Académique Documents
Professionnel Documents
Culture Documents
Hung-Yi Cheng()
E-mail: woody@access.ee.ntu.edu.tw
Graduate Institute of Communication Engineering
National Taiwan University, Taipei, Taiwan, ROC
1. Abstract
---------------- 2
2. Introduction
3.
----------------- 2
----------------- 3
-----------------4
-----------------4
-----------------5
-----------------7
------------------9
-----------------10
------------------13
------------------14
------------------15
------------------16
7. Conclusions
------------------18
8. Reference
------------------19
1. Abstract
Recently, an interesting matrix named Jacket was proposed by Prof. Moon Ho Lee
(1989, 2000, 2001, IEEE Trans. CAS), it is extended from Walsh Hadamard (1893),
but includes both orthogonal and non-orthogonal cases. In particular, we can present
the Jacket transform and a simple factor decomposition, which is used to develop
either a fast algorithm or communications for the Jacket transform. The matrix
decomposition is of the form of the matrix products Hadamard matrices and
successively lower order coefficient matrices. This decomposition very clearly leads
to a block circular sparse matrix factorization of the Jacket matrix. The main property
of Jacket is that the inverse matrices of its elements can be obtained very easily and
have a special structure. It is useful to construct space time codes, fast algorithm and
so on.
2. Introduction
2-1 What is Jacket martrix
[
where
[ J ]mxn
J11
J
m1
1/ J11
1
[ J ]mxn 1
C
1/ J m1
J1n
J mn
1/ J1n
1/ J mn
1 1 1 1
1 1 1 1 4 0 0
1 2 2 1
1 1 1 1 1 0 6 2
[WH ]4
1 2 2 1 4 1 1 1 1 0 2 6
1 1 1 1
1 1 1 1 0 0 0
0
0
0
2 2 2 2
1 1 1 1 4
1 2 1 1 2 1 1 1 1 1 0
1
[WH ]4
8 2 1 1 2 16 1 1 1 1 0
2 2 2 2
1 1 1 1 0
0 0 0
3 1 0
1 3 0
0 0 4
[WH ]N /2 [ H ]2
(1)
Where is the Kronecker product and [ H ]2 is the lowest order Hadamard matrix as
following.
1 1
[ H ]2
1 1
[WH ]8 [WH ]4 [ H ]2
1
1
1
1
1
1
1
1
1 1 1
2 2 1 1 1
1 1 1
1 1 1 1 1 1
1
2
1
2
1 1 1
1 2 2
1
2
1 1
2 1
1 2
2 1
2 2 1
2 2
1 1 1 1 1
1 1 1 1 1
1
1
1
1
1
1
1
1
1
[ H ]N [WH ]N
(2)
and the [ RC ]N is sparse matrix. We can get the properties with [RC]
[ RC ]N
([ H ]N /2 [ H ]2 )([WH ]N /2 [ H ]2 )
([ H ]N /2 [WH ]N2 ) ([ H ]2 [ H ]2 )
[ RC ]N /2 2[ I ]2
(3)
1
[ H ]N , we find the higher order WH matrix by these fast algorithm.
N
1
[ H ]N [ RC ]N
N
[WH ]N1 N [ RC ]N1[ H ]N 1
[WH ]N
(4)
1 1 1 1
1 1 1 1 4 0 0
1 2 2 1
1 1 1 1 1 0 6 2
[WH ]4
1 2 2 1 4 1 1 1 1 0 2 6
1 1 1 1
1 1 1 1 0 0 0
1
[ H ]4 [ RC ]4
4
0
0
0
The simple recursive relationship in (1) (2) (3) and (4) can be used to formulate a high
order [WH] as the fast algorithm.
As an example for N=8 can be represented as
[WH ]8
1
[ H ]8 [ RC ]8
8
1
( [ H ]4 [ H ]2 )([ RC ]4 2[ I ]2 )
8
1
([ RC ]4 [ H ]4 ) 2[ H ]2
8
The higher order of WH can be derived by the lower order of WH and lower order of
Hadamard matrix.
After calculation , we get the [WH ]8 from the fast algorithm.
1 1 1 1 4
1 1 1 1 1 0
[WH ]8
8 1 1 1 1 0
1 1 1 1 0
1 1 1 1 1
1 1 1 1 1
1 1 2 2 2
1 1 2 2 2
1 1 2 2 2
1 1 2 2 2
1 1 1 1 1
1 1 1 1 1
2 1
2 1
2
0
0
1 1
2
0
1 1
4
1
1
1
1
1
1
1
1
[ H ]N [ RC ]N N [ RC ]N1[ H ]N 1
N
([ H ]N /2 [ RC ]N /2 2[ H ]2 ) *
1
([ H ]N /2 [ RC ]N /2 1 [ H ]2 )
2
[ I ]N
Property : Products WH and inverse WH is identity matrix.
1 1 1 1
1 2 2 1
[WH ]4
1 2 2 1
1 1
[WH ]2
1 1 1 1 ,
1 2
[WH ]2
a
[ RJ ]2 M1 T
b
The matrix
M1
b
c
consists of three elements (a, b and c), which are all non-zero and
n
take of 2 , n = 0, 1,2. We define other matrices to get the [RJ].
b a
c b
M2
M4
c b ,
b a
Finally, we obtain
[ RJ ]4
, defined by
[ RJ ]4
M1
M T
2
b
a
a b
M 2 b c c b
M 4 b c c b
a b b a
signum(a * c)2
[ RJ ]2 1
det([ RJ ]2 )
1,
c b
b a , where signum( x) 0,
1,
[ RJ ]4 1
4 1/ b 1/ c 1/ c 1/ b
1/ a 1/ b 1/ b 1/ a
x0
x0
x0
We can see the elements position in the forward matrix can be replaced by its
inverse matrix, and the signs are not changed between the matrix and its inverse.
[ RJ ]2k
for the reverse jacket matrix with (2k * 2k). There are two
Property 3 :
A matrix is called the reverse jacket matrix if
,
1
[ RJ ]2k
det([ RJ ]2k )
where k>2
1 1
[ RJ ]2K 1 [ RJ ]2k [ H ]2 [ RJ ]2k
,
1 1
where k>2
[ H ]2
[ H ]
2
[ RJ ]8
[ H ]2
[ H ]2
[ H ]2
[ H ]2
2[ H ]2
2[ H ]2
2[ H ]2
2[ H ]2
[ H ]2
[ H ]2
[ H ]2
[ H ]2
[ H ]2
[ H ]2
1
[ H ]2k , we have from
2k
[ RJ ]2k 1
1
[ H ]2k 1 [ RC ]2k 1
2k 1
Therefore, the fast RJ transform is simply the fast Hadamard transform followed by a
spars matrix
1
[ RC ]2k .
2k
a b 1 1 1 1 1 2 1
M 2 [ RJ ]2
b c 1 2 2 1 1 0 3
And its inverse matrix
1 2 1 1 1 1 3 0
[ RJ ]2 1
3 1 1 6 1 1 1 2
Also, we have
1 1 1 1
1 2 2 1
,
[ RJ ]4
1 2 2 1
1 1 1 1
2 2 2 2
1 2 1 1 2
1
[ RJ ]4
8 2 1 1 2
2 2 2 2
The algorithm for the fast reverse jacket transform is similar as the WH. The FRJT
can be derived by decomposing [RJ] into a product of sparse matrices given each
rows/columns with only two non-zero elements.
The fast RJ transform for N=8,
1
[ RJ ]8 [ H ]8 [ RC ]8
8
1
([ H ]4 [ H ]2 )([ RC ]4 2[ I ]2 )
8
1
([ RJ ]4 2[ H ]2 )
2
their applications .Now we discuss about anther from for WHT in order to derive the
complex weighted Hadamard transform.
For n 2r length real vector (a0 , a1 ,
real vector ( A0 , A1 ,
, An 1 ) given by
A0
a0
A
1 [ H ] a1
n 1
An 1
an 1
that is,
n 1
A j (1) j ,i ai ,
j=1 ~ n-1
i 0
j0i0
1 n 1
( 1) j ,i Ai ,
n i 0
j=1 ~ n-1
Aj (1) j ,i ( )( ir 1 ir 2 ) ( jr 1 jr 2 ) ai ,
j 0,1, 2,
, n 1
i 0
where is any real number is the weight. The inverse transform is given by
n 1
a j (1) j ,i ( 1 ) (ir 1 ir 2 ) ( jr 1 jr 2 ) Ai ,
j 0,1, 2,
, n 1
i 0
We can know that for 1 the CWHT is same as the WHT. The 4x4 and 8x8 CWHT
matrices corresponding to the weight are
1 1
1
[W ]4
1
1 1
1
1
1
and
1
1
1
[W ]8
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
[C ]4
1 j j 1
1 1 1 1
and
1
1
1
[C ]8
1
1
1
j j
1 j
j 1
j j 1
j j
1 1 1 1
1 1
1
1
1
1
1
1
1
1
1
1
The following diagram is shown the different matrix and function with DFT,DCT
Hadamard and Jacket transform.
[ RJ ]4
b
a
a b
b c c b
b c c b
a b b a
[ RJ ]4 1
4 1/ b 1/ c 1/ c 1/ b
1/ a 1/ b 1/ b 1/ a
2b 2c
b
a 1
a b
b c c b 1
b c c b 1
a b b a 1
Assume we received the signals without noise and the demultiplexer in CDMA
system is the inverse Jacket transform. The received signal multiplex the each low of
Jacket matrix. Because the Jacket transform is orthogonal matrix, we can get the
message perfectly as following step.
We received the signal (0, -2b+2c, -2b-2c, 0) from TX with Jacket transform.
Through the demultiplexer, the signal became
1/ b 1/ a 0
1
1/ a 1/ b
1
1/ b 1/ c 1/ c 1/ b 2b 2c
1
1 4 1/ b 1/ c 1/ c 1/ b 2b 2c
1
1/ a 1/ b 1/ b 1/ a 0
7. Conclusions
The reserve Jacket matrix is a generalized form of the Weighted Hadamard and
Hadamard. The matrix has a recursive structure and symmetric characteristics.
The elements positions in the forward matrix can be replaced by its inverse matrix,
and the signs are not changed between the matrix and its inverse. The Hadamard
matrix is a special case of the Jacket matrix. The fast transform algorithm is the
matrix decomposition of the Hadamard matrices and succesively lower order reverse
Jacket transform. The [ RC ]N sparse matrix of [ RJ ]N leads to very clear
decomposition. Using the orthogonal Jacket transform, we can apply the matrix to
wireless communication such as CDMA.
8. Reference
[1]M.H. Lee, The Center Weighted Hadamard Transform, IEEE Trans.1989 AS-36,
(9),pp.1247-1249.
[2]S.-R.Lee and M.H.Lee, On the Reverse Jacket Matrix for Weighted Hadamard
Transform, IEEE Trans. on Circuit Syst.II, vol.45.no.1, pp.436-441,Mar.1998.
M.H. Lee, A New Reverse Jacket Transform and its Fast Algorithm, IEEE Trans.
Circuits Syst.-II , vol 47, pp.39-46, 2000.
[3] M.H. Lee and B.S. Rajan, A Generalized Reverse Jacket Transform, IEEE Trans.
Circuits Syst. II, Analog Digit. Signal Process., vol. 48 no.7 pp 684-691, 2001.
J. Hou, M.H. Lee and J.Y. Park, New Polynomial Construction of Jacket
Transform, IEICE Trans. Fundamentals, vol. E86-A no. 3, pp.652-659, 2003.
[4]W.P. Ma and M. H. Lee, Fast Reverse Jacket Transform Algorithms, Electronics
Letters, vol. 39 no. 18 , 2003.
[5]Moon Ho Lee, Ju Yong Park, and Jia Hou,Fast Jacket Transform Algorithm Based
on Simple Matrices Factorization, IEICE Trans. Fundamental, vol.E88-A, no.8,
Aug.2005.
[6]Moon Ho Lee and Jia Hou, Fast Block Inverse Jacket Transform, IEEE Signal
Processing Letters, vol.13. No.8, Aug.2006.
[7] Moon Ho Lee, and Ken.Finlayson, A Simple Element Inverse Jacket Transform
Coding, Information Theory Workshop 2005, ITW 2005, Proc. of IEEE ITW 2005,
28.Aug-1.Sept., New Zealand, also IEEE Signal Processing Letters, vol. 14 no. 5,
May 2007
[8]M. H. Lee,A new reverse jacket transform and its fast algorithm IEEE Trans.
Circuits Syst.II, vol. 47, no.1, pp.39-46, Jan.2000.