 ) =
P (0=1) =
P (X1 =Y0 ) =
P (1=0) =
I (X0 ; Y1 ) =
I (X1 ; Y0 ) =
P (X0 ; Y1 )
P (Y1 )
P (0; 1) 0:3 0:02
=
0:00867
P1
0:692
P (X1 ; Y0 )
P (Y0 )
P (1; 0) 0:7 0:02
=
0:04545
P0
0:308
P (X0 =Y1 )
0:00867
log2

log2
5:11 bit
P (X0 )
0:3
P (X1 =Y0 )
0:04545
log2

log2
3:945 bit;
P (X1 )
0:7
(1.65)
(1.66)
where the negative sign re ects the amount of \misinformation" as regards, for example, X0 upon receiving Y1 . In this example we informally introduced the denition of
mutual information. Let us now set out to formally exploit the benets of our deeper
insight into the eects of the noisy channel.
42
1.10.5 Information Loss via Imperfect Channels

Upon rewriting the denition of mutual information in Equation 1.56, we have:
P (Xi =Yj )
I (Xi ; Yj ) = log2
P (Xi )
1
1
= log2
log2
P (Xi )
P (Xi =Yj )
= I (Xi )
I (Xi =Yj ):
(1.67)
Following Shannon's [1{4,12,13] and Ferenczy's [10] approach and rearranging Equation 1.67 yields:
I (Xi )
I (Xi ; Yj )
= I (Xi =Yj ) :
| {z }
| {z }
| {z }
Source Inf. Inf. conveyed to rec. Inf. loss
(1.68)
Brie y returning to gure 1.18 assists the interpretation of P (Xi =Yj ) as the probability or certainty/uncertainty that Xi was transmitted, given that Yj was received,
which justies the above denition of the information loss. It is useful to observe
from this gure that, as it was stated before, P (Yj =Xi ) represents the probability of
erroneous or error-free reception. Explicitly, if j = i, then P (Yj =Xi ) = P (Yj =Xj ) is
the probability of error-free reception, while if j 6= i, then P (Yj =Xi ) is the probability
of erroneous reception.
With the probability P (Yj =Xi ) of erroneous reception in mind, we can actually
associate an error information term with it:
1
I (Yj =Xi ) = log2
:
(1.69)
P (Yj =Xi)
Let us now concentrate on the average mutual information's expression in Equation 1.60 and expand it as follows:
X
1
I (X; Y ) =
P (Xi ; Yj ) log2
P
(
Xi )
X;Y
X
1
P (Xi ; Yj ) log2
:
(1.70)
P
(
X
i =Yj )
X;Y
Considering the rst term at the right-hand side (rhs) of the above equation and
invoking Equation 1.55, we have:
"
X X
X
P (Xi ; Yj ) log2
X
1
1
= P (Xi ) log2
= H (X ):
P (Xi ) X
P (Xi )
(1.71)
Then rearranging Equation 1.70 gives:
H (X ) I (X; Y ) =
X
X;Y
P (Xi ; Yj ) log2
1
;
P (Xi =Yj )
(1.72)
43
where H (X ) is the average source information per symbol and I (X; Y ) is the average
conveyed information per received symbol.
Consequently, the rhs term must be the average information per symbol lost in the
noisy channel. As we have seen in Equation 1.67 and Equation 1.68, the information
loss is given by:
1
I (Xi =Yj ) = log2
:
(1.73)
P (Xi =Yj )
The average information loss H (X=Y ) equivocation, which Shannon [3] terms is computed as the weighted sum of these components:
XX
1
H (X=Y ) =
P (Xi ; Yj ) log2
:
(1.74)
P (Xi =Yj )
X Y
Following Shannon, this denition allowed us to express Equation 1.72 as:
H (X )
I (X; Y )
=
H (X=Y )
| {z }
| {z }
| {z }
(av. source inf/sym.) (av. conveyed inf/sym.) (av. lost inf/sym.)
(1.75)
1.10.6 Error Entropy via Imperfect Channels

Similarly to our previous approach and using the probability P (Yj =Xi ) of erroneous
reception associated with the information term of:
1
I (Yj =Xi ) = log2
(1.76)
P (Yj =Xi )
we can dene the average \error information" or error entropy. Hence, the above error
information terms in Equation 1.76 are weighted using the probabilities P (Xi ; Yj ) and
averaged for all X and Y values, dening the error entropy:
XX
1
:
(1.77)
H (Y=X ) =
P (Xi ; Yj ) log2
P (Yj =Xi )
X Y
Using Bayes' rule from Equation 1.52, we have
P (Xi =Yj ) P (Yj ) = P (Yj =Xi ) P (Xi )

P (Xi =Yj )
P (Yj =Xi )
=
:
(1.78)
P (Xi )
P (Yj )
Following from this, for the average mutual information in Equation 1.56 we have:
I (X; Y ) = I (Y; X );
(1.79)
which, after interchanging X and Y in Equation 1.75, gives:
I (Y; X ) = H (Y=X ) :
H (Y )
| {z }
| {z }
| {z }
destination entropy conveyed inf error entropy
(1.80)
44
Quantity
Source inf.
Received inf.
Joint inf.
Mutual inf.
Av. Mut. inf.
Source entropy
Destination entr.
Equivocation
Error entropy
Table 1.7:
Denition
I (Xi ) = log2 P (Xi )
I (Yj ) = log2 P (Yj )
IXi ;Yj = log2 P (Xi ; Yj )
I (Xi ; Yj ) = log2 P (PX(iX=Yi )j )
P P
i =Yj )
I (X; Y ) = X Y P (Xi Yj ) log2 P (PX(X
i)
P
H (X ) = P X P (Xi ) log2 P (Xi )
H (Y ) =
P (YP
j ) log2 P (Yj )
YP
H (X=Y ) = P X P Y P (Xi ; Yj ) log2 P (Xi =Yj )
H (Y=X ) =
X
Y P (Xi Yj ) log2 P (Yj =Xi )
Summary of Denitions c Ferenczy [10]
Subtracting the conveyed information from the destination entropy gives the error
entropy, which is nonzero, if the destination entropy and conveyed information are
not equal due to channel errors. Let us now proceed following Ferenczy's approach [10]
and summarize the most important denitions for future reference in Table 1.7 before
we attempt to augment their physical interpretations using the forthcoming numerical
example.
Example Using the BSC model of Figure 1.16, as an extension of the worked
examples of Subsections 1.10.1 and 1.10.4 and following Ferenczy's interpretation [10] of Shannon's elaborations [1{4, 12, 13], let us compute the following
range of system characteristics:
(a) The joint information, as distinct from the mutual information introduced earlier, for all possible channel input/output combinations.
(b) The entropy, i.e., the average information of both the source and the sink.
(c) The average joint information H (X; Y ).
(d) The average mutual information per symbol conveyed.
(e) The average information loss and average error entropy.
With reference to Figure 1.16 and to our introductory example from Section 1.10.1
we commence by computing further parameters of the BSC. Recall that the
source information was:
1
I (X0 ) = log2
3:322 log10 3:333 1:737 bit
0:3
1
I (X1 ) = log2
0:515 bit:
0:7
The probability of receiving a logical 0 was 0.308 and that of logical 1 was 0.692,
of whether 0 or 1 was transmitted. Hence, the information inferred upon the
reception of 0 and 1, respectively, is given by:
1
I (Y0 ) = log2
3:322 log10 3:247 1:699 bit
0:308
45
1
0:531 bit:
0:692
Observe that because of the reduced probability of receiving a logical 1 from
0:7 ! 0:692 as a consequence of channel-induced corruption, the probability of
receiving a logical 0 is increased from 0:3 ! 0:308. This is expected to increase
the average destination entropy, since the entropy maximum of unity is achieved,
when the symbols are equiprobable. We note, however, that this does not give
more information about the source symbols, which must be maximized in an
ecient communications system. In our example, the information conveyed
increases for the reduced probability logical 1 from 0:515 bit ! 0:531 bit
and decreases for the increased probability 0 from 1:737 bit ! 1:699 bit.
Furthermore, the average information conveyed is reduced, since the reduction
from 1.737 to 1.699 bit is more than the increment from 0.515 to 0.531. In the
extreme case of an error probability of 0.5 we would have P (0) = P (1) = 0:5, and
I (1) = I (0) = 1 bit, associated with receiving equiprobable random bits, which
again would have a maximal destination entropy, but a minimal information
concerning the source symbols transmitted. Following the above interesting
introductory calculations, let us now turn our attention to the computation of
the joint information.
a/ The joint information, as distinct from the mutual information introduced
earlier in Equation 1.56, of all possible channel input/output combinations is
computed from Figure 1.16 as follows:
I (Y1 ) = log2
IXi ;Yj
I00
I01
I10
I11
=
=
=
=
=
log2 P (Xi ; Yj )
(1.81)
log2 (0:3 0:98) 3:322 log10 0:294 1:766 bit
log2 (0:3 0:02) 7:381 bit
log2 (0:7 0:02) 6:159 bit
log2 (0:7 0:98) 0:544 bit:
These information terms can be individually interpreted formally as the information carried by the simultaneous occurrence of the given symbol combinations. For example, as it accrues from their computation, I00 and I11 correspond to the favorable event of error-free reception of a transmitted 0 and 1,
respectively, which hence were simply computed by formally evaluating the information terms. By the same token, in the computation of I01 and I10 , the
corresponding source probabilities were weighted by the channel error probability rather than the error-free transmission probability, leading to the corresponding information terms. The latter terms, namely, I01 and I10 , represent
low-probability, high-information events due to the low channel error probability
of 0.02.
Lastly, a perfect channel with zero error probability would render the probability
of the error-events zero, which in turn would assign innite information contents
to the corresponding terms of I01 and I10 , while I00 and I11 would be identical to
the self-information of the 0 and 1 symbols. Then, if under zero error probability
we evaluate the eect of the individual symbol probabilities on the remaining
46
joint information terms, the less frequently a symbol is emitted by the source,
the higher its associated joint information term becomes and vice versa, which
is seen by comparing I00 and I11 . Their dierence can be equalized by assuming
an identical probability of 0.5 for both, which would yield I00 =I11 = 1-bit. The
unweighted average of I00 and I11 would then be lower than in case of the
previously used probabilities of 0.3 and 0.7, respectively, since the maximum
average would be associated with the case of 0 and 1, where the associated log2
terms would be 0 and 1, respectively. The appropriately weighted average
joint information terms will be evaluted under paragraph c/ during our later
calculations. Let us now move on to evaluate the average information of the
source and sink.
b/ Calculating the entropy, that is, the average information for both the source
and the sink, is quite straightforward and ensues as follows:
H (X ) =

1
P
(
Xi )
i=1
0:3 log2 3:333 + 0:7 log2 1:429
0:5211 + 0:3605
0:8816 bit/symbol:
2
X
P (Xi ) log2
(1.82)
For the computation of the sink's entropy, we invoke Equations 1.49 and 1.50,
yielding:
1
1
+ 0:692 log2
H (Y ) = 0:308 log2
0:308
0:692
0:5233 + 0:3676
0:8909 bit/symbol:
(1.83)
Again, the destination entropy H (Y ) is higher than the source entropy H (X )

due to the more random reception caused by channel errors, approaching H (Y ) =
1 bit/symbol for a channel bit error rate of 0:5. Note, however, that unfortunately this increased destination entropy does not convey more information
about the source itself.
c/ Computing the average joint information H (X; Y ) gives:
H (X; Y ) =
=
2 X
2
X
i=1 j =1
2 X
2
X
i=1 j =1
P (Xi ; Yj ) log2 P (Xi ; Yj )

P (Xi ; Yj )IXi ;Yj :
(1.84)
Upon substituting the IXi ;Yj values calculated in Equation 1.81 into Equation 1.84, we have:
H (X; Y )
0:3 0:98 1:766 + 0:3 0:02 7:381
47
+ 0:7 0:02 6:159 + 0:7 0:98 0:544

0:519 + 0:044 + 0:086 + 0:373
1:022 bit/symbol-combination:
In order to interpret H (X; Y ), let us again scrutinize the denition given in
Equation 1.84, which weights the joint information terms of Equation 1.81 by
their probability of occurence. We have argued before that the joint information
terms corresponding to erroneous events are high due to the low error probability
of 0.02. Observe, therefore, that these high-information symbol combinations
are weighted by their low-probability of occurrence, causing H (X; Y ) to become
relatively low. It is also instructive to consider the above terms in Equation 1.84
for the extreme cases of zero and 0.5 error probabilities and for dierent source
emission probabilities, which are left for the reader to explore. Here we proceed
considering the average conveyed mutual information per symbol.
d/ The average conveyed mutual information per symbol was dened in Equation 1.60 in order to quantify the average source information acquired per received symbol, which is repeated here for convenience as follows:
XX
P (Xi =Yj )
I (X; Y ) =
P (Xi ; Yj ) log2
P ( Xi )
X Y
=
XX
X
P (Xi ; Yj ) I (Xi ; Yj ):
Using the individual mutual information terms from Equations 1.61{1.66 in

Section 1.10.4, we get the average mutual information representing the average
amount of source information acquired from the received symbols, as follows:
I (X; Y ) 0:3 0:98 1:67 + 0:3 0:02 ( 5:11)
+ 0:7 0:02 ( 3:945) + 0:7 0:98 0:502
0:491 0:03066 0:05523 + 0:3444
0:7495 bit/symbol:
(1.85)
In order to interpret the concept of mutual information, in Section 1.10.4 we
noted that the amount of information \lost" owing to channel errors was given
by the dierence between the amount of information carried by the source symbols and the mutual information gained upon inferring a particular symbol at
the noisy channel's output. These were given in Equations 1.61{1.64, yielding
(1.737 - 1.67) 0.067 bit and (0.5146 - 0.502) 0.013 bit, for the transmission of a 0 and 1, respectively. We also noted that the negative sign of the
terms corresponding to the error-events re ected the amount of misinformation
as regards, for example, X0 upon receiving Y1 . Over a perfect channel, the
cross-coupling transitions of Figure 1.16 are eliminated, since the associated error probabilities are 0, and hence there is no information loss over the channel.
Consequently, the error-free mutual information terms become identical to the
self-information of the source symbols, since exactly the same amount of information can be inferred upon reception of a symbol, as much is carried by its
appearance at the output of the source.
48
It is also instructive to study the eect of dierent error probabilities and source
symbol probabilities in the average mutual information denition of Equation 1.84 in order to acquire a better understanding of its physical interpretation
and quantitative power as regards the system's performance. It is interesting to
note, for example, that assuming an error probability of zero will therefore result
in average mutual information, which is identical to the source and destination
entropy computed above under paragraph b/. It is also plausible that I (X; Y )
will be higher than the previously computed 0.7495 bits/symbol, if the symbol
probabilities are closer to 0.5, or in general in case of q-ary sources closer to 1=q.
As expected, for a binary symbol probability of 0.5 and error probability of 0,
we have I (X; Y )=1 bit/symbol.
e/ Lastly, let us determine the average information loss and average error entropy, which were dened in Equations 1.74 and 1.80 and are repeated here for
convenience. Again, we will be using some of the previously computed probabilities from Sections 1.10.1 and 1.10.4, beginning with computation of the average
information loss of Equation 1.74:
H (X=Y ) =
XX
X
P (Xi ; Yj ) log2 P (Xi =Yj )
P (X0 ; Y0 ) log2 P (X0 =Y0) P (X0 ; Y1 ) log2 P (X0 =Y1 )

P (X1 ; Y0 ) log2 P (X1 =Y0) P (X1 ; Y1 ) log2 P (X1 =Y1 )
= P (0; 0) log2 P (0=0) + P (0; 1) log2 P (0=1)
P (1; 0) log2 P (1=0) + P (1; 1) log2 P (1=1)
0:3 0:98 log2 0:9545 0:3 0:02 log2 0:00867
0:7 0:02 log2 0:04545 0:7 0:98 log2 0:9913
0:0198 + 0:0411 + 0:0624 + 0:0086
0:132 bit/symbol:
=
In order to augment the physical interpretation of the above-average information

loss expression, let us examine the main contributing factors in it. It is expected
to decrease as the error probability decreases. Although it is not straightforward
to infer the clear eect of any individual parameter in the equation, experience
shows that as the error probability increases, the two middle terms corresponding to the error events become more dominant. Again, the reader may nd it
instructive to alter some of the parameters on a one-by-one basis and study the
way its in uence manifests itself in terms of the overall information loss.
Moving on to the computation of the average error entropy, we nd its denition
equation is repeated below, and on inspecting Figure 1.16 we have:
H (Y=X ) =
XX
X
P (Xi ; Yj ) log2 P (Yj =Xi)
P (X0 ; Y0 ) log2 P (Y0 =X0 ) P (X0 ; Y1 ) log2 P (Y1 =X0 )

P (X1 ; Y0 ) log2 P (Y0 =X1 ) P (X1 ; Y1 ) log2 P (Y1 =X1 )
P (Y0 =X0) = 0:98
=
1.11. CAPACITY OF DISCRETE CHANNELS
P (Y0 =X1)
P (Y1 =X0)
P (Y1 =X1)
H (Y=X )
49
0:02
0:02
0:98
P (0; 0) log2 P (0=0) + P (0; 1) log2 P (0=1)
P (1; 0) log2 P (1=0) + P (1; 1) log2 P (1=1)
= 0:294 log2 0:98 0:014 log2 0:02
0:006 log2 0:02 0:686 log2 0:98
0:0086 + 0:079 + 0:034 + 0:02
0:141 bit/symbol:
=
=
=
=
The average error entropy in the above expression is expected to fall as the
error probability is reduced and vice versa. Substituting dierent values into
its denition equation further augments its practical interpretation. Using our
previous results in this section, we see that the average loss of information per
symbol or equivocation denoted by H (X=Y ) is given by the dierence between
the source entropy of Equation 1.82 and the average mutual information of
Equation 1.85, yielding:
H (X=Y ) = H (X ) I (X; Y ) 0:8816 0:7495 0:132 bit/symbol;

which according to Equation 1.75, is identical to the value of H (X=Y ) computed
earlier. In harmony with Equation 1.80, the error entropy can also be computed
as the dierence of the average entropy H (Y ) in Equation 1.83 of the received
symbols and the mutual information I (X; Y ) of Equation 1.85, yielding:
H (Y ) I (X; Y ) 0:8909 0:7495 0:141 bit/symbol;

as seen above for H (Y=X ).
Having dened the fundamental parameters summarized in Table 1.7 and used in
the information-theoretical characterization of communications systems, let us now
embark on the denition of channel capacity. Initially, we consider discrete noiseless
channels, leading to a brief discussion of noisy discrete channels, and then we proceed
to analog channels, before exploring the fundamental message of the Shannon-Hartley
law.
1.11
Capacity of Discrete Channels
[3, 10]
Shannon [3] dened the channel capacity C of a channel as the maximum achievable
information transmission rate at which error-free transmission can be maintained over
the channel.
Every practical channel is noisy, but transmitting at a suciently high power the
channel error probability pe can be kept arbitrarily low, providing us with a simple
initial channel model for our further elaborations. Following Ferenczy's approach [10],
50
assume that the transmission of symbol Xi requires a time interval of ti , during which
an average of
q
X
1
bit
H (X ) = P (Xi ) log2
(1.86)
P (Xi ) symbol
i=1
information is transmitted, where q is the size of the source alphabet used. This
approach assumes that a variable-length coding algorithm, such as the previously
described Shannon-Fano or the Human coding algorithm may be used in order to
reduce the transmission rate to as low as the source entropy. Then the average time
required for the transmission of a source symbol is computed by weighting ti with the
probability of occurrence of symbol Xi ; i = 1 : : : q:
tav =
q
X
i=1
P (Xi )ti
sec

symbol
(1.87)
Now we can compute the average information transmission rate v by dividing the
average information content of a symbol by the average time required for its transmission:
H (X ) bit
v=
:
(1.88)
tav sec
The maximum transmission rate v as a function of the symbol probability P (Xi )
must be found. This is not always an easy task, but a simple case occurs when
the symbol duration is constant; that is, we have ti = t0 for all symbols. Then the
maximum of v is a function of P (Xi ) only and we have shown earlier that the entropy
H (X ) is maximized by equiprobable source symbols, where P (Xi ) = q1 . Then from
Equations 1.86 and 1.87 we have an expression for the channel's maximum capacity:
C = vmax =
H (X ) log2 q bit
=
:
tav
t0 sec
(1.89)
Shannon [3] characterized the capacity of discrete noisy channels using the previously dened mutual information describing the amount of average conveyed information, given by:
I (X; Y ) = H (Y ) H (Y=X );
(1.90)
where H (Y ) is the average amount of information per symbol at the channel's output,
while H(Y/X) is the error entropy. Here a unity symbol-rate was assumed for the
sake of simplicity. Hence, useful information is transmitted only via the channel if
H (Y ) > H (Y=X ). Via a channel with pe = 0:5, where communication breaks down,
we have H (Y ) = H (Y=X ), and the information conveyed becomes I (X; Y ) = 0.
The amount of information conveyed is maximum if the error entropy H (Y=X ) = 0.
Therefore, Shannon [3] dened the noisy channel's capacity as the maximum value of
the conveyed information I (X; Y ):
C = I (X; Y )MAX = [H (Y ) H (Y=X )]MAX ;
(1.91)
where the maximization of Equation 1.91 is achieved by maximizing the rst term
and minimizing the second term.
1.11. CAPACITY OF DISCRETE CHANNELS
P (0) = 0:5
(1 pe )
pe
pe
P (1) = 0:5
(1 pe )
Figure 1.19:
51
-3
s-
BSC model.
In general, the maximization of Equation 1.91 is an arduous task, but for the BSC
seen in Figure 1.19 it becomes fairly simple. Let us consider this simple case and
assume that the source probabilities of 1 and 0 are P (0) = P (1) = 0:5 and the error
probability is pe . The entropy at the destination is computed as:
1
1
log
2 22
while the error entropy is given by:
1
1
log = 1 bit/symbol;
2 22
H (Y ) =
H (Y=X ) =
XX
X
P (Xi ; Yj ) log2 P (Yj =Xi ):
(1.92)
In order to be able to compute the capacity of the BSC as a function of the channel's
error probability, let us substitute the required joint probabilities of:
P (0; 0)
P (0; 1)
P (1; 0)
P (1; 1)
=
=
=
=
P (0)(1 pe )
P (0)pe
P (1)pe
P (1)(1 pe ):
(1.93)
(1 pe )
pe
pe
(1 pe ):
(1.94)
and the conditional probabilities of:
P (0=0)
P (0=1)
P (1=0)
P (1=1)
=
=
=
=
into Equation 1.92, yielding:
H (Y=X ) =
[P (0)(1 pe ) log2 (1 pe ) + P (0) pe log2 pe

+P (1) pe log2 pe + P (1)(1 pe ) log2 (1 pe )]
52
C=1+(1-pe)log2(1-pe)+pelog2pe
1.0
$RCSfile: bsc-capacity-vs-pe.gle,v $
0.8
0.6
0.4
0.2
0.0 -4
10
-3
10
10
-2
-1
10
0.2
0.5
pe
Figure 1.20:
Channel capacity versus pe for the BSC.
[P (0) + P (1)](1 pe ) log2 (1 pe )

+[P (0) + P (1)]pe log2 pe
= (1 pe ) log2 (1 pe ) pe log2 pe :
=
(1.95)
Finally, upon substituting H (Y ) and H (Y=X ) from above into Equation 1.91, the
BSC's channel capacity becomes:
C = 1 + (1 pe ) log2 (1 pe ) + pe log2 pe :
(1.96)
Following Ferenczy's [10] interpretation of Shannon's lessons [1{4,12,13], the graphic

representation of the BSC's capacity is depicted in Figure 1.20 using various pe error
probabilities.
Observe, for example, that for pe = 10 2 the channel capacity is C 0:9 bit/symbol,
that is, close to its maximum of C = 1 bit/symbol; but for higher pe values it
rapidly decays, falling to C = 0:5 bit/symbol around pe = 10 1. If pe = 50%,
we have C = 0 bit/symbol; since no useful information transmission takes place,
the channel delivers random bits. Notice also that if P (0) 6= P (1) 6= 0:5, then
H (Y ) < 1 bit/symbol and hence C < Cmax = 1 bit/symbol, even if pe = 0.
1.12. SHANNON'S CHANNEL CODING THEOREM

1.12
53
Shannon's Channel Coding Theorem [7, 15]
In the previous section, we derived a simple expression for the capacity of the noisy
BSC in Equation 1.96, which was depicted in Figure 1.20 as a function of the channel's
error probability pe . In this section, we focus on Shannon's channel coding theorem,
which states that as long as the information transmission rate does not exceed the
channel's capacity, the bit error rate can be kept arbitrarily low [12, 13]. In the context of the BSC channel capacity curve of Figure 1.20, this theorem implies that noise
over the channel does not preclude the reliable transmission of information; it only
limits the rate at which transmission can take place. Implicitly, this theorem prophesies the existence of an appropriate error correction code, which adds redundancy
to the original information symbols. This reduces the system's useful information
throughput but simultaneously allows error correction coding. Instead of providing
a rigorous proof of this theorem, following the approach suggested by Abramson [7],
which was also used by Hey and Allen [15] in their compilation of Feyman's lectures,
we will make it plaussible.
The theorem is stated more formally as follows. Let us assume that a message
of K useful information symbols is transmitted by assigning it to an N -symbol socalled block code, where the symbols are binary and the error probability is pe . Then,
according to Shannon, upon reducing the coding rate R = K
N beyond every limit, the
error probability obeys the following relationship:
R=
K
N
C = 1 + (1 pe ) log (1 pe ) + pe log pe :
2
(1.97)
As Figure 1.20 shows upon increasing the bit error rate pe , the channel capacity
reduces gradually toward zero, which forces the channel coding rate R = K
N to zero
in the limit. This inequality therefore implies that an arbitrarily low BER is possible
only when the coding rate R tends to zero, which assumes an innite-length block
code and an innite coding delay. By scrutinizing Figure 1.20, we can infer that,
1
for example, for a BER of 10 1 an approximately R = K
N 2 so-called half-rate
code is required in order to achieve asymptotically perfect communications, while for
BER = 10 2 an approximately R 0:9 code is required.
Shannon's channel coding theorem does not specify how to create error correction
codes, which can achieve this predicted performance; it merely states their existence.
Hence, the error correction coding community has endeavored over the years to create
such good codes but until 1993 had only limited success. Then in that year Berrou
et al. [16] invented the family of iteratively decoded turbo-codes, which are capable
of approaching the Shannonian predictions within a fraction of a dB.
Returning to the channel coding theorem, Hey and Feynman [15] oered a witty
approach to deepening the physical interpretation of this theorem, which we brie y
highlight below. Assuming that the block-coded sequences are long, in each block on
the average there are t = pe N number of errors. In general, t number of errors can
be allocated over the block of N positions in
CNt
N
N!
=
=
t!(N t)!
t
54
dierent ways, which are associated with the same number of error patterns. The
number of additional parity bits added during the coding process is P = (N K ),
which must be suciently high for identifying all the CNt number of error patterns, in
order to allow inverting (i.e., correcting) the corrupted bits in the required positions.
Hence, we have [15]:
N!
2(N K ):
(1.98)
t!(N t)!
Upon exploiting the Stirling formula of
N ! 2N
N
e
N
= 2 N N N e
and taking the logarithm of both sides, we have:

p
1
loge N ! loge 2 + loge N + N loge N N:
2
Furthermore, when N is large, the rst and second terms are diminishingly small in
comparison to the last two terms. Thus, we have:
loge N ! N loge N
N:
Then, after taking the logarithm, the factorial expression on the left-hand side (L) of
Equation 1.98 can be written as:
L [N loge N
N ] [t loge t t] [(N
t) loge (N
t) (N
t)] :
Now taking into account that t pe N , we have [15]:
[N loge N N ] [pe N loge (pe N ) pe N ]

[(N pe N ) loge (N pe N ) (N pe N )]
[N loge N N ] [pe N loge pe + pe N loge N pe N ]
[N loge (N (1 pe )) pe N loge (N (1 pe )) (N pe N )]
[N loge N N ] [pe N loge pe + pe N loge N pe N ]
[N loge N + N loge (1 pe ) pe N loge N
pe N loge (1 pe ) (N pe N )]
N [loge N 1 pe loge pe pe loge N + pe
loge N loge (1 pe ) + pe loge N
+ pe loge (1 pe ) + 1 pe ]
N [ pe loge pe loge (1 pe ) + pe loge (1 pe )]
N [ pe loge pe (1 pe ) loge (1 pe )]:
If we consider that loge a = log2 a loge 2, then we can convert the loge terms to log2
as follows [15]:
L N loge 2[ pe log2 pe (1 pe ) log2 (1 pe )]:
1.13. CAPACITY OF CONTINUOUS CHANNELS
55
Finally, upon equating this term with the logarithm of the right-hand side expression
of Equation 1.98, we arrive at:
N loge 2[ pe log2 pe (1 pe ) log2 (1 pe )] (N

which can be simplied to:
pe log2 pe
(1 pe ) log2 (1 pe ) 1
K ) loge 2;
K
N
or to a form, identical to Equation 1.97:

K
1 + (1 pe ) log2 (1 pe ) + pe log2 pe :
N
1.13
Capacity of Continuous Channels [4, 10]
During our previous discussions, it was assumed that the source emitted discrete
messages with certain nite probabilities, which would be exemplied by an 8-bit
analog-to-digital converter emitting one of 256 discrete values with a certain probability. However, after digital source encoding and channel encoding according to
the basic schematic of Figure 1.1 the modulator typically converts the digital messages to a nite set of bandlimited analog waveforms, which are chosen for maximum
\transmission convenience." In this context, transmission convenience can imply a
range of issues, depending on the communications channel. Two typical constraints
are predominantly power-limited or bandwidth-limited channels, although in many
practical scenarios both of these constraints become important. Because of their limited solar power supply, satellite channels tend to be more severely power-limited
than bandlimited, while typically the reverse situation is experienced in mobile radio
systems.
The third part of Shannon's pioneering paper [4] considers many of these issues.
Thus, in what follows we dene the measure of information for continuous signals,
introduce a concept for the continuous channel capacity, and reveal the relationships
among channel bandwidth, channel capacity, and channel signal-to-noise ratio, as
stated by the Shannon-Hartley theorem. Finally, the ideal communications system
transpiring from Shannon's pioneering work is characterized, before concluding with
a brief discussion of the ramications of wireless channels as regards the applicability
of Shannon's results.
Let us now assume that the channel's analog input signal x(t) is bandlimited
and hence that it is fully characterized by its Nyquist samples and by its probability
density function (PDF) p(x). The analogy of this continuous PDF and that of a
discrete source are characterized by P (Xi ) p(Xi )X , which re ects the practical
way of experimentally determining the histogram of a bandlimited analog signal by
observing the relative frequency of events, when its amplitude resides in a X wide
amplitude bin-centered around Xi . As an analogy to the discrete average information
or entropy expression of:
H (X ) =
X
i
P (Xi ) log2 P (Xi );
(1.99)
56
Shannon [4] introduced the entropy of analog sources, as it was also noted and exploited, for example, by Ferenczy [10], as follows:
Z 1
H (x) =
p(x) log2 p(x)dx:
(1.100)
1
For our previously used discrete sources, we have shown that the source entropy is
maximized for equiprobable messages. The question that arises is whether this is also
true for continuous PDFs. Shannon [4] derived the maximum of the analog signal's
entropy under the constraints of:
Z 1
p(x)dx = 1
(1.101)
1
Z 1
x2 =
x2 p(x)dx = Constant
(1.102)
1
based on the calculus of variations. He showed that the entropy of a signal x(t) having
a constant variance of x2 is maximum, if x(t) has a Gaussian distribution given by:
2
2
1
p(x) = p e (x =2 ) :
(1.103)
2
Then the maximum of the entropy can be derived upon substituting this PDF
into the expression of the entropy. Let us rst take the natural logarithm of both
sides of the PDF and convert it to base two logarithm by taking into account that
loge a = log2 a loge 2, in order to be able to use it in the entropy's log2 expression.
Then the PDF of Equation 1.103 can be written as:
p
1
log2 p(x) = + log2 2 + (x2 =22 )
;
(1.104)
loge 2
and upon exploiting that loge 2 = 1= log2 e, the entropy is expressed according to
Shannon [4] and Ferenczy [10] as:
Z
Hmax (x) =
=
p(x) log2 p(x)dx
p(x) log2 2dx +
= log2 2
p(x)dx +

log e
22 2
p
log e
= log2 2 + 2
2
p
1
= log2 2 + log2 e
2
p
= log2 2e:
= log2 2 +
x2 log2 e
dx
2 2
Z
p(x)
log2 e
x2 p(x)dx
22 | {z }
2
(1.105)
57
Since the maximum of the entropy is proportional to the logarithm of the signal's
average power Sx = x2 , upon quadrupling the signal's power the entropy is increased
by one bit because the range of uncertainty as regards where the signal samples can
reside is expanded.
We are now ready to formulate the channel capacity versus channel bandwidth and
versus channel SNR relationship of analog channels. Let us assume white, additive,
signal-independent noise with a power of N via the channel. Then the received
(signal+noise) power is given by:
y2 = S + N:
(1.106)
By the same argument, the channel's output entropy is maximum if its output signal
y(t) has a Gaussian PDF and its value is computed from Equation 1.105 as:
1
1
Hmax (y) = log2 (2ey2 ) = log2 2e(S + N ):
(1.107)
2
2
We proceed by taking into account the channel impairments, reducing the amount of
information conveyed by the amount of the error entropy H (y=x) giving:
I (x; y) = H (y) H (y=x);
(1.108)
where again the noise is assumed to be Gaussian and hence:

1
(1.109)
H (y=x) = log2 (2eN ):
2
Upon substituting Equation 1.107 and Equation 1.109 in Equation 1.108, we have:

2e(S + N )
1
log
2 2
2eN

1
S
=
log2 1 +
;
2
N
I (x; y) =
(1.110)
where, again, both the channel's output signal and the noise are assumed to have
Gaussian distribution.
The analog channel's capacity is then calculated upon multiplying the information
conveyed per source sample by the Nyquist sampling rate of fs = 2 fB , yielding [12]:

C = fB log2 1 +
S
N
bit
:
sec
(1.111)
Equation 1.111 is the well-known Shannon-Hartley law,1 establishing the relationship

among the channel capacity C , channel bandwidth fB , and channel signal-to-noise
ratio (SNR).
Before analyzing the consequences of the Shannon-Hartley law following Shannon's
deliberations [12], we make it plausible from a simple practical point of view. As we
1 Comment by the Authors: Although the loose denition of capacity is due to Hartley, the
underlying relationship is entirely due to Shannon.
58
have seen, the root mean squared (RMS) value

p of the noise is N , and that of the
signal plus noise at the channel's output is S + N . The receiver has to decide from
the noisy channel's output signal what signal has been input to the channel, although
this has been corrupted by an additive Gaussian noise sample. Over an ideal noiseless
channel, the receiver would be able to identify what signal sample was input to the
receiver. However, over noisy channels it is of no practical benet to identify the
corrupted received message exactly. It is more benecial to quantify a discretized
version of it using a set of decision threshold values, where the resolution is dependent
on how corrupted the samples are. In order to quantify this SNR-dependent receiver
dynamic range resolution, let us consider the following argument.
Having very densely spaced receiver detection levelsp would often yield noiseinduced decision errors, while a decision-level spacing of N according to the RMS
noise-amplitude intuitively seems a good compromise between high information resolution and low decision
p error rate. Then assuming a transmitted sample, which
p resides
at the center of a N wide decision interval, noise samples larger than N=2 will
carry samples across the adjacent decision boundaries. According to this spacing, the
number of receiver reconstruction levels is given by:
1
S+N
S 2
q= p
= 1+
;
(1.112)
N
N
which creates a scenario similar to the transmission of equiprobable q-ary discrete
symbols via a discrete noisy channel, each conveying log2 q amount of information at
the Nyquist sampling rate of fs = 2 fB . Therefore, the channel capacity becomes [12]:

S
C = 2 fB log2 q = fB log2 1 +
;
(1.113)
N
as seen earlier in Equation 1.111.
1.13.1 Practical Evaluation of the Shannon-Hartley Law

The Shannon-Hartley law of Equation 1.111 and Equation 1.113 reveals the fundamental relationship of the SNR, bandwidth, and channel capacity. This relationship
can be further studied following Ferenczy's interpretation [10] upon referring to Figure 1.21.
Observe from the gure that a constant channel capacity can be maintained, even
when the bandwidth is reduced, if a suciently high SNR can be guaranteed. For
example, from Figure 1.21 we infer that at fB = 10 KHz and SNR = 30 dB the
channel capacity is as high as C = 100 kbps. Surprisingly, C 100 kbps can be
achieved even for fB = 5 KHz, if SNR = 60 dB is guaranteed.
Figure 1.22 provides an alternative way of viewing the Shannon-Hartley law, where
the SNR is plotted as a function of fB , parameterized with the channel capacity C .
It is important to notice how dramatically the SNR must be increased in order to
maintain a constant channel capacity C , as the bandwidth fB is reduced below 0:1 C ,
where C and fB are expressed in kbit/s and Hz, respectively. This is due to the
log2 (1 + SNR) function in Equation 1.111, where a logarithmically increasing SNR
value is necessitated to compensate for the linear reduction in terms of fB .
59
$RCSfile: graph-representation-shannon-hartley-law.gle,v $
C=fB log2(1+S/N)
dB
dB
dB
60
80
10
0d
B
150
200
dB
[Kbit/s]
200
dB
dB
50
150
40
30d
100
20dB
50
10dB
fB
Figure 1.21:
10
12
[KHz]
Graphical representation of the Shannon-Hartley law. c Ferenczy [10].
From our previous discourse, the relationship between the relative channel capacity
C=fB expressed from Equation 1.113, and the channel SNR now becomes plausible.
This relationship is quantied in Table 1.8 and Figure 1.23 for convenience.
h
i Notice that due to the logarithmic SNR scale expressed in dBs, the C=fB bps
Hz curve
becomes near-linear, allowing a near-linearly proportional relative channel capacity
improvement upon increasing the channel SNR. A very important consequence of
this relationship is that if the channel SNR is suciently high to support communications using a high number of modulation levels, the channel is not exploited to its
full capacity upon using C=fB values lower than is aorded by the prevailing SNR.
Proposing various techniques in order to exploit this philosophy was the motivation
of reference [17].
The capacity C of a noiseless channel with SNR = 1 is C = 1, although noiseless
channels do not exist. In contrast, the capacity of an ideal system with fB = 1 is
nite [8, 11]. Assuming additive white Gaussian noise (AWGN) with a double-sided
power spectral density (PSD) of =2, we have N = 2 2 fB = fB , and applying
the Shannon-Hartley law gives [8]:
C = fB log2
S
1+
fB
60
500
$RCSfile: snr-vs-fb-for-shannon-hartley-law.gle,v $
C=fB log2(1+S/N) [bit/s] for fB in Hz

C=fB log2(1+S/N) [Kbit/s] for fB in KHz
C=fB log2(1+S/N) [Mbit/s] for fB in MHz
450
400
[dB]
350
300
C
S/N
250
=4
00
200
=2
00
150
C=
10
100
C=
50
50
0
fB
Figure 1.22:
10
12
14
[Hz/KHz/MHz]
SNR versus fB relations according to the Shannon-Hartley law. c Ferenczy

[10].
SNR
C=fB
Ratio dB bit/sec/Hz
1
0
1
3
4.8
2
7
8.5
3
15
11.8
4
31
14.9
5
63
18.0
6
127 21.0
7
Table 1.8:
Relative Channel Capacity versus SNR
Figure 1.23:
61
Relative channel capacity (C=fB ) versus SNR (dB).
=
=

S

fB
S
log2 1 +
S
fB

S
S
log2 1 +

fB
fSB
(1.114)
Our aim is now to determine C1 = limfB !1 C . Upon exploiting that:

1
limx!0(1 + x) x = e
where x = S=( fB ), we have
(1.115)

S
S
C1 = limfB !1 C = log2 e = 1:45
;

(1.116)
which is the capacity of the channel with fB = 1. The practically achievable transmission rate R is typically less than the channel capacity C , although complex turbocoded modems [16] can approach its value. For example, for a telephone channel with
a signal-to-noise ratio of S=N = 103 = 30 dB and a bandwidth of B = 3:4 kHz from
Equation 1.113, we have C = 3:4 log2 (1 + 103) kbit
sec 3:4 10 = 34 kbit=s, which is
fairly close to the rate of the V.34 CCITT standard 28.8 kbit/s telephone-channel
modem that was recently standardized.
62
Source
M-ary
encoder
Signaling
waveform
generator
Waveform
storage
Figure 1.24:
AWGN
channel
Waveform
detector
M-ary
decoder
Sink
Waveform
storage
Shannon's ideal communications system for AWGN channels.
In this chapter, we have been concerned with various individual aspects of Shannon's information theory [1{4,12,13]. Drawing nearer to concluding our discourse on
the foundations of information theory, let us now outline in broad terms the main
ramications of Shannon's work [1{4].
1.13.2 Shannon's Ideal Communications System

for Gaussian Channels
The ideal Shannonian communications system shown in Figure 1.24 has the following
characteristics. The system's information-carrying capacity is given by the information rate C = fB log2 (1 + S=N ), while as regards its error rate we have pe ! 0. The
transmitted and received signals are bandlimited Gaussian random variables, which
facilitate communicating at the highest possible rate over the channel.
Information from the source is observed for T seconds, where T is the symbol
duration and encoded as equiprobable M -ary symbols with a rate of R = logT2 M .
Accordingly, the signaling waveform generator of Figure 1.24 assigns a bandlimited
AWGN representation having a maximum frequency of fB from the set of M = 2RT
possible waveforms to the source message, uniquely representing the signal x(t) to
be transmitted for a duration of T . The noisy received signal y(t) = x(t) + n(t) is
compared to all M = 2RT prestored waveforms at the receiver, and the most \similar"
is chosen to identify the most likely transmitted source message. The observation
intervals at both the encoder and decoder amount to T , yielding an overall coding
delay of 2T . Signaling at a rate equal to the channel capacity is only possible, if the
source signal's observation interval is innitely long, that is, T ! 1.
Before concluding this chapter, we oer a brief discussion of the system-architectural
ramications of transmitting over wireless channels rather than over AWGN channels.
1.14
Shannon's Message and Its

Implications for Wireless Channels
In wireless communications over power- and bandlimited channels it is always of

prime concern to maintain an optimum compromise in terms of the contradictory
1.14. SHANNON'S MESSAGE FOR WIRELESS CHANNELS
63
requirements of low bit rate, high robustness against channel errors, low delay, and
low complexity. The minimum bit rate at which distortionless communications is
possible is determined by the entropy of the speech source message. Note, however,
that in practical terms the source rate corresponding to the entropy is only asymptotically achievable as the encoding memory length or delay tends to innity. Any
further compression is associated with information loss or coding distortion. Note
that the optimum source encoder generates a perfectly uncorrelated source-coded
stream, where all the source redundancy has been removed; therefore, the encoded
symbols are independent, and each one has the same signicance. Having the same
signicance implies that the corruption of any of the source-encoded symbols results
in identical source signal distortion over imperfect channels.
Under these conditions, according to Shannon's pioneering work [1], which was
expanded, for example, by Hagenauer [18] and Viterbi [19], the best protection against
transmission errors is achieved if source and channel coding are treated as separate
entities. When using a block code of length N channel coded symbols in order to
encode K source symbols with a coding rate of R = K=N , the symbol error rate can
be rendered arbitrarily low, if N tends to innity and the coding rate to zero. This
condition also implies an innite coding delay. Based on the above considerations and
on the assumption of additive white Gaussian noise (AWGN) channels, source and
channel coding have historically been separately optimized.
Mobile radio channels are subjected to multipath propagation and so constitute a more hostile transmission medium than AWGN channels, typically exhibiting path-loss, log-normal slow fading and Rayleigh fast-fading [20]. Furthermore,
if the signaling rate used is higher than the channel's coherence bandwidth, over
which no spectral-domain linear distortion is experienced, then additional impairments are in icted by dispersion, which is associated with frequency-domain linear
distortions. Under these circumstances the channel's error distribution versus time
becomes bursty, and an innite-memory symbol interleaver is required in Figure 1.1 in
order to disperse the bursty errors and hence to render the error distribution random
Gaussian-like, such as over AWGN channels. For mobile channels, many of the above
mentioned, asymptotically valid ramications of Shannon's theorems have limited
applicability.
A range of practical limitations must be observed when designing mobile radio
speech or video links. Although it is often possible to further reduce the prevailing
typical bit rate of state-of-art speech or video codecs, in practical terms this is possible
only after a concomitant increase of the implementational complexity and encoding
delay. A good example of these limitations is the half-rate GSM speech codec, which
was required to approximately halve the encoding rate of the 13 kbps full-rate codec,
while maintaining less than quadrupled complexity, similar robustness against channel
errors, and less than doubled encoding delay. Naturally, the increased algorithmic
complexity is typically associated with higher power consumption, while the reduced
number of bits used to represent a certain speech segment intuitively implies that
each bit will have an increased relative signicance. Accordingly, their corruption
may in ict increasingly objectionable speech degradations, unless special attention is
devoted to this problem.
In a somewhat simplistic approach, one could argue that because of the reduced
64
source rate we could accommodate an increased number of parity symbols using a

more powerful, implementationally more complex and lower rate channel codec, while
maintaining the same transmission bandwidth. However, the complexity, quality, and
robustness trade-o of such a scheme may not always be attractive.
A more intelligent approach is required to design better speech or video transceivers
for mobile radio channels [18]. Such an intelligent transceiver is portrayed in Figure 1.1. Perfect source encoders operating close to the information-theoretical limits
of Shannon's predictions can only be designed for stationary source signals, a condition not satised by most source signals. Further previously mentioned limitations
are the encoding complexity and delay. As a consequence of these limitations the
source-coded stream will inherently contain residual redundancy, and the correlated
source symbols will exhibit unequal error sensitivity, requiring unequal error protection. Following Hagenauer [18], we will refer to the additional knowledge as regards
the dierent importance or vulnerability of various speech-coded bits as source significance information (SSI). Furthermore, Hagenauer termed the condence associated
with the channel decoder's decisions as decoder reliability information (DRI). These
additional links between the source and channel codecs are also indicated in Figure 1.1.
A variety of such techniques have successfully been used in robust source-matched
source and channel coding [21, 22].
The role of the interleaver and de-interleaver seen in Figure 1.1 is to rearrange
the channel coded bits before transmission. The mobile radio channel typically in icts bursts of errors during deep channel fades, which often overload the channel
decoder's error correction capability in certain speech or video segments. In contrast
other segments are not beneting from the channel codec at all, because they may
have been transmitted between fades and hence are error-free even without channel
coding. This problem can be circumvented by dispersing the bursts of errors more
randomly between fades so that the channel codec is always faced with an \averagequality" channel, rather than the bimodal faded/nonfaded condition. In other words,
channel codecs are most ecient if the channel errors are near-uniformly dispersed
over consecutive received segments.
In its simplest manifestation, an interleaver is a memory matrix lled with channel
coded symbols on a row-by-row basis, which are then passed on to the modulator on
a column-by-column basis. If the transmitted sequence is corrupted by a burst of
errors, the de-interleaver maps the received symbols back to their original positions,
thereby dispersing the bursty channel errors. An innite memory channel interleaver
is required in order to perfectly randomize the bursty errors and therefore to transform
the Rayleigh-fading channel's error statistics to that of a AWGN channel, for which
Shannon's information theoretical predictions apply. Since in interactive video or
speech communications the tolerable delay is strictly limited, the interleaver's memory
length and eciency are also limited.
A specic deciency of these rectangular interleavers is that in case of a constant
vehicular speed the Rayleigh-fading mobile channel typically produces periodic fades
and error bursts at traveled distances of =2, where is the carrier's wavelength, which
may be mapped by the rectangular interleaver to another set of periodic bursts of
errors. A range of more random rearrangement or interleaving algorithms exhibiting
a higher performance than rectangular interleavers have been proposed for mobile
1.15. SUMMARY AND CONCLUSIONS
65
channels in [23], where a variety of practical channel coding schemes have also been
portrayed.
Returning to Figure 1.1, the soft-decision information (SDI) or channel state information (CSI) link provides a measure of condence with regard to the likelihood
that a specic symbol was transmitted. Then the channel decoder often uses this information in order to invoke maximum likelihood sequence estimation (MLSE) based
on the Viterbi algorithm [23] and thereby improve the system's performance with
respect to conventional hard-decision decoding. Following this rudimentary review of
Shannon's information theory, let us now turn our attention to the characterization
of wireless communications channels.
1.15
Summary and Conclusions
An overview of Shannonian information theory has been given, in order to establish a

rm basis for our further discussions throughout the book. Initially we focussed our
attention on the basic Shannonian information transmission scheme and highlighted
the dierences between Shannon's theory valid for ideal source and channel codecs
as well as for Gaussian channels and its ramications for Rayleigh channels. We also
argued that practical nite-delay source codecs cannot operate at transmission rates
as low as the entropy of the source. However, these codecs do not have to operate
losslessly, since perceptually unobjectionable distortions can be tolerated. This allows
us to reduce the associated bit rate.
Since wireless channels exhibit bursty error statistics, the error bursts can only
be randomized with the aid of innite-length channel interleavers, which are not
amenable to real-time communications. Although with the advent of high-delay
turbo channel codecs it is possible to operate near the Shannonian performance limits
over Gaussian channels, over bursty and dispersive channels dierent informationtheoretical channel capacity limits apply.
We considered the entropy of information sources both with and without memory
and highlighted a number of algorithms, such as the Shannon-Fano, the Human and
run-length coding algorithms, designed for the ecient encoding of sources exhibiting
memory. This was followed by considering the transmission of information over noisecontaminated channels leading to Shannon's channel coding theorem. Our discussions
continued by considering the capacity of communications channels in the context of
the Shannon-Hartley law. The chapter was concluded by considering the ramications
of Shannon's messages for wireless channels.
Chapter
The Propagation
Environment
2.1
The Cellular Concept
In mobile radio systems, communications take place between a stationary base station
(BS) and a number of roaming mobile stations (MSs) or portable stations (PSs) [24{
26]. The BS's and the MS's transmitter is expected to provide a suciently high
received signal level for the far-end receivers in order to maintain the required communications integrity, which is usually ensured by power control. The geographical
area in which this condition is satised is termed a trac cell, which typically has an
irregular shape, depending on the prevailing propagation environment determined by
terrain and architectural features as well as the local paraphernalia. In theoretical
studies, a simple hexagonal cell structure is often favored for its simplicity, where the
BSs are located at the centers of the cells.
In an ideal situation, the total bandwidth available to a specic mobile radio
system could be allocated within each cell, assuming no energy spills into the adjacent
cell's coverage area. However, since wave propagation cannot be shielded at the cell
boundary, PSs near the cell edge will experience approximately the same signal energy
within their channel bandwidth from at least two BSs. This phenomenon is known as
co-channel interference. A remedy for this problem is to divide the total bandwidth
Btotal in frequency slots of Bcell = Btotal =N and assign a mutually exclusive reduced
bandwidth of Bcell = Btotal =N to each trac cell within a cluster of N cells, as
demonstrated in Figure 2.1 for N = 7. The seven-cell clusters are then tessellated
in order to provide contiguous radio coverage. Observe from the gure that the
phenomenon of co-channel interference between the black co-channel cells having
an identical frequency set is not completely eliminated, but because of the increased
co-channel BS distance or frequency reuse distance, the interference is signicantly
reduced. Note also that in analytical and simulation-based interference studies the
\second tier" of interfering cells, which are hatched, is typically neglected.
67
68
CHAPTER 2. THE PROPAGATION ENVIRONMENT

5
6
4
1
5
6
7
2
4
1
6
3
5
6
1
5
6
5
6
3
2
4
1
3
1
7
2
4
6
3
4
1
5
6
7
5
4
1
5
6
7
2
Figure 2.1:
Hexagonal cells and seven-cell clusters.
A consequence of this cellular concept, however, is that the total number of MSs
that can be supported simultaneously over a unit area is now reduced by a factor of
N . This is because assuming a simple frequency division multiple access (FDMA)
scheme, where each MS is assigned a radio frequency carrier and a user bandwidth
of Buser now only M = Bcell =Buser number of MS can be serviced rather than
N M = Btotal =Buser . This problem can be circumvented by making the clusters as
small as the original cells, which is achieved by reducing the transmitted power. Further reduction of the cell size has the advantage of serving more and more users over
a certain xed-size geographical area, since the same frequency is reused within the
same xed-size area an increased number of times for a higher number of subscribers,
while requiring a reduced transmitted signal power and hence lightweight batteries. A
further favorable eect is that the smaller the trac cell, the more benign the propagation environment due to the presence of a dominant line-of-sight (LOS) propagation
path and due to the mitigated eects of the multipath propagation. These arguments
lead to the concept of micro- and picocells [27{29], which are often conned to the
size of a railway station or airport terminal and an oce, respectively. Cluster sizes
smaller than N = 7 are often used in practice in order to enhance the system's area
spectral eciency, but such schemes require modulation arrangements that are resilient against the increased co-channel interference, an issue to be addressed in more
depth in Chapter 5. These dierent cells will have to coexist in practical systems,
where, for example, an \over-sailing" macrocell can provide an emergency handover
capability for the microcells, when the MS roams in a propagation blindspot but can-
2.1. THE CELLULAR CONCEPT
Figure 2.2:
69
Various trac cells.
not hand over to another microcell, since in the target microcell no trac channels
are available.
The various cell scenarios are exemplied by Figure 2.2, where the conventional
macrocell BS is allocated to a high tower, illuminating a large area but providing
a rather hostile propagation cell, since often there is no LOS path between the BS
and PS. Hence, the communications are more prone to fading than in the stylized
microcell illuminated by the antenna at the top of the lower buildings. The smaller
microcells typically use lower transmit power, and they channel most of the energy in
the street canyon, which mitigates the signal's variability and thus has more benign
fading. Furthermore, microcells also reduce the signal's dispersion due to path-loss
dierences of the multipath components. Finally, indoor picocells provide typically
even better channels and tend to mitigate co-channel interferences due to partitions
and ceilings.
The previously mentioned handover process is crucial with regard to the perceived
Grade of Service (GOS). A wide range of dierent complexity techniques have been
proposed for the various existing and future systems, some of which are summarized
in Figure 2.3. Explicitly, the BS and PS keep compiling the statistics of a range of
communications quality parameters shown in the gure and weight them according
to the prevalent optimization criterion, before a handover or operating mode recon-
70
Figure 2.3:
Handover control and transceiver reconguration parameters.
guration command is issued, for example, in order to recongure the transceiver in a
higher bit-rate but less error resilient modulation mode, when the channel conditions
improve. Although the PS plays an active role in monitoring the various parameters,
these are typically reported to the BS, which carries out the required decisions.
The rst-generation Public Land Mobile Radio (PLMR) systems were designed for
low trac density, and the typical cell radius was often of the order of tens of miles.
Even the second-generation GSM system [30] was contrived to be able to cope with
the hostile large-cell environment of 35 km radius rural cells. These systems had to
incorporate sophisticated and power-hungry signal processing in order to be able to
combat a wide range of channel impairments. The less robust Pan-American Digital
Advanced Phone System (DAMPS) [31] and the second-generation Japanese digital
mobile radio (JDMR) systems [32] re ect the more recent trend of moving toward
high-capacity small cells. The indoors-oriented cordless telephone or CT systems
CT2 [33] and DECT [34] also adopted this tendency.
Following this rudimentary introduction, let us now concentrate on the propagation environment.
2.2. RADIO WAVE PROPAGATION

2.2
71
Radio Wave Propagation
2.2.1 Background
The mobile radio channel has been characterized in references [20,24,25,35,36]. Here
we restrict our treatment to a rudimentary introduction so that we can arrive at the
third-generation Personal Communications Network (PCN) concept.
The propagation laws obeyed by radio waves are not dissimilar to those known
from our daily experience with light sources, whose beams propagate through direct
LOS paths, through re ection, diraction, and so on, although dierences exist because of their rather dierent wavelenghts. Just as the light of a torch diminishes
with increasing distance, so typically does the average received signal power of the
MS, when moving away from the BS in a radial direction. This phenomenon is known
as path-loss.
For the sake of illustration, let us assume that the MS orbits around the BS
on a hypothetical orbit in free space. Then the average received signal power is
constant, since no multipath propagation takes place owing to the lack of re ecting
and scattering objects. In a typical real propagation scenario, when the BS transmits a
sinusoidal radio frequency carrier, the MS receives a plethora of delayed, phase shifted,
and attenuated replicas, which add vectorially at the MS's antenna. These so-called
multipath components sometimes add constructively, sometimes destructively, which
results in rapid received signal amplitude and phase uctuations, known as fast-fading,
often in icting signal level variations up to 40 dB. This process is depicted vectorially
for the simple case of a two-path channel in Figure 2.4, where both paths have an
identical signal magnitude and one of them has a constant phase associated with a
constant vector, while the other one rotates, as the MS moves along.
The third characteristic propagation phenomenon, referred to as slow fading, obeys
medium-term path-loss variations due to the characteristic building patterns, obstructing large vehicles, and so on, which introduce shadowing eects.
When the length of the various propagation paths is approximately the same, because there are no signicant far-eld re ections due to mountains or high-rise buildings on the horizon, then the channel's impulse response is sharply decaying similarly
to a Dirac delta function. This is referred to as a nondispersive channel, since the
transmitted signaling pulses are not subjected to time dispersion along the time axis.
As the signaling rate increases, the adjacent signaling pulses are more densely spaced
and the eect of time dispersion on the reduced-duration signaling pulses becomes
relatively more signicant. When the signaling pulses start to overlap, intersymbol interference (ISI) is encountered, which rapidly degrades the communications integrity.
Note that in contrast to the high-elevation antennas of the conventional PLMR systems, the previously introduced microcellular systems have low-elevation antennas,
typically situated on lamp posts below the urban skyline. This property is advantageous in terms of channeling most of the transmitted energy into the street canyons
and hence mitigating channel dispersion and supporting higher signaling rates than
PLMR systems.
If the channel's impulse response is a sharply decaying Dirac delta, its frequencydomain transfer function given by the Fourier transform is constant across the entire
72
Figure 2.4:
Two-path fading channel.
bandwidth. In practice, there are no such innite bandwidth channels, but a lowspread impulse response is always associated with a frequency-domain transfer function, which is constant across a wide frequency range. If this near-constant passband
domain is signicantly wider than the transmitted signal's bandwidth, then the ISI
due to dispersion is insignicant and the narrowband propagation conditions apply.
Among these conditions every frequency component of the transmitted signal is subjected to the same fading, a condition that in wave propagation parlance is described
as frequency at-fading as distinct from dispersive wideband channel conditions or
frequency-selective fading, where dierent frequencies fade dierently.
If the channel's frequency-domain transfer function is not constant over the signal's bandwidth, the linear distortions produced must be equalized by a channel
equalizer, before high-integrity demodulation can take place. The employment of
channel equalizers in partial response modulation schemes [37], such as that of the
large-cell GSM PLMR system [30], is fairly common, and, somewhat surprisingly, the
wideband channel conditions often result in performance improvements because of
their inherent diversity eect in terms of providing several staggered replicas of the
transmitted signal. The channel equalizer can lock on to these staggered, dierently
fading multipath components, eliminate the dispersion introduced by them, and exploit the fact that the independently fading received paths are not likely to be in
73
fade simultaneously. This scenario is brie y considered in the context of the PanEuropean GSM system later in this chapter. In contrast, in order to avoid the use
of complex and power-thirsty equalizers, third-generation PCNs are designed to have
low-dispersion microcellular trac cells. In order to underline the importance of this
scenario for PCNs, narrowband propagation will be the subject of a more detailed
discussion in the next section.
2.2.2 Narrowband Fading Channels

Since most existing public mobile radio systems use the 0.9-1.8 GHz band, we concentrate on channels in this range. These frequencies fall in the Ultra High Frequency
(UHF) band. At these frequencies the signal level drops rapidly, once over the radio horizon, advantageously limiting co-channel interference, when the frequencies are
reused in neighboring cell clusters. At these frequencies, even if there is no line-of-sight
path between the transmitter and receiver, sucient signal power may be received by
means of wave scattering, re ection, and diraction to allow communications.
The basic factors of path-loss, slow fading, and fast-fading in uencing the wave
propagation phenomena are illustrated in Figure 2.5.
The prediction of the expected mean or median received signal power plays a
crucial role in planning the coverage area of a specic base station and in determining
the closest acceptable reuse of the propagation frequency employed. For high antenna
elevations and large rural cells, a more slowly decaying power exponent is expected
than for low elevations and densely builtup urban areas. As mentioned, the received
signal is also subjected to slow or shadow fading, which is governed mainly by terrain
and topographical features in the vicinity of the mobile receiver, such as small hills and
tall buildings. When designing the system's power budget and coverage area pattern,
the slow-fading phenomenon is taken into account by adding a shadow fading margin
to the required transmitted power.
Slow fading is compensated by increasing the transmitted power Ptx by the slowfading margin Lslow illustrated in Figure 2.5, which is usually chosen to be the 1 2%
quantile of the slow-fading probability density function (PDF), in order to minimize
the probability of unsatisfactorily low received signal power Prx . In addition, the
short-term fast signal fading due to multipath propagation is taken into account by
employing the fast-fading margin Lf ast of Figure 2.5, which is typically chosen to
be a few percent quantile of the fast-fading distribution. There remains a certain
probability that both of these fading margins are simultaneously exceeded by the
superimposed slow and fast envelope fading. This situation is often referred to as
fading margin overload and results in a very low received signal level, which may
cause call dropping in mobile telephony, if this condition persists over a period of
time. The probability of this worst-case event can be taken to be the sum of the
individual margin overload probabilities.
2.2.3 Propagation Path-loss Law

When considering the propagation path-loss of Figure 2.5, the parameter of prime
concern is normally the distance from the BS. Path-loss increases with distance due
74
Figure 2.5:
Power-budget design for mobile systems.
to the increasing illumination area of the spherical wavefront expanding from the
BS, which reduces the power ux per unit area. An in-depth treatment of pathloss calculations can be found in references by Jakes [24], Lee [25], and Parsons [36].
Suce to note here that path-loss calculations become signicantly more complex
when there is any form of obstruction between the transmitter and receiver. In large
cells, these obstructions will often be hills or undulating terrain. In modelling tools,
these obstacles are often considered to be \knife edges" at the point of the summit of
the obstruction, and knife edge diraction techniques can then be used [36] to predict
the path-loss. The analysis becomes increasingly complex if there are a number of
hills between the transmitter and receiver. An alternative approach is to determine
which Fresnel zone the hill obstructs and to perform the prediction accordingly [38].
As mentioned earlier, in urban areas modeling becomes more complex, and in the
case of large cells it is rarely possible to model each individual building, although
this can be achieved in microcells. Typically, this situation is resolved by adding
clutter-loss dependent on the density of the buildings in the urban area and also
allowing for a shadowing loss behind buildings, when computing the link budget
75
according to Figure 2.5. This approximation leads to suboptimal network design,

with overlapping cells and inecient frequency reuse. Nevertheless, it seems unlikely
that signicantly more accurate tools will become available in the immediate future.
In our probabilistic approach, it is dicult to give a worst-case path-loss exponent for
any mobile channel. However, it is possible to specify the most optimistic scenario.
This is given by propagation in free space. The free-space path-loss, Lpl , is given
by [36]:
Lpl = 10 log10 GT + 10 log10 GR
20 log10 f MHz
20 log10 dkm + 147:6dB; (2.1)
where GT and GR are the transmitter and receiver antenna gains, f is the propagation frequency in MHz, and d is the distance from the BS antenna in km. Observe
that the free-space path-loss is increased by 6 dB every time, when the propagation
frequency is doubled or the distance from the mobile is doubled. This corresponds to
a 20 dB/decade decay when moving away from the BS, and at d=1 km, f=1 MHz,
and GT = GR = 1 a path-loss of LF = 147:6 dB is encountered. Not only technological diculties, but also propagation losses discourage the deployment of higher
frequencies. Nevertheless, spectrum is usually only available at increasingly higher
frequency bands.
In practice, for the UHF mobile radio propagation channels of interest, the freespace conditions do not apply. However, a number of useful path-loss prediction
models can be adopted to derive other prediction bounds. One such case is the
\plane-earth" model. This is a two-path model constituted by a direct line-of-sight
path and a ground-re ected one, which ignores the curvature of the earth's surface.
Assuming transmitter base station (BS) and receiver mobile station (MS) antenna
m
heights of hm
BS ; hMS d expressed in meters, respectively, the plane-earth path-loss
formula [36] can be derived:
Lpl = 10 log10 GT + 10 log10 GR + 20 log10 hm

BS +
km ;
20 log10 hm
40
log
10 d
MS
(2.2)
where the dependence on propagation frequency is now absent. Observe that a 6 dB

path-loss reduction results when doubling the transmitter or receiver antenna elevations, and there is an inverse fourth power law decay with increasing the BS-MS
distance d. In the close vicinity of the transmitter antenna, where the condition hm
BS
or hm
MS d does not hold, Equation 2.2 is no longer valid.
Hata [39] developed three path-loss models for large cells, which are widely used,
forming the basis for many modeling tools. These were developed from an extensive
database derived by Okumura [40] from measurements in and around Tokyo. The
typical urban Hata model is dened as:
LHu = 69:55 + 26:16 log10 f MHz 13:82 log10 hm

BS
km [dB ];
(44:9 6:55 log10 hm
)
log
d
10
BS
a(hm
MS ) +
(2.3)
where f is the propagation frequency in MHz, hBS and hMS are the BS and MS
antenna elevations in terms of meters, respectively, a(hMS ) is a terrain-dependent
76
correction factor, while d is the BS-MS distance in km. The correction factor a(hMS )
for small- and medium-sized cities was found to be [39]:
a(hMS ) = (1:1 log10 f MHz
0:7)hm
MS
(1:56 log10 f MHz
0:8);
(2.4)
1:1 if f MHz 200MHz :

4:97 if f MHz 400MHz
(2.5)
while for large cities the connection factor is frequency-parametrized:
a(hMS ) =
2
8:29[log10 (1:54hm
MS )]
m
3:2[log10 (11:75hMS )]2
The typical suburban Hata model applies a correction factor to the urban model
yielding:
LHsuburban = LHu 2[log10 (f MHz =28)]2 5:4 [dB ]:
(2.6)
The rural Hata model modies the urban formula dierently:
LHrural = LHu
4:78(log10 f MHz )2 + 18:33 log10 f MHz
40:94 [dB ]:
(2.7)
The limitations of its parameters as listed by Hata are [39]:
f:
hBS :
hMS :
d:
150 1500MHz
30 200m
1 10m
1 20km:
For a 900 MHz PLMR system, these conditions can usually be satised, but for a
1.8 GHz typical PCN urban microcell all these limits have to be stretched.
For the sake of illustration in Figure 2.6 we have plotted a typical microcellular
path-loss regression line tting and the corresponding Hata path-loss characteristic
for an antenna elevation of 6.4 m and propagation frequency of 1.9 GHz, as a function
of the logarithmic distance from the BS. Observe that the measurement points show a
denser clustering for higher distances than for logarithmic distances around 2, which
corresponds to 100 m. This is a consequence of the limited number of streets in the
BS's vicinity, and this fact gives a higher weighting to points allocated around 1000 m,
corresponding to a logarithmic distance of 3. The path-loss exponent in this example
appears somewhat lower than that of the corresponding urban Hata model.
2.2.4 Slow-Fading Statistics

Having highlighted a few propagation path-loss prediction models, we now brie y
focus our attention on the characterization of the slow-fading phenomenon, which
constitutes the second component of the overall power budget design of mobile radio
links, as portrayed in Figure 2.5. In slow-fading analysis, the eects of fast-fading and
path-loss have to be removed. The fast-fading uctuations are removed by averaging
the signal level over a distance of typically some 20 wavelengths. The slow-fading
uctuations are separated by subtracting the best-t path-loss regression estimate
from each individual 20-wavelength-spaced averaged received signal value. A slowfading histogram derived this way is depicted in Figure 2.7. The gure suggests
a log-normal distribution in terms of dBs caused by normally distributed random
77
PATH-LOSS REGRESSION
160
Path-loss (dB)
150
140
130
120
110
100
90
1.5
2.5
3.5
Log. of Distance from BS (m)

Measured
Figure 2.6:
Hata
Regression
Typical microcellular path-loss regression line tting and the corresponding

Hata path-loss characteristic for an antenna elevation of 6.4 m and propagation
frequency of 1.9 GHz [20]. c Wiley 1999, Greenwood & Hanzo.
shadowing eects. Indeed, when subjected to rigorous distribution tting using the
log-normal hypothesis, the hypothesis is conrmed at a high condence level. The
associated standard deviation from the expected path-loss that is, from the long-term
mean value in this particular case, is 6.5 dB.
2.2.5 Fast-Fading Statistics

Regardless of the distribution of the numerous individual constituent propagation
paths of both the in-phase and quadrature component (ai ; aq ) of the received signal,
their distribution will be normal or Gaussian due to the central limit theorem. Then
the complex baseband equivalent signal's amplitude and phase characteristics are
given by:
q
a2i (k) + a2q (k);

a(k) =
(k) = arctan[aq (k)=ai (k)]:
(2.8)
(2.9)
Our aim is to determine the distribution of the amplitude a(k), if ai (k) and aq (k) are
known to have a normal or Gaussian distribution.
78
Figure 2.7:
Typical microcellular slow-fading histogram [20]. c Wiley 1999, Greenwood &

Hanzo.
In general, for n normally distributed random constituent processes with means
ai and identical variances 2 , the resultant process y =

with a PDF given below [41]:
p(y) =
where
1 y (n
22 s2
2)=4
e
(s2 +y )=2 2
y0
and
s2 =
n
X
i=1
(ai )2
n
X
i=1
I n=
(
a2i has a 2 distribution
2) 1
py s
2
(2.10)
(2.11)
(2.12)
is the noncentrality parameter computed from the rst moments of the component
processes a1 an . If the constituent processes have zero means, the 2 distribution
is central; otherwise it is noncentral. Each of these processes has a variance of 2 ,
79
and Ik (x) is the modied Kth order Bessel function of the rst kind, given by
1 (x=2)k+2j
X
Ik (x) =
; x 0:
(2.13)
j ! (k + j + 1)
j =0
The
function is dened as
(p) =
Z
0
tp 1 e tdt
if p > 0
(p) = (p 1)!
if p > 0 integer
p
p
3
1
:
( ) = ; ( )=
2
2
2
(2.14)
In our case we have two quadrature components,

that is, n = 2; s2 = (ai )2 + (aq )2 ,
q
p
the envelope is computed as a = y = a2i + a2q ; a2 = y; p(a)da = p(y)dy, and
hence p(a) = p(y)dy=da = 2ap(y), yielding the Rician PDF [41]:
as
2
2
2
a
pRice (a) = 2 e (a +s )=2 Io 2
a 0:
(2.15)

Formally introducing the Rician K -factor as
K = s2 =22
(2.16)
renders the Rician distribution's PDF dependent on one parameter only, yielding [41]:

a p
a2
a
22 e K Io
2
K
;
(2.17)

e

2

where K physically represents the ratio of the power received in the direct line-of-sight
path to the total power received via indirect scattered paths. Therefore, if there is no
dominant propagation path, K = 0, e K = 1 and I0 (0) = 1, yielding the worst-case
Rayleigh PDF [41]:
a a2
pRayleigh(a) = 2 e 22 :
(2.18)

Conversely, in the clear direct line-of-sight situation with no scattered power, K =
1, yielding a Dirac delta-shaped PDF, representing a step-function-like cumulative
distribution function (CDF). The signal at the receiver then has a constant amplitude
with a probability of one. Such a channel is referred to as a Gaussian channel. This is
because although no fading is present, the receiver will still generate thermal additive
white Gaussian noise (AWGN). If the K -factor is known, the fast-fading envelope's
distribution is completely described. The set of Rician PDFs for K = 0, 1, 2, 4,
10, and 15 is plotted at the top of Figure 2.8, while the corresponding time-domain
waveforms are portrayed in Figure 2.9.
Observe in the gures that for the worst-case Rician channel constituted by the
Rayleigh fading channel we have K = 0 or K = 1 dB , although K = 40 dB
practically corresponds to conditions similar to Rayleigh fading. In contrast, the best
Rician channel is the Gaussian channel, where there is no envelope fading at all,
pRice (a) =
80
Figure 2.8:
Rician PDFs. c Webb & Hanzo [17].
81
fading-mag-ph.gle
Rayleigh magnitude (dB)
Rayleigh phase (Deg.)

200
20
0
0
-20
-40
0.0
0.25
Time (s)
0.5
-200
0.0
Ricean k=0dB magnitude (dB)
0.25
Time (s)
0.5
Ricean k=0dB phase (Deg.)

200
20
0
0
-20
-40
0.0
0.25
Time (s)
0.5
-200
0.0
0.25
Time (s)
0.5

200
20
0
0
-20
-40
0.0
0.25
Time (s)
0.5
-200
0.0
0.25
Time (s)
0.5

200
20
0
0
-20
-40
0.0
0.25
Time (s)
0.5
-200
0.0
0.25
Time (s)
0.5

200
20
0
0
-20
-40
0.0
Figure 2.9:
0.25
Time (s)
0.5
-200
0.0
0.25
Time (s)
0.5
Typical Rayleigh and Ricean fast-fading and phase proles for a MS speed of
30 mph and carrier frequency of 900 MHz.
82
since the received signal level is constant and the only impairment is the additive
white Gaussian noise (AWGN). Accordingly, the fading PDF of an AWGN channel
is a Dirac function at the average received signal level, while the receiver noise is
represented by the usual Gaussian bell-shaped PDF. Expressed in terms of the CDF,
the fading CDF of the AWGN channel is a step-function positioned at the mean
received signal level, which also corresponds to the constant instantaneous received
signal level. For the AWGN channel we have K = 1, although Rician channels
having K -factors in excess of about 15 can be considered near-Gaussian.
A range of typical Rician fading channels with K -factors between 0 and 15 are
also shown in Figure 2.9, where for high K -factors the envelope fading is seen to
be conned to a few dB, while the phase is virtually linear and restricted to [].
The probability density function (PDF), cumulative distribution function (CDF),
and logarithmic CDF of a set of Rayleigh and Rician channels can be compared in
Figure 2.8, where the K -factors are now expressed in actual ratios rather than in the
logarithmic dB-domain. Observe at the bottom of Figure 2.8 that the Rician CDFs
are plotted on a convenient logarithmic vertical probability scale, which reveals the
enormous dierence in terms of deep fades for the K values considered. When choosing
the fading margin overload probability of Figure 2.5, dened as the probability of
receiving a signal level below the designed received signal level of a system, the bottom
curves of Figure 2.8 are particularly useful, since they expand the tails of the CDFs,
where, for example, for a Rician CDF with K = 1 the 15 dB fading margin overload
probability is seen to be approximately 10 2 .
Note that the relationship between the variances and means of the Rayleigh and
the component Gaussian distributions is given by [41]:
R2 = (2 =2) 2
and
mR = 2 =2 = =2;
respectively. The Rayleigh distribution's variance is not twice that of the composite
Gaussian processes, since we are adding the squares of the Gaussian processes, not
the processes themselves.
The Rician CDF takes the shape of [41]
X
1 s m
a s
a2

Im 2
CRice (a) = 1 e K + 22
a

m=0
!m
p
p !

1
2
a 2K
K + 2a2 X 2K
= 1 e
Im :
a
m=0
(2.19)
This formula is more dicult to evaluate than the PDF of Equation 2.17 because of the
summation of an innite number of terms, requiring double or quadruple computation
precision, and it is avoided in numerical evaluations, if possible. However, in practical
terms it is sucient to increase m to a value, where the last term's contribution
becomes less than 0.1%. Having characterized the path-loss as well as the slow-fading
83
INCIDENT
WAVES FROM
DISTANT BS
DIRECTION OF MOTION
Figure 2.10:
Relative path-length change due to the MS's movement. c Webb &

Hanzo [17].
and fast-fading statistics of narrowband mobile channels, let us now consider the rate
or frequency of fast-fading, which are explicitly related to the velocity of motion. In
the next section, we will introduce the Doppler frequency, which is a function of the
vehicular speed, and highlight the in uence of vehicular velocity on the correlation
between consecutive Rayleigh-distributed fading samples.
2.2.6 Doppler Spectrum

Having described the fading statistics, let us now concentrate on the eects of the
Doppler shift and consider again the transmission of an unmodulated carrier frequency
fc from a BS. Consider a MS traveling in a direction at an angle of i with respect to
the signal received on the ith path, as seen in Figure 2.10, which advances a distance
of d = v 4t during 4t, when traveling at a velocity of v. This introduces a relative
carrier phase change of 4 = 2, if the ight of the wave is shortened by an amount
of 4l = , where the wavelength can be computed from = c=fc, with c being the
velocity of light.
For an arbitrary 4l value we have 4 = 2 4l=, where the negative sign
implies that the carrier wave's phase delay is reduced, if the MS is traveling toward
the BS. From the simple geometry of Figure 2.10 we have 4l = d cos i ; therefore
the phase change becomes:
4 =
2v4t cos i
:

(2.20)
The Doppler frequency can be dened as the phase change due to the movement of
84
the MS during the innitesimal interval 4t:

1 4
fD =
:
(2.21)
2 4t
When substituting Equation 2.20 into 2.21 we get:
v
fD = cos i = fm cos i ;
(2.22)

where fm = v= = vfc=c is the maximum Doppler frequency deviation from the
transmitted carrier frequency due to the MS's movement. Notice that a Doppler
frequency can be positive or negative depending on i and that the maximum and
minimum Doppler frequencies are fm. These extreme frequencies correspond to
i = 0 and 180, when the ray is aligned with the street that the MS is traveling
along, and correspond to the ray coming toward or from behind the MS, respectively.
It is analogous to the change in the frequency of a whistle from a train perceived by
a person standing on a railway line, when the train is bearing down or receding from
the person, respectively.
According to Equation 2.22 and assuming that i is uniformly distributed, the
Doppler frequency has a random cosine distribution. The received power in an angle
of d around i is given by p(i )d, where p(i ) is the probability density function
(PDF) of the received power, which is assumed to be uniformly distributed over
the range of 0 i 2, giving p(i ) = 1=(2). The Doppler power spectral
density S (fD ) can be computed using Parseval's theorem by equating the incident
received power p(i )d in an angle d with the Doppler power S (fD )dfD , yielding
S (fD ) = di =(2 dfD ). Upon expressing i from fD = fm cos i and exploiting that
d cos 1 x
=
dx
1 x2
we then have:
S (fD ) =
d(cos 1 fD =fm)
1
di
p
=
=
2dfD
2dfD
2fm 1 (fD =fm)2
(2.23)
The incident received power at the MS depends on the power gain of the antenna and
the polarization used. Thus, the transmission of an unmodulated carrier is received
as a \smeared" signal whose spectrum is not a single carrier frequency fc but contains
frequencies up to fc fm . In general, we can express the received RF spectrum S (fD )
for a particular MS speed, propagation frequency, antenna design, and polarization
as [42]:
C
S (fD ) = p
;
(2.24)
1 (fD =fm)2
where C is a constant that absorbs the 1=2fm multiplier in Equation 2.23. Notice
that the Doppler spectrum of Equation 2.24 becomes S (fD = 0) = C at fD = 0,
while S (fD = fm ) = 1, when fD = fm . Between these extreme values S (f ) has a
U-shaped characteristic, as it is portrayed in a stylized form in the simulation model
of Figure 2.11, which will be developed in the next subsection.

n (t)
AWGN
S(f)
85
Mag.=SQRT[I + R ]
-fm
fm
RAYLEIGH
MAGNITUDE
PHASE
n (t)
AWGN
Phase=arctan(I/R)
-fm
Figure 2.11:
S(f)
fm
Baseband Rayleigh-fading simulation model. c Webb & Hanzo [17].
2.2.7 Simulation of Narrowband Fading Channels

The Rayleigh-fading channel model used in the previous subsection can be represented
by the quadrature arrangement shown in Figure 2.11, where the distribution of both
received quadrature components (ai ; aq ) is normal due to the central limit theorem.
These can be modeled as uncorrelated normally distributed AWGN sources. The outputs from the AWGN sources are applied to low-pass-type lters, having \U"-shaped
frequency-domain transfer functions that represent the eects of Doppler frequency
shifts, as will be demonstrated.
Observe that the maximum Doppler frequency fm = v= = v fc =c depends on
the product of the speed v of the MS and the propagation frequency fc. Therefore,
the higher the speed or the propagation frequency, the wider the frequency band over
which the received carrier is \smeared." The eect of employing the Doppler lter
in Figure 2.11 is that the originally uncorrelated quadrature components (nI ; nQ ) are
now eectively low-pass ltered, which introduces some intersample correlation by
restricting the maximum rate of change of (nI ; nQ ), while retaining their Gaussian
distribution. This is true, because passing an AWGN process through a linear system,
such as a low-pass lter, will reduce the variance of the signal but will retain the
Gaussian nature of the distribution for both quadrature components. Having no lowpass Doppler lter at all would be equivalent to fm = 1, implying an innitely high
vehicular speed and hence allowing the reception of arbitrary uncorrelated frequencies,
when transmitting the unmodulated carrier frequency of fc .
The simulation of the Rayleigh-distributed fast-fading envelope is based on the
concept portrayed in Figure 2.11, where the Rayleigh-distributed magnitude and uniformly distributed phase are then computed from the ltered real and imaginary
AWGN components R and I as follows: Magnitude=SQRT[I 2 + R2 ], Phase=arctan
(I=R). We note in closing that the above narrowband Rayleigh distributed fading
channel model can be implemented in both time- and frequency-domain, as will be
86
detailed in the next two paragraphs.
2.2.7.1 Frequency-Domain Fading Simulation

In the frequency-domain approach, one could exploit the fact that the Fourier transform of an AWGN process is another AWGN process. Hence the frequency-domain
AWGN sequence generated by the Box-Muller algorithm to be described below can
be multiplied by the frequency-domain Doppler transfer function of Equation 2.24
and then transformed back by IFFT to the time-domain in order to generate the
real and imaginary components (R; I ) of the ltered complex Gaussian process. The
main diculty associated with this frequency-domain approach is that the Doppler
lter's bandwidth fm is typically much lower than the sampling frequency, at which
the faded samples are produced. The faded samples have to be generated exactly at
the signaling rate at which modulation symbols are transmitted in order to supply a
fading envelope and phase rotation value for their corruption.
A typical Doppler bandwidth fm at a carrier frequency of fc = 1:8 GHz and
vehicular speed of v = 30mph = 13:3m=s is given by
fm =
v fc 13:3 ms 1:8 109 1s

=
c
3 108 ms
80 Hz;
where c = 3 108m=s is the speed of light. This fm value becomes about 40 Hz at fc =

900 MHz and 8 Hz at 1.8 GHz and a pedestrian speed of 3 mph. Typical multiuser
signaling rates in state-of-the-art Time Division Multiple Access (TDMA) systems are
in excess of fs = 40KBd. Therefore, the pedestrian relative Doppler bandwidth Dr
becomes Dr = 8 Hz=40 KHz = 2 10 4 , suggesting that only a negligible proportion
of the input samples will fall in the Doppler lter's passband. This fact may severely
aect the statistical soundness of this approach, unless a suciently high number
of frequency-domain AWGN samples is retained after removing those outside the
Doppler lter's bandwidth.
2.2.7.2 Time-Domain Fading Simulation

When using the time-domain approach, the AWGN samples generated by the BoxMuller algorithm to be described in the next paragraph must be convolved with the
impulse response hD (t) of the Doppler lter S (f ) of Equation 2.24 in both quadrature
arms of Figure 2.11. Similarly to the frequency-domain method, we are faced with a
dicult notch-ltering problem. Since the lter bandwidth is typically very narrow
in comparision to the sampling frequency, the impulse response hD (t) will be very
slowly decaying, requiring tens of thousands of tap values to be taken into account
in the convolution in case of a low Doppler bandwidth. This problem is aggravated
by the U-shaped nature of the Doppler's transfer function S (f ). The consequence
of these diculties is that the condence measures delivered by rigorous goodnessof-t techniques such as the 2 -test or the Kolmogorov-Smirnov test described, for
example, in Chapter 2 of reference [26] might become low, although the resulting
fading envelope and phase trajectories are acceptable for simulation studies of mobile
radio systems.
87
2.2.7.3 Box-Muller Algorithm of AWGN Generation

The previously mentioned Box-Muller algorithm is formulated as follows [43]:
1. Generate two random variables u1 ; u2 and let:
s = u21 + u22 :
2. While s 1, discard s and recompute u1 ; u2 and s.
3. If s < 1 is satised, compute the I and Q components of the
noise as follows:
r
22
loge s
s
r
22
uQ = u2
loge s;
s
where is the standard deviation of the AWGN. This algorithm can be used for
generating both the Rayleigh-faded signal envelope and the thermal AWGN.
In conclusion, with regard to the narrowband propagation issues, the Doppler lter of Figure 2.11 acts as a low-pass lter, which simply limits the rate of change of
both AWGN sources, while retaining a Gaussian distribution. Explicitly, for a Gaussian distributed source signal, the output of linear systems, such as the Doppler lter,
also exhibits Gaussian distribution; only the variance of the process is altered by this
linear system. We note, however, that this is only true for the Gaussian distribution
in the context of linear systems. Furthermore, due to limiting the rate of change at
their outputs, the Doppler lters introduce correlation into the originally uncorrelated
AWGN sources. In other words, the complex Gaussion process of Figure 2.11 generates a Rayleigh distribution both with and without the Doppler lters. In the former
case, however, there is no time-domain correlation between these Rayleigh-distributed
samples, and hence the eect of the vehicular speed is not taken into account. By contrast, when incorporating the Doppler lters, the required velocity-dependent amount
of adjacent-sample correlation is introduced in the Rayleigh-fading model.
As stated before, the impulse response of the nondispersive at Rayleigh-fading
mobile radio channel is constituted by a Dirac delta function whose weight has a
Rayleigh PDF. This channel is encountered when there are no signicant far-eld
re ectors on the horizon and hence no dispersion occurs, but a plethora of multipath
components having similar delays is superimposed on the receiver antenna. As argued
before, in this context the Gaussian channel may be represented by a unity-weight
Dirac delta impulse response, where AWGN is superimposed. Before concluding our
discourse on channel properties, let us brie y consider the properties of the wideband
mobile channels in the next subsection.
uI = u1
2.2.8 Wideband Channels

2.2.8.1 Modeling of Wideband Channels
The impulse response of the at Rayleigh-fading mobile radio channel consists of a single delta function, whose weight has a Rayleigh PDF. This occurs because all the mul-
88
tipath components arrive almost simultaneously and combine to have a Rayleigh PDF.
If the signal's transmission bandwidth is narrower than the channel's coherence bandwidth Bc = 1=(2 ), where represents the time-dispersion interval over which
signicant multipath components are received, then all transmitted frequency components encounter nearly identical fading attenuations across the frequency-domain.
Therefore, the so-called narrowband condition is met, and the signal is subjected to
nonselective or at fading. The channel's coherence bandwidth (Bc ) is dened as the
frequency separation, where the correlation of two dierent-frequency received signal
components' attenuation becomes less than 0.5.
The eect of multipath propagation is to spread the received symbols in the timedomain. If the path-delay dierences are longer than the symbol duration, several
echoes of the same transmitted modulated symbol are received over a number of
consecutive symbol periods. This is equivalent to saying that in wideband channels
the symbol rate is suciently high that each symbol is spread into adjacent symbols,
causing intersymbol interference (ISI). In order for the receiver to remove the ISI
and regenerate the symbols correctly, it must determine the impulse response of the
mobile radio channel by a process termed channel sounding. This response must be
frequently measured, since the mobile channel may change rapidly in both time and
space.
The magnitude of a typical impulse response is described by a continuous waveform, when plotting received amplitude against time-delay. If we partition the timedelay axis into equal-delay segments, usually called delay bins, then there will be, in
general, a number of received signals in each bin corresponding to the dierent paths
whose times of arrival are within the bin duration. These signals, when vectorially
combined, can be represented by a delta function occurring in the center of the bin
having a magnitude that is Rayleigh distributed. Impulses, which are suciently
small in amplitude, such that they are of no signicance to the receiver, can then be
discarded.
As an example here, we describe a set of frequently used typical wideband channel
impulses specied by the Group Speciale Mobile Committee during the denition of
the Pan-European mobile radio system known as GSM [30]. These impulse responses
describe typical urban, rural, and hilly terrain environments, as well as an articially
contrived equalizer test response.
The wideband propagation channel is the superposition of a number of dispersive fading paths, suering from various attenuations and delays, aggravated by the
phenomenon of Doppler shift caused by the MS's movement. The maximum Doppler
shift (fm ) is given by fm = v= = v fc=c, where v is the vehicular speed, is
the wavelength of the carrier frequency fc, and c is the velocity of light. The momentary Doppler shift fD depends on the angle of incidence i , which is uniformly
distributed, that is, fD = fm cos i , which therefore has a random cosine distribution with a Doppler spectrum limited to fm < fD < fm. Because of time-frequency
duality, this \frequency dispersive" phenomenon results in \time-selective" behavior
and the wider the Doppler spread (i.e., the higher the vehicular speed), the faster is
the time-domain impulse response uctuation.
The set of 6-tap GSM impulse responses is depicted in Figure 2.12. In simple
terms, the wideband channel's impulse response is measured by transmitting an im-
Figure 2.12:
89
Typical GSM channel impulse responses.
pulse and detecting the received echoes at the channel's output in every D-spaced
delay bin. In some bins, no delayed and attenuated multipath component is received,
while in others signicant energy is detected, depending on the typical re ecting objects and their distance from the receiver. The path-delay can be easily related to the
distance of the re ecting objects, since radio waves are traveling at the speed of light.
For example, at a speed of 300 000 km/s a re ecting object situated at a distance of
0.15 km yields a multipath component at a round-trip delay of 1 s, corresponding
to a path-length of 0.3 km.
The Typical Urban (TU) impulse response of Figure 2.12 spreads over a delay
interval of 5 s. Therefore, it denitely spills energy into adjacent signaling intervals
for signaling rates in excess of 200 ksymbols/s, resulting in serious ISI. In practical
terms, the transmissions can be considered nondispersive if the excess path-delay does
not exceed 10% of the signaling interval, which in our example would correspond to 20
ksymbols/s or 20 kBaud. The Hilly Terrain (HT) model of Figure 2.12 has a sharply
decaying short-delay section due to local re ections and a long-delay part around
15 s due to distant re ections. Therefore, in practical terms, it can be considered a
two- or three-path model having re ections associated with a path-length of 3 108
m/s 15 s = 4.5 km. The Rural Area (RA) response seems the least hostile among
all standardized responses, decaying fast within 1 s; hence up to signaling rates of
100 kBaud, this environment can be treated as a at-fading narrowband channel.
90
Figure 2.13:
Generation of the baseband in-phase received signal via wideband channels.
The last one of the standardized GSM impulse responses in Figure 2.12 was articially contrived in order to test the channel equalizer's performance, and it is constituted by six equidistant unit-amplitude impulses representing six equal-powered
independent Rayleigh-fading paths with a delay-spread over 16 s. Note that in practice such an impulse response would be unrealistic, since the delayed paths have a
higher path-length, which is associated with a higher path-attenuation. With these
impulse responses in mind, the required channel can be simulated by summing the
appropriately delayed and weighted received signal components. In all but one case,
the individual components are assumed to have Rayleigh amplitude distribution. In
the RA model, the main tap at zero delay is supposed to have Rician distribution
with the presence of a dominant path.
Once the impulse response of the channel is known, we can model a wideband impulse response using the simple nite-impulse response schematic of Figure 2.13, where
the lter coecients are determined by the impulse response and the shift-register delay D corresponds to the sampling frequency used in generating the impulse response.
Accordingly, those impulse response coecients, which were set to zero, since they
were below the thresholding value will have to be set to zero in the schematic. We
note again, however, that all the nonzero lter tap-weights correspond to a signicant
fading multipath component. When amalgamating the schematics of Figures 2.11
and 2.13, we arrive at the wideband channel simulator scheme of Figure 2.14.
The in-phase modulated signal sI (t) is applied to a series of delays equal to the
width of a delay bin, namely D. At each delay n D, SI (t) is multiplied by the
magnitude of the wideband channel impulse response at that delay, which is then
faded obeying the Rayleigh distribution in the worst-case non-LOS scenario. An
identical arrangement to that for SI (t) in Figure 2.14 is used for the quadrature
component sQ (t), and the appropriate convolutional terms are combined to obtain
the received signal's quadrature components rI (t) and rQ (t).
Figure 2.14:
91
Generation of the baseband in-phase received signal via wideband channels [26]. c 1992, Steele.
92
2.3

Summary and Conclusions
In this chapter, we have introduced the cellular concept and demonstrated how radio coverage can be provided in propagation cells of various sizes. Specically, we
have argued that macrocells constitute a a hostile propagation environment, since the
propagation properties vary across the geographic area of the cell. This results in
less predictable propagation scenarios than those experienced in the typically smaller
microcells often beneting from LOS wave propagation. Finally, indoor coverage is
often provided by picocells, where the LOS scenario is predominant. The required
transmitted power is reduced in line with the cell size. This results in the reduction of far-eld re ections, which ultimately reduces the channel-induced dispersion.
Hence the eects of multipath propagation become less distructive, potentially supporting higher bit rates in picocells or allowing the system to dispense with employing
high-complexity channel equalisers.
The range of parameters that may be used for controlling handovers and transceiver reconguration has also been reviewed. Path-loss, shadow fading, and fastfading have been identied as the three most important factors in uencing the power
budget design of wireless systems. The Doppler phenomenon has been discussed
brie y and a model has been introduced for the simulation of both narrowband and
wideband Rayleigh and Rician fading channels.
Having characterized the mobile radio channel, we are now equipped to consider
how the detrimental eects of this hostile transmission medium can be counteracted
using channel coding techniques.
Part II
Video Systems Based on

Proprietary Video Codecs
Part III
High-Resolution Image
Coding
Part IV
Video Systems Based on

Standard Video Codecs
Glossary
16CIF
Sixteen Common Intermediate Format Frames are sixteen times as

big as CIF frames and contain 1408 pixels vertically and 1152 pixels
horizontally
2G
3G
3GPP
4CIF
Second generation
AB
ACCH
ACELP
ACF
ACL
ACO
Access burst
Associated control channel
Algebraic Code Excited Linear Predictive (ACELP) Speech Codec
ACTS
Advanced Communications Technologies and Services. The fourth

framework for European research (1994{98). A series of consortia
consisting of universities and industrialists considering future communications systems.
ADC
ADPCM
AGCH
AI
AICH
ANSI
ARIB
ARQ
Analog{to{Digital Converter
Third generation
Third Generation Partnership Project
Four Common Intermediate Format Frames are four times as big as
CIF frames and contain 704 pixels vertically and 576 pixels horizontally
autocorrelation function
Autocorrelation
Augmented Channel Occupancy matrix, which contains the channel
occupancy for the local and surrounding base stations. Often used
by locally distributed DCA algorithms to aid allocation decisions.
Adaptive Dierential Pulse Coded Modulation

Access grant control channel
Acquisition Indicator
Acquisition Indicator CHannel
American National Standards Institute
Association of Radio Industries and Businesses
Automatic Repeat Request, Automatic request for retransmission of
corrupted data
1023
1024
ATDMA
ATM
AUC
AV.26M
GLOSSARY
Advanced Time Division Multiple Access
Asynchronous Transfer Mode
Authentication center
A draft recommendation for transmitting compressed video over errorprone channels, based on the H.263 [191] video codec
AWGN
B-ISDN
BbB
BCCH
BCH
Additive White Gaussian Noise

Broadband ISDN
Burst-by-Burst
BCH Codes
BER
BN
BPSK
BS
BSIC
BTC
CBER
CBP
Bose-Chaudhuri-Hocquenghem (BCH) Codes
CBPB
Broadcast control channel

Bose-Chaudhuri-Hocquenghem, a class of forward error correcting
codes (FEC)
Bit error rate, the fraction of the bits received incorrectly
Bit number
Binary Phase Shift Keying
A common abbreviation for base station
Base station identier code
Block Truncation Coding
Channel bit error rate, the bit error rate before FEC correction
Coded block pattern, a H.261 video codec symbol that indicates which
of the blocks in the macroblock are active
A xed-length codeword used by the H.263 video codec to convey the
coded block pattern for bidirectionally predicted (B) blocks
CBPY
A variable-length codeword used by the H.263 video codec to indicate

the coded block pattern for luminance blocks
CC
CCCH
CCITT
CCL
CD
Convolutional Code
Commom control channel
Now ITU, standardization group
CDF
Cumulative density function, the integral of the probability density

function (PDF)
CDMA
CELL BAR ACCESS
CIF
Cross-correlation
Code Division, a multiplexing technique whereby signals are coded
and then combined in such a way that they can be separated using
the assigned user signature codes at a later stage
Code Division Multiple Access

Boolean ag to indicate whether the MS is permitted
Common Intermediate Format Frames containing 352 pixels vertically
and 288 pixels horizontally
CIR
COD
Carrier to Interference Ratio, same as SIR
CPICH
Common PIlot CHannel
A one-bit codeword used by the H.263 video codec that indicates

whether the current macroblock is empty or nonempty
GLOSSARY
CT2
CWTS
DAB
DAC
DAMPS
DB
DC
DCA
DCH
DCS1800
DCT
DECT
DL
DPCCH
DPCH
DPCM
DPDCH
DQUANT
DS{CDMA
DSMA-CD
DTTB
DVB-T
EIR
EMC
EOB
EREC
ERPC
ETSI
EU
FA
1025
British Second Generation Cordless Phone Standard
China Wireless Telecommunication Standard
Digital Audio Broadcasting
Digital-to-Analog Convertor
Pan-American Digital Advanced Phone System, IS-54
Dummy burst
Direct Current, normally used in electronic circuits to describe a
power source that has a constant voltage, as opposed to AC power
in which the voltage is a sine wave. It is also used to describe things
that are constant, and hence have no frequency component.
Dynamic Channel Allocation
Dedicated transport CHannel
A digital mobile radio system standard, based on GSM but operating
at 1.8 GHz at a lower power
A discrete cosine transform that transforms data into the frequency
domain. Commonly used for video compression by removing highfrequency components of the video frames
A Pan-European digital cordless telephone standard
Down-link
Dedicated Physical Control CHannel
Dedicated Physical CHannel
Dierential Pulse Coded Modulation
Dedicated Physical Data CHannel
A xed-length coding parameter used to dierential change the current quantizer used by the H.263 video codec
Direct Sequence Code Division Multiple Access
Digital Sense Multiple Access-Collision Detection
Digital Terrestrial Television Broadcast
Terrestrial Pan-European Digital Video Broadcast Standard
Equipment identity register
Electromagnetic Compatibility
An end-of-block variable-length symbol used to indicate the end of
the current block in the H.261 video codec
Error-Resilient Entropy Coding. A coding technique that improves
the robustness of variable-length coding by allowing easier resynchronization after errors
Error-Resilient Position Coding, a relative of the coding scheme known
as Error-Resilient Entropy Coding (EREC)
European Telecommunications Standards Institute
European Union
First Available, a simple centralized DCA scheme that allocates the
rst channel found that is not reused within a given preset reuse
distance
1026
FACCH
FACH
FAW
FBER
FCA
FCB
FCCH
FD
GLOSSARY
Fast associated control channel
Forward Access CHannel
Frame Alignment Word
Feedback error ratio, the ratio of feedback acknowledgment messages
that are received in error
Fixed Channel Allocation
Frequency correction burst
Frequency Correction Channel
Frequency Division, a multiplexing technique whereby dierent frequencies are used for each communications link
FDD
Frequency-Division Duplex, a multiplexing technique whereby the

forward and reverse links use a dierent carrier frequency
FDM
FDMA
Frequency Division Multiplexing
FEC
FEF
FER
FH
FIFO
Forward Error Correction

Frame Error Flag
Frequency Division multiple access, a multiple access technique whereby

frequency division (FD) is used to provide a set of access channels
Frame error rate

Frequency hopping
First-In First-Out, a queuing strategy in which elements that have
been in the queue longest are served rst
FN
FPLMTS
fps
FRAMES
GBSC
TDMA frame number

Future Public Land Mobile Telecommunication System
GEI
Functions similar to PEI but in the GOB layer of the H.261 video
codec
A xed-length codeword used by H.263 video codec to aid correct
resynchronization after an error
GFID
Frames per second

Future Radio Wideband Multiple Access System
Group of blocks (GOB) start code, used by the H.261 and H.263 video
codecs to regain synchronization, playing a similar role to PSC
GMSK
Gaussian Mean Shift Keying, a modulation scheme used by the PanEuropean GSM standard by virtue of its spectral compactness
GN
Group of blocks number, an index number for a GOB used by the

H.261 and H.263 video codecs
Group of blocks, a term used by the H.261 and H.263 video codecs,
consisting of a number of macroblocks
GOB
GOS
GP
GPS
GQUANT
Grade of Service, a performance metric to describe the quality of a

mobile radio network
Guard Period
Global Positioning System
Group of blocks quantizer, a symbol used by the H.261 and H.263
video codecs to modify the quantizer used for the GOB
GLOSSARY
GSM
GSPARE
H.261
H.263
HC
HCA
HCS
HDTV
HLR
HO
HTA
IF
IMSI
IMT-2000
IMT2000
1027
A Pan-European digital mobile radio standard, operating at 900MHz
Functions similar to PSPARE but in the GOB layer of the H.261
video codec
A video coding standard [499], published by the ITU in 1990
A video coding standard [191], published by the ITU in 1996
Human Coding
Hybrid Channel Allocation, a hybrid of FCA and DCA
Hierarchical Cell Structure
High-Denition Television
Home location register
Handover
Highest interference below Threshold Algorithm, a distributed DCA
algorithm also known as MTA. The algorithm allocates the most interfered channel, whose interference is below the maximum tolerable
interference threshold.
Intermediate Frequency
International mobile subscriber identity
International Mobile Telecommunications-2000
Intelligent Mobile Telecommunications in the Year 2000, Japanese
Initiative for 3rd Generation Cellular Systems
IS-54
IS-95
ISDN
Pan-American Digital Advanced Phone System, IS-54
ISI
ITU
Intersymbol Interference, Inter-Subcarrier Interference
ITU-R
International Mobile Telecommunication Union { Radiocommunication Sector

Japanese Digital Cellular Standard
JDC
JPEG
LFA
LIA
LODA
LOLIA
North American mobile radio standard, that uses CDMA technology

Integrated Services Digital Network, digital replacement of the analogue telephone network
International Telecommunications Union, formerly the CCITT, standardization group
\Lossy" DCT-based Still Picture Coding Standard

Lowest Frequency below threshold Algorithm, a distributed DCA algorithm that is a derivative of the LTA algorithm, the dierence being
that the algorithm attempts to reduce the number of carrier frequencies being used concurrently
Least Interference Algorithm, a distributed DCA algorithm that assigns the channel with the lowest measured interference that is available.
Locally Optimized Dynamic Assignment, a centralized DCA scheme,
which bases its allocation decisions on the future blocking probability
in the vicinity of the cell
Locally Optimized Least Interference Algorithm, a locally distributed
DCA algorithm that allocates channels using a hybrid of the LIA and
an ACO matrix
1028
LOMIA
LP ltering
LP-DDCA
LPF
LSB
LSR
LTA
GLOSSARY
Locally Optimized Most Interference Algorithm, a locally distributed
DCA algorithm that allocates channels using a hybrid of the MTA
and an ACO matrix
Low-pass ltering
Local Packing Dynamic Distributed Channel Assignment, a locally
distributed DCA algorithm that assigns the rst channel available
that is not used by the surrounding base stations, whose information
is contained in an ACO matrix
Low-pass lter
Least signicant bit
Linear Shift Register
Least interference below Threshold Algorithm, a distributed DCA algorithm that allocates the least interfered channel, whose interference
is below a preset maximum tolerable interference level
LTI
MA
Linear Time-invariant
Abbreviation for Miss America, a commonly used head and shoulders
video sequence referred to as Miss America
Macroblock
A grouping of 8 by 8 pixel blocks used by the H.261 and H.263 video

codecs. Consists of four luminance blocks and two chrominance blocks
Multiple Access Interference
MAI
MAP
MB
MBA
MBS
MC
MCBPC
MCER
MDM
MF-PRMA
MFlop
MODB
Maximum{A{Posteriori
Macroblock
Macroblock address symbol used by the H.261 video codec, indicating
the position of the macroblock in the current GOB
Mobile Broadband System
Motion Compensation
A variable-length codeword used by the H.263 video codec to convey
the macroblock type and the coded block pattern for the chrominance
blocks
Motion Compensated Error Residual
Modulation Division Multiplexing
Multi-Frame Packet Reservation Multiple Access
Mega Flop, 1 million oating point operations per second
A variable-length coding parameter used by the H.263 video codec to
indicate the macroblock mode for bidirectionally predicted (B) blocks
MPEG
Motion Picture Expert Group, also a video coding standard designed

by this group that is widely used
MPG
MQUANT
Multiple Processing Gain
MS
MSC
MSE
A H.261 video codec symbol that changes the quantizer used by current and future macroblocks in the current GOB
A common abbreviation for Mobile Station
Mobile switching center
Mean Square Error
GLOSSARY
1029
MSQ
Mean Square centralized DCA algorithm that attempts to minimize

the mean square distance between cells using the same channel
MTA
Most interference below Threshold Algorithm, a distributed DCA algorithm also known as HTA. The algorithm allocates the most interfered channel, whose interference is below the maximum tolerable
interference level.
H.261 video codec symbol that contains information about the macroblock, such as coding mode, and ags to indicate whether optional
modes are used, like motion vectors, and loop ltering
MTYPE
MV
MVD
MVDB
Motion vector, a vector to estimate the motion in a frame
NB
NCC
NLF
NMC
NN
Normal burst
Normalized Channel Capacity
NN+1
Nearest-Neighbor-plus-one centralized DCA algorithm; allocates a

channel used by the nearest cell, which is at least the reuse distance
plus one cell radius away
OFDM
OMC
OVSF
P-CCPCH
PCH
PCM
PCN
PCPCH
PCS
Orthogonal Frequency Division Multiplexing
PDF
PDSCH
PEI
Probability Density Function
PFU
PGZ
PHP
PI
PICH
PLMN
Partial Forced Update
Motion vector data symbol used by H.261 and H.263 video codecs
A variable-length codeword used by the H.263 video codec to convey
the motion vector data for bidirectionally predicted (B) blocks
Nonlinear ltering
Network management center
Nearest-Neighbor centralized DCA algorithm; allocates a channel used
by the nearest cell, which is at least the reuse distance away
Operation and maintenance center

Orthogonal Variable Spreading Factor
Primary Common Control Physical CHannel
Paging CHannel
Pulse code modulation
Personal Communications Network
Physical Common Packet CHannel
Personal Communications System, a term used to describe thirdgeneration mobile radio systems in North America
Physical Down-link Shared CHannel
Picture layer extra insertion bit, used by the H.261 video codec, indicating that extra information is to be expected
Peterson-Gorenstein-Zierler (PGZ) Decoder
Japanese Personal Handyphone Phone System
Page Indicator
Page Indicator CHannel
Public land mobile network
1030
GLOSSARY
PLMN PERMITTED
PLMR
PLR
PP
PQUANT
Boolean ag to indicate, whether the MS is permitted

Public Land Mobile Radio
Packet-Loss Ratio
Partnership Project
A xed-length codeword used by the H.263 video codec to indicate
the quantizer to use for the next frame
PRACH
PRMA
Physical Random Access CHannel

Packet Reservation Multiple Access, a statistical multiplexing arrangement contrived to improve the eciency of conventional TDMA
systems, by detecting inactive speech segments using a voice activity detector, surrendering them and allocating them to subscribers
contending to transmit an active speech packet
PRMA System allowing contention only in the so-called contention
slots, which protect the information slots from contention and collisions
Pilot symbol-assisted modulation, a technique whereby known symbols (pilots) are transmitted regularly. The eect of channel fading
on all symbols can then be estimated by interpolating between the
pilots.
Picture start code, a preset sequence used by the H.261 and H.263
video codecs, that can be searched for to regain synchronization after
an error
Power Spectral Density
Peak Signal to Noise Ratio, noise energy compared to the maximum
possible signal energy. Commonly used to measure video image quality
PRMA++
PSAM
PSC
PSD
PSNR
PSPARE
PSTN
PTYPE
QAM
QCIF
QMF
QN
QoS
QT
RACE
RACH
RC ltering
RF
RFCH
RFN
Picture layer extra information bits, indicated by a PEI symbol in

H.261 video codec
Public switched telephone network
Picture layer information, used by H.261 and H.263 video codecs to
transmit information about the picture, e.g. Resolution, etc.
Quadrature Amplitude Modulation
Quarter Common Intermediate Format Frames containing 176 pixels
vertically and 144 pixels horizontally
Quadrature Mirror Filtering
Quater bit number
Quality of Service
Quad-Tree
Research in Advanced Communications Equipment Programme in
Europe, from June 1987 to December 1995
Random Access CHannel
Raised-cosine ltering
Radio frequency
Radio frequency channel
Reduced TDMA frame number in GSM
GLOSSARY
RING
RLC
RPE
RS Codes
RSSI
1031
A centralized DCA algorithm that attempts to allocate channels in
one of the cells, which is at least the reuse distance away that forms
a \ring" of cells
Run-Length Coding
Regular pulse excited
Reed-Solomon (RS) codes
Received Signal Strength Indicator, commonly used as an indicator
of channel quality in a mobile radio network
RTT
RXLEV
RXQUAL
S-CCPCH
SAC
Radio Transmission Technology
SACCH
SB
SCH
SCS
Slow associated control channel

Synchronization burst
SDCCH
SF
SINR
Received signal level: parameter used in hangovers

Received signal quality: parameter used in hangovers
Secondary Common Control Physical CHannel
Syntax-based arithmetic coding, an alternative to variable-length coding, and a variant of arithmetic coding
Synchronization CHannel
Sequential Channel Search distributed DCA algorithm that searches
the available channels in a predetermined order, picking the rst channel found, which meets the interference constraints
Stand-alone dedicated control channel
Spreading Factor
Signal-to-Interference plus Noise ratio, same as signal-to-noise ratio
(SNR) when there is no interference.
SIR
SNR
SPAMA
SQCIF
Signal-to-Interference ratio
Signal-to-Noise Ratio, noise energy compared to the signal energy
SSC
TA
TB
TC
TCH
TCH/F
TCH/F2.4
TCH/F4.8
TCH/F9.6
TCH/FS
TCH/H
TCH/H2.4
Secondary Synchronization Codes
Statistical Packet Assignment Multiple Access

Sub-Quarter Common Intermediate Format Frames containing 128
pixels vertically and 96 pixels horizontally
Timing advance
Tailing bits
Trellis Coded
Trac channel
Full-rate trac channel
Full-rate 2.4 kbps data trac channel
Full-rate speech trac channel
Half-rate trac channel
Half-rate 2.4 kbps data trac channel
1032
TCH/H4.8
TCM
TCOEFF
TD
TDD
TDMA
TFCI
TIA
TN
TPC
TR
TS
TTA
TTC
TTIB
UHF
UL
UMTS
UTRA
VA
VAD
VAF
VE
VL
VLC
VLR
VQ
W-CDMA
WARC
WATM
WLAN
WN
WWW
GLOSSARY
Half-rate 4.8 kbps data trac channel
Trellis code modulation
An H.261 and H.263 video codec symbol that contains the transform
coecients for the current block
Time Division, a multiplexing technique whereby several communications links are multiplexed onto a single carrier by dividing the
channel into time periods, and assigning a time period to each communications link
Time-Division Duplex, a technique whereby the forward and reverse
links are multiplexed in time.
Time Division Multiple Access
Transport-Format Combination Indicator
Telecommunications Industry Association
Time slot number
Transmit Power Control
Temporal reference, a symbol used by H.261 and H.263 video codecs
to indicate the real-time dierence between transmitted frames
Technical Specications
Telecommunications Technology Association
Telecommunication Technology Committee
Transparent tone in band
Ultra high frequency
Up-link
Universal Mobile Telecommunications System, a future Pan-European
third-generation mobile radio standard
Universal Mobile Telecommunications System Terrestrial Radio Access
Viterbi Algorithm
Voice activity detection
Voice activity factor, the fraction of time the voice activity detector
of a speech codec is active
Viterbi equalizer
Variable length
Variable-length coding/codes
Visiting location register
Vector Quantization
Wideband Code Division Multiple Access
World Administrative Radio Conference
Wireless Asynchronous Transfer Mode (ATM)
Wireless Local Area Network
White noise
World Wide Web, the name given to computers that can be accessed
via the Internet using the HTTP protocol. These computers can
provide information in a easy-to-digest multimedia format using hyperlinks.
Bibliography
[1] C. Shannon, Mathematical Theory of Communication. University of Illinois Press, 1963.

[2] C. Shannon, \A mathematical theory of communication | part I," Bell Systems Technical
Journal, pp. 379{405, 1948.
[3] C. Shannon, \A mathematical theory of communication | part II," Bell Systems Technical
Journal, pp. 405{423, 1948.
[4] C. Shannon, \A mathematical theory of communication | part III," Bell Systems Technical
Journal, vol. 27, pp. 623{656, 1948.
[5] H. Nyquist, \Certain factors aecting telegraph speed," Bell System Technical Journal, p. 617,
April 1928.
[6] R. Hartley, \Transmission of information," Bell Systems Technical Journal, p. 535, 1928.
[7] N. Abramson, Information Theory and Coding. New York: McGraw-Hill, 1975.
[8] A. Carlson, Communication Systems. New York: McGraw-Hill, 1975.
[9] H. Raemer, Statistical Communication Theory and Applications. Englewood Clis, NJ:
Prentice-Hall, 1969.
[10] P. Ferenczy, Telecommunications Theory, in Hungarian: Hirkozleselmelet. Budapest, Hungary: Tankonyvkiado, 1972.
[11] K. Shanmugam, Digital and Analog Communications Systems. New York: John Wiley and
Sons, 1979.
[12] C. Shannon, \Communication in the presence of noise," Proceedings of the I.R.E., vol. 37,
pp. 10{22, January 1949.
[13] C. Shannon, \Probability of error for optimal codes in a Gaussian channel," Bell Systems
Technical Journal, vol. 38, pp. 611{656, 1959.
[14] A. Jain, Fundamentals of Digital Image Processing. Englewood Clis, NJ: Prentice-Hall, 1989.
[15] A. Hey and R. Allen, eds., R.P. Feynman: Feynman Lectures on Computation. Reading, MA:
Addison-Wesley, 1996.
[16] C. Berrou, A. Glavieux, and P. Thitimajshima, \Near shannon limit error-correcting coding
and decoding: Turbo codes," in Proceedings of the International Conference on Communications, (Geneva, Switzerland), pp. 1064{1070, May 1993.
[17] L. Hanzo, W. Webb, and T. Keller, Single- and Multi-Carrier Quadrature Amplitude Modulation: Principles and Applications for Personal Communications, WLANs and Broadcasting.
IEEE Press, 2000.
[18] J. Hagenauer, \Quellengesteuerte kanalcodierung fuer sprach- und tonuebertragung im mobilfunk," Aachener Kolloquium Signaltheorie, pp. 67{76, 23{25 March 1994.
1033
1034
BIBLIOGRAPHY
[19] A. Viterbi, \Wireless digital communications: A view based on three lessons learned," IEEE
Communications Magazine, pp. 33{36, September 1991.
[20] D. Greenwood and L. Hanzo, \Characterisation of mobile radio channels," in Steele and Hanzo
[26], ch. 2, pp. 92{185.
[21] L. Hanzo and J. Woodard, \An intelligent multimode voice communications system for indoor communications," IEEE Transactions on Vehicular Technology, vol. 44, pp. 735{748,
November 1995.
[22] L. Hanzo, R. Salami, R. Steele, and P. Fortune, \Transmission of digitally encoded speech at
1.2 Kbaud for PCN," IEE Proceedings, Part I, vol. 139, pp. 437{447, August 1992.
[23] K. Wong and L. Hanzo, \Channel coding," in Steele and Hanzo [26], ch. 4, pp. 347{488.
[24] W. Jakes, ed., Microwave Mobile Communications. New York: John Wiley and Sons, 1974.
[25] W. Lee, Mobile Cellular Communications. New York: McGraw-Hill, 1989.
[26] R. Steele and L. Hanzo, eds., Mobile Radio Communications. Piscataway, NJ: IEEE Press,
1999.
[27] R. Steele, \Towards a high capacity digital cellular mobile radio system," Proceedings of the
IEE, vol. 132, Part F, pp. 405{415, August 1985.
[28] R. Steele and V. Prabhu, \High-user density digital cellular mobile radio system," IEE Proceedings, vol. 132, Part F, pp. 396{404, August 1985.
[29] R. Steele, \The cellular environment of lightweight hand-held portables," IEEE Communications Magazine, pp. 20{29, July 1989.
[30] L. Hanzo and J. Stefanov, \The Pan-European Digital Cellular Mobile Radio System | known
as GSM," in Steele and Hanzo [26], ch. 8, pp. 677{765.
[31] P. Mermelstein, \The IS-54 digital cellular standard," in Gibson [291], ch. 26, pp. 419{429.
[32] K. Kinoshita and M. Nakagawa, \Japanese cellular standard," in Gibson [291], ch. 28, pp. 449{
461.
[33] L. Hanzo, \The British cordless telephone system: CT2," in Gibson [291], ch. 29, pp. 462{477.
[34] S. Asghar, \Digital European Cordless Telephone," in Gibson [291], ch. 30, pp. 478{499.
[35] J. Parsons and J. Gardiner, Mobile Communication Systems. London: Blackie, 1989.
[36] D. Parsons, The Mobile Radio Propagation Channel. London: Pentech Press, 1992.
[37] I. Wassell and R. Steele, \Wideband systems," in Steele and Hanzo [26], ch. 6, pp. 523{600.
[38] R. Edwards and J. Durkin, \Computer prediction of service area for VHF mobile radio networks," Proc of IRE, vol. 116, no. 9, pp. 1493{1500, 1969.
[39] M. Hata, \Empirical formula for propagation loss in land mobile radio," IEEE Transactions
on Vehicular Technology, vol. 29, pp. 317{325, August 1980.
[40] Y. Okumura, E. Ohmori, T. Kawano, and K. Fukuda, \Field strength and its variability in
VHF and UHF land mobile service," Review of the Electrical Communication Laboratory,
vol. 16, pp. 825{873, September{October 1968.
[41] J. Proakis, Digital Communications. New York: McGraw-Hill, 3rd ed., 1995.
[42] M. Gans, \A power-spectral theory of propogation in the mobile-radio environment," IEEE
Transactions on Vehicular Technology, vol. 21, pp. 27{38, February 1972.
[43] W. Press, S. Teukolsky, W. Vetterling, and B. Flannery, \Random numbers," in Numerical
Recipes in C [147], ch. 7, pp. 289{290.
[44] R. Hamming, \Error detecting and error correcting codes," Bell System Technical Journal,
vol. 29, pp. 147{160, 1950.
[45] P. Elias, \Coding for noisy channels," IRE Conv. Rec. pt.4, pp. 37{47, 1955.
[46] J. Wozencraft, \Sequential decoding for reliable communication," IRE Natl. Conv. Rec., vol. 5,
pt.2, pp. 11{25, 1957.
BIBLIOGRAPHY
1035
[47] J. Wozencraft and B. Reien, Sequential Decoding. MIT Press, 1961.
[48] R. Fano, \A heuristic discussion of probabilistic coding," IEEE Transactions on Information
Theory, vol. IT-9, pp. 64{74, April 1963.
[49] J. Massey, Threshold Decoding. Cambridge, MA: MIT Press, 1963.
[50] A. Viterbi, \Error bounds for convolutional codes and an asymphotically optimum decoding
algorithm," IEEE Transactions on Information Theory, vol. IT-13, pp. 260{269, April 1967.
[51] G. Forney, \The Viterbi algorithm," Proceedings of the IEEE, vol. 61, pp. 268{278, March
1973.
[52] J. Heller and I. Jacobs, \Viterbi decoding for satellite and space communication," IEEE Transactions on Communication Technology, vol. COM-19, pp. 835{848, October 1971.
[53] L. Bahl, J. Cocke, F. Jelinek, and J. Raviv, \Optimal decoding of linear codes for minimising
symbol error rate," IEEE Transactions on Information Theory, vol. 20, pp. 284{287, March
1974.
[54] A. Hocquenghem, \Codes correcteurs d'erreurs," Chires (Paris), vol. 2, pp. 147{156, September 1959.
[55] R. Bose and D. Ray-Chaudhuri, \On a class of error correcting binary group codes," Information and Control, vol. 3, pp. 68{79, March 1960.
[56] R. Bose and D. Ray-Chaudhuri, \Further results on error correcting binary group codes,"
Information and Control, vol. 3, pp. 279{290, September 1960.
[57] W. Peterson, \Encoding and error correction procedures for the Bose-Chaudhuri codes," IRE
Trans. Inform. Theory, vol. IT-6, pp. 459{470, September 1960.
[58] D. Gorenstein and N. Zierler, \A class of cyclic linear error-correcting codes in pm synbols,"
J. Soc. Ind. Appl. Math., vol. 9, pp. 107{214, June 1961.
[59] I. Reed and G. Solomon, \Polynomial codes over certain nite elds," J. Soc. Ind. Appl. Math.,
vol. 8, pp. 300{304, June 1960.
[60] E. Berlekamp, \On decoding binary Bose-Chaudhuri-Hocquenghem codes," IEEE Transactions on Information Theory, vol. 11, pp. 577{579, 1965.
[61] E. Berlekamp, Algebraic Coding Theory. New York: McGraw-Hill, 1968.
[62] J. Massey, \Step-by-step decoding of the Bose-Chaudhuri-Hocquenghem codes," IEEE Transactions on Information Theory, vol. 11, pp. 580{585, 1965.
[63] J. Massey, \Shift-register synthesis and BCH decoding," IEEE Transactions on Information
Theory, vol. IT-15, pp. 122{127, January 1969.
[64] Consultative Committee for Space Data Systems, Blue Book: Recommendations for Space
Data System Standards: Telemetry Channel Coding, May 1984.
[65] W. Peterson, Error Correcting Codes. Cambridge, MA: MIT Press, 1st ed., 1961.
[66] W. Peterson and E. Weldon Jr., Error Correcting Codes. Cambridge, MA: MIT. Press, 2nd ed.,
August 1972.
[67] G. Clark Jr. and J. Cain, Error Correction Coding for Digital Communications. New York:
Plenum Press, May 1981.
[68] A. Michelson and A. Levesque, Error Control Techniques for Digital Communication. New
York: John Wiley and Sons, 1985.
[69] R. Blahut, Theory and Practice of Error Control Codes. Reading, MA: Addison-Wesley, 1983.
[70] S. Lin and D. Constello Jr., Error Control Coding: Fundamentals and Applications. Englewood
Clis, NJ: Prentice-Hall, October 1982.
[71] V. Pless, Introduction to the Theory of Error-correcting Codes. New York: John Wiley and
Sons, 1982.
[72] I. Blake, ed., Algebraic Coding Theory: History and Development. Dowden, Hutchinson and
Ross, 1973.
1036
BIBLIOGRAPHY
[73] K. Wong, Transmission of Channel Coded Speech and Data over Mobile Channels. PhD thesis,
University of Southampton, 1989.
[74] R. Lidl and H. Niederreiter, Finite Fields. Cambridge: Cambridge University Press, October
1996.
[75] D. Gorenstein and N. Zierler, \A class of error-correcting codes in pm symbols," J. Soc. Ind.
Appl. Math., no. 9, pp. 207{214, 1961.
[76] J. Makhoul, \Linear prediction: A tutorial review," Proceedings of the IEEE, vol. 63, pp. 561{
580, April 1975.
[77] R. Blahut, Fast Algorithms for Digital Signal Processing. Reading MA: Addison-Wesley, 1985.
[78] J. Schur, \Ueber potenzreihen, die im innern des einheits- kreises beschraenkt sind," Journal
fuer Mathematik, pp. 205{232. Bd. 147, Heft 4.
[79] R. Chien, \Cyclic decoding procedure for the Bose-Chaudhuri-Hocquenghem codes," IEEE
Transactions on Info. Theory, vol. 10, pp. 357{363, October 1964.
[80] A. Jennings, Matrix Computation for Engineers and Scientists. New York: John Wiley and
Sons Ltd., 1977.
[81] G. Forney Jr, \On decoding BCH codes," IEEE Transactions on Information Theory, vol. IT11, pp. 549{557, 1965.
[82] Y. Sugiyama, M. Kasahara, S. Hirasawa, and T. Namekawa, \A method for solving key equation for decoding goppa codes," Information Control, no. 27, pp. 87{99, 1975.
[83] S. Golomb, Shift Register Sequences. Laugana Hills, CA: Aegean Park Press, 1982.
[84] L. Hanzo, W. Webb, and T. Keller, Single- and Multi-carrier Quadrature Amplitude Modulation. New York: IEEE Press-John Wiley, April 2000.
[85] R. Steele and L. Hanzo, eds., Mobile Radio Communications. New York: IEEE Press-John
Wiley, 2nd ed., 1999.
[86] J. Cavers, \Variable rate transmission for rayleigh fading channels," IEEE Transactions on
Communications Technology, vol. COM-20, pp. 15{22, February 1972.
[87] W. Webb and R. Steele, \Variable rate QAM for mobile radio," IEEE Transactions on Communications, vol. 43, pp. 2223{2230, July 1995.
[88] L. Hanzo, \Bandwidth-ecient wireless multimedia communications," Proceedings of the
IEEE, vol. 86, pp. 1342{1382, July 1998.
[89] S. Nanda, K. Balachandran, and S. Kumar, \Adaptation techniques in wireless packet data
services," IEEE Communications Magazine, vol. 38, pp. 54{64, January 2000.
[90] T. Keller and L. Hanzo, \Adaptive multicarrier modulation: A convenient framework for timefrequency processing in wireless communications," Proceedings of the IEEE, vol. 88, pp. 611{
642, May 2000.
[91] L. Hanzo, C. Wong, and P. Cherriman, \Channel-adaptive wideband video telephony," IEEE
Signal Processing Magazine, vol. 17, pp. 10{30, July 2000.
[92] L. Hanzo, F. Somerville, and J. Woodard, \Voice compression and communications: Principles
and applications for xed and wireless channels." 2001 (For detailed contents, please refer to
http://www-mobile.ecs.soton.ac.uk.).
[93] P. Cherriman, E. Kuan, and L. Hanzo, Patent: Transceiver. Mobile VCE, Basingstoke, 20th
May 1999. Patent Application Number 9911777.2.
[94] P. Cherriman, E. Kuan, and L. Hanzo, \Turbo coded burst-by-burst adaptive joint-detection
CDMA based video telephony." Submitted to IEEE Transactions on Circuits Systems and
Video Technology, June 1999.
[95] L. Hanzo, P. Cherriman, and E. Kuan, \Interactive cellular and cordless video telephony:
State-of-the-art, system design principles and expected performance," Proceedings of IEEE,
September 2000.
BIBLIOGRAPHY
1037
[96] P. Cherriman, E. Kuan, and L. Hanzo, \Multi-mode joint-detection CDMA/H.263 based video
telephony," in Proceeding of ACTS Mobile Communication Summit '99 [673], pp. 715{720.
[97] E. Kuan, C. Wong, and L. Hanzo, \Burst-by-burst adaptive joint detection CDMA," in Proceeding of VTC'99 (Spring) [665].
[98] G. Forney Jr, R. Gallager, G. Lang, F. Longsta, and S. Qureshi, \Ecient modulation for
band-limited channels," IEEE Journal on Selected Areas in Communications, vol. 2, pp. 632{
647, September 1984.
[99] K. Feher, \Modems for emerging digital cellular mobile systems," IEEE Transactions on Vehicular Technology, vol. 40, pp. 355{365, May 1991.
[100] W. Webb, L. Hanzo, and R. Steele, \Bandwidth-ecient QAM schemes for rayleigh-fading
channels," IEE Proceedings, vol. 138, pp. 169{175, June 1991.
[101] A. Wright and W. Durtler, \Experimental performance of an adaptive digital linearized power
amplier," IEEETransactions on Vehicular Technology, vol. 41, pp. 395{400, November 1992.
[102] M. Faulkner and T. Mattson, \Spectral sensitivity of power ampliers to quadrature modulator
misalignment," IEEE Transactions on Vehicular Technology, vol. 41, pp. 516{525, November
1992.
[103] P. Kenington, R. Wilkinson, and J. Marvill, \Broadband linear amplier design for a PCN
base-station," in Proceedings of IEEE Vehicular Technology Conference (VTC'91), (St. Louis,
MO), pp. 155{160, IEEE, 19{22 May 1991.
[104] R. Wilkinson et al., \Linear transmitter design for MSAT terminals," in Proceedings of 2nd
International Mobile Satellite Conference, June 1990.
[105] S. Stapleton and F. Costescu, \An adaptive predistorter for a power amplier based on adjacent
channel emissions," IEEE Transactions on Vehicular Technology, vol. 41, pp. 49{57, February
1992.
[106] S. Stapleton, G. Kandola, and J. Cavers, \Simulation and analysis of an adaptive predistorter
utilizing a complex spectral convolution," IEEE Transactions on Vehicular Technology, vol. 41,
pp. 387{394, November 1992.
[107] Y. Kamio, S. Sampei, H. Sasaoka, and N. Morinaga, \Performance of modulation-level-control
adaptive-modulation under limited transmission delay time for land mobile communications,"
in Proceedings of IEEE Vehicular Technology Conference (VTC'95) [669], pp. 221{225.
[108] J. Torrance and L. Hanzo, \Upper bound performance of adaptive modulation in a slow
Rayleigh fading channel," Electronics Letters, vol. 32, pp. 718{719, 11 April 1996.
[109] J. Torrance and L. Hanzo, \Optimisation of switching levels for adaptive modulation in a slow
Rayleigh fading channel," Electronics Letters, vol. 32, pp. 1167{1169, 20 June 1996.
[110] J. Torrance and L. Hanzo, \Demodulation level selection in adaptive modulation," Electronics
Letters, vol. 32, pp. 1751{1752, 12 September 1996.
[111] J. Torrance and L. Hanzo, \Performance upper bound of adaptive QAM in slow Rayleighfading environments," in Proceedings of IEEE ICCS'96/ISPACS'96 [675], pp. 1653{1657.
[112] J. Torrance and L. Hanzo, \Adaptive modulation in a slow Rayleigh fading channel," in Pro-
ceedings of IEEE International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC'96) [681], pp. 497{501.
[113] J. Torrance and L. Hanzo, \Latency considerations for adaptive modulation in a slow Rayleigh
fading channel," in Proceedings of IEEE VTC'97 [679], pp. 1204{1209.
[114] K. Feher, ed., Digital Communications|Satellite/Earth Station Engineering. Englewood
Clis, NJ: Prentice-Hall, 1983.
[115] Y. Chow, A. Nix, and J. McGeehan, \Analysis of 16-APSK modulation in AWGN and rayleigh
fading channel," Electronic Letters, vol. 28, pp. 1608{1610, November 1992.
[116] B. Sklar, Digital Communications|Fundamentals and Applications. Englewood Clis, NJ:
Prentice-Hall, 1988.
1038
BIBLIOGRAPHY
[117] J. Torrance, Adaptive Full Response Digital Modulation for Wireless Communications Systems. PhD thesis, Department of Electronics and Computer Science, University of Southampton, 1997.
[118] J. Cavers, \An analysis of pilot symbol assisted modulation for rayleigh fading channels,"
IEEE Transactions on Vehicular Technology, vol. 40, pp. 686{693, November 1991.
[119] F. Adachi, \Error rate analysis of dierentially encoded and detected 16APSK under rician
fading," IEEE Transactions on Vehicular Technology, vol. 45, pp. 1{12, February 1996.
[120] J. McGeehan and A. Bateman, \Phase-locked transparent tone in band (TTIB): A new spectrum conguration particularly suited to the transmission of data over SSB mobile radio networks," IEEE Transactions on Communications, vol. COM-32, no. 1, pp. 81{87, 1984.
[121] A. Bateman and J. McGeehan, \Feedforward transparent tone in band for rapid fading protection in multipath fading," in IEE International Conference on Communications, vol. 68,
pp. 9{13, 1986.
[122] A. Bateman and J. McGeehan, \The use of transparent tone in band for coherent data
schemes," in IEEE International Conference on Communications, (Boston, MA), 1983.
[123] A. Bateman, G. Lightfoot, A. Lymer, and J. McGeehan, \Speech and data transmissions
over a 942MHz TAB and TTIB single sideband mobile radio system," IEEE Transactions on
Vehicular Technology, vol. VT-34, pp. 13{21, February 1985.
[124] A. Bateman and J. McGeehan, \Data transmissions over UHF fading mobile radio channels,"
Proceedings of IEE, vol. 131, no. Pt.F, pp. 364{374, 1984.
[125] J. McGeehan and A. Bateman, \A simple simultaneous carrier and bit synchronisation system
for narrowband data transmissions," Proceedings of IEE, vol. 132, no. Pt.F, pp. 69{72, 1985.
[126] J. McGeehan and A. Bateman, \Theoretical and experimental investigation of feedforward
signal regeneration," IEEE Transactions on Vehicular Technology, vol. VT-32, pp. 106{120,
1983.
[127] A. Bateman, \Feedforward transparent tone in band: Its implementation and applications,"
IEEE Transactions on Vehicular Technology, vol. 39, pp. 235{243, August 1990.
[128] J. Torrance and L. Hanzo, \Comparative study of pilot symbol assisted modem schemes,"
in Proceedings of IEE Conference on Radio Receivers and Associated Systems (RRAS'95),
(Bath), pp. 36{41, IEE, 26{28 September 1995.
[129] M. Moher and J. Lodge, \TCMP|a modulation and coding strategy for rician fading channels," IEEE Journal on Selected Areas in Communications, vol. 7, pp. 1347{1355, December
1989.
[130] S. Sampei and T. Sunaga, \Rayleigh fading compensation method for 16-QAM in digital land
mobile radio channels," in Proceedings of IEEE Vehicular Technology Conference (VTC'89),
(San Francisco, CA), pp. 640{646, IEEE, 1{3 May 1989.
[131] S. Haykin, Adaptive Filter Theory. Englewood Clis, NJ: Prentice-Hall, 1996.
[132] J. Cavers, \The performance of phase locked transparent tone in band with symmetric phase
detection," IEEE Transactions on Communications, vol. 39, pp. 1389{1399, September 1991.
[133] R. Steele and W. Webb, \Variable rate QAM for data transmission over Rayleigh fading
channels," in Proceeedings of Wireless '91, (Calgary, Alberta), pp. 1{14, IEEE, 1991.
[134] K. Arimochi, S. Sampei, and N. Morinaga, \Adaptive modulation system with discrete power
control and predistortion-type non-linear compensation for high spectral ecient and high
power ecient wireless communication systems," in Proceedings of IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, PIMRC'97 [674], pp. 472{
477.
[135] M. Naijoh, S. Sampei, N. Morinaga, and Y. Kamio, \ARQ schemes with adaptive modulation/TDMA/TDD systems for wireless multimedia communication systems," in Proceedings
of IEEE International Symposium on Personal, Indoor and Mobile Radio Communications,

PIMRC'97 [674], pp. 709{713.
BIBLIOGRAPHY
1039
[136] S.-G. Chua and A. Goldsmith, \Variable-rate variable-power mQAM for fading channels," in
Proceedings of IEEE VTC'96 [678], pp. 815{819.
[137] A. Goldsmith and S. Chua, \Variable-rate variable-power MQAM for fading channels," IEEE
Transactions on Communications, vol. 45, pp. 1218{1230, October 1997.
[138] A. Goldsmith, \The capacity of downlink fading channels with variable rate and power," IEEE
Transactions on Vehicular Technolgy, vol. 46, pp. 569{580, August. 1997.
[139] A. Goldsmith and P. Varaiya, \Capacity of fading channels with channel side information,"
IEEE Transactions on Information Theory, vol. 43, pp. 1986{1992, November 1997.
[140] M.-S. Alouini and A. Goldsmith, \Capacity of rayleigh fading channels under dierent adaptive transmission and diversity-combining techniques," to appear in IEEE Transactions on
Vehicular Technology, 1999. http://www.systems.caltech.edu.
[141] M.-S. Alouini and A. Goldsmith, \Area spectral eciency of cellular mobile radio systems," to
appear IEEE Transactions on Vehicular Technology, 1999. http://www.systems.caltech.edu.
[142] A. Goldsmith and S. Chua, \Adaptive coded modulation for fading channels," IEEE Transactions on Communications, vol. 46, pp. 595{602, May 1998.
[143] D. Pearce, A. Burr, and T. Tozer, \Comparison of counter-measures against slow Rayleigh
fading for TDMA systems," in IEE Colloquium on Advanced TDMA Techniques and Applications, (London), pp. 9/1{9/6, IEE, 28 October 1996. digest 1996/234.
[144] W. Lee, \Estimate of channel capacity in Rayleigh fading environment," IEEE Transactions
[145] N. Morinaga, \Advanced wireless communication technologies for achieving high-speed mobile
radios," IEICE Transactions on Communications, vol. 78, no. 8, pp. 1089{1094, 1995.
[146] J. Woodard and L. Hanzo, \A low delay multimode speech terminal," in Proceedings of IEEE
VTC'96 [678], pp. 213{217.
[147] W. Press, S. Teukolsky, W. Vetterling, and B. Flannery, Numerical Recipes in C. Cambridge:
Cambridge University Press, 1992.
[148] J. Torrance and L. Hanzo, \Statistical multiplexing for mitigating latency in adaptive
modems," in Proceedings of IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, PIMRC'97 [674], pp. 938{942.
[149] J. Torrance, L. Hanzo, and T. Keller, \Interference resilience of burst-by-burst adaptive
modems," in Proceeding of ACTS Mobile Communication Summit '97 [668], pp. 489{494.
[150] C. Wong, T. Liew, and L. Hanzo, \Turbo coded burst by burst adaptive wideband modulation
with blind modem mode detection," in Electronic copy [673], pp. 303{308.
[151] C. Wong and L. Hanzo, \Upper-bound of a wideband burst-by-burst adaptive modem," in
Proceeding of VTC'99 (Spring) [665], pp. 1851{1855.
[152] K. Narayanan and L. Cimini, \Equalizer adaptation algorithms for high speed wireless communications," in Proceedings of IEEE VTC'96 [678], pp. 681{685.
[153] J. Wu and A. Aghvami, \A new adaptive equalizer with channel estimator for mobile radio
communications," IEEE Transactions on Vehicular Technology, vol. 45, pp. 467{474, August
1996.
[154] Y. Gu and T. Le-Ngoc, \Adaptive combined DFE/MLSE techniques for ISI channels," IEEE
Transactions on Communications, vol. 44, pp. 847{857, July 1996.
[155] A. Clark and R. Harun, \Assessment of kalman-lter channel estimators for an HF radio link,"
IEE Proceedings, vol. 133, pp. 513{521, October 1986.
[156] R. Chang, \Synthesis of band-limited orthogonal signals for multichannel data transmission,"
Bell Systems Technical Journal, vol. 46, pp. 1775{1796, December 1966.
[157] M. Zimmermann and A. Kirsch, \The AN/GSC-10/KATHRYN/variable rate data modem for
HF radio," IEEE Transactions on Communication Technology, vol. CCM{15, pp. 197{205,
April 1967.
1040
BIBLIOGRAPHY
[158] E. Powers and M. Zimmermann, \A digital implementation of a multichannel data modem," in

Proceedings of the IEEE International Conference on Communications, (Philadelphia), 1968.
[159] B. Saltzberg, \Performance of an ecient parallel data transmission system," IEEE Transactions on Communication Technology, pp. 805{813, December 1967.
[160] R. Chang and R. Gibby, \A theoretical study of performance of an orthogonal multiplexing
data transmission scheme," IEEE Transactions on Communication Technology, vol. COM{16,
pp. 529{540, August 1968.
[161] S. Weinstein and P. Ebert, \Data transmission by frequency division multiplexing using the
discrete fourier transform," IEEE Transactions on Communication Technology, vol. COM{19,
pp. 628{634, October 1971.
[162] A. Peled and A. Ruiz, \Frequency domain data transmission using reduced computational
complexity algorithms," in Proceedings of International Conference on Acoustics, Speech, and
Signal Processing, ICASSP'80 [680], pp. 964{967.
[163] B. Hirosaki, \An orthogonally multiplexed QAM system using the discrete fourier transform,"
IEEE Transactions on Communications, vol. COM-29, pp. 983{989, July 1981.
[164] L. Cimini, \Analysis and simulation of a digital mobile channel using orthogonal frequency
division multiplexing," IEEE Transactions on Communications, vol. 33, pp. 665{675, July
1985.
[165] K. Kammeyer, U. Tuisel, H. Schulze, and H. Bochmann, \Digital multicarrier transmission
of audio signals over mobile radio channels," European Transactions on Telecommunications,
vol. 3, pp. 243{253, May{June 1992.
[166] F. Mueller-Roemer, \Directions in audio broadcasting," Journal Audio Engineering Society,
vol. 41, pp. 158{173, March 1993.
[167] G. Plenge, \DAB | a new radio broadcasting system | state of development and ways for
its introduction," Rundfunktech. Mitt., vol. 35, no. 2, 1991.
[168] M. Alard and R. Lassalle, \Principles of modulation and channel coding for digital broadcasting
for mobile receivers," EBU Review, Technical No. 224, pp. 47{69, August 1987.
[169] Proceedings of 1st International Symposium,DAB, (Montreux, Switzerland), June 1992.
[170] I. Kalet, \The multitone channel," IEEE Transactions on Communications, vol. 37, pp. 119{
124, February 1989.
[171] H. Kolb, \Untersuchungen uber ein digitales mehrfrequenzverfahren zur datenubertragung," in
Ausgewahlte Arbeiten uber Nachrichtensysteme, no. 50, Universitat Erlangen-Nurnberg, 1982.
[172] H. Schussler, \Ein digitales Mehrfrequenzverfahren zur Datenubertragung," in ProfessorenKonferenz, Stand und Entwicklungsaussichten der Daten und Telekommunikation, (Darmstadt, Germany), pp. 179{196, 1983.
[173] K. Preuss, \Ein Parallelverfahren zur schnellen Datenubertragung Im Ortsnetz," in Ausgewahlte Arbeiten uber Nachrichtensysteme, no. 56, Universitat Erlangen-Nurnberg, 1984.
[174] R. Ruckriem, \Realisierung und messtechnische Untersuchung an einem digitalen Parallelverfahren zur Datenubertragung im Fernsprechkanal," in Ausgewahlte Arbeiten uber Nachrichtensysteme, no. 59, Universitat Erlangen-Nurnberg, 1985.
[175] ETSI, Digital Audio Broadcasting (DAB), 2nd ed., May 1997. ETS 300 401.
[176] ETSI, Digital Video Broadcasting (DVB), 1.1.2 ed., August 1997. EN 300 744.
[177] J. Lindner et al, \OCDM | ein ubertragungsverfahren fur lokale funknetze," in Codierung
fuer Quelle, Kanal und Uebertragung, no. 130 in ITG Fachbericht, pp. 401{409, VDE Verlag,
26{28 October 1994.
[178] K. Fazel and G. Fettweis, eds., Multi-Carrier Spread-Spectrum. Dordrecht: Kluwer, 1997.
[179] S. Nanda, D. Goodman, and U. Timor, \Performance of PRMA: A packet voice protocol for
cellular systems," IEEE Transactions on Vehicular Technology, vol. 40, pp. 584{598, August
1991.
BIBLIOGRAPHY
1041
[180] W. Webb, R. Steele, J. Cheung, and L. Hanzo, \A packet reservation multiple access assisted
cordless telecommunications scheme," IEEE Transactions on Vehicular Technology, vol. 43,
pp. 234{245, May 1994.
[181] R. Salami, C. La amme, J.-P. Adoul, and D. Massaloux, \A toll quality 8 kb/s speech codec
for the personal communications system (PCS)," IEEE Transactions on Vehicular Technology,
pp. 808{816, August 1994.
[182] M. Frullone, G. Riva, P. Grazioso, and C. Carciofy, \Investigation on dynamic channel allocation strategies suitable for PRMA schemes," IEEE International Symposium on Circuits and
Systems, pp. 2216{2219, May 1993.
[183] M. Frullone, G. Falciasecca, P. Grazioso, G. Riva, and A. Serra, \On the performance of packet
reservation multiple access with xed and dynamic channel allocation," IEEE Transactions
on Vehicular Technology, vol. 42, pp. 78{86, February 1993.
[184] J. Brecht, M. del Buono, and L. Hanzo, \Multiframe packet reservation multiple access using oscillation-scaled histogram-based markov modelling of video codecs," Signal Processing:
Image Communications, vol. 12, pp. 167{182, 1998.
[185] J. Brecht and L. Hanzo, \Statistical packet assignment multiple access for wireless asynchronous transfer mode systems," in Proceeding of ACTS Mobile Communication Summit
'97 [668].
[186] J. Leduc and P. Delogne, \Statistic for variable bit-rate digital television sources," Signal
Processing: Image Communication, vol. 8, pp. 443{464, July 1996.
[187] S. Lam, S. Chow, and D. Yau, \A lossless smoothing algorithm for compressed video,"
IEEE/ACM Transactions on networking, vol. 4, pp. 697{708, October 1996.
[188] J. Cosmas, G. Petit, R. Lehenert, C. Blondia, K. Kontovassilis, O. Casals, and T. Theimer, \A
review of voice, data and video trac models for ATM," European Transactions on Telecommunications, vol. 5, pp. 139{154, March-April 1994.
[189] O. Rose and M. Frater, \A comparison of models for VBR trac sources in B-ISDN," in
Proceedings of the IFIP TC6 Second International Conference on Broadband Communications,
(Paris, France), pp. 275{287, Chapman and Hall Ltd, London, 2{4 March 1994.
[190] D. Heymann and T. Lakshman, \Source models for VBR broadcast-video trac," IEEE/ACM
Transactions on networking, vol. 4, pp. 40{48, February 1996.
[191] ITU-T, Recommendation H.263: Video Coding for Low Bitrate communication, March 1996.
[192] M. Whybray and W. Ellis, \H.263 - video coding recommendation for PSTN videophone and
multimedia," in IEE Colloquium (Digest), pp. 6/1{6/9, IEE, June 1995.
[193] M. Khansari, A. Jalali, E. Dubois, and P. Mermelstein, \Low bit-rate video transmission
over fading channels for wireless microcellular systems," IEEE Transactions on Circuits and
Systems for Video Technology, vol. 6, pp. 1{11, February 1996.
[194] N. Farber, E. Steinbach, and B. Girod, \Robust H.263 video transmission over wireless channels," in Proceedings of International Picture Coding Symposium (PCS), (Melbourne, Australia), pp. 575{578, March 1996.
[195] P. Cherriman and L. Hanzo, \H.261 and H.263-based programmable video transceivers," in
Proceedings of IEEE ICCS'96/ISPACS'96 [675], pp. 1369{1373.
[196] P. Cherriman and L. Hanzo, \Robust H.263 video transmission over mobile channels in
interference-limited environments," in Proc of First International Workshop on Wireless Image/Video Communications, (Loughborough), pp. 1{7, 4{5 September 1996.
[197] P. Cherriman and L. Hanzo, \Power-controlled H.263-based wireless videophone performance
in interference-limited scenarios," in Proceedings of IEEE International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC'96) [681], pp. 158{162.
[198] P. Cherriman and L. Hanzo, \Programmable H.263-based wireless video transceivers for
interference-limited environments," IEEE Transactions on Circuits and Systems for Video
Technology, vol. 8, pp. 275{286, June 1998.
1042
BIBLIOGRAPHY
[199] P. Cherriman, T. Keller, and L. Hanzo, \Orthogonal frequency division multiplex transmission of H.263 encoded video over wireless ATM networks," in Proceeding of ACTS Mobile
Communication Summit '97 [668], pp. 276{281.
[200] A. Tanenbaum, \Introduction to queueing theory," in Computer Networks, pp. 631{641, Englewood Clis, NJ: Prentice-Hall, 2nd ed., 1989.
[201] P. Skelly, M. Schwartz, and S. Dixit, \A histogram-based model for video trac behavior in a
ATM multiplexer," IEEE/ACM Transactions Networking, vol. 1, pp. 446{459, August 1993.
[202] D. Habibi, S. Gabrielsson, and Z. Man, \A multiplexed four layers markov model for queueing
studies of MPEG trac," in Proceedings of IEEE ICCS'96/ISPACS'96 [675], pp. 1180{1184.
[203] F. Panken, \Multiple-access protocols over the years: a taxonomy and survey," in 1996 IEEE
International Conference on Communication Systems (ICCS), pp. 2.1.1{2.1.5, November 1996.
[204] V. Li and X. Qiu, \Personal communications systems," Proceedings of the IEEE, vol. 83,
pp. 1210{1243, September 1995.
[205] D. Goodman and S. Wei, \Eciency of packet reservation multiple access," IEEE Transactions. on Vehicular Technology, vol. 40, pp. 170{176, February 1991.
[206] J. Dunlop, D. Robertson, P. Cosimi, and J. D. Vile, \Development and optimisation of a
statistical multiplexing mechanism for ATDMA," in Proceedings of IEEE VTC '94 [676],
pp. 1040{1044.
[207] F. Delli Priscoli, \Adaptive parameter computation in a PRMA, TDD based medium access
control for ATM wireless networks," in Proceeding of IEEE Global Telecommunications Conference, Globecom 96 [672], pp. 1779{1783.
[208] A. Acampora, \Wireless ATM: a perspective on issues and prospects," IEEE Personal Communications, vol. 3, pp. 8{17, August 1996.
[209] \Group speciale mobile (GSM) recommendation," April 1988.
[210] V. Claus, ed., Duden zur Informatik. Mannheim: Dudenverlag, 1993.
[211] M. Schwartz, Broadband Integrated Networks. Englewood Clis, NJ: Prentice-Hall, March
1996.
[212] A. Safak, \Optimal channel reuse in cellular radio systems with multiple correlated log-normal
interferers," IEEE Transactions on Vehicular Technology, vol. 43, pp. 304{312, May 1994.
[213] L.-C. Wang and C.-T. Lea, \Incoherent estimation on co-channel interference probability for
microcellular systems," IEEE Transactions on Vehicular Technology, vol. 45, pp. 164{173,
February 1996.
[214] C. Lee and R. Steele, \Signal-to-interference calculations for modern TDMA cellular communication systems," IEE Proceedings on Communication, vol. 142, pp. 21{30, February 1995.
[215] P. Brady, \A technique for investigating on-o patterns of speech," Bell Systems Technical
Journal, vol. 44, pp. 1{22, January 1965.
[216] P. Brady, \A model for generating on-o speech patterns in two-way conversation," Bell Systems Technical Journal, vol. 48, pp. 2445{2472, September 1969.
[217] D. Appleby and Y. Ko, \Frequency hopping," in Steele and Hanzo [26], ch. 7, pp. 601{676.
[218] C. Lee, CDMA for Cellular Mobile Radio Systems. PhD thesis, Department of Electronics
and Computer Science, University of Southampton, November 1994.
[219] W. Lee, \Spectrum eciency in cellular," IEEE Transactions on Vehicular Technology, vol. 38,
pp. 69{75, May 1989.
[220] R. Gejji, \Channel eciency in digital cellular capacity," in Proceedings of IEEE VTC '92,
vol. 2, (Denver, CO), pp. 1005{1007, 10{13 May 1992.
[221] M. Chiani, V. Tralli, and R. Verdone, \Outage and spectrum eciency analysis in microcellular
systems," in Proceedings of IEEE VTC '93 [687], pp. 598{601.
[222] R. Haas and J.-C. Belore, \Spectrum eciency limits in mobile cellular systems," IEEE
Transmissions on Vehicular Technology, vol. 45, pp. 33{40, February 1996.
BIBLIOGRAPHY
1043
[223] M. Zorzi and S. Pupolin, \Outage probability in multiple access packet radio networks in the
presence of fading," IEEE Transactions on Vehicular Technology, vol. 43, pp. 604{610, August
1994.
[224] N. Senarath and D. Everitt, \Combined analysis of transmission and trac characteristics
in micro-cellular mobile communications systems," in Proceedings of IEEE VTC '93 [687],
pp. 577{580.
[225] A. Abu-Dayya and N. Beaulieu, \Outage probability in the presence of correlated lognormal
interferers," IEEE Tr. on Vehicular Technology, vol. 43, pp. 33{39, February 1994.
[226] Q. Zhang, \Outage probability in cellular mobile radio due to nakagami signal and interferers
with arbitrary parameters," IEEE Transactions on Vehicular Technology, vol. 45, pp. 364{372,
May 1996.
[227] P. Cherriman, F. Romiti, and L. Hanzo, \Channel allocation for third-generation mobile radio
systems," in Proceeding of ACTS Mobile Communication Summit '98 [670], pp. 255{261.
[228] J. Blogh, P. Cherriman, and L. Hanzo, \Comparative study of dynamic channel allocation
algorithms." to appear in IEEE Transactions on Vehicular Technology, 1999.
[229] S. Wales, \Technique for cochannel interference suppression in TDMA mobile radio systems,"
IEE Proceedings on Communication, vol. 142, no. 2, pp. 106{114, 1995.
[230] J. Litva and T. Lo, Digital Beamforming in Wireless Communications. London: Artech House,
1996.
[231] L. Godara, \Applications of antenna arrays to mobile communications, part I: Performance improvement, feasibility, and system considerations," Proceedings of the IEEE, vol. 85, pp. 1029{
1060, July 1997.
[232] L. Godara, \Applications of antenna arrays to mobile communications, part II: Beam-forming
and direction-of-arrival considerations," Proceedings of the IEEE, vol. 85, pp. 1193{1245, August 1997.
[233] E. Sourour, \Time slot assignment techniques for TDMA digital cellular systems," IEEE
Transactions on Vehicular Technology, vol. 43, pp. 121{127, February 1994.
[234] D. Wong and T. Lim, \Soft handos in CDMA mobile systems," IEEE Personal Communications, vol. 4, pp. 6{17, December 1997.
[235] S. Tekinay and B. Jabbari, \A measurement-based prioritisation scheme for handovers in
mobile cellular networks," IEEE Journal on Selected Areas of Communications, vol. 10, no. 8,
pp. 1343{1350, 1992.
[236] G. Pollini, \Trends in handover design," IEEE Communications Magazine, vol. 34, pp. 82{90,
March 1996.
[237] R. Bernhardt, \Timeslot re-assignment in a frequency reuse TDMA portable radio system,"
[238] A. Viterbi, CDMA: Principles of Spread Spectrum Communication. Reading MA: AddisonWesley, June 1995.
[239] R. Prasad, CDMA for Wireless Personal Communications. London: Artech House, May 1996.
[240] S. Glisic and B. Vucetic, Spread Spectrum CDMA Systems for Wireless Communications.
London: Artech House, April 1997.
[241] S. Glisic and P. Leppanen, eds., Wireless Communications : TDMA versus CDMA. Dordrecht:
Kluwer Academic Publishers, June 1997.
[242] A. Ross and K. Gilhousen, \CDMA technology and the IS-95 north american standard," in
Gibson [291], ch. 27, pp. 430{448.
[243] ETSI, Universal Mobile Telecommunications Systems (UMTS); UMTS Terrestrial Radio Access (UTRA); Concept evaluation, December 1997. TR 101 146 V3.0.0.
[244] I. Katzela and M. Naghshineh, \Channel assignment schemes for cellular mobile telecommunication systems: A comprehensive survey," IEEE Personal Communications, pp. 10{31, June
1996.
1044
BIBLIOGRAPHY
[245] S. Tekinay and B. Jabbari, \Handover and channel assignment in mobile cellular networks,"
IEEE Communications Magazine, pp. 42{46, November 1991.
[246] B. Jabbari, \Fixed and dynamic channel assignment," in Gibson [291], ch. 83, pp. 1175{1181.
[247] J. Zander, \Radio resource management in future wireless networks: Requirements and limitations," IEEE Communications Magazine, pp. 30{36, August 1997.
[248] D. Everitt, \Trac engineering of the radio interface for cellular mobile networks," Proceedings
of the IEEE, vol. 82, pp. 1371{1382, September 1994.
[249] J. Dahlin, \Ericsson's multiple reuse pattern for DCS1800," Mobile Communications International, November 1996.
[250] M. Madfors, K. Wallstedt, S. Magnusson, H. Olofsson, P.-O. Backman, and S. Engstrom,
\High capacity with limited spectrum in cellular systems," IEEE Communications Magazine,
vol. 35, pp. 38{45, August 1997.
[251] H. Jiang and S. Rappaport, \Prioritized channel borrowing without locking: a channel sharing strategy for cellular communications," IEEE/ACM Transactions on Networking, vol. 43,
pp. 163{171, April 1996.
[252] L. Anderson, \A simulation study of some dynamic channel assignment algorithms in a high
capacity mobile telecommunications system," IEEE Transactions on Communications, vol. 21,
pp. 1294{1301, November 1973.
[253] J. Engel and M. Peritsky, \Statistically optimum dynamic server assignment in systems with
interfering servers," IEEE Transactions on Vehicular Technology, vol. 22, pp. 203{209, November 1973.
[254] M. Zhang and T.-S. Yum, \Comparisons of channel assignment strategies in cellular mobile telephone systems," IEEE Transactions on Vehicular Technology, vol. 38, pp. 211{215,
November 1989.
[255] S. Elnoubi, R. Singh, and S. Gupta, \A new frequency channel assignment algorithm in high capacity mobile communications systems," IEEE Transactions on Vehicular Technology, vol. 31,
pp. 125{131, August 1982.
[256] M. Zhang and T.-S. Yum, \The non-uniform compact pattern allocation algorithm for cellular
mobile systems," IEEE Transactions on Vehicular Technology, vol. 40, pp. 387{391, May 1991.
[257] S. Kuek and W. Wong, \Ordered dynamic channel assignment scheme with reassignment in
highway microcell," IEEE Transactions on Vehicular Technology, vol. 41, pp. 271{277, August
1992.
[258] T.-S. Yum and W.-S. Wong, \Hot spot trac relief in cellular systems," IEEE Journal on
selected areas in Communications, vol. 11, pp. 934{940, August 1993.
[259] J. Tajima and K. Imamura, \A strategy for exible channel assignment in mobile communication systems," IEEE Transactions on Vehicular Technology, vol. 37, pp. 92{103, May 1988.
[260] ETSI, Digital European Cordless Telecommunications (DECT), 1st ed., October 1992. ETS
300 175-1 { ETS 300 175-9.
[261] R. Steele, \Digital European Cordless Telecommunications (DECT) systems," in Steele and
Hanzo [26], ch. 1.7.2, pp. 79{83.
[262] H. Ochsner, \The digital european cordless telecommunications specication, DECT," in Tuttlebee [263], pp. 273{285.
[263] W. Tuttlebee, ed., Cordless telecommunications in Europe: the evolution of personal communications. London: Springer-Verlag, 1990.
[264] A. Law and L. Lopes, \Performance comparison of DCA call assignment algorithms within
DECT," in Proceedings of IEEE VTC'96 [678], pp. 726{730.
[265] H. Salgado-Galicia, M. Sirbu, and J. Peha, \A narrowband approach to ecient PCS spectrum
sharing through decentralized DCA access policies," IEEE Personal Communications, pp. 24{
34, February 1997.
BIBLIOGRAPHY
1045
[266] R. Steele, J. Whitehead, and W. Wong, \System aspects of cellular radio," IEEE Communications Magazine, vol. 33, pp. 80{86, January 1995.
[267] J.-I. Chuang, \Performance issues and algorithms for dynamic channel assignment," IEEE
Journal on Selected Areas of Communications, vol. 11, pp. 955{963, August 1993.
[268] D. Cox and D. Reudink, \The behavior of dynamic channel-assignment mobile communications
systems as a function of number of radio channels," IEEE Transactions on Communications,
vol. 20, pp. 471{479, June 1972.
[269] D. Dimitrijevic and J. Vuceric, \Design and performance analysis of the algorithms for channel
allocation in cellular networks," IEEE Transactions on Vehicular Technology, vol. 42, pp. 526{
534, November 1993.
[270] D. Cox and D. Reudink, \Increasing channel occupancy in large scale mobile radio systems: Dynamic channel reassignment," IEEE Transactions on Vehicular Technology, vol. 22, pp. 218{
222, November 1973.
[271] D. Cox and D. Reudink, \A comparison of some channel assignment strategies in large-scale
mobile communications systems," IEEE Transactions on Communications, vol. 20, pp. 190{
195, April 1972.
[272] M. Cheng and J.-I. Chuang, \Performance evaluation of distributed measurement-based dynamic channel assignment in local wireless communications," IEEE Journal on Selected Areas
of Communications, vol. 14, pp. 698{710, May 1996.
[273] S. Grandhi, R. Yates, and D. Goodman, \Resource allocation for cellular radio systems," IEEE
Transactions on Vehicular Technology, vol. 46, pp. 581{587, August 1997.
[274] J.-I. Chuang and N. Sollenberger, \Performance of autonomous dynamic channel assignment
and power control for TDMA/FDMA wireless access," IEEE Journal on Selected Areas of
Communications, vol. 12, pp. 1314{1323, October 1994.
[275] M. Serizawa and D. Goodman, \Instability and deadlock of distributed dynamic channel allocation," in Proceedings of IEEE VTC'93, (Secaucus, NJ), pp. 528{531, May 18{20 1993.
[276] Y. Akaiwa and H. Andoh, \Channel segregation | a self-organized dynamic channel allocation
method: Application to TDMA/FDMA microcellular system," IEEE Journal on Selected
Areas in Communications, vol. 11, pp. 949{954, August 1993.
[277] A. Baiocchi, F. Delli Priscoli, F. Grilli, and F. Sestini, \The geometric dynamic channel allocation as a practical strategy in mobile networks with bursty user mobility," IEEE Transactions
on Vehicular Technology, vol. 44, pp. 14{23, February 1995.
[278] F. Delli Priscoli, N. Pio Magnani, V. Palestini, and F. Sestini, \Application of dynamic channel allocation strategies to the GSM cellular network," IEEE Journal on Selected Areas in
[279] I. ChihLin and C. PiHui, \Local packing | distributed dynamic channel allocation at cellular
base station," in Proceedings of the IEEE Global Telecommunications Conference 1993 [667],
pp. 293{301.
[280] E. Del Re, R. Fantacci, and G. Giambene, \Handover and dynamic channel allocation techniques in mobile cellular networks," IEEE Transactions on Vehicular Technology, vol. 44,
pp. 229{237, May 1995.
[281] T. Kahwa and N. Georganas, \A hybrid channel assignment scheme in large-scale, cellularstructured mobile communication systems," IEEE Transactions on Communications, vol. 26,
pp. 432{438, April 1978.
[282] J. Sin and N. Georganas, \A simulation study of a hybrid channel assignment scheme for cellular land-mobile radio systems with erlang-c service," IEEE Transactions on Communications,
vol. 29, pp. 143{147, February 1981.
[283] S.-H. Oh and D.-W. Tcha, \Prioritized channel assignment in a cellular radio network," IEEE
Transactions on Communications, vol. 40, pp. 1259{1269, July 1992.
[284] D. Hong and S. Rappaport, \Trac model and performance analysis for cellular mobile radio
telephone systems with prioritizes and nonprioritized hando procedures," IEEE Transactions
1046
BIBLIOGRAPHY
[285] R. Guerin, \Queueing-blocking system with two arrival streams and guard channels," IEEE
Transactions on Communications, vol. 36, pp. 153{163, February 1988.
[286] S. Grandhi, R. Vijayan, D. Goodman, and J. Zander, \Centralized power control in cellular
radio systems," IEEE Transactions Vehicular Technology, vol. 42, pp. 466{468, November
1993.
[287] J. Zander, \Performance of optimum transmitter power control in cellular radio systems,"
IEEE Transactions on Vehicular Technology, vol. 41, pp. 57{62, February 1992.
[288] J. Zander, \Distributed cochannel interference control in cellular radio systems," IEEE Transactiosn on Vehicular Technology, vol. 41, pp. 305{311, August 1992.
[289] D. Cox and D. Reudink, \Eects of some nonuniform spatial demand proles on mobile radio
system performance," IEEE Transactions on Vehicular Technology, vol. 21, pp. 62{67, May
1972.
[290] M. Sonka, V. Hlavac, and R. Boyle, \Image pre-processing," in Image Processing, Analysis
and Machine Vision, ch. 4, pp. 59{61, London: Chapman and Hall, 1st ed., 1993.
[291] J. Gibson, ed., The Mobile Communications Handbook. Boca Raton FL: CRC Press and IEEE
Press, 1996.
[292] \Special issue: The European Path Toward UMTS," IEEE Personal Communications: The
magazine of nomadic communications and computing, vol. 2, February 1995.
[293] European Commission, Advanced Communications Technologies and Services (ACTS), August 1994. Workplan DGXIII-B-RA946043-WP.
[294] Telcomm. Industry Association (TIA), Washington, DC, Dual-mode subscriber equipment |
Network equipment compatibility specication, Interim Standard IS-54, 1989.
[295] Telcomm. Industry Association (TIA), Washington, DC, Mobile station | Base station com-
patibility standard for dual-mode wideband spread spectrum cellular system, EIA/TIA Interim
Standard IS-95, 1993.
[296] H. Alikhani, R. Bohnke, and M. Suzuki, \BDMA-band division multiple access; a new airinterface for 3rd generation mobile system, UMTS, in europe," in Proceeding of ACTS Mobile
Communication Summit '97 [668], pp. 482{488.
[297] P. Baier, P. Jung, and A. Klein, \Taking the challenge of multiple access for third-generation
cellular mobile radio systems | a European view," IEEE Communications Magazine, vol. 34,
pp. 82{89, February 1996.
[298] F. Adachi, M. Sawahashi, and K. Okawa, \Tree-structured generation of orthogonal spreading
codes with dierent lengths for forward link of DS-CDMA mobile radio," Electronic Letters,
vol. 33, pp. 27{28, January 1997.
[299] F. Adachi, K. Ohno, A. Higashi, T. Dohi, and Y. Okumura, \Coherent multicode DS-CDMA
mobile Radio Access," IEICE Transactions on Communications, vol. E79-B, pp. 1316{1324,
September 1996.
[300] F. Adachi and M. Sawahashi, \Wideband wireless access based on DS-CDMA," IEICE Transactions on Communications, vol. E81-B, pp. 1305{1316, July 1998.
[301] F. Adachi, M. Sawahashi, and H. Suda, \Wideband DS-CDMA for next-generation mobile
communications systems," IEEE Communications Magazine, vol. 36, pp. 56{69, September
1998.
[302] D. Knisely, S. Kumar, S. Laha, and S. Nanda, \Evolution of wireless data services : IS-95 to
cdma2000," IEEE Communications Magazine, vol. 36, pp. 140{149, October 1998.
[303] Telecommunications Industry Association (TIA), The cdma2000 ITU-R RTT Candidate Submission, 1998.
[304] D. Knisely, Q. Li, and N. Rames, \cdma2000: A third generation radio transmission technology," Bell Labs Technical Journal, vol. 3, pp. 63{78, July{September 1998.
[305] \Feature topic: Wireless personal communications," IEEE Personal Communications, vol. 2,
April 1995.
BIBLIOGRAPHY
1047
[306] \Feature topic: Universal telecommunications at the beginning of the 21st century," IEEE
Communications Magazine, vol. 33, November 1995.
[307] \Feature topic: Software radios," IEEE Communications Magazine, vol. 33, pp. 24{68, May
1995.
[308] \Feature topic: Wireless personal communications," IEEE Communications Magazine, vol. 33,
January 1995.
[309] \Feature topic: European research in mobile communications," IEEE Communications Magazine, vol. 34, February 1996.
[310] R. Cox and P. Kroon, \Low bit-rate speech coders for multimedia communications," IEEE
Communications Magazine, pp. 34{41, December 1996.
[311] A. Kucar, \Mobile radio: An overview," in Gibson [291], ch. 15, pp. 242{262.
[312] C. Li, C. Zheng, and C. Tai, \Detection of ECG characteristic points using wavelet transforms,"
IEEE Transactions in Biomedical Engineering, vol. 42, pp. 21{28, January 1995.
[313] \Personal handy phone (PHP) system." RCR Standard, STD-28, Japan.
[314] ETSI, Sophia Antipolis, Cedex, France, Digital Cellular Telecommunications System (Phase
2+); High Speed Circuit Switched Data (HSCSD) | Stage 1 (GSM 02.34 Version 5.2.1), July
1997.
[315] A. Noerpel, Y.-B. Lin, and H. Sherry, \PACS: Personal access communications system | a
tutorial," IEEE Personal Communications Magazine, vol. 3, pp. 32{43, June 1996.
[316] P. Vary and R. Sluyter, \MATS-D speech codec: Regular-pulse excitation LPC," in Proceedings
of the Nordic Seminar on Digital Land Mobile Radio Communications (DMRII), (Stockholm,
Sweden), pp. 257{261, October 1986.
[317] R. Salami, L. Hanzo, R. Steele, K. Wong, and I. Wassell, \Speech coding," in Steele and
Hanzo [26], ch. 3, pp. 186{346.
[318] J. Rapeli, \UMTS:targets, system concept, and standardization in a global framework," IEEE
Personal Communications, vol. 2, pp. 20{28, February 1995.
[319] P.-G. Andermo and L.-M. Ewerbring, \A CDMA-based radio access design for UMTS," IEEE
Personal Communications, vol. 2, pp. 48{53, February 1995.
[320] E. Nikula, A. Toskala, E. Dahlman, L. Girard, and A. Klein, \FRAMES multiple access for
UMTS and IMT-2000," IEEE Personal Communications, vol. 5, pp. 16{24, April 1998.
[321] T. Ojanpera and R. Prasad, \An overview of air interface multiple access for IMT2000/UMTS," IEEE Communications Magazine, vol. 36, pp. 82{95, September 1998.
[322] A. Toskala, J. Castro, E. Dahlman, M. Latva-aho, and T. Ojanpera, \FRAMES FMA2
wideband-CDMA for UMTS," European Transactions on Telecommunications, vol. 9, pp. 325{
336, July{August 1998.
[323] E. Berruto, M. Gudmundson, R. Menolascino, W. Mohr, and M. Pizarroso, \Research activities on UMTS radio interface, network architectures, and planning," IEEE Communications
Magazine, vol. 36, pp. 82{95, February 1998.
[324] M. Callendar, \Future public land mobile telecommunication systems," IEEE Personal Communications, vol. 12, no. 4, pp. 18{22, 1994.
[325] W. Lee, \Overview of cellular CDMA," IEEE Transactions on Vehicular Technology, vol. 40,
pp. 291{302, May 1991.
[326] K. Gilhousen, I. Jacobs, R. Padovani, A. Viterbi, L. Weaver Jr., and C. Wheatley III, \On the
capacity of a cellular CDMA system," IEEE Transactions on Vehicular Technology, vol. 40,
pp. 303{312, May 1991.
[327] R. Pickholtz, L. Milstein, and D. Schilling, \Spread spectrum for mobile communications,"
IEEE Transactions on Vehicular Technology, vol. 40, pp. 312{322, May 1991.
[328] R. Kohno, R. Meidan, and L. Milstein, \Spread spectrum access methods for wireless communication," IEEE Communications Magazine, vol. 33, pp. 58{67, January 1995.
1048
BIBLIOGRAPHY
[329] V. Garg, K. Smolik, J. Wilkes, and K. Smolik, Applications of CDMA in Wireless/Personal

Communications. Englewood Clis NJ: Prentice-Hall, 1996.
[330] R. Price and E. Green Jr., \A communication technique for multipath channels," Proceedings
of the IRE, vol. 46, pp. 555{570, March 1958.
[331] B. Sklar, \Rayleigh fading channels in mobile digital communication systems part I : Characterization," IEEE Communications Magazine, vol. 35, pp. 90{100, July 1997.
[332] B. Sklar, \Rayleigh fading channels in mobile digital communication systems part II : Mitigation," IEEE Communications Magazine, vol. 35, pp. 148{155, July 1997.
[333] F. Amoroso, \Use of DS/SS signaling to mitigate rayleigh fading in a dense scatterer environment," IEEE Personal Communications, vol. 3, pp. 52{61, April 1996.
[334] M. Nakagami, \The m-distribution-a general formula of intensity distribution of fading," Statistical Methods in Radio Wave Propagation, 1960. W.C. Homan, Ed. New York: Pergamon.
[335] H. Suzuki, \A statistical model for urban multipath propagation," IEEE Transactions on
Communications, vol. COM-25, pp. 673{680, July 1977.
[336] \COST 207: Digital land mobile radio communications, nal report." Oce for Ocial Publications of the European Communities, 1989. Luxembourg.
[337] M. Whitmann, J. Marti, and T. Kurner, \Impact of the power delay prole shape on the
bit error rate in mobile radio systems," IEEE Transactions on Vehicular Technology, vol. 46,
pp. 329{339, May 1997.
[338] T. Eng, N. Kong, and L. Milstein, \Comparison of diversity combining techniques for Rayleighfading channels," IEEE Transactions on Communications, vol. 44, pp. 1117{1129, September
1996.
[339] M. Kavehrad and P. McLane, \Performance of low-complexity channel coding and diversity
for spread spectrum in indoor, wireless communications," AT&T Technical Journal, vol. 64,
pp. 1927{1965, October 1985.
[340] K.-T. Wu and S.-A. Tsaur, \Selection diversity for DS-SSMA communications on nakagami
fading channels," IEEE Transactions on Vehicular Technology, vol. 43, pp. 428{438, August
1994.
[341] L.-L. Yang and L. Hanzo, \Serial acquisition techniques for DS-CDMA signals in frequencyselective multi-user mobile channels," in Proceeding of VTC'99 (Spring) [665].
[342] L.-L. Yang and L. Hanzo, \Serial acquisition of DS-CDMA signals in multipath fading mobile
channels." submitted to IEEE Transactions on Vehicular Technology, 1998.
[343] R. Ziemer and R. Peterson, Digital Communications and Spread Spectrum System. New York:
Macmillan Publishing Company, 1985.
[344] R. Pickholtz, D. Schilling, and L. Milstein, \Theory of spread-spectrum communications | a
tutorial," IEEE Transactions on Communications, vol. COM-30, pp. 855{884, May 1982.
[345] S. Rappaport and D. Grieco, \Spread-spectrum signal acquisition: Methods and technology,"
IEEE Communications Magazine, vol. 22, pp. 6{21, June 1984.
[346] E. Strom, S. Parkvall, S. Miller, and B. Ottersten, \Propagation delay estimation in asynchronous direct-sequence code division multiple access systems," IEEE Transactions on Communications, vol. 44, pp. 84{93, January 1996.
[347] R. Rick and L. Milstein, \Optimal decision strategies for acquisition of spread-spectrum signals in frequency-selective fading channels," IEEE Transactions on Communications, vol. 46,
pp. 686{694, May 1998.
[348] J. Lee, CDMA Systems Engineering Handbook. London: Artech House Publishers, 1998.
[349] M. Varanasi and B. Aazhang, \Multistage detection in asynchronous code-division multipleaccess communications," IEEE Transactions on Communications, vol. 38, pp. 509{519, April
1990.
[350] S. Moshavi, \Multi-user detection for DS-CDMA communications," IEEE Communications
Magazine, vol. 34, pp. 124{136, October 1996.
BIBLIOGRAPHY
1049
[351] S. Verdu, \Minimum probability of error for asynchronous gaussian multiple-access channel,"
IEEE Transactions on Communications, vol. 32, pp. 85{96, January 1986.
[352] E. Kuan and L. Hanzo, \Joint detection CDMA techniques for third-generation transceivers,"
in Proceeding of ACTS Mobile Communication Summit '98 [670], pp. 727{732.
[353] E. Kuan, C. Wong, and L. Hanzo, \Upper-bound performance of burst-by-burst adaptive joint
detection CDMA." submitted to IEEE Communications Letters, 1998.
[354] S. Verdu, Multiuser Detection. Cambridge: Cambridge University Press, 1998.
[355] F. Simpson and J. Holtzman, \Direct sequence CDMA power control, interleaving, and coding," IEEE Journal on Selected Areas in Communications, vol. 11, pp. 1085{1095, September
1993.
[356] M. Pursley, \Performance evaluation for phase-coded spread-spectrum multiple-access
communication-part I: System analysis," IEEE Transactions on Communications, vol. COM25, pp. 795{799, August 1977.
[357] R. Morrow Jr., \Bit-to-bit error dependence in slotted DS/SSMA packet systems with random
signature sequences," IEEE Transactions on Communications, vol. 37, pp. 1052{1061, October
1989.
[358] J. Holtzman, \A simple, accurate method to calculate spread-spectrum multiple-access error
probabilities," IEEE Transactions on Communications, vol. 40, pp. 461{464, March 1992.
[359] U.-C. Fiebig and M. Schnell, \Correlation properties of extended m-sequences," Electronic
Letters, vol. 29, pp. 1753{1755, September 1993.
[360] F. Davarian, \Mobile digital communications via tone calibration," IEEE Transactions on
Vehicular Technology, vol. VT-36, pp. 55{62, May 1987.
[361] G. Irvine and P. McLane, \Symbol-aided plus decision-directed reception for PSK/TCM modulation on shadowed mobile satellite fading channels," IEEE Journal on Selected Areas in
[362] A. Baier, U.-C. Fiebig, W. Granzow, W. Koch, P. Teder, and J. Thielecke, \Design study for a
CDMA-based third-generation mobile system," IEEE Journal on Selected Areas in Communications, vol. 12, pp. 733{743, May 1994.
[363] P. Rapajic and B. Vucetic, \Adaptive receiver structures for asynchornous CDMA systems,"
IEEE Journal on Selected Areas in Communications, vol. 12, pp. 685{697, May 1994.
[364] M. Benthin and K.-D. Kammeyer, \In uence of channel estimation on the performance of a
coherent DS-CDMA system," IEEE Transactions on Vehicular Technology, vol. 46, pp. 262{
268, May 1997.
[365] M. Sawahashi, Y. Miki, H. Andoh, and K. Higuchi, \Pilot symbol-assisted coherent multistage
interference canceller using recursive channel estimation for DS-CDMA mobile radio," IEICE
Transactions on Communications, vol. E79-B, pp. 1262{1269, September 1996.
[366] S. Sampei and T. Sunaga, \Rayleigh fading compensation for QAM in land mobile radio
communications," IEEE Transactions on Vehicular Technology, vol. 42, pp. 137{147, May
1993.
[367] K. Yen and L. Hanzo, \Third-generation wireless systems," in Steele and Hanzo [85], ch. 10,
pp. 897{964.
[368] The 3GPP1 website. http://www.3gpp.org.
[369] The 3GPP2 website. http://www.3gpp2.org.
[370] T. Ojanpera and R. Prasad, Wideband CDMA for Third Generation Mobile Communications.
London: Artech House, 1998.
[371] E. Dahlman, B. Gudmundson, M. Nilsson, and J. Skold, \UMTS/IMT-2000 based on wideband
CDMA," IEEE Communications Magazine, vol. 36, pp. 70{80, September 1998.
[372] T. Ojanpera, \Overview of research activities for third generation mobile communications," in
Glisic and Leppanen [241], ch. 2 (Part 4), pp. 415{446.
1050
BIBLIOGRAPHY
[373] European Telecommunications Standards Institute, The ETSI UMTS Terrestrial Radio Access
(UTRA) ITU-R RTT Candidate Submission, June 1998. ETSI/SMG/SMG2.
[374] Association of Radio Industries and Businesses, Japan's Proposal for Candidate Radio Transmission Technology on IMT-2000: W-CDMA, June 1998.
[375] A. Sasaki, \Current situation of IMT-2000 radio transmission technology study in Japan,"
IEICE Transactions on Communications, vol. E81-B, pp. 1299{1304, July 1998.
[376] J. Schwarz da Silva, B. Barani, and B. Arroyo-Fernandez, \European mobile communications
on the move," IEEE Communications Magazine, vol. 34, pp. 60{69, February 1996.
[377] F. Ovesjo, E. Dahlman, T. Ojanpera, A. Toskala, and A. Klein, \FRAMES multiple access
mode 2 | wideband CDMA," in Proceedings of IEEE International Symposium on Personal,
Indoor and Mobile Radio Communications, PIMRC'97 [674].
[378] The UMTS Forum website. http://www.umts-forum.org/.
[379] M. Sunay, Z.-C. Honkasalo, A. Hottinen, H. Honkasalo, and L. Ma, \A dynamic channel allocation based TDD DS CDMA residential indoor system," in IEEE 6th International Conference
on Universal Personal Communications, ICUPC'97, (San Diego, CA), pp. 228{234, October
1997.
[380] C. Berrou and A. Glavieux, \Near optimum error correcting coding and decoding: turbo
codes," IEEE Transactions on Communications, vol. 44, pp. 1261{1271, October 1996.
[381] A. Fujiwara, H. Suda, and F. Adachi, \Turbo codes application to DS-CDMA mobile radio,"
IEICE Transactions on Communications, vol. E81A, pp. 2269{2273, November 1998.
[382] M. Juntti, \System concept comparison for multirate CDMA with multiuser detection," in
Proceedings of IEEE Vehicular Technology Conference (VTC'98) [666], pp. 18{21.
[383] S. Ramakrishna and J. Holtzman, \A comparison between single code and multiple code
transmission schemes in a CDMA system," in Proceedings of IEEE Vehicular Technology
Conference (VTC'98) [666], pp. 791{795.
[384] M. Simon, J. Omura, R. Scholtz, and B. Levitt, Spread Spectrum Communications Handbook.
New York: McGraw-Hill, 1994.
[385] T. Kasami, Combinational Mathematics and its Applications. University of North Carolina
Press, 1969.
[386] A. Brand and A. Aghvami, \Multidimensional PRMA with prioritized bayesian broadcast
| a MAC strategy for multiservice trac over UMTS," IEEE Transactions on Vehicular
Technology, vol. 47, pp. 1148{1161, November 1998.
[387] R. Ormondroyd and J. Maxey, \Performance of low rate orthogonal convolutional codes in
DS-CDMA," IEEE Transactions on Vehicular Technology, vol. 46, pp. 320{328, May 1997.
[388] A. Chockalingam, P. Dietrich, L. Milstein, and R. Rao, \Performance of closed-loop power
control in DS-CDMA cellular systems," IEEE Transactions on Vehicular Technology, vol. 47,
pp. 774{789, August 1998.
[389] R. Gejji, \Forward-link-power control in CDMA cellular-systems," IEEE Transactions on Vehicular Technology, vol. 41, pp. 532{536, November 1992.
[390] K. Higuchi, M. Sawahashi, and F. Adachi, \Fast cell search algorithm in DS-CDMA mobile
radio using long spreading codes," in Proceedings of IEEE VTC'97 [679], pp. 1430{1434.
[391] M. Golay, \Complementary series,," IRE Transactions on Information Theory, vol. IT-7,
pp. 82{87, 1961.
[392] V. Tarokh, H. Jafarkhani, and A. Calderbank, \Space-time block codes from orthogonal designs," IEEE Transactions on Information Theory, vol. 45, pp. 1456{1467, May 1999.
[393] W. Lee, Mobile Communications Engineering. New York: McGraw-Hill, 2nd ed., 1997.
[394] H. Wong and J. Chambers, \Two-stage interference immune blind equaliser which exploits
cyclostationary statistics," Electronics Letters, vol. 32, pp. 1763{1764, September 1996.
[395] C. Lee and R. Steele, \Eects of Soft and Softer Handos on CDMA System Capacity," IEEE
Transactions on Vehicular Technology, vol. 47, pp. 830{841, August 1998.
BIBLIOGRAPHY
1051
[396] M. Gustafsson, K. Jamal, and E. Dahlman, \Compressed mode techniques for inter-frequency
measurements in a wide-band DS-CDMA system," in Proceedings of IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, PIMRC'97 [674], pp. 231{
235.
[397] Y. Okumura and F. Adachi, \Variable-rate data transmission with blind rate detection for
coherent DS-CDMA mobile radio," IEICE Transactions on Communications, vol. E81B,
pp. 1365{1373, July 1998.
[398] M. Raitola, A. Hottinen, and R. Wichman, \Transmission diversity in wideband CDMA," in
Proceeding of VTC'99 (Spring) [665], pp. 1545{1549.
[399] J. Liberti Jr. and T. Rappaport, \Analytical results for capacity improvements in CDMA,"
[400] J. Winters, \Smart antennas for wireless systems," IEEE Personal Communications, vol. 5,
[401] T. Lim and L. Rasmussen, \Adaptive symbol and parameter estimation in asynchronous multiuser CDMA detectors," IEEE Transactions on Communications, vol. 45, pp. 213{220, February 1997.
[402] T. Lim and S. Roy, \Adaptive lters in multiuser (MU) CDMA detection," Wireless Networks,
vol. 4, pp. 307{318, June 1998.
[403] L. Wei, \Rotationally-invariant convolutional channel coding with expanded signal space, part I
and II," IEEE Transactions on Selected Areas in Comms, vol. SAC-2, pp. 659{686, September
1984.
[404] T. Lim and M. Ho, \LMS-based simplications to the kalman lter multiuser CDMA detector," in Proceedings of IEEE Asia-Pacic Conference on Communications/International
Conference on Communication Systems, (Singapore), November 1998.
[405] D. You and T. Lim, \A modied blind adaptive multiuser CDMA detector," in Proceedings of
IEEE International Symposium on Spread Spectrum Techniques and Application (ISSSTA'98)
[685], pp. 878{882.

[406] S. Sun, L. Rasmussen, T. Lim, and H. Sugimoto, \Impact of estimation errors on multiuser
detection in CDMA," in Proceedings of IEEE Vehicular Technology Conference (VTC'98)
[666], pp. 1844{1848.
[407] Y. Sanada and Q. Wang, \A co-channel interference cancellation technique using orthogonal
convolutional codes on multipath rayleigh fading channel," IEEE Transactions on Vehicular
Technology, vol. 46, pp. 114{128, February 1997.
[408] P. Patel and J. Holtzman, \Analysis of a simple successive interference cancellation scheme in
a DS/CDMA system," IEEE Journal on Selected Areas in Communications, vol. 12, pp. 796{
807, June 1994.
[409] P. Tan and L. Rasmussen, \Subtractive interference cancellation for DS-CDMA systems," in
Proceedings of IEEE Asia-Pacic Conference on Communications/International Conference

on Communication Systems, (Singapore), November 1998.
[410] K. Cheah, H. Sugimoto, T. Lim, L. Rasmussen, and S. Sun, \Performance of hybrid interference
canceller with zero-delay channel estimation for CDMA," in Proceeding of Globecom'98 [671],
pp. 265{270.
[411] S. Sun, L. Rasmussen, and T. Lim, \A matrix-algebraic approach to hybrid interference cancellation in CDMA," in Proceedings of IEEE International Conference on Universal Personal
Communications '98, (Florence, Italy), pp. 1319{1323, October 1998.
[412] A. Johansson and L. Rasmussen, \Linear group-wise successive interference cancellation in
CDMA," in Proceedings of IEEE International Symposium on Spread Spectrum Techniques
and Application (ISSSTA'98) [685], pp. 121{126.
[413] S. Sun, L. Rasmussen, H. Sugimoto, and T. Lim, \A hybrid interference canceller in CDMA,"
in Proceedings of IEEE International Symposium on Spread Spectrum Techniques and Application (ISSSTA'98) [685], pp. 150{154.
1052
BIBLIOGRAPHY
[414] D. Guo, L. Rasmussen, S. Sun, T. Lim, and C. Cheah, \MMSE-based linear parallel interference cancellation in CDMA," in Proceedings of IEEE International Symposium on Spread
Spectrum Techniques and Application (ISSSTA'98) [685], pp. 917{921.
[415] L. Rasmussen, D. Guo, Y. Ma, and T. Lim, \Aspects on linear parallel interference cancellation in CDMA," in Proceedings of IEEE International Symposium on Information Theory,
(Cambridge, MA), p. 37, August 1998.
[416] L. Rasmussen, T. Lim, H. Sugimoto, and T. Oyama, \Mapping functions for successive interference cancellation in CDMA," in Proceedings of IEEE Vehicular Technology Conference
(VTC'98) [666], pp. 2301{2305.
[417] S. Sun, T. Lim, L. Rasmussen, T. Oyama, H. Sugimoto, and Y. Matsumoto, \Performance
comparison of multi-stage SIC and limited tree-search detection in CDMA," in Proceedings of
IEEE Vehicular Technology Conference (VTC'98) [666], pp. 1854{1858.
[418] H. Sim and D. Cruickshank, \Chip based multiuser detector for the downlink of a DS-CDMA
system using a folded state-transition trellis," in Proceeding of VTC'99 (Spring) [665], pp. 846{
850.
[419] M. Barnsley, \A better way to compress images," BYTE, pp. 215{222, January 1988.
[420] J. Beaumont, \Image data compression using fractal techniques," BT Technology, vol. 9,
pp. 93{109, October 1991.
[421] A. Jacquin, \Image coding based on a fractal theory of iterated contractive image transformations," IEEE Transactions on Image Processing, vol. 1, pp. 18{30, January 1992.
[422] D. Monro and F. Dudbridge, \Fractal block coding of images," Electronic Letters, vol. 28,
pp. 1053{1055, May 1992.
[423] D. Monro, D. Wilson, and J. Nicholls, \High speed image coding with the bath fractal transform," in Damper et al. [677], pp. 23{30.
[424] B. Ramamurthi and A. Gersho, \Classied vector quantization of images," IEEE Transactions
on communications, vol. COM-34, pp. 1105{1115, November 1986.
[425] J. Streit and L. Hanzo, \A fractal video communicator," in Proceedings of IEEE VTC '94 [676],
pp. 1030{1034.
[426] W. Welsh, \Model based coding of videophone images," Electronic and Communication Engineering Journal, pp. 29{36, February 1991.
[427] J. Ostermann, \Object-based analysis-synthesis coding based on the source model of moving
rigid 3D objects," Signal Processing: Image Communication, vol. 6, pp. 143{161, 1994.
[428] M. Chowdhury, \A switched model-based coder for video signals," IEEE Transactions on
Circuits and Systems, vol. 4, pp. 216{227, June 1994.
[429] G. Bozdagi, A. Tekalp, and L. Onural, \3-D motion estimation and wireframe adaptation
including photometric eects for model-based coding of facial image sequences," IEEE Transactions on circuits and Systems for Video Technology, vol. 4, pp. 246{256, June 1994.
[430] Q. Wang and R. Clarke, \Motion estimation and compensation for image sequence coding,"
Signal Processing: Image Communications, vol. 4, pp. 161{174, 1992.
[431] H. Gharavi and M. Mills, \Blockmatching motion estimation algorithms | new results," IEEE
Transactions on Circuits and Systems, vol. 37, pp. 649{651, May 1990.
[432] J. Jain and A. Jain, \Displacement measurement and its applications in inter frame image
coding," IEEE Transactions on Communications, vol. 29, December 1981.
[433] B. Wang, J. Yen, and S. Chang, \Zero waiting-cycle hierarchical block matching algorithm and
its array architectures," IEEE Transactions on Circuits and Systems for Video Technology,
vol. 4, pp. 18{27, February 1994.
[434] P. Strobach, \Tree-structured scene adaptive coder," IEEE Transactions on Communications,
vol. 38, pp. 477{486, April 1990.
[435] B. Liu and A. Zaccarin, \New fast algorithms for the estimation of block motion vectors,"
IEEE Transactions on Circuits and Systems, vol. 3, pp. 148{157, April 1993.
BIBLIOGRAPHY
1053
[436] R. Li, B. Zeng, and N. Liou, \A new three step search algorithm for motion estimation," IEEE
Transactions on Circuits and Systems, vol. 4, pp. 439{442, August 1994.
[437] L. Lee, J. Wang, J. Lee, and J. Shie, \Dynamic search-window adjustment and interlaced
search for block-matching algorithm," IEEE Transactions on Circuits and Systems for Video
[438] B. Girod, \Motion-compensating prediction with fractional-pel accuracy," IEEE Transactions
on Communications, vol. 41, pp. 604{611, April 1993.
[439] J. Huang et al, \A multi-frame pel-recursive algorithm for varying frame-to-frame displacement estimation," in Proceedings of International Conference on Acoustics, Speech, and Signal
Processing, ICASSP'92 [684], pp. 241{244.
[440] N. Efstratiadis and A. Katsaggelos, \Adaptive multiple-input pel-recursive displacement estimation," in Proceedings of International Conference on Acoustics, Speech, and Signal Processing, ICASSP'92 [684], pp. 245{248.
[441] C. Huang and C. Hsu, \A new motion compensation method for image sequence coding using hierarchical grid interpolation," IEEE Transactions on Circuits and Systems for Video
[442] J. Nieweglowski, T. Moisala, and P. Haavisto, \Motion compensated video sequence iinterpolation using digital image warping," in Proceedings of the IEEE International Conference on
Acoustics, Speech and Signal Processing (ICASSP'94) [682], pp. 205{208.
[443] C. Papadopoulos and T. Clarkson, \Motion compensation using second-order geometric transformations," IEEE Transactions on Circuits and Systems for Video Technology, vol. 5,
pp. 319{331, August 1995.
[444] C. Papadopoulos, The use of geometric transformations for motion compensation in video
data compression. PhD thesis, University of London, 1994.
[445] M. Hoetter, \Dierential estimation based on object oriented mapping parameter estimation,"
Signal Processing, vol. 16, pp. 249{265, March 1989.
[446] S. Karunaserker and N. Kingsbury, \A distortion measure for blocking artifacts in images based
on human visual sensitivity," IEEE Transactions on Image Processing, vol. 6, pp. 713{724,
June 1995.
[447] D. Pearson and M. Whybray, \Transform coding of images using interleaved blocks," IEE
Proceedings, vol. 131, pp. 466{472, August 1984.
[448] J. Magarey and N. Kingsbury, \Motion estimation using complex wavelets," in Proceedings of
the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP'96)
[683], pp. 2371{2374.

[449] R. Young and N. Kingsbury, \Frequency-domain motion estimation using a complex lapped
transform," IEEE Transactions on Image Processing, vol. 2, pp. 2{17, January 1993.
[450] R. Young and N. Kingsbury, \Video compression using lapped transforms for motion estimation/compensation and coding," in Proceedings of the SPIE Communication and Image
Processing Conference, (Boston, MA), pp. 1451{1463, SPIE, November 1992.
[451] K. Rao and P. Yip, Discrete Cosine Transform: Algorithms, Advantages and Applications.
New York: Academic Press Ltd., 1990.
[452] A. Sharaf, Video coding at very low bit rates using spatial transformations. PhD thesis,
Department of Electronic and Electrical Engineering, Kings College, London, 1997.
[453] R. Clarke, Transform Coding of Images. New York: Academic Press, 1985.
[454] A. Palau and G. Mirchandani, \Image coding with discrete cosine transforms using ecient
energy-based adaptive zonal ltering," in Proceedings of the IEEE International Conference
on Acoustics, Speech and Signal Processing (ICASSP'94) [682], pp. 337{340.
[455] H. Yamaguchi, \Adaptive DCT coding of video signals," IEEE Transactions on Communications, vol. 41, pp. 1534{1543, October 1993.
[456] K. Ngan, \Adaptive transform coding of video signals," IEE Proceedings, vol. 129, pp. 28{40,
February 1982.
1054
BIBLIOGRAPHY
[457] R. Clarke, \Hybrid intra-frame transform coding of image data," IEE Proceedings, vol. 131,
[458] F.-M. Wang and S. Liu, \Hybrid video coding for low bit-rate applications," in Proceedings of
[682], pp. 481{484.

[459] M. Ghanbari and J. Azari, \Eect of bit rate variation of the base layer on the performance of
two-layer video codecs," IEEE Transactions on Communications for Video Technology, vol. 4,
[460] N. Jayant and P. Noll, Digital Coding of Waveforms, Principles and Applications to Speech
and Video. Englewood Clis, NJ: Prentice-Hall, 1984.
[461] N. Cheng and N. Kingsbury, \The ERPC: an ecient error-resilient technique for encoding positional information of sparse data," IEEE Transactions on Communications, vol. 40,
pp. 140{148, January 1992.
[462] M. Narasimha and A. Peterson, \On the computation of the discrete cosine transform," IEEE
Transactions on Communications, vol. 26, pp. 934{936, June 1978.
[463] R. M. Pelz, \An un-equal error protected px8 kbit/s video transmission for DECT," in Proceedings of IEEE VTC '94 [676], pp. 1020{1024.
[464] L. Hanzo, R. Stedman, R. Steele, and J. Cheung, \A portable multimedia communicator
scheme," in Damper et al. [677], pp. 31{54.
[465] R. Stedman, H. Gharavi, L. Hanzo, and R. Steele, \Transmission of subband-coded images via
mobile channels," IEEE Transactions on Circuits and Systems for Video Technology, vol. 3,
[466] L. Hanzo and J. Streit, \Adaptive low-rate wireless videophone systems," IEEE Transactions
on Circuits and Systems for Video Technology, vol. 5, pp. 305{319, August 1995.
[467] ETSI, GSM Recommendation 05.05, Annex 3, November 1988.
[468] G. Djuknic and D. Schilling, \Performance analysis of an ARQ transmission scheme for meteor burst communications," IEEE Transactions on Communications, vol. 42, pp. 268{271,
February/March/April 1994.
[469] L. de Alfaro and A. Meo, \Codes for second and third order GH-ARQ schemes," IEEE Transactions on Communications, vol. 42, pp. 899{910, February{April 1994.
[470] T.-H. Lee, \Throughput performance of a class of continuous ARQ strategies for burst-error
channels," IEEE Transactions on Vehicular Technology, vol. 41, pp. 380{386, November 1992.
[471] S. Lin, D. Constello Jr., and M. Miller, \Automatic-repeat-request error-control schemes,"
IEEE Communications Magazine, vol. 22, pp. 5{17, December 1984.
[472] Research and Development Centre for Radio Systems, Japan, Public Digital Cellular (PDC)
Standard, RCR STD-27.
[473] A. Gersho and R. Gray, Vector Quantization and Signal Compression. Dordrecht: Kluwer
Academic Publishers, 1992.
[474] L. Torres, J. Casas, and S. deDiego, \Segmetation based coding of textures using stochastic vector quantization," in Proceedings of the IEEE International Conference on Acoustics,
Speech and Signal Processing (ICASSP'94) [682], pp. 597{600.
[475] M. Jaisimha, J. Goldschneider, A. Mohr, E. Riskin, and R. Haralick, \On vector quantization for fast facet edge detection," in Proceedings of the IEEE International Conference on
Acoustics, Speech and Signal Processing (ICASSP'94) [682], pp. 37{40.
[476] P. Yu and A. Venetsanopoulos, \Hierarchical nite-state vector quantisation for image coding,"
IEEE Transactions on Communications, vol. 42, pp. 3020{3026, November 1994.
[477] C.-H. Hsieh, K.-C. Chuang, and J.-S. Shue, \Image compression using nite-state vector quantization with derailment compensation," IEEE Transactions on Circuits and systems for Video
Technology, vol. 3, pp. 341{349, October 1993.
BIBLIOGRAPHY
1055
[478] N. Nasrabadi, C. Choo, and Y. Feng, \Dynamic nite-state vector quantisation of digital
images," IEEE Transactions on Communications, vol. 42, pp. 2145{2154, May 1994.
[479] V. Sitaram, C. Huang, and P. Israelsen, \Ecient codebooks for vector quantisation image
compression with an adaptive tree search algorithm," IEEE Transactions on Communications,
vol. 42, pp. 3027{3033, November 1994.
[480] W. Yip, S. Gupta, and A. Gersho, \Enhanced multistage vector quantisation by joint codebook
design," IEEE Transactions on Communications, vol. 40, pp. 1693{1697, November 1992.
[481] L. Po and C. Chan, \Adaptive dimensionality reduction techniques for tree-structured vector
quantisation," IEEE Transactions on Communications, vol. 42, pp. 2246{2257, June 1994.
[482] L. Lu and W. Pearlman, \Multi-rate video coding using pruned tree-structured vector quantization," in Proceedings of the IEEE International Conference on Acoustics, Speech and Signal
Processing (ICASSP'93), vol. 5, (Minneapolis, MN), pp. 253{256, IEEE, 27{30 April 1993.
[483] F. Bellifemine and R. Picco, \Video signal coding with DCT and vector quantisation," IEEE
[484] K. Ngan and K. Sin, \HDTV coding using hybrid MRVQ/DCT," IEEE Transactions on
Circuits and Systems for Video Technology, vol. 3, pp. 320{323, August 1993.
[485] D. Kim and S. Lee, \Image vector quantiser based on a classication in the DCT domain,"
IEEE Transactions on Communications, pp. 549{556, April 1991.
[486] L. Torres and J. Huguet, \An improvement on codebook search for vector quantisation," IEEE
[487] J. Streit and L. Hanzo, \Dual-mode vector-quantised low-rate cordless videophone systems
for indoors and outdoors applications," IEEE Transactions on Vehicular Technology, vol. 46,
pp. 340{357, May 1997.
[488] X. Zhang, M. Cavenor, and J. Arnold, \Adaptive quadtree coding of motion -compensated
image sequences for use on the broadband ISDN," IEEE Transactions on Circuits and Systems
for Video Technology, vol. 3, pp. 222{229, June 1993.
[489] J. Vaisey and A. Gersho, \Image compression with variable block size segmentation," IEEE
Transactions on Signal Processing, vol. 40, pp. 2040{2060, August 1992.
[490] M. Lee and G. Crebbin, \Classied vector quantisation with variable block-size DCT models,"
IEE Proceedings, Vision, Image and Signal Processing, pp. 39{48, February 1994.
[491] E. Shustermann and M. Feder, \Image compression via improved quadtree decomposition
algorithms," IEEE Transactions on Image Processing, vol. 3, pp. 207{215, March 1994.
[492] F. DeNatale, G. Desoli, and D. Giusto, \A novel tree-structured video codec," in Proceedings of
[493]
[494]
[495]
[496]
[497]
[498]
[682], pp. 485{488.

M. Hennecke, K. Prasad, and D. Stork, \Using deformable templates to infer visual speech dynamics," in Proceedings of the 28th Asilomar Conference on Signals, Systems and Computers,
vol. 1, (Pacic Grove, CA), pp. 578{582, 30 October { 2 November 1994.
G. Wolf et al., \Lipreading by neural networks: Visual preprocessing, learning and sensory
integration," Proceedings of the neural information processing systems, vol. 6, pp. 1027{1034,
1994.
J. Streit and L. Hanzo, \Quad-tree based parametric wireless videophone systems," IEEE
Transactions Video Technology, vol. 6, pp. 225{237, April 1996.
E. Biglieri and M. Luise, \Coded modulation and bandwidth-ecient transmission," in Proceedings of the Fifth Tirrenia Internatioanal Workshop, (Netherlands), 8{12 September 1991.
L.-F. Wei, \Trellis-coded modulation with multidimensional constellations," IEEE Transactions on Information Theory, vol. IT-33, pp. 483{501, July 1987.
ITU-T, ISO/IEC-CD-11172 | Coding of moving pictures and associated audio for digital
storage.
1056
BIBLIOGRAPHY
[499] ITU-T, Recommendation H.261: Video codec for audiovisual services at p x64 Kbit/s, March
1993.
[500] D. Redmill and N. Kingsbury, \Improving the error resilience of entropy encoded video signals,"
in Proceedings of the Conference on Image Processing: Theory and Applications (IPTA),
(Netherlands), pp. 67{70, Elsevier, 1993.
[501] S. Emani and S. Miller, \DPCM picture transmission over noisy channels with the aid of a
markov model," IEEE Transactions on Image Processing, vol. 4, pp. 1473{1481, November
1995.
[502] M. Chan, \The performance of DPCM operating on lossy channels with memory," IEEE
Transactions on Communications, vol. 43, pp. 1686{1696, April 1995.
[503] N. Jayant, \Adaptive quantization with a one-word memory," Bell System Technical Journal,
vol. 52, pp. 1119{1144, September 1973.
[504] L. Zetterberg, A. Ericsson, and C. Couturier, \DPCM picture coding with two-dimensional
control of adaptive quantisation," IEEE Transactions on Communications, vol. 32, no. 4,
pp. 457{642, 1984.
[505] C. Hsieh, P. Lu, and W. Liou, \Adaptive predictive image coding using local characteristics,"
IEE Proceedings, vol. 136, pp. 385{389, December 1989.
[506] P. Wellstead, G. Wagner, and J. Caldas-Pinto, \Two-dimensional adaptive prediction, smoothing and ltering," Proceedings of the IEE, vol. 134, pp. 253{266, June 1987.
[507] O. Mitchell, E. Delp, and S. Carlton, \Block truncation coding: A new approach to image
compression," in IEEE International Conference on Communications (ICC), pp. 12B.1.1{
12B.1.4, 1978.
[508] E. Delp and O. Mitchell, \Image compression using block truncation coding," IEEE Transactions on Communications, vol. 27, pp. 1335{1342, September 1979.
[509] D. Halverson, N. Griswold, and G. Wiese, \A generalized block truncation coding algorithm
for image compression," IEEE Transactions Acoustics, Speech and Signal Processing, vol. 32,
pp. 664{668, June 1984.
[510] G. Arce and N. Gallanger, \BTC image coding using median lter roots," IEEE Transactions
on Communications, vol. 31, pp. 784{793, June 1983.
[511] M. Noah, \Optimal Lloyd-Max quantization of LPC speech parameters," in Proceedings of International Conference on Acoustics, Speech, and Signal Processing, ICASSP'84, (San Diego,
CA), pp. 1.8.1{1.8.4, IEEE, 19{21 March 1984.
[512] R. Crochiere, S. Webber, and J. Flangan, \Digital coding of speech in sub-bands," Bell System
Technology Journal, vol. 52, pp. 1105{1118, 1973.
[513] R. Crochiere, \On the design of sub-band coders for low bit rate speech communication," Bell
System Technology Journal, vol. 56, pp. 747{770, 1977.
[514] J. Woods and S. O'Neil, \Subband coding of images," IEEE Transactions on Accoustic, Sound
and Signal Processing, vol. 34, pp. 1278{1288, October 1986.
[515] J. Woods, ed., Subband Image Coding. Dordrecht: Kluwer Academic Publishers, March 1991.
[516] H. Gharavi and A. Tabatabai, \Subband coding of digital images using two-dimensional
quadrature mirror ltering," in Proceedings of SPIE, 1986.
[517] H. Gharavi and A. Tabatabai, \Subband coding of monochrome and color images," IEEE
Transactions on Circuits and Systems, vol. 35, pp. 207{214, February 1988.
[518] H. Gharavi, \Subband coding algorithms for video applications: Videophone to HDTVconferencing," IEEE Transactions on Circuits and Systems for Video Technology, vol. 1,
pp. 174{183, February 1991.
[519] A. Alasmari, \An adaptive hybrid coding scheme for HDTV and digital video sequences,"
IEEE Transactions on consumer electronics, vol. 41, no. 3, pp. 926{936, 1995.
[520] K. Irie et al., \High-quality subband coded for HDTV transmission," IEEE Transactions on
Circuits and Systems for Video Technology, vol. 4, pp. 195{199, April 1994.
BIBLIOGRAPHY
1057
[521] E. Simoncelli and E. Adelson, \Subband transforms," in Woods [515], pp. 143{192.
[522] K. Irie and R. Kishimoto, \A study on perfect reconstructive subband coding," IEEE Transactions on Circuits and Systems for Video Technology, vol. 1, pp. 42{48, January 1991.
[523] J. Woods and T. Naveen, \A lter based bit allocation scheme for subband compression of
HDTV," IEEE Transactions on Image Processing, vol. 1, pp. 436{440, July 1992.
[524] D. Esteban and C. Galand, \Application of quadrature mirror lters to split band voice coding scheme," in Proceedings of International Conference on Acoustics, Speech, and Signal
Processing, ICASSP'77, (Hartford, CT), pp. 191{195, IEEE, 9{11 May 1977.
[525] J. Johnston, \A lter family designed for use in quadrature mirror lter banks," in Proceedings
of International Conference on Acoustics, Speech, and Signal Processing, ICASSP'80 [680],
pp. 291{294.
[526] H. Nussbaumer, \Complex quadrature mirror lters," in Proceedings of International Conference on Acoustics, Speech, and Signal Processing, ICASSP'83, (Boston, MA), pp. 221{223,
IEEE, 14{16 April 1983.
[527] C. Galand and H. Nussbaumer, \New quadrature mirror lter structures," IEEE Transactions
on Acoustic Speech Signal Processing, vol. ASSP-32, pp. 522{531, June 1984.
[528] R. Crochiere and L. Rabiner, Multirate Digital Processing. Englewood Clis, NJ: Prentice-Hall,
1993.
[529] S. Aase and T. Ramstad, \On the optimality of nonunitary lter banks in subband coders,"
IEEE Transactions on Image Processing, vol. 4, pp. 1585{1591, December 1995.
[530] V. Nuri and R. Bamberger, \Size limited lter banks for subband image compression," IEEE
Transactions on Image Processing, vol. 4, pp. 1317{1323, September 1995.
[531] H. Gharavi, \Subband coding of video signals," in Woods [515], pp. 229{271.
[532] O. Egger, W. Li, and M. Kunt, \High compression image coding using an adaptive morphological subband decomposition," Proceedings of the IEEE, vol. 83, pp. 272{287, February 1995.
[533] P. Westerink and D. Boekee, \Subband coding of color images," in Woods [515], pp. 193{228.
[534] Q. Nguyen, \Near-perfect-reconstruction pseudo-QMF banks," IEEE Transactions on signal
processing, vol. 42, pp. 65{76, January 1994.
[535] S.-M. Phoong, C. Kim, P. Vaidyanathan, and R. Ansari, \A new class of two-channel biorthogonal lter banks and wavelet bases," IEEE Transactions on Signal Processing, vol. 43, pp. 649{
665, March 1995.
[536] E. Jang and N. Nasrabadi, \Subband coding with multistage VQ for wireless image communication," IEEE Transactions in Circuit and Systems for Video Technology, vol. 5, pp. 347{253,
June 1995.
[537] P. Cosman, R. Gray, and M. Vetterli, \Vector quantisation of image subbands: A survey,"
IEEE Transactions on Image Processing, vol. 5, pp. 202{225, February 1996.
[538] ITU, Joint Photographic Experts Group ISO/IEC, JTC/SC/WG8, CCITT SGVIII. JPEG
technical specications, revision 5. Report JPEG-8-R5, January 1990.
[539] P. Franti and O. Nevalainen, \Block truncation coding with entropy coding," IEEE Transcations on Communications, vol. 43, no. 4, pp. 1677{1685, 1995.
[540] V. Udpikar and J. Raina, \BTC image coding using vector quantisation," IEEE Transactions
on Communications, vol. 35, pp. 353{359, March 1987.
[541] International Standards Organization, ISO/IEC 11172 MPEG 1 International Standard, 'Cod-
ing of moving pictures and associated audio for digital storage media up to about 1.5 Mbit/s,
Parts 1{3.
[542] International Standards Organization, ISO/IEC CD 13818 MPEG 2 International Standard,
Information Technology, Generic Coding of Moving Video and Associated Audio Information,
Parts 1{3.
[543] Telenor Research and Development, P.O.Box 83, N-2007 Kjeller, Norway, Video Codec Test
Model 'TMN 5', ITU Study Group 15, Working Party 15/1.
1058
BIBLIOGRAPHY
[544] D. Choi, \Frame alignment in a digital carrier system | a tutorial," IEEE Communications
Magazin, vol. 28, pp. 46{54, February 1990.
[545] ITU (formerly CCITT), ITU Recommendation X25, 1993.
[546] M. Al-Subbagh and E. Jones, \Optimum patterns for frame alignment," IEE Proceedings,
vol. 135, pp. 594{603, December 1988.
[547] T. Turletti, \A H.261 software codec for videoconferencing over the internet," Tech. Rep. 1834,
INRIA, 06902 Sophia-Antipolis, France, January 1993.
[548] N. Kenyon and C. Nightingale, eds., Audiovisual Telecommunications. London: Chapman and
Hall, 1992.
[549] N. MacDonald, \Transmission of compressed video over radio links," BT technology Journal,
vol. 11, pp. 182{185, April 1993.
[550] M. Khansari, A. Jalali, E. Dubois, and P. Mermelstein, \Robust low bit-rate video transmission
over wireless access systems," in Proceedings of International Communications Conference
(ICC), pp. 571{575, 1994.
[551] N. Cheng, Error resilient video coding for Noisy Channels. PhD thesis, Department of Engineering, University of Cambridge, 1991.
[552] D. Redmill, Image and Video Coding for Noisy Channels. PhD thesis, Signal Processing and
Communication Laboratory, Department of Engineering, University of Cambridge, November
1994.
[553] Y. Matsumura, S. Nakagawa, and T. Nakai, \Very low bit rate video coding with error resilience," in VLBV'95 [686], pp. L{1.
[554] K. Ngan and D. Chai, \Enhancement of image quality in VLBR coding," in VLBV'95 [686],
pp. L{3.
[555] K. Ngan and D. Chai, \Very low bit rate video coding using H.263 coder," IEEE Transactions
on Circuits and Systems for Video Technology, vol. 6, pp. 308{312, June 1996.
[556] W. Webb and L. Hanzo, \Square QAM," in Modern Quadrature Amplitude Modulation: Principles and Applications for Wireless Communications [17], ch. 5, pp. 156{169.
[557] IBM Corp., White Plains, NY, General Information: Binary Synchronous Communication,
IBM Publication GA27-3004, 1969.
[558] S. Sampei, S. Komaki, and N. Morinaga, \Adaptive modulation/TDMA scheme for large
capacity personal multi-media communication systems," IEICE Transactions on Communications (Japan), vol. E77-B, pp. 1096{1103, September 1994.
[559] W. Webb and L. Hanzo, \Variable rate QAM," in Modern Quadrature Amplitude Modulation:
Principles and Applications for Wireless Communications [17], ch. 13, pp. 384{406.
[560] M. Ghanbari and V. Seferidis, \Cell-loss concealment in ATM video codecs," IEEE Transactions on Circuits and Systems for Video Technology, vol. 3, pp. 238{247, June 1993.
[561] W. Chung, F. Kossentini, and M. Smith, \An ecient motion estimation technique based on a
rate-distortion criterion," in Proceedings of the IEEE International Conference on Acoustics,
[562] P. Howard and J. Vitter, \Arithmetic coding for data compression," Proceedings of the IEEE,
vol. 82, pp. 857{865, June 1994.
[563] Telenor Research and Development, P.O.Box 83, N-2007 Kjeller, Norway, H.263 Software
Codec. http://www.nta.no/brukere/DVC.
[564] W. Ding and B. Liu, \Rate control of MPEG video coding and recording by rate-quantization
modeling," IEEE Transactions on Circuits and Systems for Video Technology, vol. 6, pp. 12{
20, February 1996.
[565] G. Schuster and A. Katsaggelos, \A video compression scheme with optimal bit allocation
between displacement vector eld and displaced frame dierence," in Proceedings of the IEEE
International Conference on Acoustics, Speech and Signal Processing (ICASSP'96) [683],
pp. 1967{1970.
BIBLIOGRAPHY
1059
[566] F. Martins, W. Ding, and E. Feig, \Joint control of spatial quantization and temporal sampling
for very low bitrate video," in Proceedings of the IEEE International Conference on Acoustics,
[567] T. Wiegand, M. Lightstone, and D. Mukherjee, \Rate-distortion optimized mode selection
for very low bit rate video coding and the emerging H.263 standard," IEEE Transactions on
Circuits and Systems for Video Technology, vol. 6, pp. 182{190, April 1996.
[568] A. Paulraj, \Diversity techniques," in Gibson [291], ch. 12, pp. 166{176.
[569] A. Mammela, Diversity receivers in a fast fading multipath channel. PhD thesis, Department
of Electrical Engineering, University of Oulu, Finland, 1995.
[570] P. Crespo, R. M. Pelz, and J. Cosmas, \Channel error prole for DECT," IEE Proceedings on
Communications, vol. 141, pp. 413{420, December 1994.
[571] M. Zorzi, \Power control and diversity in mobile radio cellular systems in the presence of
rician fading and log-normal shadowing," IEEE Transactions on Vehicular Technology, vol. 45,
pp. 373{382, May 1996.
[572] Y.-W. Leung, \Power control in cellular networks subject to measurement error," IEEE Transsactions on Communications, vol. 44, pp. 772{775, July 1996.
[573] S. Ariyavisitakul and L. Chang, \Signal and interference statistics of a CDMA system with
feedback power control," IEEE Transactions on Communications, vol. 41, pp. 1626{1634,
November 1993.
[574] R. Pichna and Q. Wang, \Power control," in Gibson [291], ch. 23, pp. 370{380.
[575] T.-H. Lee, J.-C. Lin, and Y. Su, \Downlink power control algorithms for cellular radio systems,"
[576] M. Austin and G. Stuber, \In-service signal quality estimation for TDMA cellular systems,"
Wireless Personal Communications, vol. 2, pp. 245{254, 1995. Kluwer Academic Publishers.
[577] J.-I. Chuang and N. Sollenberger, \Uplink power control for TDMA portable radio channels,"
[578] P. Kumar, R. Yates, and J. Holtzman, \Power control based on bit error rate (BER) measurements," in Proceedings of the Military Communications Conference (MILCOM), (San Diego,
CA), 5{8 November 1995.
[579] W. Press, S. Teukolsky, W. Vetterling, and B. Flannery, \Minimization or maximization of
functions," in Numerical Recipes in C [147], ch. 10, pp. 394{455.
[580] L. Hanzo, W. Webb, and T. Keller, Single- and Multi-carrier Quadrature Amplitude Modulation. New York: John Wiley-IEEE Press, April 2000.
[581] C. Douillard, A. Picart, M. Jezequel, P. Didier, C. Berrou, and A. Glavieux, \Iterative correction of intersymbol interference: Turbo-equalization," European Transactions on Communications, vol. 6, pp. 507{511, 1995.
[582] M. Gertsman and J. Lodge, \Symbol-by-symbol MAP demodulation of CPM and PSK signals
on Rayleigh at-fading channels," IEEE Transactions on Communications, vol. 45, pp. 788{
799, July 1997.
[583] I. Marsland, P. Mathiopoulos, and S. Kallel, \Non-coherent turbo equalization for frequency
selective Rayleigh fast fading channels," in Proceedings of the International Symposium on
Turbo Codes & Related Topics, (Brest, France), pp. 196{199, 3{5 September 1997.
[584] Q. Dai and E. Shwedyk, \Detection of bandlimited signals over frequency selective
Rayleigh fading channels," IEEE Transactions on Communications, pp. 941{950, February/March/April 1994.
[585] G. Bauch, H. Khorram, and J. Hagenauer, \Iterative equalization and decoding in mobile communications systems," in European Personal Mobile Communications Conference, pp. 301{
312, 1997.
[586] M. Moher, \Decoding via cross-entropy minimization," in Proceedings of the IEEE Global
Telecommunications Conference 1993 [667], pp. 809{813.
1060
BIBLIOGRAPHY
[587] G. Bauch and V. Franz, \Iterative equalisation and decoding for the GSM-system," in Proceedings of IEEE Vehicular Technology Conference (VTC'98) [666], pp. 2262{2266.
[588] D. Raphaeli and Y. Zarai, \Combined turbo equalization and turbo decoding," IEEE Communications Letters, vol. 2, pp. 107{109, April 1998.
[589] K. Narayanan and G. Stuber, \A serial concatenation approach to iterative demodulation and
decoding," IEEE Transactions on Communications, vol. 47, pp. 956{961, July 1999.
[590] B. Yeap, T. Liew, J. Hamorsky, and L. Hanzo, \Comparative study of turbo equalisers using
convolutional codes and block-based turbo-codes for GMSK modulation," in Proceeding of
VTC'99 (Fall), (Amsterdam, Netherlands), pp. 2974{2978, IEEE, 19{22 September 1999.
[591] H. Matsuoka, S. Sampei, N. Morinaga, and Y. Kamio, \Adaptive modulation system with
variable coding rate concatenated code for high quality multi-media communications systems,"
in Proceedings of IEEE VTC'96 [678], pp. 487{491.
[592] A. Goldsmith and S. Chua, \Adaptive coded modulation for fading channels," in Proceedings
of IEEE International Conference on Communications, vol. 3, (Montreal, Canada), pp. 1488{
1492, 8{12 June 1997.
[593] J. Cheung, Adaptive Equalisers for Wideband TDMA Mobile Radio. PhD thesis, Department
of Electronics and Computer Science, University of Southampton, 1991.
[594] N. Farber, E. Steinbach, and B. Girod, \Robust H.263 compatible transmission for mobile
video server access," in Proc of First International Workshop on Wireless Image/Video Communications, (Loughborough), pp. 122{124, 4{5 September 1996.
[595] A. Sadka, F. Eryurtlu, and A. Kondoz, \Improved performance H.263 under erroneous transmission conditions," Electronics Letters, vol. 33, pp. 122{124, 16 January 1997.
[596] A. Klein, R. Pirhonen, J. Skoeld, and R. Suoranta, \FRAMES multiple access mode 1 | wideband TDMA with and without spreading," in Proceedings of IEEE International Symposium
on Personal, Indoor and Mobile Radio Communications, PIMRC'97 [674], pp. 37{41.
[597] J. Torrance and L. Hanzo, \Latency and networking aspects of adaptive modems over slow
indoors rayleigh fading channels," IEEE Transactions on Vehicular Technology, vol. 48, no. 4,
pp. 1237{1251, 1998.
[598] L. Hanzo, P. Cherriman, and J. Streit, \Video compression and communications over wireless
channels: From second to third generation systems, WLANs and beyond." IEEE Press, 2001
(For detailed contents please refer to http://www-mobile.ecs.soton.ac.uk.).
[599] P. Cherriman, C. Wong, and L. Hanzo, \Multi-mode H.263-assisted video telephony using
wideband adaptive burst-by-burst modems," Core programme deliverable | terminals work
area, Mobile VCE, 1998.
[600] P. Cherriman, T. Keller, and L. Hanzo, \Orthogonal frequency division multiplex transmission
of H.263 encoded video over highly frequency-selective wireless networks," IEEE Transactions
on Circuits and Systems for Video Technology, vol. 9, pp. 701{712, August 1999.
[601] M. Sandell, J.-J. van de Beek, and P. Borjesson, \Timing and frequency synchronisation in
OFDM systems using the cyclic prex," in Proceedings of International Symposium on Synchronisation, (Essen, Germany), pp. 16{19, 14{15 December 1995.
[602] J. D. Marca and N. Jayant, \An algorithm for assigning binary indices to the codevectors of a
multi-dimensional quantizer," in Proceedings of IEEE International Conference on Communications 1987, (Seattle, WA), pp. 1128{1132, IEEE, 7{10 June 1987.
[603] T. Keller and L. Hanzo, \Orthogonal frequency division multiplex synchronisation techniques
for wireless local area networks," in Proceedings of IEEE International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC'96) [681], pp. 963{967.
[604] K. Pehkonen, H. Holma, I. Keskitalo, E. Nikula, and T. Westman, \Performance analysis of
TDMA and CDMA based air interface solutions for UMTS high bit rate services," in Proceed-
ings of IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, PIMRC'97 [674], pp. 22{26.
BIBLIOGRAPHY
1061
[605] K. Pajukoski and J. Savusalo, \Wideband CDMA test system," in Proceedings of IEEE In-
ternational Symposium on Personal, Indoor and Mobile Radio Communications, PIMRC'97
[674], pp. 669{673.

[606] R. Blahut, \Transform techniques for error control codes," IBM Journal on Research and
Development, vol. 23, pp. 299{315, May 1979.
[607] J. Torrance, L. Hanzo, and T. Keller, \Interference aspects of adaptive modems over slow
rayleigh fading channels," IEEE Transactions on Vehicular Technology, vol. 48, pp. 1527{
[608] T. Liew, C. Wong, and L. Hanzo, \Block turbo coded burst-by-burst adaptive modems," in
Proceedings of Microcoll'99, Budapest, Hungary, pp. 59{62, 21{24 March 1999.
[609] V. Lau and M. Macleod, \Variable rate adaptive trellis coded QAM for high bandwidth eciency applications in rayleigh fading channels," in Proceedings of IEEE Vehicular Technology
Conference (VTC'98) [666], pp. 348{352.
[610] M. Yee and L. Hanzo, \Upper-bound performance of radial basis function decision feedback equalised burst-by-burst adaptive modulation," in Proceedings of ECMCS'99, (Krakow,
Poland), 24{26 June 1999.
[611] T. Keller and L. Hanzo, \Adaptive orthogonal frequency division multiplexing schemes," in
Proceeding of ACTS Mobile Communication Summit '98 [670], pp. 794{799.
[612] E. Kuan, C. Wong, and L. Hanzo, \Burst-by-burst adaptive joint detection CDMA," in Proceeding of VTC'99 (Spring) [665].
[613] A. Czylwik, \Adaptive OFDM for wideband radio channels," in Proceeding of IEEE Global
Telecommunications Conference, Globecom 96 [672], pp. 713{718.
[614] R. Fischer and J. Huber, \A new loading algorithm for discrete multitone transmission," in
Proceeding of IEEE Global Telecommunications Conference, Globecom 96 [672], pp. 713{718.
[615] P. Chow, J. Cio, and J. Bingham, \A practical discrete multitone transceiver loading algorithm for data transmission over spectrally shaped channels," IEEE Transactions on Communications, vol. 48, pp. 772{775, 1995.
[616] H. Rohling and R. Grunheid, \Peformance of an OFDM-TDMA mobile communication system," in Proceeding of IEEE Global Telecommunications Conference, Globecom 96 [672],
pp. 1589{1593.
[617] K. Fazel, S. Kaiser, P. Robertson, and M. Ruf, \A concept of digital terrestrial television
broadcasting," Wireless Personal Communications, vol. 2, pp. 9{27, 1995.
[618] H. Sari, G. Karam, and I. Jeanclaude, \Transmission techniques for digital terrestrial TV
broadcasting," IEEE Communications Magazine, pp. 100{109, February 1995.
[619] J. Borowski, S. Zeisberg, J. Hubner, K. Koora, E. Bogenfeld, and B. Kull, \Performance of
OFDM and comparable single carrier system in MEDIAN demonstrator 60GHz channel," in
Proceeding of ACTS Mobile Communication Summit '97 [668], pp. 653{658.
[620] P. Cherriman, T. Keller, and L. Hanzo, \Constant-rate turbo-coded and block-coded orthogonal frequency division multiplex videophony over UMTS," in Proceeding of Globecom'98 [671],
pp. 2848{2852.
[621] J. Woodard, T. Keller, and L. Hanzo, \Turbo-coded orthogonal frequency division multiplex
transmission of 8 kbps encoded speech," in Proceeding of ACTS Mobile Communication Summit '97 [668], pp. 894{899.
[622] Y. Li and N. Sollenberger, \Interference suppression in OFDM systems using adaptive antenna
arrays," in Proceeding of Globecom'98 [671], pp. 213{218.
[623] F. Vook and K. Baum, \Adaptive antennas for OFDM," in Proceedings of IEEE Vehicular
Technology Conference (VTC'98) [666], pp. 608{610.
[624] ETSI, Digital Video Broadcasting (DVB); Framing structure, channel coding and modulation
for digital terrestrial television, August 1997. EN 300 744 V1.1.2.
for cable systems, December 1997. EN 300 429 V1.2.1.
1062
BIBLIOGRAPHY
for 11/12 GHz Satellite Services, August 1997. EN 300 421 V1.1.2.
[627] A. Michelson and A. Levesque, Error Control Techniques for Digital Communication. New
York: Wiley-Interscience, 1985.
[628] S. O'Leary and D. Priestly, \Mobile broadcasting of DVB-T signals," IEEE Transactions on
Broadcasting, vol. 44, pp. 346{352, September 1998.
[629] W.-C. Lee, H.-M. Park, K.-J. Kang, and K.-B. Kim, \Performance analysis of viterbi decoder
using channel state information in COFDM system," IEEE Transactions on Broadcasting,
vol. 44, pp. 488{496, December 1998.
[630] S. O'Leary, \Hierarchical transmission and COFDM systems," IEEE Transactions on Broadcasting, vol. 43, pp. 166{174, June 1997.
[631] L. Thibault and M. Le, \Performance evaluation of COFDM for digital audoo broadcasting
Part I: parametric study," IEEE Transactions on Broadcasting, vol. 43, pp. 64{75, March
1997.
[632] B. Haskell, A. Puri, and A. Netravali, Digital Video: An Introduction To MPEG-2. Digital
Multimedia Standards Series, London: Chapman and Hall, 1997.
[633] ISO/IEC 13818-2: Information Technology | Generic Coding of Moving Pictures and Associated Audio Information | Part 2: Video, March 1995.
[634] L. Hanzo and J. Woodard, \An intelligent voice communications system for indoors communications," in Proceedings of IEEE Vehicular Technology Conference (VTC'95) [669], pp. 735{
749.
[635] P. Shelswell, \The COFDM modulation system: the heart of digital audio broadcasting,"
Electronics & Communication Engineering Journal, vol. 7, pp. 127{136, June 1995.
[636] P. Robertson, E. Villebrun, and P. Hoeher, \A comparison of optimal and sub-optimal MAP
decoding algorithms operating in the log domain," in Proceedings of the International Conference on Communications, pp. 1009{1013, June 1995.
[637] S. Wicker, Error Control Systems for Digital Communication and Storage. Englewood Clis,
NJ: Prentice-Hall, 1994.
[638] A. Barbulescu and S. Pietrobon, \Interleaver design for turbo codes," IEE Electronic Letters,
pp. 2107{2108, December 1994.
[639] M. Failli, \Digital land mobile radio communications COST 207," tech. rep., European Commission, 1989.
[640] H. Gharavi and S. Alamouti, \Multipriority video transmission for third-generation wireless
communication system," in Gharavi and Hanzo [642], pp. 1751{1763.
[641] A. Aravind, M. Civanlar, and A. Reibman, \Packet loss resilience of MPEG-2 scalable video
coding algorithms," IEEE Transaction on Circuits And Systems For Video Technology, vol. 6,
pp. 426{435, October 1996.
[642] H. Gharavi and L. Hanzo, eds., Proceedings of the IEEE, vol. 87, October 1999.
[643] G. Reali, G. Barua, S. Cacopardi, and F. Frescura, \Enhancing satellite broadcasting services
using multiresolution modulations," IEEE Transactions on Broadcasting, vol. 44, pp. 497{506,
December 1998.
[644] Y. Hsu, Y. Chen, C. Huang, and M. Sun, \MPEG-2 spatial scalable coding and transport
stream error concealment for satellite TV broadcasting using Ka-band," IEEE Transactions
on Broadcasting, vol. 44, pp. 77{86, March 1998.
[645] L. Atzori, F. D. Natale, M. D. Gregario, and D. Giusto, \Multimedia information broadcasting using digital TV channels," IEEE Transactions on Broadcasting, vol. 43, pp. 383{392,
December 1997.
[646] W. Sohn, O. Kwon, and J. Chae, \Digital DBS system design and implementation for TV and
data broadcasting using Koreasat," IEEE Transactions on Broadcasting, vol. 44, pp. 316{323,
September 1998.
BIBLIOGRAPHY
1063
[647] J. Griths, Radio Wave Propagation and Antennas | An Introduction. Englewood Clis,
NJ: Prentice-Hall, 1987.
[648] M. Karaliopoulos and F.-N. Pavlidou, \Modelling the land mobile satellite channel: a review,"
Electronics and Communication Engineering Journal, vol. 11, pp. 235{248, October 1999.
[649] J. Goldhirsh and W. Vogel, \Mobile satellite system fade statistics for shadowing and multipath
from roadside trees at UHF and L-band," IEEE Transactions on Antennas and Propagation,
vol. 37, pp. 489{498, April 1989.
[650] W. Vogel and J. Goldhirsh, \Multipath fading at L band for low elevation angle, land mobile
satellite scenarios," IEEE Journal on Selected Areas in Communications, vol. 13, pp. 197{204,
February 1995.
[651] W. Vogel and G. Torrence, \Propagation measurements for satellite radio reception inside
buildings," IEEE Transactions on Antennas and Propagation, vol. 41, pp. 954{961, July 1993.
[652] W. Vogel and U. Hong, \Measurement and modelling of land mobile satellite propagation at
UHF and L-band," IEEE Transactions on Antennas and Propagation, vol. 36, pp. 707{719,
May 1988.
[653] S. Saunders, C. Tzaras, and B. Evans, \Physical statistical propagation model for mobile
satellite channel," tech. rep., European Commission, 1998.
[654] S. Saunders, Antennas and Propagation for Wireless Communication Systems Concept and
Design. New York: John Wiley and Sons, 1999.
[655] K. Wesolowsky, \Analysis and properties of the modied constant modulus algorithm for blind
equalization," European Transactions on Telecommunication, vol. 3, pp. 225{230, May{June
1992.
[656] M. Goursat and A. Benveniste, \Blind equalizers," IEEE Transactions on Communications,
vol. COM{28, pp. 871{883, August 1984.
[657] G. Picchi and G. Prati, \Blind equalization and carrier recovery using a \stop{and{go"
decision{directed algorithm," IEEE Transactions on Communications, vol. COM{35, pp. 877{
[658] A. Polydoros, R. Raheli, and C. Tzou, \Per{survivor processing: a general approach to MLSE
in uncertain environments," IEEE Transactions on Communications, vol. COM{43, pp. 354{
364, February{April 1995.
[659] D. Godard, \Self{recovering equalization and carrier tracking in two{dimensional data communication systems," IEEE Transactions on Communications, vol. COM{28, pp. 1867{1875,
November 1980.
[660] Y. Sato, \A method of self{recovering equalization for multilevel amplitude{modulation systems," IEEE Transactions on Communications, vol. COM{23, pp. 679{682, June 1975.
[661] Z. Ding, R. Kennedy, B. Anderson, and R. Johnson, \Ill-convergence of Godard blind equalizers
in data communications systems," IEEE Transactions on Communications, vol. COM-39,
pp. 1313{1327, September 1991.
[662] Y.-Q. Zhang, F. Pereira, T. Sikora, and C. Reader (Guest Editors), \Special issue on MPEG4," IEEE Transactions on Circuits and Systems for Video Technology, vol. 7, February 1997.
[663] L. Chiariglione, \MPEG and multimedia communication," IEEE Transaction On Circuits And
Systems For Video Technology, vol. 7, pp. 5{18, February 1997.
[664] T. Sikora, \The MPEG-4 video standard verication model," IEEE Transaction On Circuits
And Systems For Video Technology, vol. 7, pp. 19{31, February 1997.
[665] IEEE, Proceeding of VTC'99 (Spring), (Houston, TX), 16{20 May 1999.
[666] IEEE, Proceedings of IEEE Vehicular Technology Conference (VTC'98), (Ottawa, Canada),
18{21 May 1998.
[667] Proceedings of the IEEE Global Telecommunications Conference 1993, (Houston, TX), 29
November { 2 December 1993.
1064
BIBLIOGRAPHY
[668] ACTS, Proceeding of ACTS Mobile Communication Summit '97, (Aalborg, Denmark), 7{10
October 1997.
[669] IEEE, Proceedings of IEEE Vehicular Technology Conference (VTC'95), (Chicago), 15{28
July 1995.
[670] ACTS, Proceeding of ACTS Mobile Communication Summit '98, (Rhodes, Greece), 8{11 June
1998.
[671] IEEE, Proceeding of Globecom'98, (Sydney, Australia), 8{12 November 1998.
[672] IEEE, Proceeding of IEEE Global Telecommunications Conference, Globecom 96, (London),
18{22 November 1996.
[673] ACTS, Proceeding of ACTS Mobile Communication Summit '99, (Sorrento, Italy), 8{11 June
1999.
[674] IEEE, Proceedings of IEEE International Symposium on Personal, Indoor and Mobile Radio
Communications, PIMRC'97, (Marina Congress Centre, Helsinki, Finland), 1{4 September
1997.
[675] IEEE, Proceedings of IEEE ICCS'96/ISPACS'96, (Singapore), 25{29 November 1996.
[676] IEEE, Proceedings of IEEE VTC '94, (Stockholm, Sweden), 8{10 June 1994.
[677] R. Damper, W. Hall, and J. Richards, eds., Proceedings of IEEE International Symposium of
Multimedia Technologies and Future Applications, (London), Pentech Press, April 1993.
[678] IEEE, Proceedings of IEEE VTC'96, (Atlanta, GA), 28 April{1 May 1996.
[679] IEEE, Proceedings of IEEE VTC'97, (Phoenix, AZ), 4{7 May 1997.
[680] IEEE, Proceedings of International Conference on Acoustics, Speech, and Signal Processing,
ICASSP'80, (Denver, CO), 9{11 April 1980.
[681] IEEE, Proceedings of IEEE International Symposium on Personal, Indoor, and Mobile Radio
Communications (PIMRC'96), (Taipei, Taiwan), 15{18 October 1996.
[682] IEEE, Proceedings of the IEEE International Conference on Acoustics, Speech and Signal
Processing (ICASSP'94), (Adelaide, Australia), 19{22 April 1994.
[683] IEEE, Proceedings of the IEEE International Conference on Acoustics, Speech and Signal
Processing (ICASSP'96), (Atlanta, GA), 7{10 May 1996.
[684] IEEE, Proceedings of International Conference on Acoustics, Speech, and Signal Processing,
ICASSP'92, March 1992.
[685] IEEE, Proceedings of IEEE International Symposium on Spread Spectrum Techniques and
Application (ISSSTA'98), (Sun City, South Africa), September 1998.
[686] Proceedings of International Workshop on Coding Techniques for Very Low Bit-rate Video
(VLBV'95), (Shinagawa, Tokyo, Japan), 8{10 November 1995.
[687] IEEE, Proceedings of IEEE VTC '93, (Secaucus, NJ), 18{20 May 1993.
About the Authors

Lajos Hanzo received his degree in electronics in 1976 and his doctorate in 1983.
During his 25-year career in telecommunications he has held various research and
academic posts in Hungary, Germany and the UK. Since 1986 he has been with the
Department of Electronics and Computer Science, University of Southampton, UK,
where he holds the chair in telecommunications. He has co-authored ve books on mobile radio communications, published over 300 research papers, organised and chaired
conference sessions, presented overview lectures and been awarded a number of distinctions. Currently he is managing an academic research team, working on a range
of research projects in the eld of wireless multimedia communications sponsored
by industry, the Engineering and Physical Sciences Research Council (EPSRC) UK,
the European IST Programme and the Mobile Virtual Centre of Excellence (VCE),
UK. He is an enthusiastic supporter of industrial and academic liaison and he oers
a range of industrial courses. He is also an IEEE Distinguished Lecturer. For further information on research in progress and associated publications please refer to
http://www-mobile.ecs.soton.ac.uk
Peter J. Cherriman graduated in 1994 with an M.Eng. degree in Information En-
gineering from the University of Southampton, UK. Since 1994 he has been with the
Department of Electronics and Computer Science at the University of Southampton,
UK, working towards a Ph.D. in mobile video networking which was completed in
1999. Currently he is working on projects for the Mobile Virtual Centre of Excellence, UK. His current areas of research include robust video coding, microcellular
radio systems, power control, dynamic channel allocation and multiple access protocols. He has published ve journal papers, 13 conference papers, and holds three
patents. Further information and a list of publications can be found on http://wwwmobile.ecs.soton.ac.uk/peter
Jurgen Streit received his Dipl.-Ing. Degree in electronic engineering from the
Aachen University of Technology in 1993 and his Ph.D. in image coding from the
Department of Electronics and Computer Science, University of Southampton, UK, in
1995. From 1992 to 1996 Dr. Streit had been with the Department of Electronics and
Computer Science working in the Communications Research Group. His work there
led to numerous publications. Since then he has joined a management consultancy
working as an information technology consultant.
1065

Video Book Ch1 3+5

Transféré par

Informations du document

Description originale:

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Video Book Ch1 3+5

Transféré par

Droits d'auteur :

Formats disponibles

Wireless Video Communications:

Second to Third Generation Systems and

c L. Hanzo, P.J. Cherriman, J. Streit

Department of Electronics and Computer Science,

1.1 Issues in Information Theory . . . . . . . . . . . . . . . . . . . . . . .

1.8.3 First-Order Markov Model . . . . . . . . . . . . . . . . . . . .

2 The Propagation Environment

2.1 The Cellular Concept . . . . . . . . . . . . . . . . . . . . . .

3 Convolutional Channel Coding

Brief Channel Coding History . . . . . . . . . . . .

4 Block-Based Channel Coding

5 Modulation and Transmission Techniques

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

5.1.2.6 Di erentially Detected QAM [84] . . . . . . . . . . . .

6 Video Trac Modeling and Multiple Access

6.1 Video Trac Modeling . . . . . . . . . . . . . . . . . . . .

7.7 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 284

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287

9 Second-Generation Mobile Systems

9.1 The Wireless Communications Scene . . . . . . . . .

10 CDMA Systems: Third-Generation and Beyond

10.3.2.3.2.1 Common Physical Channels of the FDD

10.3.2.10.2 Inter-Frequency Handover or Hard Handover 436

II Video Systems Based on Proprietary Video Codecs

11.1 Fractal Principles . . . . . . . . . .

12.1 Video Codec Outline . . . . . . . . . . . . . . . . . . . . . . .

12 Very Low Bit-Rate DCT Codecs

12.2.3.2 Grid Interpolation Techniques . . . . . . . .

13 VQ Codecs and Multimode Video Transceivers

14 Low Bit-Rate Parametric Quad-Tree-Based Codecs and

III High-Resolution Image Coding

15.1 Introduction and Video Formats . . . . . . . . . . . . . . . .

16 High-Resolution DCT Coding

IV Video Systems Based on Standard Video Codecs

17 An ARQ-Assisted H.261-Based Recon gurable Multilevel

17.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 675

17.2.2 Source Encoder . . . . . . . . . . . . . . . . . . . . . . . . . . .

18 Comparison of the H.261 and H.263 Codecs

19 A H.263 Videophone System for Use over Mobile Channels

20 Error Rate Based Power Control

20.9 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 859

21 Adaptive Single-Carrier, Multicarrier, and CDMA-based Video

21.1 Turbo-equalised H.263-based videophony for GSM/GPRS . . . . . . .

21.5.8 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . .

About the Authors

Contradictory system design requirements of various video communications systems.

channels. Hence, they do not employ channel interleavers, which is advantageous

Part II is dedicated to a host of xed but arbitrarily programmable rate video

Video Trac Modelling and Multiple Access;

Chapter 21 is partially based on our research sponsored by the Virtual Center of

Issues in Information Theory

The ultimate aim of telecommunications is to communicate information between two

CHAPTER 1. INFORMATION THEORY

1.1. ISSUES IN INFORMATION THEORY

Basic transmission model of information theory.

CHAPTER 1. INFORMATION THEORY

What is the true information generation rate of our information sources? If

Is the information emission rate the only important characteristic of a source,

1.2. ADDITIVE WHITE GAUSSIAN NOISE CHANNEL

Additive White Gaussian Noise Channel

CHAPTER 1. INFORMATION THEORY

Ideal, distortion-free channel model having a linear phase and a at magnitude

1.2.2 Practical Gaussian Channels

5.1.2.6 Dierentially Detected QAM [84] . . . . . . . . . . . .

17 An ARQ-Assisted H.261-Based Recongurable Multilevel

The ACF is symmetric: R( ) = R( ).

pk 1, then as regards the

log2 (x) = log10 (x) log2 (10) 3:322 log10 (x)

where is the so-called Lagrange multiplier. Following the standard procedure of

1.5.2 Human Coding

Human coding algorithm (see also Algorithm 2 and Table 1.2).

Symbol Probability BCD Human Code

Human Coding Example Summary of Table 1.2

Tree-based Human coding example.

H 0:8 0:469 + 0:2 0:469 0:469 bit/symbol: