Académique Documents
Professionnel Documents
Culture Documents
International Journal of
ISSN 2225-658X (Online)
DIGITAL INFORMATION AND WIRELESS COMMUNICATIONS
Volume 4 Issue 4
2014
TABLE OF CONTENTS
Original Articles
PAPER TITLE
AUTHORS
Imran Jattala, Junaid Farooqi,
WIRELESS SENSOR NETWORK (WSN) BASED AUTOMATIC FIRING
Shakeel Durrani, Nassar Ikram,
PRACTICE SYSTEM (AFPS) FOR TRAINING OF LAW ENFORCEMENT
Irfan Hussian, Muazzam A.
AGENCIES (LEAS)
Goraya, Obaid B. Zakria
INFORMATION TRANSPARENCY IN ADMISSION SELECTION OF
STATE ISLAMIC HIGHER EDUCATION IN INDONESIA USING Linda Salma Angreani
INFORMATION TECHNOLOGY APPROACH
HYBRID BOUNDARY DETECTION METHOD FOR IMAGE WITH Syaiful Anam,
APPLICATION TO CORONARY PLAQUE
Noriaki Suetake
Eiji
Atef
408
418
Uchino,
PAGES
428
438
Harb,
447
458
474
486
493
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 408-417
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Wireless Sensor Network (WSN) based Automatic Firing Practice System (AFPS) for
Training of Law Enforcement Agencies (LEAs)
Imran Jattala, Junaid Farooqi, Shakeel Durrani, Nassar Ikram
Horizon Technologies
Islamabad, Pakistan
imran.jattala@gmail.com, shakeel.durranii@gmail.com, junid.farooqi@gmail.com,
dr_nassar_ikram@yahoo.com
Irfan Hussian, Muazzam A. Goraya, Obaid B. Zakria
National University of Science & Technology (NUST)
Islamabad, Pakistan
irfanmechatronics@gmail.com, dmagoraya@hotmail.com, obaid2002@yahoo.com
ABSTRACT
The critical importance of an efficient infantryman in
special operations force, tactical paramilitary and Law
Enforcement Agencies (LEAs) is insurmountable. One
of the many vital aspects of an effective solider is
excellent marksmanship which requires extensive
training at sophisticated firing ranges. Modern firing
ranges are supported by Automatic Firing Practice
Systems (AFPS) and this paper presents the design and
development of such a system based on WSN. AFPS
provide an automatic bullet-impact count during firing
training session and is modular scalable in design for
multiple of eight concurrent shooters. The system is
versatile and flexible allowing for different small-arms
and firing training modes and supports night firing
exercise. AFPS comprises of two major components,
the automatic target-box and a commander console.
Automatic target-box has a motor & gear assembly,
target sheet, bullet-impact sensor, control board, and
WiFi communication module. Commander console is a
ruggedized sunlight readable 10.4 Tablet PC, which
with a built-in WiFi acts an access point. The
automatic target-boxes equipped with embedded WiFi
modules form sensor nodes of a WSN. The paper
presents the complete System Development Life Cycle
(SDLC) of the firing practice system and associated
WSN. The AFPS and bullet-impact sensor was
extensively tested on Firing Ranges for accuracy of
bullet-impact count. The results showed a bulletimpact count accuracy of over 97 percent.
KEYWORDS
WSN, WiFi, Bullet-impact Sensor,
Console, Firing Practice System
INTRODUCTION
The infantryman still plays a critical role in the
outcome of battles in modern warfare especially
with the advent of weapons of mass destruction
(WMD) where wars between equally adversaries
are high improbably. Proficiency in effective
shooting and excellent marksmanship is one of the
many vital characters of a well-disciplined and
efficient solider. Instilling this character into new
recruits require the need of dynamic shooting
ranges for training of small-arms Law
Enforcement Agencies (LEAs), paramilitary and
military [1]. Firing practice systems used in
developing countries are predominantly manual or
semi-automatic systems [2]. These semi-automatic
systems have inaccurate bullet-impact count and
had to be manually and tediously counted by the
on-duty staff. Automated shooting practice
systems developed by leading defense companies
are cost intensive and are beyond financial
budgets estimate of most of LEAs in the
developing world [3]. The proposed automatic
firing practice system has been developed to
achieve the following objectives:
Commander
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 408-417
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Commander Console acts as an access point (AP)
while 24 Field Boxes (FBs) form WiFi sensor
nodes for AFPS-WSN.
This paper presents the complete System
Development Life Cycle (SDLC) of the
Automated Firing Practice System (AFPS) from
concept to development and manufacturing. The
paper is divided into four parts, description of
AFPS, development of Target box assembly,
commander console (CCon) & WSN, and finally
the on-site test results of AFPS.
10
11 12
DESCRIPTION OF AFPS
Automatic Firing Practice System (AFPS)
provides recruits of LEAs with better training
environment and achieves a higher degree of skill
more rapidly. Since AFPS is reconfigurable to
different shooting training module that can be
incorporated via software from the CCon, AFPS
has a robust structure that allows it to adapt to
different terrains and environments. The
architecture of AFPS is shown in Fig. 1.
13
14 15
16
17
18
19
20 21
22
23 24
WiFi Antenna
Commander Console
Figure 1: Architecture of AFPS
The main features of AFPS are accurate bulletimpact count, rejection of false-impact (a ricochet
or stone), wireless connectivity and a ruggedized
CCon. The main modules of AFPS are defined as:
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 408-417
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
410
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 408-417
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
The finalized environmental tested and the final
target box assembly is shown in Fig. 3.
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 408-417
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
412
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 408-417
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
413
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 408-417
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
The main features of CCon software are:
The software/GUI is divided into eight subareas. Each sub-area represents each target unit
A bullet-impact is immediately reported to the
CCon and shooter score is update on the GUI
in real-time
GUI shows individual score of each target plate
AFPS CCon supports night firing mode.
The CCon is 10.4 ruggedized tablet MARS3100R by Advantech and provides portability to
the commander. CCon has a supporting stand to
place it on table with a USB keyboard and mouse
for conventional Desktop like usage. The
commander console is shown in Fig. 10.
1.9
WiFi Communication Protocol
AFPS WiFi Communication Protocol is defined as:
Configuration: CCon assign IDs to address of
targets a, b, c w, for 24 Targets
Get Target Status: Commander reads the status
of Target
Set Target Status: Commander sets the current
status of Target to Up or Down with LED
Light On or Off
Get Bullet-Impact: Target immediately reports
to Commander the Bullet Impact
Bullet-Impact
Acknowledgement:
Commanders Acknowledgement of the
reception of Bullet-Impact
Hit Down Mode: Command to put Target in
Hit Down Firing Practice Mode. Target will go
down on every hit and will remain in such
position until command for popup is issued
Pop Up Mode: Command to put Target in Popup Firing Practice Mode. The target will go
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 408-417
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
A software de-bouncer was added in the bulletimpact detection logic. Software de-bouncer
brought a significant improvement further to the
system. Fourth firing test showed even greater
improvement with an average accuracy of 97.6%.
Results of fourth field firing of AFPS are shown in
Fig. 15.
415
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 408-417
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
average Software Count Accuracy to a astonishing
figure of 97.6%. The overall achievement from a
82.5% to 97.6% mark shows the success of the
development procedure and cycle. The graph of
average Software Count Accuracy percentage is
shown in Fig. 16.
CONCLUSIONS
A firing practice system has been designed and
developed to automate the existing manual or
semi-automatic firing ranges. WSN based AFPS
provided real-time bullet-impact count wirelessly
adding a greater level of flexibility for the
Commander for conduct of dynamic firing exercise
scenarios. The complete SDLC of the system has
been presented including mechanical and electrical
design of different parts of AFPS with revision and
improvement through different development
stages. The improvement process continued during
the field trials phase as well, till a 97.6% Bulletimpact count accuracy was achieved. The system is
controlled through a wireless (WiFi) commander
console. The power to the target units is powered
by a DC power supply. The next version of the
AFPS will be battery-powered, wirelessly
controlled, completely portable system for versatile
shooting practice scenarios.
REFERENCES
[1] S. Li, "The Design and Implementation of Shooting
System
Simulation
Platform
for
Police
College," International
Conference
on Scalable
Computing and Communications; Eighth International
Conference on Embedded Computing, 2009.
416
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 408-417
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Computing, Information and Control (ICICIC), 2009,
pp.454-457, 7-9 Dec. 2009.
[14] J. Xi, G. Liao, and W. Yang, "Study of Stepping Motor
Subdivision Driver," International Conference on
Intelligent Computation Technology and Automation
(ICICTA), 2010, vol.3, pp.443-446, 11-12 May 2010.
[15] F. Yuan, L. Tsai, V. Prakash, D. P. Dandekar, and A.M.
Rajendran, "Dynamic shear strength of S2 glass fiber
reinforced
polymer
composites
under
shock
compression," Journal of Applied Physics , vol.103,
no.10, pp.103537,103537-8, May 2008.
[16] C. Huang, S. Wang, C. Chang, C. Huang, Y. Cheng and
L. Fu, "An air combat simulator in the virtual reality
with the visual tracking system and force-feedback
components," IEEE International Conference on
Control Applications, 2004. Proceedings of the 2004,
vol.1, pp.515-520, 2-4 Sep. 2004.
[17] R. Mirshak, "Collateral damage risk: Quantifying the
water
surface
bullet
ricochet
problem," International Waterside Security Conference
(WSS), 2010, pp.1-7, 3-5 Nov. 2010.
[18] T. Matsuno, Y. Shima, T. Motoyoshi, K. Koyanagi, T.
Oshima, and T. Fukuda, "Basic characteristics analysis
of tuning fork structured piezoelectric vibration sensor
and application for robot control," International
Conference on Mechatronics and Automation (ICMA),
2011, pp.1532-1537, 7-10 Aug. 2011.
[19] X. Xiumei, and P. Jinfeng, "The simulation of
temperature and humidity control system based on
PROTEUS," International Conference on Mechatronic
Science, Electric Engineering and Computer (MEC),
2011, pp.1896-1898, 19-22 Aug. 2011.
[20] M. S. Lee, and H. C. Lee, "Dummy Gate-Assisted nMOSFET Layout for a Radiation-Tolerant Integrated
Circuit," IEEE Transactions on Nuclear Science,
vol.60, no.4, pp.3084-3091, Aug. 2013.
[21] R.D. Snow, J.R. Garner, and J.M. Whitaker.
"Developing Load Cell Monitoring on the Mock UF6
Feed and Withdrawal System." ORNL/TM-2013, no.34,
2013.
[22] www.ruggedpcreview.com/3_slates_advantech_3100r.h
tml
[23] M. Radek, and J. Zidek. "Application of synthetic
instrumentation that applies the trend of software-based
approach for measuring on the field of modern wireless
transfer systems." International Journal of Digital
Information
and
Wireless
Communications
(IJDIWC), vol.2, no.3, pp.208-221, 2012.
[24] E. Wang, S. Zhang, and Z. Zhang, "A Study on GPS
Signal Performance Analysis Platform Based on
LabVIEW," 8th International Conference on Wireless
Communications, Networking and Mobile Computing
(WiCOM), 2012, pp.1-4, 21-23 Sep. 2012.
417
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 418-427
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
ABSTRACT
The right to information is one of fundamental
human right essential. The transparency, availability
and accessibility of information enable people to hold
policy-makers to account and participate in the
decision making process, promoting better
governance, and reducing inefficiency and
corruption. In Indonesia, there is an integrated system
of student admission to State Islamic Higher
Education, named SPMB-PTAIN. This kind of
integrated admission selection model to enter State
Islamic Higher Education in Indonesia through
written examination conducted simultaneously
nationwide involving 53 Islamic Higher Education
Institution and coordinated by the Ministry of
Religious Affairs of Republic of Indonesia. In 2013,
this model are used by for about 200 thousands
applicants. However, there is no such model that
provides information about exams result tracking for
users. These condition lead to some critical issues
related to the transparency in SPMB-PTAIN exams
results. The subject of this paper is proposing the
monitoring system as part of system improvements of
SPMB-PTAIN (named SMUPDA) in terms of
monitoring and transparency. SMUPDA has been
tested in SPMB-PTAIN 2013 and was made such a
clear form of continuous improvement in data
processing of SPMB-PTAIN. SMUPDA can also
provide transparency of data processing on the
implementation of SPMB-PTAIN. Therefore the
similar model may be applied to other systems
related to public services to provide transparency and
gain the public trust of public services institutions
itself.
KEYWORDS
1. BACKGROUND
The right to information is recognized as a
fundamental human right essential for
meaningful democracy and the advancement of
other rights [1]. The transparency, availability
and accessibility of information enable people to
hold policy-makers to account and participate in
the decision making process, promoting better
governance, and reducing inefficiency and
corruption [2]. The Freedom of Information Act
in Indonesia adopted in 2008 named Undangundang No.14 Tahun 2008 tentang Keterbukaan
Informasi Publik, and it came into effect on 1
May 2010 [3].
In Indonesia, there is an integrated system of
student admission to State Islamic Higher
Education, named SPMB-PTAIN. This kind of
integrated admission selection model to enter
State Islamic Higher Education in Indonesia
through written examination. SPMB-PTAIN
conducted simultaneously nationwide involving
53 Islamic Higher Education Institution and
coordinated by the Ministry of Religious Affairs
of Republic of Indonesia. In 2013, this model are
used by for about 200 thousands applicants.
Figure 1 shown the new student admission
process published by SPMB-PTAIN Committee.
418
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 418-427
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
However, there is no such model that provides
information about exams result tracking for
users. Therefore, users only have information
about exams final result whether their pass or
fail the exams - as published by the admission
committee, without knowing the detail results of
the exams, especially the cause that make them
fail the exams. These condition lead to some
critical issues related to the transparency in
SPMB-PTAIN student admission exams results.
The subject of this paper is proposing the
monitoring system as part of system
improvements of SPMB-PTAIN system in terms
of monitoring and transparency. This monitoring
system has been tested in SPMB-PTAIN 2013
and was made such a clear form of continuous
improvement in data processing of SPMBPTAIN.
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 418-427
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
In SPMB-PTAIN, there are at least 10 (ten)
steps, as shown in figure 2, that are managed by
data processing team of SPMB-PTAIN [6].
Issues
There is a delay in delivery of LJU and disrupting
the flow or process and queue of the next process of
data processing.
There are some errors in the recording report by
exam supervisors in field, such as:
a. The number of participants presence in the
exams location was not match with the
actual situation.
b. Exams supervisors are fill biographical
data of exams participants that did not
attend in the Exam Answer Sheet (LJU)
There is some discrepancy between the physical
LJU and number of LJU reported that make the
matching control of LJU is not running
There are some participants who make a mistake to
fulfill biographical data, including a crucial data
such as:
a.
Participant Test Number
b.
Exams Code
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 418-427
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
main factors that causes some issues in table 1 is
came from the large number of documents that
handled by exams implementer team in field
(such as number of LJU, statement of purpose,
participant attendance list, etc.), and large people
involved in the data processing. This make a data
matching control in data processing are needed.
Here are the recommendations and tools needed
to control data processing issues [7]:
1. Recommendation: Better planning
coordination are needed in delivery of LJU to
reduce the possibility shipping delays.
Tools: LJU delivery schedules and time
arrival estimation with delivery reporting
features are needed.
2. Recommendations: for evaluation No. 2 and
3 (table 1) is Develop and Implement Exams
Reporting Applications with benefits:
a. Find out more detailed information early
about exams participant attendance list
b. Attendance list report can be used as a
data control of physical LJU received
and processed
c. Simplify the process of verification and
validation of LJU.
d. Detection and reporting of unfairly
information.
e. Perform grouping of unfairly acts as a
justification for graduation meeting.
f. Find out more information early about
mismatches amount of LJU and perform
data tracking.
g. Avoid the occurrence of errors and
irregularities in LJU information, such
as no/incompatibility examinee number,
etc.
Tools needed: Develop and Implement
Reporting Exam applications, with features:
a. Online apps, and fulfill as soon as
possible after the exam ends
b. Have notification function when it was
not fulfill yet after the exams done and
sent it to the relevant parties.
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 418-427
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
9. Number of participants who pass the exams
5. EXAMS
MONITORING
AND
DATA
5.2. People
One of successful key of SMUPDA
implementation is the discipline of people who
run and manage the system to follow their role
and responsibilities. Based on SMUPDA
workflow (figure 3), the role and responsibilities
of every actor in the system are:
1. Actor: Exams Local Committee (Panlok)
Role and responsibilities:
422
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 418-427
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
2.
3.
4.
5.
6.
423
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 418-427
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
424
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 418-427
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
With track shanges and tracking function, each
sub section of data processing team can perform
physical check of LJU number with the data that
stored in the system. If there is a shortage of LJU
that was sent to the next team than fixing process
can be done immediately. This process can also
be followed by the other parties related to
SMUPDA system. Figure 6 show SMUPDA
looks in Lahta area [10].
4.
5.
6.
7.
6. SUMMARY
1. Data processing monitoring system
(SMUPDA) that implemented in SPMBPTAIN 2013 have been reviewed and
can be implement in next SPMB-PTAIN
process.
2. The implementation of SMUPDA as
Data Processing improvement in SPMBPTAIN 2013 is part of governance
activities.
425
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 418-427
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
3. SMUPDA are also part of transparency
in public service activities.
4. SMUPDA provide detail information of
examinees and any obstacles to be
reported in field in real time.
5. SMUPDA are also act as tracking model
in each phase of data processing process.
6. SMUPDA have functionality of early
detection of discrepancies.
7. SMUPDA
act
as
information
recapitulation report in real time on each
phase of data processing.
8. SMUPDA made the control and
monitoring function are based on
system, rather than on personal.
9. SMUPDA is a model in realization of
continues improvement of the previous
year activities.
7. DISCUSSION AND NEXT RESEARCH
Although the improvements model has been
implemented in the SPMB-PTAIN 2013, in fact
the system is still in the testing phase. The full
implementation of such a system is also needs to
be improved especially with the human
resources that involved in data processing
activities.
Therefore, as a continuation of this study, the
evaluation process are needed on the successful
implementation of the system in at least 2 things.
First, from SMUPDA application, whether it is
ready meet the data processing needs of SPMBPTAIN, in this case is the control of information
and data. Secondly, the evaluation of any factors
that related to successful implementation of the
system due to capability of people who using the
system.
Finally, SMUPDA is intended as a solution to
some issues that happened earlier as part of
improvement process. Moreover this system can
also be used as a model that provides
transparency of information. Similar systems can
be developed to provide transparency of
REFERENCES
426
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 418-427
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
] Manual Book for Panlok, Malang: SPM-PTAIN
Committee, 2013.
[10SPMB-PTAIN 2013 Data Processing, SMUPDA
] Manual Book for Lahta, Malang: SPMB-PTAIN
Committee, 2013.
427
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 428-437
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Hybrid Boundary Detection Method for Image with Application to Coronary Plaque
Syaiful Anama,b, Eiji Uchinoa,c and Noriaki Suetakea
a
ABSTRACT
This paper proposes a hybrid boundary detection
method for image based on a new modified level set
method and a fuzzy model. It is applied to a boundary
detection problem of coronary plaque. Level set
method has been applied widely in image processing. It
however does not work well for an intravascular
ultrasound (IVUS) image because an image gradient,
commonly used for calculating a speed function in the
level set method, cannot detect an image boundary
well. The level set method and the weighted image
separability proposed by the authors in the past were
applied for a coronary plaque boundary detection
problem. The level set method could not however
detect the plaque boundary in several regions. One
problem was that the candidates of the plaque
boundary detected by the weighted separability were
unclear in several regions. The other problem was that
the IVUS image often becomes shadowed and it
contains no texture information there, which is caused
by the presence of the guide wire. To overcome this
problem, we propose a new modified level set, and we
further propose a hybrid boundary detection method
based on the new modified level set and the Takagi
Sugeno (T-S) fuzzy model for detecting a coronary
plaque boundary. The boundary detection accuracy of
the proposed method was significantly better than
those of the previous methods we proposed in the past.
KEYWORDS
Coronary plaque, intravascular ultrasound image,
hybrid boundary detection method, new modified level
set method, fuzzy model.
1 INTRODUCTION
Acute coronary syndrome (ACS) happens when
the heart is not getting enough blood. If the
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 428-437
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Boundary detection is one of the tasks in
computer vision which has wide applications such
as feature extraction, object recognition and image
segmentation [5]. The boundary detection is to find
the lines separating homogeneous regions.
Active contour models have been applied for
detecting image boundaries [6-11]. The active
contour models have several advantages over the
classical image segmentation methods, e.g., edge
detection, thresholding, and region growth.
The first advantage of the active contour models
is that they can achieve sub-pixel accuracy of
object boundaries [9]. The second advantage is that
the active contour models can be easily formulated
under a principled energy minimization
framework, and allow incorporation of various
prior knowledge, such as shape and intensity
distribution for robust segmentation [10]. The third
advantage is that they can give smooth and closed
contours as a segmentation result, which is
necessary for many applications.
Active contour models can be classified into
two categories, i.e., the classical snakes and the
level set method. In this paper, we use the level set
method, introduced by Osher and Sethian [12],
because it is a highly robust and accurate method
for tracking interfaces moving under complex
motions, which meets our requirements.
The level set method has been applied
successfully in many cases in image segmentation.
It has several advantages over other segmentation
methods such as the snake method, region growth
and thresholding. The advantages of the level set
method over the snake method are that the curve
may break or merge naturally during an evolution,
and its topological changes are thus automatically
handled.
The authors have proposed a method for
detecting a plaque boundary by using the weighted
image separability method and the level set method
[13]. An image gradient is commonly used for
calculating the speed function in the level set
method, but it cannot work well in the IVUS
image. Therefore, in [13], the image gradient in the
speed function was substituted by the weighted
image separability. However it failed to detect the
plaque boundary in some regions of the image as
shown in Fig 1.
429
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 428-437
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
proposed by the authors in the past.
(a)
I
div (c( x, y, t )I )
t
c( x, y, t )I c( x, y, t )I ,
(1)
c( x, y, t ) g ( I ( x, y, t ) )
(2)
where
(4)
I s( n 1) I s( n )
g (I
(n)
s, p
) I s(,np) ,
(5)
I s(,np) I p( n ) I s( n ) ,
(6)
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 428-437
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
value at the regions where the intensity gradients
are high.
2.2 Image Separability
The candidates of the plaque boundaries in IVUS
image are given by using a statistical discriminant
measure of the image separability [16]. The
advantages of the image separability are:
1) Insensitive to noisy and blurred edges,
2) Able to differentiate the edges between
texture regions.
The weighted separability is modified by
considering the conditions peculiar of IVUS
image. The weighted image separability for pixel
h in Fig. 3 is calculated by the difference of the
two regions as follows:
I I
I
h max A B
I max
I max
w
h
(7)
n A ( I A I ) 2 nB ( I B I ) 2
S
(I
k 1
I)
(8)
431
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 428-437
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
the left and on the right hand side of the guide
wire shadow region.
In the first step of the present hybrid boundary
detection method, the guide wire shadow region is
detected. Its procedure is as follows:
1) Convert the B-mode image of Fig. 2(a) in the
cartesian coordinate system to the polar
coordinate system of Fig. 2(b).
2) Detect the position of the guide wire shadow.
From Fig. 2 it can be seen that the guide wire
shadow is located in the small gray-level area.
The mean gray-level of every column of the
polar coordinate system is calculated to
determine the guide wire shadow. The mean
gray-level of every column is given by:
/ t F ,
(11)
which is referred to as a level set evolution
equation. In this paper, we use the level set (x)
Y ( x)
I ( x, y )
y 1
(9)
(10)
1023
255
Intensity
C / t F N ,
511
0
Angle Index
432
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 428-437
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
1
511
Intensity
1023
2)
3)
0
Angle Index
4)
5)
Intensity
1023
511
0
Angle Index
Figure 8. Initial contour of the new modified level set.
( )div ( g / ) g ( ),
(12)
g 1/(1 (G * I ),
7)
8)
9)
defined by:
/ t div (d p ( ) )
6)
(13)
1
.
g
w
1
(14)
433
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 428-437
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
left and on the right of hand side the guide wire
shadow region as shown in Fig. 10.
The plaque boundary is inferred by the series of
the following fuzzy rules:
If xi is Au then f u ( xi ) au xi bu ,
(15)
y ( xi ) u ( xi ) fu ( xi ) u 1 ( xi ) fu 1 ( xi ).
(16)
E ( yi y i ( xi )) 2 ,
(17)
( m)
Method
Hybrid Boundary Detection
Method (1st Experiment)
Hybrid Boundary Detection
Method (2nd Experiment)
Hybrid Boundary Detection
Method (3rd Experiment)
Hybrid Boundary Detection
Method (4th Experiment)
Hybrid Boundary Detection
Method (5th Experiment)
Hybrid Boundary Detection
Method (Average)
T-S Fuzzy Model [2]
T-S Fuzzy Model Optimized by
Particle Swarm Optimization [3]
LB
AB
9.1
20.4
9.3
20.0
9.2
20.1
9.3
19.9
9.3
20.1
9.2
20.1
13.7
28.4
12.2
28.5
434
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 428-437
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
fuzzy model [2], and the T-S fuzzy model
optimized by particle swarm optimization (PSO)
[3].
If the initial contour of the level set is different,
the proposed method will produce a different
plaque boundary. Therefore the experiment for
each image was repeated 5 times with different
initial contours. The desired boundaries (correct
boundaries) were decided by expert by using the
difference between the image brightness.
(a)
Table 2. RMSEs of boundary detection results for image 2.
( m)
Method
Hybrid Boundary Detection
Method (1st Experiment)
Hybrid Boundary Detection
Method (2nd Experiment)
Hybrid Boundary Detection
Method (3rd Experiment)
Hybrid Boundary Detection
Method (4th Experiment)
Hybrid Boundary Detection
Method (5th Experiment)
Hybrid Boundary Detection
Method (Average)
T-S Fuzzy Model [2]
T-S Fuzzy Model Optimized by
Particle Swarm Optimization [3]
LB
AB
11.8
24.9
12.8
17.3
12.7
17.6
12.6
17.2
12.7
17.5
12.5
18.9
28.0
35.4
23.8
33.2
(b)
Figure 12. Comparisons of the plaque boundary detection
methods. (a) The IVUS image to be processed. (b) The
plaque boundary detection results. The solid white lines
show the desired (correct) boundaries and the solid black
lines show the boundaries detected by the proposed
method. The dashed black lines and the dashed white lines
show the boundaries detected by the T-S Fuzzy model [2]
and the T-S Fuzzy model optimized by particle swarm
optimization [3], respectively.
( m)
Method
Hybrid Boundary Detection
Method (1st Experiment)
Hybrid Boundary Detection
Method (2nd Experiment)
Hybrid Boundary Detection
Method (3rd Experiment)
Hybrid Boundary Detection
Method (4th Experiment)
Hybrid Boundary Detection
Method (5th Experiment)
Hybrid Boundary Detection
Method (Average)
T-S Fuzzy Model [2]
T-S Fuzzy Model Optimized by
Particle Swarm Optimization [3]
LB
AB
15.3
12.0
15.6
13.1
15.5
21.8
15.9
12.5
15.7
12.3
15.6
14.3
20.0
30.2
19.9
40.1
435
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 428-437
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
show the boundaries detected by the T-S fuzzy
model [2] and the T-S fuzzy model optimized by
particle swarm optimization (PSO) [3],
respectively.
It can be observed from Fig. 12(b) that the
boundaries detected by the proposed method are
closer to the desired (correct) boundaries than
those by the previous methods [2], [3].
5 CONCLUSIONS
We have proposed a hybrid boundary detection
method for detecting a coronary plaque in an
IVUS image. It incorporates a newly modified
level set method with the Takagi-Sugeno fuzzy
model. The present method was compared with
the previous methods [2], [3], and the boundary
detection accuracy of the present method was
significantly better.
Acknowledgments: This work was supported by
the Grant-in-Aid for Scientific Research (B) of the
Japan Society for Promotion of Science (JSPS)
under the Contract No. 23300086. Syaiful Anam
would also like to thank DGHE postgraduate
scholarship of Indonesia for supporting his stay at
Yamaguchi University, Japan.
6 REFERENCES
[1] Potkin, B. N., Bartorelli, A. L., Gessert, J. M., Neville,
R. F., Almagor, Y., Roberts, W. C., Leon, M. B.:
Coronary Artery Imaging with Intravascular HighFrequency Ultrasound. Circulation, vol.81, pp.1575
1585 (1990).
[2] Koga, T., Ichiyama, S., Uchino, E., Suetake, N., Hiro,
T., Matsuzaki, M.: Fully Automatic Boundary
Extraction of Coronary Plaque in IVUS Image by
Anisotropic Diffusion and T-S Type Fuzzy Inference.
Advances in Intelligent and Soft Computing, SpringerVerlag, AISC 75, pp.139147 (2010).
[3] Anam, S., Misawa, H., Uchino, E., Suetake, N.:
Parameter Tuning by PSO for Fuzzy Inference-Based
Coronary Plaque Extraction in IVUS Image. In: Proc.
of the Joint 6th International Conference on Soft
Computing and Intelligent Systems and 13th
International Symposium on Advanced Intelligent
Systems (SCIS & ISIS 2012), pp.1426-1429, Japan,
(2012).
436
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 428-437
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
[14] Anam, S., Uchino, E., Suetake, N.: Image Boundary
Detection Using the Modified Level Set Method and a
Diffusion Filter. In: Proc. of the 17th International
Conference on Knowledge-Based and Intelligent
Information & Engineering (KES 2013), pp.161-169,
Japan (2013).
[15] Perona, P., Malik, J.: Scale-Space and Edge Detection
Using Anisotropic Diffusion. IEEE Transactions on
Pattern Analysis and Machine Intelligence, vol.12, pp.
629-639 (1990).
[16] Fukui, K.: Edge Extraction Method Based on
Separability of Image Features. IEICE Transactions on
Information System, vol. E78-D, pp.1533-1538 (1995).
[17] Koga, T., Uchino, E., Suetake, N.: Automated
Boundary Extraction and Visualization System for
Coronary Plaque in IVUS Image by Using Fuzzy
Inference-based Method. In: Proc. of the International
Conference on Fuzzy Systems, pp.1966 1973,
Taiwan (2011).
437
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 438-446
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
1,2
ABSTRACT
A wireless sensor network consisting of heterogeneous
sensors is capable of dealing with more than one type
of information about any region at a time. Sensor
applications that cater to disaster relief require faster
data delivery and effective connectivity than the other
monitoring applications. Moreover, usually, sensors
collect a single type of data. Our approach focuses on
generating a tree based communication architecture
that is made up of three different types of sensors and
ensures that at least one sensor of each type is
connected to the communication backbone; thereby
forming a MULTI-TREE. By responding to query
initiated by the sink for one type of data at a time, the
heterogeneous sensors ensure reduced amount of
carried load and latency in data transmission. Our
strategy implemented for structural health monitoring
application shows that when compared to a minimum
spanning tree under the same simulation constraints,
performs better as it reduces the carried load and delay
to one third under the existing set of ad hoc protocols.
KEYWORDS
Wireless sensor network, Minimum Spanning Tree,
Multi tree, Carried load, Energy consumption, Delay.
1 INTRODUCTION
The main aim of a wireless sensor network is to be
able to deliver the required functionality with
unattended operation for the longest possible time
without sacrificing the major constraints.
Irrespective of the type of application scenario, the
main requirement for nodes is to be self-powered.
This implies they will either have to contain
enough stored energy to last for years, or they will
have to be able to scavenge energy from the
environment. Till date, a wireless sensor network
has been subject to a vast number of application
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 438-446
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
connections. This are true to a greater degree for
ad-hoc (or infrastructure-less) communication.
Once sensors have been deployed, they need to
form some kind of network for communication.
Since the sensors have limited energy, the network
used should be efficient enough to minimize the
energy usage.
In a clustered network, many sensors connect to a
single sensor, known as cluster head, forming a
cluster, which joins to the sink to complete the
communication backbone network. All sensors
send their data to their cluster head, from where
these data are transferred to the sink. The
communication between the member and cluster
heads and the corresponding inter cluster head
communication usually happens in multi tiers. So,
this architecture resembles a hierarchy similar to a
tree which we analyze in our approach. Network
energy consumption is reduced as the number of
nodes involved in long distance transmission gets
reduced. But in this type of architecture, when
cluster head goes out of order, the whole cluster
becomes useless rendering reconfiguration. This
situation is analogous to a tree, where, to use the
sensors of such cluster, tree restructuring can be
done for that part of the network. Hence, we
advocate the use of a tree-based network for the
aforementioned purpose. In a tree based network,
we free the node connections from any loop or
cycle, which ensures the non redundancy and
transmission of data in one direction only. Here,
each sensor node is connected to sink directly or
via other nodes depending on their distance from
the sink and all the sensors can send their data to
the sink via single or multi-hops. The issue that
arises is to minimize these transmission distances
by using different available tree routing algorithm
and protocol. Some of the available tree based
architecture in wireless sensor networks are:
clustered tree network [1], Minimum Steiner tree
[2], Minimum spanning tree, Minimum-hot-spot
query trees [3], Query routing tree [4]. MinimumHot-Spot Query Trees are query routing trees with
three properties; decreased collisions during data
transmission, decreased query response time, and
increased system lifetime and coverage. [4] uses a
query routing tree, which is formed in two phases
1) node discovery phase and 2) node balancing
phase. It tries to make a near_balanced tree on the
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 438-446
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
network by assigning privileges to the traffic.
ViTAMin [15] offers a hierarchical backbone tree
algorithm for energy efficiency and sufficient
network lifetime. While Localized area spanning
tree (LAST) protocols for wireless short range
sensor networks optimizes the energy cost and the
interference imposed by the structure [16], a BFS
based tree rooted at the base station offers shortest
path traversal for each data message which utilizes
the sensor resources efficiently by employing a
local repairing approach for the crashing nodes
thereby increasing the lifetime. [17]. CTP is a
routing protocol implemented in TinyOS-2. x,
offers 90-99.9% packet delivery in highly
dynamic environments while sending up to 73%
fewer control packets than existing approaches.
[18], [19], [20]. Tree based strategies reduce the
burden of retransmissions and hence can be used
for congestion management. [14]. Thus, it can be
believed that a tree structure is very popular as in
wireless sensor network structure, in most of
applications when we have one sink and too many
sender nodes.
To address this problem we discuss a query
routing multi-tree approach on a set of
heterogeneous sensors. By using different sensors
for different purpose we aim to reduce the burden
of the nodes in terms of packets generated,
forwarded and aggregated by one third. This also
affects the overall delay and the final throughput
of the network. We implement our algorithm for
structural
health
monitoring
applications
specifically as it does not require a large number
of sensor nodes or a dense deployment scheme
and needs three different sensors to aptly monitor
the structure. The Requirements of Structural
Health Monitoring are enlisted as follows: SHM is based on monitoring buildings and
bridges which requires different classes of
data to be collected. Our multi-tree
approach
tries
to
facilitate
this
requirement.
Data redundancy is acceptable.
Data security is not that important.
Doesnt require dense deployment.
Saving energy is not the main constraint as
redeployment is possible.
2.1 Assumptions
Once the sensors are connected, they start
communicating in order to receive and send data.
We assume that the sensor network has the
following properties:
Sensor nodes are heterogeneous (three
types of sensors) and location aware; and
Sensor
nodes
have
limited
and
irreplaceable battery power.
The tree is formed under a constraint that
each sensor will connect to only one sensor
440
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 438-446
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
2.2 Procedure
Step 1: Choose any element r and a set S =
{r} where S and E = (Taking r as the root
node of the spanning tree.)
Step 2: Find the lightest edge (here the weight is
chosen on the basis of the shortest distance and the
associated sensor class type) such that one end
point is in S and the other point is in V\S. Add
the edge to E and the other point to S.
Step 3: If V\S = then stop and output the
spanning tree (S, E) otherwise go to step 1.
Notations
N
Node_List[]
Array_T
Array_A
Array_S
Dist[]
S
i
n
k
EDi,j
Description
total number of nodes
deployed of each
type
Array to store all
deployed nodes
Array
to
store
Temperature/Pressure
sensor nodes
Array
to
store
Accelerometer sensor
nodes
Array
to
store
Stress/Strain sensor
nodes
Array
to
store
Euclidean distance
between i and j
Euclidean distance
between i and j
2.3 Algorithm
/*Tree Generation*/
Figure 2. Multi-Tree
A: Accelerometer
S: Stress/Strain,
T: Temperature/Pressure
In our architecture, sink links to one sensor of
each type, as shown in the figure 2, restricting its
children to be a maximum of three. This results in
a balanced tree construction, in which none of the
nodes are overburdened. When the sink requests a
data type, only those sensor nodes that belong to
its class respond to the query, this reduces the
1. Begin
2. Node_List[0] = Sink;
3. K=1;
4. For(i=0 to N-1)
5. {
6. For(j=i to N-1)
7. {
8. Dist[j] = Calculate EDi,j;
9. }
10. Sort Dist[i,n];
11. Connect Node_List[i] and Array_T[i];
12. Node_List[k]=Array_T[i];
13. K++;
14. Repeat STEP 5 to STEP 13 for Array_A[];
15. Repeat STEP 5 to STEP 13 for Array_S[];
441
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 438-446
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
16. }
17. End
For query processing, whenever sink needs to get
any type of data, it sends a query to all its
children, which is further forwarded to rest of
sensor nodes by next level nodes in the tree till it
reaches to all the sensor nodes. Sensors after
receiving the query responds by transmitting its
collected data to the upper level nodes in the
network, which is further forward till it reaches
sink. All the sensors respond to the query of their
type only and discard other queries, which results
in maximum of one-third sensors of the entire
network to send its data. This reduces the load on
the entire network by significant amount, thus
increasing the lifetime of the network. As the sink
node starts the tree construction, for which it
calculates the distances of all the neighboring
nodes the complexity is O(n). However, for
sorting the distances the complexity is O(n 2). So,
the overall complexity of Multi-Tree is O(n3).
The performance has been evaluated for Zigbee
application between the nodes in Qualnet
simulator as shown in figures 3 and 4. Figure 3
depicts the snapshot of backbone using a multi
tree with three different node types (represented
by different colors) while figure 4 depicts the
MST backbone with homogeneous nodes. The
results presented are obtained by Qualnet and
stand alone C packages as per the following
parameters:
9.
Energy model
10.
11.
12.
13.
Node Type
Items to Send
Item Size
Application Type
Linear gradient
model
MICAZ motes
100
127
Zigbee
S.No
1.
2.
3.
4.
5.
6.
7.
8.
Parameters
Radio type
Transmission power
Node classes
Values
802.15.4
3.0 dbm
Stress/strain,
pressure
and
temperature,
Accelerometer
type
Number of nodes
13 nodes of each
type
Packet
reception PHY 802.15.4
model
Reception model
Modulation scheme O-QPSK
CCA Mode
Carrier sense
Noise factor
10.0
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 438-446
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
for the same set of simulation parameters and for
four routing protocols namely; AODV, Bellman
Ford, Dymo and Fisheye. The following figures
show the throughput of the proposed algorithm,
the offered load at the transport layer ,the average
carried load, the number of packet drops, the
average energy consumed in transmitting and
receiving, the delay and the jitter. The results are
analyzed for the Network and MAC Layer as the
Mac Layer protocols that traditionally manage
power saving are designed to be application aware
to some degree, for example they provide service
differentiation for data, query and management
packets. The factors considered for validating the
performance of our approach over the minimum
spanning tree is as follows:
a) Throughput at the application layer:
Throughput can be calculated as: Total bytes sent
or received*8 / (Simulation time - time first packet
is sent or received).
b) The Carried load: Indicates the total workload
that an individual node needs to process in terms
of data communicated.
c) The Average energy consumed by sensor
nodes in transmitting and receiving as any
wireless sensor network has to take energy
constraints into consideration. Energy consumed is
calculated using radio energy model [7].
The energy consumed to send a k-bit message over
distance d is:
fs
d0
Es (k , d ) k.Eelec k. fs .d 2 d d 0
(1)
Es (k , d ) k.Eelec k. mp .d 4 d d 0
(2)
443
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 438-446
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Figure 8 and 9 shows that multi-tree consumes
less energy as compared to the corresponding
MST in transmission as well as in receiving for
AODV, DYMO and Fisheye while for Bellman
Ford protocol both tree consumes almost same
energy.
444
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 438-446
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
4 CONCLUSIONS
Figure 10. Average Delay (in seconds) per node for the
protocols
Figure 11. Average Jitter (in seconds) per node for the
protocols
2.
3.
4.
445
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 438-446
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
446
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 447-457
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
ABSTRACT
Most exchange rates are volatile and mainly rely on
the principle of supply and demand. Millions of people
around the world are influenced, one way or another,
by the variation in exchange rates. In this research we
demonstrate that the Artificial Intelligence, specifically
Artificial Neural Networks (ANN), can improve the
accuracy of forecasting exchange rates compared to
statistical techniques such as regression. When we
compared the results from regression and artificial
neural network, it was clear that the ANN
outperformed regression in forecasting exchange rates.
Moreover, it became clear that using ANNs instead of
regression for forecasting exchange rates is rewarding
and necessary because the average error given by an
ANN is smaller than the average error given by
regression. Accuracy in forecasting became a major
issue and not a minor detail. It was the combination
between Artificial Intelligence and Macro Economics
that made these two models come into reality, making
it possible to use computer sciences and engineering
fields in the service of an economical problem.
The results of our research showed that using ANN
with the right parameters and variables rather than
using a regression model will yield a result with a
lower error margin.
KEYWORDS
Artificial Neural Network, Regression, Exchange Rate,
Forecasting, Inflation, Interest Rate, Multilayer
Perceptron;
I. INTRODUCTION
Most exchange rates are volatile and mainly rely
on the principle of supply and demand. Millions of
people around the world are influenced, one way
or another, by the variation in exchange rates. All
447
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 447-457
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
II.
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 447-457
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
amounts of data are to be treated and analyzed in
order to reach a final number that represents a
forecasted exchange rate.
449
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 447-457
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Because all known forecasting exchange rate
models have approximately the same efficiency,
we can find a lot of them. This shows the amount
of difficulty and complexity of generating an
efficient and reliable exchange rate forecasting
model.
In his book entitled Exchange Rate
Determination: Models and Strategies for
Exchange Rate Forecasting, Michael Rosenberg
explained that globalization and cross-border
interaction contributed largely to redefining the
worldwide business arena. He mentioned as well
that success for the companies dealing with
business in the arena is the result of forecasting
and determining accurately and thoroughly the
exchange rates.
Authors in [14] worked on estimating and
selecting feedforward and recurrent networks in a
careful way, and they also wanted to evaluate the
forecasting performance of selected networks in
different periods, thats why they have proposed a
two-step procedure. Different forecasting results
were gathered, and they were not similar. Among
five series which were evaluated, in only two out
of them, networks with significant market timing
ability (sign predictions) and/or significantly lower
out-of-sample MSPE (relative to the random walk
model) were found. The forecasting performance
in the other remaining series is not as satisfying as
in these two. According to their results, PSC is
seen to be sensible in selecting networks and the
two-step procedure that had been used could be as
a standard network construction procedure in other
applications. The results reveal that nonlinearity in
exchange rates may be exploited in order to
improve both point and sign forecasts. Although
some of the results reported are quite motivating,
they provide only limited evidence supporting the
usefulness of neural network models.
Walczak [15] speaks about neural networks
saying that they have been shown to be a
promising tool for forecasting financial time
series. Several design factors influenced
significantly the accuracy of neural network
forecasts. These factors include selection of input
variables, architecture of the network, and quantity
of training data. The issues about input variable
selection and system architecture design have been
widely researched, but regarding the information
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 447-457
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
previously mentioned (e.g. interest rate, oil price,
etc).
B. Independent Variables
The first, and most important independent
variable, is the interest rate. This variable is by far
the most influential factor that affects the
exchange rate. We are using to interest rates,
representing two independent variables: One Year
USD LIBOR, and One Year Euribor.
The second one is the inflation. It is well known
that inflation rates affect a countrys currency
value. We are using to inflation rates, representing
two independent variables: US Core Inflation Rate
and Euro Area Core Inflation Rate. When a
countrys inflation rate rises relatively to that of
another country, decreased exports and increased
imports depress the high-inflation countrys
currency. High inflation rates increases the foreign
exchange rates and hence weakens the local
currency. Notice here that we used core inflation
rates which excludes the prices of energy, to avoid
double counting because oil price is counted is an
independent variable.
Oil price is the third independent variable. In
general, oil prices have noticeable influence on
USD, and hence it has influence on the Euro/Dollar
exchange rate. Figure 3 shows the main factors
affecting the exchange rates used in this model.
451
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 447-457
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Coefficients
(1)
Model
1 (Constant)
USInflation
Unstandardized
Standard.
Std. Error
Beta
1.111
.025
.060
.010
0.020 * USDLibor1Year
.563
Sig.
44.071
.000
6.291
.000
EUInfaltion
-.166
.017
-.890
-9.579
.000
OilPrice
.004
.000
.882
18.227
.000
Euribor1Year
.049
.009
.401
5.514
.000
.007
-.221
-3.047
.003
USDLibor1Year -.020
Sum of
Squares
Df
Mean
Square F
Sig.
Regression
2.220
.444
.000b
Residual
.690
126
.005
Total
2.910
131
81.104
Euribor1Year, USInflation
Statistics
USInfl
EUInfl
Oil
Euribor
USDLibor
Exchg.
ation
ation
Price
1Year
1Year
Rate
Valid
132
132
132
132
132
132
Missing
Mean
2.4000 2.1265
66.11
2.65656 2.50564
1.28
Std. Dev.
1.3993 .80113
29.37
1.2194
1.618903
.1490
Skewness
-.677
.230
.553
.681
-.696
-.892
Std. Error of
.211
.211
.211
.211
.211
.211
1.101
2.297
-.992
-.617
-.987
.693
.419
.419
.419
.419
.419
.419
Skewness
Kurtosis
Std. Error of
Kurtosis
Model Summary
Model
1
R
a
.873
Adjusted R
Square
Square
Estimate
.763
.754
.073988
452
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 447-457
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
The parameter for OilPrice is positive
(+0.004), meaning that when the oil price
increases, the Euro Dollar Exchange rate
increases. This is a bit confusing but it can be
explained as follows: Since the United States
is the most important oil importer around the
world, and since its economy is highly related
to energy; it could be clearly said that any
rising in oil prices, would first affect or even
damage the U.S. economy.
The parameter for Euribor1Year is positive
(+0.049), meaning that when the Euro Zones
interest rates increase, the Euro appreciates,
and the Euro Dollar Exchange rate increases.
This is also a logical result.
The parameter for USDLibor1Year is
negative (-0.020), meaning that when the US
Dollars interest rates increase, the US Dollar
appreciates, and the Euro Dollar Exchange
rate decreases. This is also a logical result.
Table V shows the results of plugging the data of
2013 (first eight months) into our regression
equation.
Oil
Price
Eur
ibor
1Y
USD
Libor
1Y
Exch.
Rate
Reg.
Rate
Regr.
%
Err.
USInf.
EUInf.
Jan-13
1.6
1.97
105.04
0.5753
0.814
1.33
1.31
1.38
Feb-13
1.84
107.66
0.5942
0.762
1.34
1.37
2.57
Mar-13
1.5
1.73
102.61
0.545
0.735
1.30
1.34
3.09
Apr-13
May13
1.1
1.18
98.85
0.5284
0.717
1.30
1.39
6.57
1.4
1.42
99.35
0.4838
0.694
1.30
1.37
5.30
Jun-13
1.8
1.6
99.74
0.5071
0.684
1.32
1.36
3.49
Jul-13
1.61
105.21
0.5254
0.684
1.31
1.40
6.80
Aug-13
1.5
1.34
108.06
0.5423
0.668
1.33
1.42
6.95
Av.%
Error
4.52
Para
meter
0.06
B)
Para
meter
-0.17
Para
meter
0.004
Para
meter
0.049
Para
meter
-0.02
Cons
tant
1.11
453
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 447-457
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
There are many different activation functions,
some of the most commonly used are:
Threshold:
1 if x t 0
g(x) =
(2)
0 if x t 0
Sigmoid:
g(x) =
(3)
1 e 2s( x t )
Hyperbolic tangent:
g(x) = tanh(s(x t))
sinh( s ( x t ))
cosh(s ( x t ))
e s( x t ) e s( x t )
e s( x t ) e s( x t )
e 2( s ( x t )) 1
(4)
e 2( s ( x t )) 1
Dependent variables
Nominal
Ordinal
Scale
The method for rescaling covariates
Standardized
Normalized
Adjusted Normalized
None
Partition
Training
Test
Holdout
Architecture
Number of Hidden Layers
Output layer activation function
Identity
Softmax
Hyperbolic tangent
Sigmoid
Activation Function: The activation function
"links" the weighted sums of units in a layer to
the values of units in the succeeding layer
Hyperbolic tangent
Sigmoid
Type of Training: The training type
determines how the network processes the
records
Batch
Online
Mini-batch
Optimization Algorithm: This is the method
used to estimate the synaptic weights
Scaled conjugate gradient
Gradient descent
Training options for the scaled conjugate
gradient algorithm
Initial Lambda
Initial Sigma
Interval Center and Interval Offset
Training options for the gradient descent
algorithm
Initial Learning Rate
Lower Boundary of Learning Rate
Momentum
Learning rate reduction, in Epochs
Stopping Rules: These are the rules that
determine when to stop training the neural
network
Maximum steps without a decrease in
error
Maximum training time
454
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 447-457
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Maximum Training Epochs
Minimum relative change in training error
Minimum relative change in training error
ratio
When using an MLP Neural Network, there is an
important option called Independent variable
importance analysis. This option allows us to
evaluate each independent variable according to its
importance and influence on the dependent
variable. It performs a sensitivity analysis, which
computes the importance of each predictor in
determining the neural network. The analysis is
based on the combined training and testing samples
or only on the training sample if there is no testing
sample. This creates a table and a chart displaying
the importance and normalized importance for each
predictor. Note that sensitivity analysis is
computationally expensive and time-consuming if
there are large numbers of predictors or cases.
Figure 7 represents the normalized importance of
the independent variables.
Reg. Rate
Regr. % Err.
ANN Rate
ANN % Err.
Jan-13
1.31
1.38
1.29
3.30
Feb-13
1.37
2.57
1.31
2.21
Mar-13
1.34
3.09
1.30
0.04
Apr-13
1.39
6.57
1.32
1.57
May-13
1.37
5.30
1.31
0.76
Jun-13
1.36
3.49
1.31
0.94
Jul-13
1.40
6.80
1.32
0.85
Aug-13
Aver% Error
1.42
6.95
4.52
1.34
0.69
1.29
455
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 447-457
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Concerning the Regression Model, the
average percentage error over the 8 sets of
data is 4.52%.
Concerning the ANN Model, the average
percentage error over the 8 sets of data is
1.29%. The ANN model is very accurate, and
it was able to approximate the relation
between the dependent and independent
variable within a 1.29% average error margin.
This error margin is considered very low and
the system is hence reliable for forecasting.
Now lets calculate the average error for both
methods. Referring to the tables, we can calculate
the error as follows:
For the regression model, we can calculate the
percent error for each month between January
2002 and December 2012 using the following
formula:
Regression Monthly Error =
ABS(ExchangeRate - Regression)
ExchangeRate
(5)
ABS(ExchangeRate - ANN)
ExchangeRa te
(6)
And the average error for the same data set is:
Average ANN Error = 2.17%
456
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 447-457
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
o The network performance.
All these parameters, in addition to many others,
have to be carefully selected and fine-tuned in
order to get an optimized result. The fine tuning
effectively consumes a lot of time, using trial and
error, and experimenting many combinations.
VII. CONCLUSION
A comparison between two methods the
artificial neural network and the regression method
in forecasting exchange rates is detailed in this
paper. It was a serious effort toward finding an
adequate system that can forecast the
Euro/Dollar exchange rate with an acceptable error
margin. Both methods were set according to our
thorough research in the fields of forecasting,
macro-economics, statistical techniques, and
artificial intelligence.
The obtained result shows that exchange rates
can be forecasted with a minimum margin of error,
and therefore our ANN model can be used in
forecasting. It can have an impact on many
economical aspects. International companies can
no longer pretend that they have no idea about the
future rate of currencies.
This implementation also has many advantages.
First, future international transactions will become
safer and less prone to exchange rate fluctuations
since it is now feasible to know the trend of
currencies. Second, international companies will
suffer from fewer losses when doing international
transactions or when signing future contracts in
different international currencies. This all leads to
more profit for these companies.
On the other hand, this method has a slight
disadvantage because companies and individuals
might completely rely on such methods in order to
forecast exchange rates, disregarding other
external factors that may affect currency rates. To
prevent such inconvenience, fine tuning and
enhancements to this method are essential to
minimize the error margin and maximize
accuracy.
REFERENCES
1. Pacelli, V., "Forecasting Exchange Rates: a Comparative
Analysis", International Journal of Business and Social Science,
vol. 3, no. 10, 2012.
2. Sandler, Ralph, Forecasting Exchange Rates Within the
Context of an International Financial Management Course,
Advances in Financial Education. Vol. 6, pp. 119-135, 2008.
457
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
ABSTRACT
The mobile Virtual Private Network (MVPN) of the
Internet Engineering Task Force (IETF) is not designed
to support NEtwork MObility (NEMO) and is not
suitable for real-time applications. Therefore,
architecture and protocols to support VPN in NEMO
are needed. In this paper, we propose a cost-effective
and secure mobility management scheme (SeSIP) that
is based on session initiation protocol (SIP) and
designed for real-time applications with VPN. Our
scheme to support MVPN in NEMO enables the
session to be well maintained during movement of the
entire network. Further, in order to reduce the
authentication delay time in handoff operations, the
signaling time which occurs to maintain the session is
shortened through our proposed handoff scheme which
adopts authentication using HMAC-based one-time
password (HOTP). Our performance analysis results
show our proposed scheme provides improvement in
average handoff performance time relative to existing
schemes.
KEYWORDS
NEMO, MVPN, SIP, HOTP, Mobility Management.
1 INTRODUCTION
As the coverage area of wireless LAN (WLAN)
expands, the demand from users is growing for
access to the Internet anytime and anywhere. To
satisfy this requirement, technologies that enable
access to the Internet on trains, busses, ships, and
other modes of transportations have come into the
limelight. One such technology is NEMO, an IP
network mobility technology [1-3]. The acronyms
used in this paper is listed in Table 1. NEMO
458
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Although IETF proposed a VPN architecture that
supports mobility, this solution did not consider
mobile equipment groups and is only applicable
for a single node, and, furthermore, it is based on
MIP, which is not suited for real-time applications.
MVPN of IETF uses one IPSec [4] tunnels and
two MIP tunnels. These three tunnels are major
contributors to overhead during the real-time
packet transfer. Thus, a new architecture and
protocol are required to support the MVPN in safe
NEMO. In addition, the complexity of the
authentication procedure and multiple signaling
messages that may occur in various nodes due to
the movement of the mobile equipment group are
also major contributors to overhead.
This paper proposes a Cost Effective and
Secure Mobility Management Scheme (SeSIP)
based on the SIP (Session Initiation Protocol)
which is suitable for real-time application on
MVPN and which shortens the signaling time.
This design maintains the session continuously as
the overall network moves. It integrates SIP-based
MVPN and NEMO to provide efficient group
mobility for high security and real-time services.
Additionally, all SIP clients can directly
communicate with each other, bypassing the
mobile agent such as the Home Agent (HA) in
MIP. Thus, the path is optimized. This is useful
for real-time applications such as IP-based voice
communications (VoIP) and video streaming, and
it does not require an IPSec tunnel or MIP tunnel.
Hence, a single NEMO VPN gateway can support
an entire mobile network upon the address request
of a mobile network that has changed its
connection location address, resulting in
considerable reduction of signaling overhead.
Moreover, this approach reduces the signaling
numbers since all CNs connection addresses are
combined in a URL list and integrated in a single
INVITE message for transfer. Further, this design
adopts an authentication method based on HMACbased One Time Password (HOTP) [5] to shorten
the authentication time, a significant element of
delay during hand-off, thereby improving the
ongoing signaling time to maintain the session.
Moreover, this approach integrates the generation
signals of multiple nodes inside the mobile
network to reduce signaling time.
Description
ALG
AVP
CN
Correspondent Node
CoA
Care of Address
HA
Home Agent
HMAC
HOTP
IKE
i-HA
Internal HA
i-MIP
Internal MIP
MAA
Multimedia-Auth-Answer
MAR
Multimedia-Auth-Request
MIDCOM
Middlebox Communication
MIKEY
MIP
Mobile IP
MN
Mobile Node
MR
Mobile Router
NEMO
Network Mobility
OTP
RTCP
RTP
SA
Security Association
SDP
SIP
SIP-NVG
SRTP
TEK
TGK
UAA
User-Authorization-Answer
UAR
User-Authorization-Request
VPN
VPN-TIA
X-HA
external HA
X-MIP
external MIP
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
effective and secure. Section 4 discusses the
analytical model to evaluate the functioning of the
proposed scheme. Section 5 describes the
numerical results for the analysis presented in
Section 4. Finally, conclusions are drawn in
Section 6.
2 RELATED WORK
IETF has previously defined the architecture and
protocol for MVPN [6]; it is shown in Fig. 1. Here,
the internal HA (i-HA) and external HA (x-HA)
are present in the intranet and Internet and the two
HAs. A new care-of address (CoA) is first
obtained from the dynamic host configuration
protocol (DHCP) server or foreign agent (FA)
when the MN moves out of the intranet. This CoA
is registered in x-HA. Then, MN creates a VPN
gateway and IPSec tunnel using its external home
address (x-HoA). An IPSec tunnel is created by
using internet key exchange (IKE) [7]. Fig. 1
shows the three tunnels (x-MIP, IPSec, and i-MIP).
Mobile Gateway
DHCP
X-HA
AAAF
AAAH
Ipsec-VPN
i-HA
CN
DHCP
802.11b
AP
AAAH
802.11b
AP
AMR
AMR
i-HA
HAR
MN
x-HA
HAR
HAA
Subnet
Subnet
HAA
AMA
DHCP/i-FA
DMZ
Reg-Reply (x-MIP)
AMA
Register to x-HA
(CoA mode)
Intranet
Internet
x-FA
HDR, SA
Subnet
1. Register to x-HA by using CoA
HDR, SA
HDR, KE, Ni
Subnet
2. Establish Ipsec tunnel with VPN by using x-HoA
HDR, KE, Nr
HDR*, KE, Idii, HASH_I
Three Tunnels
802.11b
AP
MN
x-MIP
(x-HA to MN Coa)
IPsec
(VPN GW to x-HoA)
Reg-Req(i-MIP)
i-MIP
(i-HA to VPN-TIA)
Original Packet
Reg-Reply(i-MIP)
Register to i-HA
Terminate Ipsec tunnel
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
end-to-end latency. On the other hand, SIP has
been proposed to provide host mobility and
session continuity. However, by adopting SIP into
NEMO, it may increase signaling cost during
network handoff. So, I propose architecture and
protocols to support VPN in NEMO, which is
called Cost Effective and Secure Mobility
Management Scheme [40]. The proposed SeSIP
comprises SIP, secure real-time transport protocol
(SRTP) [14], multimedia internet keying
(MIKEY) [15], and a Diameter server [16] to
provide VPN services in NEMO. Fig. 3 depicts the
architecture of the proposed SeSIP.
Foreign Network 2
Diameter Server
AP/BS
VPN Gateway
MN
SIP Proxy 1
Intranet
Firewall
SIP Proxy 2
(SIP registrar)
ALG
Foreign Network 1
CN
SIP-NVG
AP/BS
Mobile Network
MN
Signaling traffic
Data traffic
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Request (UAR) / User-Authorization-Answer
(UAA) and Multimedia-Auth-Request (MAR) /
Multimedia-Auth-Answer (MAA) to process SIP
REGISTER and INVITE messages. The
authentication is done by the Diameter server
rather than by delegating to a SIP server. HOTPbased authentication is adopted in the proposed
SeSIP to reduce authentication time, an element of
delay time during handoff. HOTP is an OTP
creation algorithm based on event synchronization,
and the client and authentication server share the
secret key K. It uses C, the increasing counter
value, and HMAC-SHA-1 hash algorithm to
create the password. The increased value (C +1) is
used to create a new password (6 digits) during the
following authentication. The OTP mechanism
creates the single user password based on three
parameters: hash algorithm, secret key, and
challenge/counter. HOTP creates the password
using the authentication number (counter) - which
is remembered between the authentication server
and the user - as the input value of OTP, and the
authentication is performed only when the counter
value matches. The counter parameter has the
characteristics of synchronization OTP (HOTP
[5]), and the client creates a new password without
receiving the item beforehand from the
authentication server. HOTP performs client
authentication through one handshake, using the
OTP creation algorithm based on the event
synchronization method.
As discussed above, the SIP-NVG is the mobile
networks gateway to other networks. When a
mobile network roams among different IP subnets,
the SIP-NVG not only keeps ongoing sessions
unbroken, but also transmits data in a secure
manner. There are two types of interfaces owned
by SIP-NVG: egress interface and ingress
interface. A SIP-NVG attaches to the Internet
through an egress interface. Once a mobile
network moves to a new IP subnet, the egress
interface of the SIP-NVG will get a new IP
address. On the other hand, when an MN wants to
join a mobile network, it attaches to the ingress
interface of the SIP-NVG. In our design, each
mobile network has only one SIP-NVG which
essentially is an MR with SIP capability. The
proposed SIP-NVG is able to route SIP messages
and data traffic between its egress interface and
SIP-NVG
Diameter
Server
SIP Proxy 1
SIP Proxy 2
(SIP registrar)
REGISTER
REGISTER
UAR
UAA
REGISTER
MAR
MAA
200 OK
200 OK
200 OK
REGISTER
REGISTER
462
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
responses for the corresponding commands. When
the ALG receives a special incoming RTP stream
from the home network to an MN in the Internet,
it replaces the whole IP/UDP/RTP header with a
new one, transforms the new RTP packet into
SRTP format, and delivers the SRTP stream to the
destination. In the reverse direction, the ALG
receives the SRTP stream from the Internet, and
the ALG decrypts it and verifies it to decide
whether the SRTP packet is valid. If the SRTP
packet is decrypted and verified successfully, the
RTP payload is carried by a new RTP header. The
new RTP packet is then transmitted to the home
network.
Each session in the ALG requires sufficient
external and internal resources. For example, the
external resource may include an external listening
address, external listening port, external
destination address, and external destination port.
Destination addresses and ports are provided by
SIP Proxy 2. Only when all resources are ready,
does the session in the ALG start. When either the
external or internal resource is reserved
successfully, the ALG will reply with the reserved
listening address and port to SIP Proxy 2.
3.2 Operation Procedures
In the architecture shown in Fig. 3, the entire
mobile network may move from one IP subnet to
another. This is called network handoff. It is also
possible that an MN moves into or moves out of
the mobile network. This is called node handoff.
Fig. 5 depicts the flow when an MN moves into
a mobile network which is located inside a foreign
network. First, the MN registers with the SIPNVG and the SIP registrar, as discussed in Section
3.2. Then, the MN must re-invite the CN, if there
are active sessions between them. For the INVITE
request, in addition to translating the CONTACT
field from the MNs address into the SIP-NVGs
URI address, the SIP-NVG also adds the
RECORD-ROUTE field where the SIP-NVGs
URI address is inserted. Therefore, subsequent
messages of the existing sessions will be routed by
the SIP-NVG. In this case, the signaling messages
will need to go through SIP Proxy Server 1 and be
authenticated by the Diameter Server before they
reach SIP Proxy Server 2. When the mobile
SIP Proxy2
(SIP registar)
SIP-NVG
Diameter
Server
CN
INVITE
INVITE
MAR
MAA
INVITE
200 OK
200 OK
200 OK
ACK
ACK
ACK
RTP
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
messages in the wireless links. In order to reduce
the signaling overhead, the SIP-NVG combines all
contact addresses of CNs into a URI list. The URI
list is then conveyed by the SDP embedded in one
INVITE message. The re-INVITE contains the
new contact address of the SIP-NVG. This is sent
to SIP Proxy 1, which routes the message to SIP
Proxy 2, assuming that SIP Proxy 1 has been
informed that SIP Proxy 2 is responsible for
verification of ongoing sessions.
MN
SIP-NVG
SIP Proxy 1
Diameter Server
SIP Proxy 2
(SIP registrar)
ALG
CN
INVITE
with a URI list
MAR
MAA
Allocate Resource
Allocate Reply
INVITE
:
200 OK
:
Allocate Resource
with URI list
200 OK
with a URI list
200 OK
with a URI list
Allocate Reply
with URI list
200 OK
:
ACK
:
ACK
with a URI list
ACK
with a URI list
ACK
:
RTP(SRTP)
RTP
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
password included in the Call-ID to the calculated
password to determine authentication of the user
agent. If the two values are the same, the Diameter
server completes the user authentication through
one handshake. The Diameter server informs SIP
Proxy 2 of each sessions verification result, and it
generates one pair of TGK and MIKEY for each
verified session. Thereafter, the MAA has all the
session verification result codes. To establish
TGK and MIKEY message, pre-shared key, which
is one of the most efficient ways to handle key
transport, is used. The pairs of the TGK and
MIKEY messages are transmitted to the SIP Proxy
2 by MAA.
If the SIP-NVG is granted access to the intranet,
then SIP Proxy 2 must allocate enough resources
to guarantee that the sessions will be protected.
Thus, SIP Proxy 2 orders the ALG to reserve an
internal receiving address and receiving port for
each ongoing session. The command from SIP
Proxy 2 may include the Data SA and the TGK for
SRTP protection, and the CNs original listening
addresses and listening ports. The ALG responds
with the reserved addresses and ports. SIP Proxy 2
inserts the listening addresses and listening ports
in the SDP to reproduce INVITE requests based
on the URI list. It then routes them to each CN
individually. Each CN sends back a 200 OK if it
agrees with the SDP of the re-INVITE. After
receiving the 200 OK from each CN, SIP Proxy 2
again orders the ALG to reserve the external
receiving addresses and receiving ports. The ALG
responds with the reserved external receiving
addresses and receiving ports. If all allocated
resources are ready, SIP Proxy 2 replaces the
listening addresses and listening ports in the SDP
of each 200 OK, and it inserts a MIKEY Initiator
message into the SDP to transport the TGK. Then,
SIP Proxy 2 sends each 200 OK with the new SDP
to the SIP-NVG, which forwards it to the MN.
The ALG will then start to function when both
internal and external resources have been acquired.
When each MN attaching to the mobile network
receives the 200 OK, it must process the included
MIKEY Initiator message and extract the shared
TGK. The MN is then required to send an ACK
with SDP which includes the MIKEY responder
message. After the MN sends the ACK, it will
start all transport sessions.
MN
SIP-NVG
SIP Proxy 1
Diameter Server
SIP Proxy 2
(SIP registrar)
ALG
CN
INVITE
with a URI list
INVITE
with a URI list
MAR
MAA
Allocate Resource
with a URI list
200 OK
with a URI list
200 OK
with a URI list
200 OK
:
ACK
:
ACK
with a URI list
ACK
with a URI list
RTP(SRTP)
RTP
SIP-NVG
Diameter Server
SIP Proxy 2
(SIP registrar)
ALG
CN
INVITE
with a URI list
MAR
MAA
Deallocate Resource
with a URI list
Deallocate Reply
with a URI list
INVITE
:
200 OK
:
200 OK
with a URI list
200 OK
:
ACK
:
ACK
with a URI list
ACK
:
RTP
465
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Figure 9. Message flow when the mobile network roams
from a foreign network back to its home network
N subnets
Mobile network
Mobile network
Mobile network
k
A new session arrives or an existing session departs
Home Network
Foreign Network
H fh
H ff
H hf
f m* ( s) e st f m (t )dt.
t 0
1
1
Definition
Number of networks a mobile network visits before it
goes back to the intranet.
Session arrival rate for a mobile network
Average session service time
Average network residence time
466
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Maximum number of ongoing sessions
in a mobile network
f ( s)
.
s
*
m
E tSi
1
,
i c,
i
1
(1)
(1 f m* ( i ))
1
,
k 0,
ai (k )
1 f * ( ) 2 f * ( ) k 1 , k 0.
m
i m
i
i
(2)
(1 gi )2 N j q
( gi ) gi yz j x q ,
i gi
(4)
Hence, we obtain:
gi f m* ( i )
.
i
(5)
(3)
Shfi R f Lhf iI hf ,
S iff R f L ff iI ff ,
(6)
S Rh L fh iI fh .
i
fh
467
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
During E tS , we assume that the mobile
network crosses k subnets as shown in Fig. 10. We
define as the number of subnets the mobile
network moved across from the time it visited the
intranet until the time the last event occurred.
When 0 < < N, the total signaling cost in the
mobile network during E tS can be derived as :
i
Definition
Average handoff cost when a mobile network moves from
the home network to a foreign network with i ongoing
sessions.
Average handoff cost when a mobile network moves from a
foreign network to another foreign network with i ongoing
sessions.
Average handoff cost when a mobile network moves from a
foreign network to the home network with i ongoing
sessions.
Average registration cost of a mobile network (sent by SIPNVG) when the mobile network enters its home network.
Average registration cost of a mobile network (sent by SIPNVG) when the mobile network enters a foreign network.
Average cost for the first part of re-INVITE when a mobile
network moves from its home network to a foreign network.
Average cost for the first part of re-INVITE when a mobile
network moves from a foreign network to another foreign
network.
Average cost for the first part of re-INVITE when a mobile
network moves from a foreign network to its home network.
Average cost for the second part of re-INVITE of a session
when a mobile network moves from its home network to a
foreign network.
Average cost for the second part of re-INVITE of a session
when a mobile network moves from a foreign network to
another foreign network.
Average cost for the second part of re-INVITE of a session
when a mobile network moves from a foreign network to its
home network.
S hfi
S iff
S ifh
Rh
Rf
Lhf
L ff
L fh
I hf
I ff
I fh
Ci ( N , i , , )
k 1
i k
N S fh N
k 0
k 1 k
S iff k
ai (k )
N N
i
hf
i
i
i
i
N 1
( Shf gi S ff gi S fh S ff )(1 gi ) gi
S iff
. (7)
i
(1 giN ) gi i
1 2e ( N 1)/2 e1
e
,
1 e1
( N 1)/2
e1
e ( N ) 1 2e
1 e1
N 1
2
N 1
N 1.
2
(8)
Therefore,
Ci _ exponential ( N , i , )
N 1
C ( N , , , ) S
i
ff
(1 gi )
i i (1 giN )
1
i
i
i
i
N 1
(1 e )( Shf S ff ( S fh S ff ) gi )
1 2e ( N 1) 2 e1
1 2e ( N 1) 2 e1
N 1
N 1 2 e
e N
g
g
i
1 i N 1 2
(9)
i ! i x 0 x ! x
(10)
C
i 0
P i .
i _ exponential i
(11)
468
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Var
(12)
Definition
ax
bx
Ax , y
Bx , y
Shfi mR f iLhf iI hf ,
M ,
S iff mR f iL fh iI ff ,
(13)
(14)
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
uses UAR/UAA and MAR/MAA commands for
the Diameter server for user authentication and
authorization when the client of the SIP server
transmits SIP REGISTER and INVITE messages.
Therefore, the cost incurred in the use of HTTP
digest authentication can be calculated as follows:
Mx
M i o
Rh 2anvg 2areg 4 Anvg ,reg 2 M ,
M i i
Test
Tter
dx
ex
Wx , y
2 Breg , alg 2 M ,
Z x, y
2U 2 M ,
Lhf 3bnvg 5bpro 5breg 2balg 5 Breg , pro 5 B pro ,reg
2 M ,
4000
3500
3000
Definition
The x-MIP registration cost.
The i-MIP registration cost when MN is located
outside the intranet.
The i-MIP registration cost when MN is located inside
the intranet.
The establishment cost of IPsec tunnel.
The termination cost of IPsec tunnel.
The processing cost for MIP registration at Node x.
The processing cost for IPsec message at Node x.
The transmission cost of MIP registration between
Node x and Node y.
The transmission cost of IPsec message between Node
x and Node y.
The total cost for x-HA to process and transmit
HAR/HAA and AMR/AMA messages to AAAF and
AAAH.
IETF-MVPN
With SIP-NVG(use HTTP)
Without SIP-NVG(use HTTP)
2500
2000
1500
1000
500
0
10
Average Residence Time
470
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
requires time to establish the three tunnels.
Compared to the mobile network without SIPNVG, the method with SIP-NVG reduces handoff
signaling cost significantly, since SIP-NVG
performs registration in the SIP Registrar on
behalf of the entire mobile network when it moves
to a new subnet, whereas, without SIP-NVG, all
MNs must update their locations individually.
4000
With SIP-NVG and use HTTP
SeSIP(With SIP-NVG and HOTP)
3500
450
3000
400
2500
350
2000
1500
1000
500
0
300
250
200
150
100
10
Average Residence Time
10
Average Residence Time(1/gamma)
450
400
350
300
250
200
150
100
1
10
Average Residence Time(1/gamma)
5 CONCLUSIONS
Although the IETF standard has proposed a
mobile VPN architecture, it is designed for the
movement of a signal node only. In addition, IETF
MVPN has large overhead for transmitting realtime packets, because it requires one IPsec tunnel
and two MIP tunnels. On the other hand, there has
been no efficient way to support mobile VPN in
NEMO, even though NEMO supports network
mobility. This paper presents a novel method for
supporting MVPN in NEMO that ensures that the
471
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
session is maintained continuously when the
whole network moves, and it proposes using the
HOTP-based authentication method to shorten the
processing time of the signaling that continuously
occurs to maintain the session. In addition,
security is enhanced in our design through the
integration of NEMO and VPN.
We analyzed the design and performance of our
proposed design, and results indicate that the
proposed SeSIP based on SIP is well suited to
real-time service. Although SIP-based mobility
management can easily support routing
optimization, there may be an upswing in the
handoff signaling costs, because many signaling
messages are transmitted to maintain the session
in progress with SIP in NEMO. In the proposed
SeSIP, a URI list is used to signify the SIP proxy
server instead of transmitting signaling messages
individually to each node. Therefore, the signaling
cost is reduced. User authentication using the
existing HTTP digest authentication method
requires many handshakes, increasing the
signaling cost. In contrast, the proposed SeSIP
using HOTP-based authentication considerably
reduces the number of handshakes needed with the
authentication server, thus reducing the signaling
cost. The SIP proxy server and the Diameter
server are responsible for authentication and
authorization. Also, the ALG receives a command
from the SIP proxy server to process the security
information for the data transmission, depending
on MIDCOM architecture. ALG is responsible for
converting and relaying the protected and
unprotected data. Thus, unauthorized data cannot
pass the ALG in the Internet. This paper examined
a method for efficient management of group
mobility and cost savings for real-time services
through the integration of mobile VPN and
NEMO. NEMO, currently in the early stage of
research, is expected to be further realized through
the convergence of various technologies, policies,
and methods, such as the path optimization
method for efficient services, multi-homing
technology, and methods for services in the
inclusive mobility network.
These need to be researched in the period
ahead. For commercial service, research should
also be conducted to develop technology to enable
fast detection of movement.
6 ACKNOWLEDGEMENT
This research was supported by Basic Science
Research Program through the National Research
Foundation of Korea (NRF) funded by the
Ministry of Education (NRF-2010-0024695).
This article is a revised and expanded version of a
paper entitled Secure SIP-based Mobility
Management Scheme for Cost-Optimized NEMO
Environments presented at The International
Conference on Digital Information, Networking,
and Wireless Communications (DINWC2014)
held on June 24-66, 2014 at Ostrava, Czech
Republic.
7 REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
472
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 458-473
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
473
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
ABSTRACT
Fall detection is one of the major issues in health care
filed.
Falls can cause serious injury both in
physiology and psychology, especially to the old
people. A reliable fall detector can provide rapid
emergency medical care for the fallen down people.
Thus, a reliable and effectively fall detection system is
necessary. In this paper, we propose a system which
utilizing mobile phones as a detector to detect the
falling. When fall accident occurs, the system has
three response procedures for help.
The first
procedure is transmitting the emergency message to
the related people for help. The second procedure
shows the users status and location on the map of
webpage, according to users GPS location and status.
The third procedure makes the alarm sound; its
purpose is to let the person who nearby the user can be
noticed that the user needs help. First, using a
waist-mounted mobile phone to capture accelerometer
of the human body and adopt the DCT (Discrete
Cosine Transform) to analyze the value of
accelerometer to distinguish the activities of daily
living (ADL) and falls. ADL consist of walking,
standing and sitting.
We utilized a tri-axial
accelerometer in mobile phone to capture the signal
and transmit it to the server by way of Internet. We
adapt two judgments achieved in Server, first judgment
is based on an adaptive threshold for detecting the
energy by DCT; the setting of adaptive threshold
include height, weight and gender. The second
judgment is according to the tilt of smart phone.
Experimental results show that this method can detect
the falls effectively; in addition, it is more portable
than other devices as well.
KEYWORDS
Fall detection; health care; mobile phone; DCT;
accelerometer;
1 INTRODUCTION
bbandydd@hotmail.com
The average age of the population of the world is
increasing because the increment in the average life
span. The number of persons with older ages are
increasing faster the higher the age is considered.
According to Taiwan population statistics, the
population rate of aged 65 or over is expected to rise
from 10.4% in 2008 to approximately 37.5% in 2056.
Among older persons, approximate 25~35% of them
experienced fall-related injury more than one time per
year.
The fall in the elderly persons is one of the major
public health problems, due to it causes a serious injury
and obstacle to their independent living. If the
emergency medical treatments were not in time, these
injuries may result in disability, paralysis and even
death. Once the falling is occurred, specific persons
such as family, colleagues, and friends should be
notified to assist the people as soon as possible. It is
believed that the earlier the fall is notified; the lower is
the rate of morbidity-mortality. Thus, reliable, fast
and effective fall detection notified system is getting
more important. A typical fall detection system has
two major parts: the detection part and notify part.
The detection part is able to detect the fall efficiently
and accurately. Thus it needs a notify part to notify
the related persons. In this paper, we use message,
GPS and alarm sound to achieve the notification. A
reliable fall detection system cant just detect the fall;
it must have the ability to help the user from a danger
situation. The necessary functions in fall detection
system should include computing ability (compute the
energy by DCT for detecting the fall), media player
(make an alarm sound so that the person who nearby
the user can be notice that user needs help), Wi-Fi or
3G (transmit the accelerometer value to server) and
communication (automatic send a text message to
related person); these functions contain the detection
part and notify part. Due to mobile phone naturally
combines the two parts; therefore, we utilize mobile
phone as the major part in our fall detection system.
1.1 Background
474
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
2 RELATED WORKS
There are numbers of different approaches for the
fall detection have revealed in recent years. Recent
fall detection systems can be classified to four classes,
which are user activated alarms, visual detection,
wearable sensor and mobile phone. We give these
four classes as follows, 2.1 describes the user activated
alarms, 2.2 describes the visual detection, 2.3 describes
the wearable sensor and 2.4 describes the mobile
phone.
475
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
They use a 3D camera to extract images and combine
with a context model. Body orientation can be
obtained from extracting the body posture. The
context model is using for finding visual fall detection,
depending on the location and time of when the fall
events happened. 6 proposes a visual-based fall
detection system, which uses an omni-directional
camera for a better detection perspective. They
propose a simple algorithm of threshold and decision
tree based on the body angle and length variation.
In this class, they use image processing technique
for capturing the images of body and then detecting
visual falls events. In this way, it has limitations on
detecting the fall events, the body is a moving object
but the equipment is fixed. Moreover, due to the
privacy issues, its not suit to monitor persons for a
long time.
Fig. 1
476
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
values, the system calculates the energy by DCT
to determine whether the energy exceed than threshold.
If the energy exceeds the threshold and the phone is tilt
then client will make an alarm sound. If the alarm
sound alert exceeds 15 seconds, mobile phone will
automatic send an emergency message for help and
shows the location and situation of user on webpage.
(2)
effectively algorithm.
However; refer to Fig. 2, it is
obvious that it is difficult to distinguish ADL and fall
by a threshold only. The ADL are defined as walking,
standing and sitting.
The fatal concern in fall
detection is to make it distinguishable from various
ADL.
In this paper, we utilized a well-known transform
method Discrete Cosine Transform (DCT) in digital
signal processing to analyze the values of SVM.
Transform coding is a kernel component of
contemporary image/video processing applications.
Transform coding relies on the premise that signals
exhibits a certain level of correlation with their
neighboring signals, so that we utilized DCT transform
the spatial domain into frequency domain into find the
features of SVM.
In this paper, we use the
one-dimensional DCT 16 to transform SVM to get its
feature for classification furthermore.
The most
common DCT definition of a 1-D sequence of length N
is given as (3) :
1
1
(2 + 1)
() = () () [
]
2
2
=0
() =
20
30
0
AC
20
SVM
15
if u=0,
Walking
Sitting
15
10
25
2
,
2
c(u)=1,
o.w.
for u=0,1,2,,N-1, N=50 in our experiment.
F(0) denotes the DC and the others are refer to as ACs.
(3)
Standing
Fall
-5
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49
-10
-15
-20
Data
10
Fig. 3
0
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49
Data
Fig. 2
50 SVM values in four cases.
Fig. 2 contains four curves which show the 50
SVM values in the four cases of walking, standing,
sitting and fall, respectively. X-axis denotes 50 data
in the last 5 seconds and Y-axis denotes the SVM
values. In 15, it claimed that if SVM > 1.8g, then the
situation must be fall, and then proposed a simple and
477
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
mobile phone mounted on the user and starts a timer.
Once the user does not manually turn off the alarm
within 15 seconds, the system will send an emergency
message for help automatically. In addition, the
mobile phone will make an alarm sound so that the
persons who nearby the user can be noticed that user
needs help; therefore, they can quickly help the user
who is injured for medical care. At the same time,
the map of webpage will present that the situation of
user is dangerous and the location of accident
occurred.
Fig. 4
DC values of the four cases
Refer to Fig. 4, the DC is hard to distinguish so that
we do not take it into consideration for classification.
Due to the DCT owns excellent property of the energy
centralization; most applications of DCT are data
compression. In this paper, we use DCT to get the
energy of SVM to be the feature for classification.
We utilized an array that consists of the last 50 samples
of the SVM. When a new sample is added to the
array, all other samples will be shifted one place and
the oldest sample is eliminated. Then we use (4) to
calculate 49 AC values to get the energy of SVM.
2
= 49
(4)
=1 ()
Notice that, the DC coefficients are removed in the
four cases. DC represents the average of the samples.
Refer to Table 1, in which we list the energies of the
four cases by choosing different amount ACs. The
energy is used to distinguish the fall and ADL. The
selection of threshold will be addressed in the next
section.
Table 1.
25 ACs
20 ACs
15 ACs
10 ACs
9 ACs
Walking
306.59
261.01
85.35
26.56
25.50
Standing
0.57
0.53
0.31
0.26
0.26
Sitting
171.93
154.96
149.06
124.09
103.11
Fall
1070.45
997.36
838.56
280.79
204.53
Energy
8 ACs
7 ACs
6 ACs
5 ACs
4 ACs
Walking
25.43
25.42
25.42
23.13
16.59
Standing
0.23
0.23
0.18
0.17
0.15
Sitting
91.25
70.04
50.25
38.46
20.82
Fall
122.83
117.63
116.80
111.09
77.10
4 EXPERIMENTAL RESULTS
4.1 The training phase
In this section, we will depict how the system
designs and works. In our experiment, the mobile
phone is mounted on the waist. We implement the
system in Java, with Eclipse and Android 2.1 SDK.
We collect data in the four cases including walking,
standing, sitting and fall to get the thresholds.
Because the fall may cause serious injury, we didnt
test falls with real elderly people.
Detailed data are given in Table 2.
Table 2. Volunteer data
#Volunteer
Gender
Age
Height (m)
Weight (kg)
#Volunteer
Gender
Age
Height(m)
Weight(kg)
#Volunteer
Gender
Age
Height(m)
Weight(kg)
a
male
23
1.70
70
h
male
23
1.70
70
o
female
21
1.60
60
b
male
24
1.75
55
i
male
23
1.70
100
p
female
37
1.60
45
c
male
23
1.72
65
j
male
22
1.67
75
q
female
36
1.60
50
d
male
25
1.76
60
k
male
21
1.83
97
r
female
36
1.55
49
e
male
23
1.62
52
l
female
24
1.57
50
f
male
23
1.76
106
m
female
19
1.50
42
g
male
21
1.78
55
n
female
20
1.60
50
478
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
20
First fall
Second fall
15
10
5
AC
0
-5
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49
-10
-15
Data
Fig. 5
Fig. 9
Fig. 10
Fig. 6
Fig. 7
479
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Fig. 11
Fig. 12
Fig. 13
Fig. 14
Fig. 15
Selecting the first 4 AC values
Refer to Fig. 6 to Fig. 15, they show that selecting
the first 6 AC values to get the energy of SVM as given
in (5) will achieve a better accuracy and rapid
computing time.
= 6=1 ()2
(5)
Detailed data of the E values in four cases for each
volunteers are given in Table 3. Each person tests three
times, sequence is 1, 2 and 3, such as a1, a2 and a3. 0
shows the curve of fixed threshold in dash line,
Table 3. Fixed threshold and energy of each
volunteer
480
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
height and weight to improve the performance. The
adaptive threshold is divided into two types, one is for
male as given in (6) and another is for female in (7).
It is well-known that the strength of woman is lower
than man, so the adaptive threshold for female is lower
than male; the adaptive thresholds calculated by (6)
and (7) and the adaptive threshold for every volunteer
person are given in Table 4. We can see that each
threshold of person is different, due to the mass of
body on everyone is different.
=
1200
Walking
Standing
Sitting
1000
800
Energy
600
400
200
Table 4.
#Volunteer
threshold
#Volunteer
threshold
#Volunteer
threshold
#Volunteer
threshold
#Volunteer
threshold
9.8
2
9.8
=
(6)
0.75
(7)
b
88
f
167.67
j
131.77
n
71.77
r
74.95
c
107.65
g
85.058
k
141.92
o
86.13
d
94.912
h
118.68
l
74.54
p
64.59
0
a b c d e f g Volunteers
h i j k lm n o pq r
481
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Fig. 18 Wearing on the waist
0 shows that the user wears the phone on the
waist. When the phone is horizontal to the ground,
only the gravity of X-axis forces the gravity of ground.
Once the fall is occurred, the phone tilt status will be
changed.
35
Walking
30
25
20
SVM
15
10
5
0
a b c d e f g Volunteers
h i j k lm n o pq r
1200
Walking
Standing
Sitting
1000
800
Energy
600
400
200
0
a b c d e f g Volunteers
h i j k lm n o pq r
482
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Users must key in their gender, height and weight to
get the adaptive threshold.
Table 7. Tester test result
Number
of tests
#Volunteer
False detection
T1
T2
T3
T4
Standing
10
Walking
10
Sitting
10
Fall
10
False
Reason
Accuracy
No
100%
large strength
96%
50
large strength
98%
50
No
100%
of tests
detecti
on
Standing
50
Walking
50
Sitting
Fall
Table 6.
Tester data
False detection
T5
T6
T7
T8
Standing
10
Walking
10
Sitting
10
Fall
10
5.SYSTEM INTERFACE
5.1 The interface of server
#Volunteer
T1
T2
T3
T4
Gender
female
female
male
male
Age
23
23
30
51
Height (m)
1.7
1.54
1.65
1.72
Weight (kg)
68
47
65
62
Threshold
86.471
72.831
116.988
102.691
#Volunteer
T5
T6
T7
T8
Gender
male
female
male
male
Age
53
22
37
37
Height (m)
1.62
1.5
1.74
1.78
Weight (kg)
60
37
76
95
Threshold
112.026
60.433
123.002
146.92
483
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
The nine values from top to the bottom are 1) the
gender of the user, 2) the height of the user, 3) the
weight of the user, 4) the value of SVM, 5) the
accelerometer value of X-axis, 6) the accelerometer
value of Y-axis, 7) the accelerometer value of Z-axis, 8)
the latitude, 9) the longitude. 1), 2) and 3) are used
for adaptive threshold. Calculating 4) to obtain the
energy. 5), 6) and 7) are used for judgment the phone
tilt. 8) and 9) show the users location.
6 CONCLUSION
484
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 474-485
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
practical necessity for elderly person in daily life. We
also plan to build a system based on Google Map,
which can collect and show the locations of falls are
occurred. It can provide information that whether the
place is dangers or not, allows relevant agencies to
adopt suitable actions. By monitoring multiple smart
phones, medical center can know all of the patients at
the same time.
7 REFERENCES
1.
2.
12. Zhang, T., Wang J., Liu P. and Hou J. Fall detection
by embedding an accelerometer in cellphone and using
KFD algorithm, International Journal of Computer
Science and Network Security, vol. 6, pp. 277-284.
(2006)
13. Jiangpeng D., Xiaole B., Zhimin Y., Zhaohui S., Dong
14. Yunji L., Xingshe Z., Zhiwen Y., Bin G. and Yue Y.
3.
4.
5.
6.
7.
8.
9.
485
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 486-492
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
ABSTRACT
Automated handwritten character recognition
seems to be necessary due to the increasing number
of Persian/Arabic handwritten documents. A new
approach for Persian/Arabic handwritten digit
recognition has been proposed in this paper. This
approach employs Local Binary Pattern (LBP) operator
as the base feature extraction method. Although this
operator has shown great performance in research areas
such as context and object recognition, but it has not
been used in Persian/Arabic handwritten digit
recognition problem. First step in the proposed approach
involves smoothing, converting black and white input
image to grayscale intensity image and resizing it to a
fixed size. In the next step, input image is divided into
several blocks. LBP operator is applied to each block to
extract features. Finally, these features are used to train
a multi-layer perceptron neural network with circular
approach. Empirical results shows that the proposed
approach has a very good generalization accuracy
(99.72%) on Hoda dataset with 60000 train and 20000
test samples. This accuracy is the best among the stateof-the-art methods.
KEYWORDS
Persian/Arabic Handwritten Digit Recognition, Local
Binary pattern, Multi-Layer Perceptron, Binary to
Grayscale Conversion, Hoda DataSet
1 INTRODUCTION
Automated handwritten character recognition
seems to be necessary due to the increasing number
of Persian/Arabic handwritten documents.
Handwritten character recognition research for
Latin script family started over 50 years ago, but
its rather a new emerging research area for
Persian/Arabic language with great achievement in
the past few years. Persian and Arabic digit
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 486-492
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
context, object and face recognition problems. High
success rate of LBP is a good inspiration to use it in
Persian handwritten digit recognition. This
operator is the main feature extraction method in
the proposed approach. Also Multi-Layer
Perceptron (MLP) has been used as the base
learner. Experimental results show that our
proposed approach is more accurate and faster than
the previous ones.
The rest of this paper is organized as follows.
Section two gives a brief review of LBP operator.
The proposed approach is explained in section
three. Section four includes the experimental
results and the final section concludes the paper.
(1)
1 x 0
s ( x)
0 x 0
(2)
P 1
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 486-492
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
methods. The second advantage which make our
method easy to use in real application, is its
simplicity. This algorithm can be implemented very
fast and also has a high speed. Runtime report is
presented in the Experimental results section.
3.1 PREPROCESSING
3.1.1 SMOOTHING AND CONVERTING TO
GRAYSCALE
488
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 486-492
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
3.2 FEATURE EXTRACTION
LBP abilities in Persian handwritten digit
recognition have not been evaluated yet. Uniform
LBP has been employed to obtain more compact
and efficient feature vector. This operator is applied
in two forms (figure 7). These two obtained
histograms are put along to get the final feature
vector which is used in the training process.
1. The uniform LBP operator is applied to whole
image to get a histogram with 59 bins.
2. After dividing the image into 4 equal blocks,
Uniform LBP is applied to each block and the
resulting histograms are concatenated.
4 EXPERIMENTAL RESULTS
Multi-region LBP has been used because, singleregion LBPs output histogram is not discriminant
enough. For example consider figure 8 which
shows two Persian digits, seven (7) and eight (8)
which have reverse form of each other. Almost all
the edges in one image exist in the other one. If we
apply LBP to the whole image, it will produce
similar histograms. To overcome this problem, the
input image is divided into some blocks and LBP is
applied to them, therefore the results will no longer
be similar (figure 8).
489
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 486-492
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
4.2 PARS DATASET
Method
Ref. [5]
Ref. [14]
Ref. [15]
Ref. [6]
Ref. [15] (5-fold)
Proposed Method
(Circular training)
Proposed Method
(At once training)
Precision (%)
95.30
97.99
98.71
99.02
99.37
99.59
99.72
Rank
4
3
2
1
Team
Shahid Beheshti University
Amirkabir University
Birjand & Hormozgan
University
Proposed Method
Precision (%)
94.63
95.94
97.39
99.59
Rank
4
3
2
1
Team
Shahid Beheshti University
Amirkabir University
Birjand & Hormozgan
University
Proposed Method
Precision (%)
96.07
97.81
98.94
99.53
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 486-492
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
proposed. Experimental results show that LBP
operator can extract very discriminating features
for handwritten digit recognition. Extraordinary
recognition rate (99.72%) of the proposed approach
makes it one of the best methods that has been
proposed in this research area.
Advantages of this method can be categorized as
follows:
High discriminative power of the extracted
features
Easy implementation
Very low feature extraction time
Very low computational overhead
On the other hand, one of the drawbacks of this
system can be the number of its features. Feature
vector length can be reduced to 50% of the current
length without noticeable performance fall. This
helps the system to operate about two times faster.
Methods like PCA, LDA and evolutionary
algorithms can be considered as good feature
reduction and selection methods. Also LBP
operator can be modified to be more suitable for
Persian handwritten digit recognition. We think it
will be a good idea to extend LBP and combine it
with a feature selection method.
[6]
[7]
[8]
[9]
[10]
[11]
[12]
7 REFERENCES
[13]
[1]
[2]
[3]
[4]
[5]
[14]
[15]
[16]
[17]
491
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 486-492
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
LDA-based compound distances," Pattern
Recognition, vol. 41, pp. 3442-3451, 2008.
[18] L. S. Oliveira, R. Sabourin, F. Bortolozzi, and C.
Y. Suen, "Feature selection using multi-objective
genetic algorithms for handwritten digit
recognition," in Pattern Recognition, 2002.
Proceedings. 16th International Conference on,
2002, pp. 568-571.
492
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 493-498
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
1. ABSTRACT
For large area surveillance, a number of UAVs are
preferred. In such situation, navigation becomes a
prime factor, which reduces the chance of collision.
UAVs primarily depend on GPS, which has an order of
inaccuracy. Conventional correction methods like
differential GPS needs extra hardware and cannot
improve stochastic errors like multipath error and
receiver error. In this paper we demonstrate a novel
technique to reduce the GPS error, in a swarm of
UAVs.
2. KEYWORDS
GPS Error, Navigation Error, GPS Error Reduction,
UAV Swarm, Position Correction, Error offset, Intra
Swarm Communication;
3. INTRODUCTION
A group of cooperative UAVs, also called a
swarm of UAVs sometimes very effective for
certain tasks like mapping, target searching etc.
Portability,
lesser
take-off
and
landing
requirements, easy handling along with low
manufacturing and maintenance costs make it
more desirable than expensive and large UAVs.
However, the low cost and small size principle
puts a limit to the onboard sensors, resulting in a
significant inaccuracy in navigation with others.
IMU (Inertial Measurement Unit) is a popular
solution for such problem, but precise IMUs are
expensive and heavy. On the other hand, GPS,
which is a less expensive solution, comes with a
lower accuracy. Commercial GPS provides an
accuracy of the order 15 to 100 meters.
#
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 493-498
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
measurement is improved using the error in
relative distance between every member UAV in
the swarm.
4. INTRA-SWARM COMMUNICATION
To reduce the error in position measurement by
correlating the GPS data with other members in
the swarm, it is necessary to determine the relative
distance between the members as well as to get the
GPS provided data for every UAV.
Figure 2: A timeline showing the various delays in packet
transmission
Where,
is response pulse receiving time
is request pulse transmitting time and
is predefined delay time.
Therefore, distance between two UAVs is,
(
)
( )
where
4.2. Navigation Error Deduction
494
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 493-498
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
where
and
are distance measured and
distance calculated respectively. Errors between
all UAVs are calculated by the above equation.
For a network of n number of UAVs,
number of error equations can be formed and error
for each UAV can be calculated.
5. CORRECTION
( )
( )
th
where
is the sum of the distances of the j
UAV from all the UAVs and
is its distance
th
from the i UAV. From the above calculation the
middle UAV (whose value of
is minimum) is
determined. Once this is done, the criterion is
applied as follows:
( )
( )
5.2. Preliminary Calculation
Before the correction process, certain preliminary
calculations must be done in order to facilitate
them. As we are considering only multipath error
( )
( )
( )
(7)
Where is the unit vector in the direction of
and ( ) is the error from the calculated distance
equation between and .
Since ( ) is always less than GPS error radius
( ) for any pair of members
and , there will
exist a region of overlap of the two probability
distribution functions, which collectively give rise
to a local sum distribution function as shown in
Figure 4.
495
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 493-498
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
5.4. Correction Stage II
Stage II correction is performed on the middle
UAV ( ) after all the members have corrected
centers ( ) after execution of stage I.
New error values are calculated between the
calculated distances between members corrected
centers( ), and their measured distances
( )
( )
( )
( )
( )
Let there be a point represented by the position
vector with respect to ( )
( )
( )
Figure 4: The figure shows the reconstruction of the
probability distribution function of the location of
near
and the region of overlap
( )
{
(
(
( )
)
( ))
( )
( )
(
(
( ( )
))
( ))
( )
( )
After the stage II correction, the error in location
has been reduced to its minimum as reducible by
our algorithm. Hence, this location is considered
to be the true location.
5.5. Correction Stage III
Now that
( )
( )
Where
( )
Where
is the direction angle of
as provided
by the antenna, and and are the unit vectors
along the positive and axes.
If is the position vector of the location of
as calculated from the current centre, then the
position vector of its accurate location will be
given by:
()
( )
Therefore, the GPS center of each non-middle
member
of the swarm is shifted by () to
complete the error reduction process for the whole
swarm.
496
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 493-498
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
6. TEST RESULTS
In order to test the algorithms accuracy, five GPS
receivers were placed at different locations (whose
location coordinates are known) across the Amity
University campus and the distances between
these locations were calculated. The data was
provided to the algorithm for error reduction. The
results obtained are as shown in figures 5, 6 and 7.
497
International Journal of Digital Information and Wireless Communications (IJDIWC) 4(4): 493-498
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2225-658X)
Location
Indicated
position
True position
Corrected
position
77
77
77
77
77
77
77
77
77
7
7
77
Abs.
error
3.
77
77
77
7
7
77
77
4.
5.
7. CONCLUSION
6.
7.
8.
9.
10.
11.
12.
8. ACKNOWLEDGEMENT
This research was done during post graduation
phase, as a part of UAV Systems in Institute of
Space Science and Technology, Amity University
Uttar-Pradesh, INDIA.
13.
9. REFERENCES
1.
2.
498