Вы находитесь на странице: 1из 522

International Journal of Innovative Computing, Information and Control

Editor-in-Chief

Peng Shi, Faculty of Advanced Technology, University of Glamorgan, Pontypridd, CF37 1DL, United Kingdom (E-mail: pshi@glam.ac.uk)

Co-Editor-in-Chief

Jeng-Shyang Pan, Department of Electronic Engineering, National Kaohsiung University of Applied Sciences, Kaohsiung 807, Taiwan (E-mail: jspan@cc.kuas.edu.tw)

Executive Editor

Yan Shi, School of Industrial Engineering, Tokai University, 9-1-1, Toroku, Kumamoto 862-8652, Japan (E-mail: yshi@ktmail.tokai-u.jp)

Advisory Board

Ramesh Agarwal, USA Chin-Chen Chang, Taiwan Tom Heskes, Netherlands Jerry M. Mendel, USA Witold Pedrycz, Canada Junzo Watada, Japan Naoyoshi Yubazaki, Japan

Associate Editors

Michael V. Basin, Mexico Liya Ding, Macau Tzung-Pei Hong, Taiwan Hsiang-Cheh Huang, Taiwan Bin Jiang, China Lei Li, Japan Magdi Mahmoud, Saudi Arabia Jagadis C. Misra, India Sing Kiong Nguang, New Zealand Pavel Pakshin, Russia Xiangshi Ren, Japan Kok-lay Teo, Australia Yuanqing Xia, China Jianqiang Yi, China

2011 ICIC INTERNATIONAL

Steve P. Banks, UK Hanfu Chen, China Lakhmi C. Jain, Australia Masaharu Mizumoto, Japan Bion L. Pierson, USA Xiaopeng Wei, China Lotfi A. Zadeh, USA

Shyi-Ming Chen, Taiwan Vasile Dragan, Romania Chao-Hsing Hsu, Taiwan Yunfu Huo, China Katsuari Kamei, Japan Bin-Yih Liao, Taiwan Anatolii Martynyuk, Ukraine Saeid Nahavandi, Australia Ngoc Thanh Nguyen, Poland Charles E. M. Pearce, Australia Hirofumi Sasaki, Japan Shaocheng Tong, China Mingqing Xiao, USA Lixian Zhang, China

ISSN 1349-4198

Tianyou Chai, China Wai-Chi Fang, USA V. Lakshmikantham, USA Chengsheng Pan, China Jianrong Tan, China Takeshi Yamakawa, Japan

Ozer Ciftcioglu, Netherlands Hugang Han, Japan Xiangpei Hu, China Gerardo Iovane, Italy Jien Kato, Japan Wenyuan Liu, China Angelo Miele, USA Masahiro Nakano, Japan Nikos Nikolaidis, Greece Fuji Ren, Japan M. N. S. Swamy, Canada Shuoyu Wang, Japan Edwin Engin Yaz, USA Lindu Zhao, China

PRINTED IN JAPAN

IJICIC

ISSN 1349-4198

Volume 7, Number 3, March 2011

International Journal of Innovative Computing, Information & Control

Sponsored by

Science Communication Institute, China Dalian University, China

Published by ICIC International http://www.ijicic.org

International Journal of Innovative Computing, Information and Control

Volume 7, Number 3, March 2011

CONTENTS

The Improvement of all-Digital Amplitude-Locked Loop Separation Analysis Combined MIMO System Gwo-Jia Jong and Gwo-Jiun Horng

1001

Value Based Intelligent Requirement Prioritization (VIRP): Expert Driven Fuzzy Logic Based Prioritization Technique Muhammad Ramzan, M. Arfan Jaffar and Arshad Ali Shahid

1017

Generalization of Proxy Signature Based on Factorization Cheng-Chi Lee, Tzu-Chun Lin, Shiang-Feng Tzeng and Min-Shiang Hwang

1039

A Novel High Capacity 3D Steganographic Algorithm Meng-Tsan Li, Nien-Ching Huang and Chung-Ming Wang

1055

Real-Time Decentralized Neural Block Control: Application to a Two DOF Robot Manipulator R. Garcia-Hernandez, E. N. Sanchez, E. Bayro-Corrochano, V. Santibanez and Jose A. Ruz-Hernandez

1075

A Supervising Authenticated Encryption Scheme for Multilevel Security Chien-Lung Hsu, Liang-Peng Chang and Tzong-Chen Wu

1087

Aesthetic Emotions in Human-Robot Interaction. Implications on Interaction Design of Robotic Artists Catalin Buiu and Nirvana Popescu

1097

Modeling, Design and Experiment of Improved Super-Mini Underwater Robot Xin Song, Zao-Jian Zou and Jia-Wei Ye

1109

Design and Approximation Capabilities Analysis of Time-Variant Fuzzy Systems Degang Wang, Wenyan Song and Hongxing Li

1121

Encoding Prior Knowledge into Data Driven Design of Interval Type-2 Fuzzy Logic Systems Chengdong Li, Jianqiang Yi and Tiechao Wang

1133

Generation and Application of Decision Rules within Dominance-Based Rough Set Approach to Multicriteria Sorting Liping An, Zengqiang Chen and Lingyun Tong

1145

Automatic Classification of Driving Mental Fatigue with EEG by Wavelet Packet Energy and KPCA-SVM Chunlin Zhao, Chongxun Zheng, Min Zhao, Jianping Liu and Yaling Tu

1157

3D Triangle Mesh Compression Based on Vector Quantization with k -Ring Vector Prediction Lin-Lin Tang, Zhe-Ming Lu, Fa-Xin Yu, Ping-Hui Wang and Zhen Li

1169

(

Continued )

2011 ICIC INTERNATIONAL

ISSN 1349-4198

PRINTED IN JAPAN

International Journal of Innovative Computing, Information and Control

Volume 7, Number 3, March 2011

CONTENTS ( Continued )

An Efficient Simulated Annealing with a Valid Solution Mechanism for TDMA Broadcast Scheduling Problem Gon Kim, Sung Soo Kim, Il-Hwan Kim, Dong Hoi Kim, V. Mani and Jae-Ki Moon

1181

Scalable and Systolic Dual Basis Multiplier over GF(2 m ) Liang-Hwa Chen, Po-Lun Chang, Chiou-Yng Lee and Ying-Kuei Yang

 

1193

A

Fuzzy-Based Rough Sets Classifier for Forecasting Quarterly PGR in the Stock

 

Market (Part II) Ching-Hsue Cheng and Yoush-Yang Chen

 

1209

Re-Evaluation of Adaptive

X

Control Charts: A Cost-Effectiveness Perspective

1229

 

Yan-Kwang Chen, Hung-Chang Liao and Hsu-Hwa Chang

 

Inventory Management in a (Q,r) Inventory Model with Two Demand Classes and Flexible Delivery

 

1243

 

Shuo Huang, Bo Liao and David J. Robb

 

A

Novel Text Watermarking Algorithm Using Image Watermark

 

1255

Zunera Jalil, M. Arfan Jaffar and Anwar M. Mirza

 

An Efficient and Secure Three-Pass Authenticated Key Agreement Elliptic Curve Based Protocol

 

1273

 

Zeyad Mohammad, Chien-Lung Hsu, Yaw-Chung Chen and Chi-Chun Lo

 

Indicator of the Quality of the Sensor Set Up: A Study Using Surface EMG on Sub-Band ICA

 

1285

 

Ganesh R Naik and Dinesh K Kumar

 

H Filtering for Markovian Jump Systems with Time-Varying Delays Jinliang Liu, Zhou Gu and Songlin Hu

 

1299

Surface Texture Characterization of Fibers Using Fractional Brownian Motion Model

1311

 

Jiun-Jian Liaw, Chuan-Pin Lu and Lin-Huang Chang

 

Power Grid Node and Line Delta Centrality Measures for Selection of Critical Lines

 

in

Terms of Blackouts with Cascading Failures

 

1321

Hwachang Song, Rodel D. Dosano and Byongjun Lee

 

Optimizing MIMO-SDMA Smart Antennas by Using Linear Array Phase Perturbations Based on Particle Swarm Optimization

1331

 

Chao-Hsing Hsu

An Environmental Visual Features Based Navigation Method for Autonomous Mobile Robots

 

1341

 

Fairul Azni Jafar, Yasunori Suzuki, Yuki Tateno, Kazutaka Yokota and Takeshi Matsuoka

 
 

(

Continued )

2011 ICIC INTERNATIONAL

ISSN 1349-4198

PRINTED IN JAPAN

International Journal of Innovative Computing, Information and Control

Volume 7, Number 3, March 2011

CONTENTS ( Continued )

Observer-Based Iterative Learning Control with Evolutionary Programming Algorithm for MIMO Nonlinear Systems

1357

 

Yan-Yi Du, Jason Sheng-Hong Tsai, Shu-Mei Guo, Te-Jen Su and Chia-Wei Chen

Reversible Data Hiding for VQ Indices Using Prediction Errors

1375

 

Zhi-hui Wang, Chin-Chen Chang, Huynh Ngoc Tu and Ming-Chu Li

A

Novel Technique for Designing Decentralized Stabilizers for Robust Control

in

Power Systems Using an H Criterion Alexandre C. Castro, Jose M. Araujo, Eduardo T. F. Santos, Fabricio G. S. Silva and Clivaldo S. de Araujo

1387

A

Study of Three Novel Line-Based Techniques for Multi-Target Selection

1397

Jibin Yin and Xiangshi Ren

A

Novel Modal Series Representation Approach to Solve a Class of Nonlinear

Optimal Control Problems

1413

 

Amin Jajarmi, Naser Pariz, Ali Vahidian Kamyad and Sohrab Effati

Adaptive Frame Length Method for Hardware Context-Switching in Dynamic Partial Self-Reconfigurable Systems

1427

 

Trong-Yen Lee, Che-Cheng Hu, Yang-Kun Huang and Chia-Chun Tsai

An ID-Based Access Control in a Hierarchical Key Management for Mobile Agent

1443

 

Chia-Hui Liu, Yu-Fang Chung, Tzer-Shyong Chen and Sheng-De Wang

Meaningful Shadow Based Multiple Gray Level Visual Cryptography without Size Expansion

1457

 

Pei-Yan Pai, Chin-Chen Chang, Yung-Kuan Chan and Chi-Cheng Liao

Service-Oriented Routing and Clustering Strategies for Vehicle Infotainment Dissemination

1467

 

Chenn-Jung Huang, Chin-Fa Lin, Ching-Yu Li, Che-Yu Lee, Heng-Ming Chen Hung-Yen Shen, You-Jia Chen and I-Fan Chen

An Optimal Fuzzy Controller for a High-Performance Drilling Process Implemented Over an Industrial Network

1481

 

Rodolfo E. Haber, Agustin Gajate, Steven Y. Liang, Rodolfo Haber-Haber and Raul M. del Toro

New Secret Key Traitor Tracing Scheme with Dispute Settlement from Bilinear Maps

1499

Kuo-Yu Tsai, Tzong-Chen Wu and Chien-Lung Hsu

2011 ICIC INTERNATIONAL

ISSN 1349-4198

PRINTED IN JAPAN

International Journal of Innovative Computing, Information and Control

AIMS AND SCOPE OF IJICIC

The primary aim of the International Journal of Innovative Computing, Information and Control (IJICIC) is to publish high-quality papers of new developments and trends, novel techniques and approaches, innovative methodologies and technologies on the theory and applications of intelligent systems, information and control. The IJICIC is a peer-reviewed English language journal and is published monthly.

The scope of the IJICIC is in the broad areas of intelligent systems, information and control, includes but not limited to:

1. Intelligent Systems and Innovative Computing

Artificial intelligence and expert systems, approximate reasoning, fuzzy systems and rough sets, fuzzy optimization and decision making, evolutionary computation, intelligent information systems, image and speech signal processing, human-machine learning, neural networks, knowledge discovery, data mining, clustering and data analysis, decision support systems, intelligent fault detection and identification, knowledge-based systems, hybrid intelligent systems, soft computing, management science

2. Information and Control

Information theory and applications, optimal, adaptive, predictive, robust, fuzzy and reliable control and filtering on linear and nonlinear systems, stochastic systems, hybrid systems, supervisory systems, guidance and navigation systems, real-time systems, discrete event systems, signal and image processing, system identification, optimization techniques, robustness analysis, learning systems, fault detection, robotics and motion control, manufacturing systems, power systems, networked control systems, environmental and bio-engineering, complex and large scale systems, bioinformatics, control education, transportation systems

SUBSCRIPTION

Library/Institutional/ Individual Rates:

Hard copy (Printing): JPY66,000 CD: JPY33,000 Hard copy with CD: JPY90,000

The charge includes postage, packing, handling and surface delivery of all issues.

IJICIC Editorial Office Tokai University, Kumamoto Campus 9-1-1, Toroku, Kumamoto 862-8652, Japan Phone: +81-96-386-2666 Fax: +81-96-386-2666 E-mail: office@ijicic.org

2011 ICIC INTERNATIONAL

ISSN 1349-4198

PRINTED IN JAPAN

International Journal of Innovative Computing, Information and Control Volume 7, Number 3, March 2011

ICIC International c 2011 ISSN 1349-4198

pp. 1001–1015

THE IMPROVEMENT OF ALL-DIGITAL AMPLITUDE-LOCKED LOOP SEPARATION ANALYSIS COMBINED MIMO SYSTEM

Gwo-Jia Jong and Gwo-Jiun Horng

Department of Electronic Engineering National Kaohsiung University of Applied Sciences Chien Kung Campus 415, Chien Kung Road, Kaohsiung 807, Taiwan gjjong@cc.kuas.edu.tw

Received July 2009; revised June 2010

Abstract. The mobile communication is often interfered by any type of noises. The all digital phase-locked loop (ADPLL) system has been successfully used for decades in order to track the carrier phase of a frequency modulation (FM) signal. In this paper, we com- bine the ADPLL and the all-digital amplitude-locked loop (ADALL) system structure for modulation signals of the separation co-channel transmission system. It is demonstrated co-channel separation signals from FM system with the results of simulation experiments. The least mean square (LMS) theory is introduced more efficiently to eliminate the noise interference and to separate the multi-channel signals. This paper is also shown and demonstrated the multi-channel signal in FM system which is combined the ADPLL and the ADALL algorithms with an adaptive finite impulse response (FIR) filter. The multi- channel transmission with the additive white Gaussian noise (AWGN) interference has been solved by using the proposed algorithm. The demodulated signals are operated and demodulated by the LMS algorithm. The separation signals of the multi-channel can be obtained the prefect performance and process for high speed and low cost. The proposed system of this paper is the virtue and advantage to implement the communication security field. In general words, it can be substituted and replaced the encryption (or decryption) system. Keywords: ADPLL, ADALL, FM, FIR, LMS, Communication security, Encryption

1. Introduction. Over the last decades, digital communication has become one of the basic technologies for out modern life. Only when using digital transmission, information can be transported with moderate power consumption, high flexibility and especially over long distance, with much higher reliability than by using conventional analog modulation. Thus, the communication world has been going digital [1]. When regarding digital transmission, we have to consider two dominant impairments. Firstly, the signal is corrupted by additive white Gaussian noise environment, which can be thermal noise of the receiver front-end or crosstalk caused by other users transmitting in the same frequency band. Secondly, the transmission media is dispersive. This situation can be described as a linear system with some specific transfer function, where attenuation and phase coefficients are varied with frequency. This property causes the frequency components to be affected differently. The channel transmission signal is distorted for the transmitted pulses in the time domain [2-6]. The ADPLL system has been successfully used for decades in order to track the car- rier phase of an FM signal. It is proposed to analysis suppress the in-band interference problem efficiently. The ADPLL demodulated outputs contain large in-band spikes and get some unintelligible turbulence. In this paper, a novel separation algorithm is pro- posed for canceling co-channel interference (CCI) communication system by the ADALL

1001

1002

G.-J. JONG AND G.-J. HORNG

method. This technique is adopted to eliminate the interference of the ADPLL in wire- less transmission channel. It includes to solve many extrinsic factors such as inter-symbol interference (ISI), multi-path interference, flat-fading channel and AWGN channel. The ADALL system can separate the dominant and the subdominant signals each other. This scheme is obtained got the low complex computation and compressed the CCI effect.

2. A Digital Procedure of All-Digital PLL. Phase-locked loops (PLL) are widely used in modern electronic systems. The applications include synchronization, tracking demodulation, ranging, etc. The continued progress in increasing performance, speed and the simultaneous in size and cost of integrated circuits have resulted in strong interest in the implementation of the ADPLL, which can be made more flexible, versatile, compact, stable and reliable than its analog counterpart. The classic PLL consists of three major function units: a phase detector (PD), a loop filter (LF) and a voltage-controlled oscillator (VCO). For ADPLL, the voltage-controlled oscillator is named numerically-controlled oscillator (NCO) or digitally-controlled oscilla- tor (DCO) [7]. The synchronization concept plays a major role in the fields of digital data transmission, radar and navigation system. The phase detection is an important topic in the synchro- nization problems of digital communication systems. Based on the maximum likelihood estimation principle, the conventional analog PLL system has been developed to treat the nonlinear phase detection problem. The ADPLL has been also studied over year. However, this conventional ADPLL is a gradient-based searching algorithm to solve the highly nonlinear problem, it may lead to a local result. Furthermore, the conventional ADPLL is sensitive to the initial conditions, variation of parameters [8].

A. Basic Model of All-Digital Phase-Locked Loop. A block diagram of ADPLL and its phase model are shown in Figures 1 and 2, respectively [7]. The loop’s output signal is desired and named by the phase error φ = θ i θ o .

named by the phase error φ = θ i − θ o . Figure 1. Block

Figure 1. Block diagram of ADPLL

θ i − θ o . Figure 1. Block diagram of ADPLL Figure 2. Phase model

Figure 2. Phase model of ADPLL

The discrete input and output signals of Figure 1 can be written as

THE IMPROVEMENT OF ADALL SEPARATION ANALYSIS

1003

(2)

where ω 0 is the central angular frequency of the input signal to be tracked, V i (k) and V o (k) are the maximum amplitudes of the signals sampled by k. The phase detector considered is a multiplier whose output is given by

V o (k) = V o cos [ω 0 Tk + θ o (k)] ,

k = 0, 1, 2,

V d (k) = K d V i (k) V o (k)

K d V i V o

=

2

{sin [θ i (k) θ o (k)] + sin [2ω o Tk + θ i (k) + θ o (k)]} ,

k = 0, 1, 2,

(3)

where K d is represented the phase detector gain. Equation (3) is defined a second-harmonic term added to a nonlinear function of ϕ (k). It is a common assumption that the second-harmonic term of (3) and Figure 2 will be filtered by the discrete-time loop filter. The discrete-time loop filter is considered and characterized by a first-order transfer function, shown as:

(4)

Since the order of loop is the order of its loop filter plus one. The ADPLL is a second-

order loop whose dynamics can be described by the following difference equation:

C 2

1

z 1

D(z) = C 1 +

ϕ (k + 2) 2ϕ (k + 1) + ϕ (k) = θ (k + 2) 2θ (k + 1) + θ (k) rK 1 sin [ϕ (k + 1)]

+ K 1 sin [ϕ (k)] ,

k = 0, 1, 2,

(5)

digitally-controlled

oscillator gain. When the phase error is small, then the approximation sin ϕ ϕ is valid and the sinusoidal nonlinear can be removed. ADPLLs are being implemented in the coherent communication systems for a long time in carrier regenerating and clock time recovery purposes. The theory and the design techniques of an ADPLL are well known. ADPLLs are drawing the attention of research workers till now because they are basically nonlinear feedback control loops in the discrete time domain and their dynamical behavior is not completely known [9].

where K 1 = K o K d C 1 , r = 1 + C 2 /C 1 and K o is represented the

3. All-Digital Amplitude-Locked Loop of CCI.

A. Frequency Modulation (FM). A narrowband digital FM with frequency detection

has been widely used. A conventional [10-12] PLL has been often operated as a frequency detector. The PLL detector (PLD) has the native frequency tracking ability as a tracking filter and is possible to track very rapid center frequency variation of the input FM signal such as Doppler shifts from low earth orbit (LEO) satellites. We adopt the FM scheme to analyze the FM separation, it can be rewritten as follows:

y (t) = A c cos [ω c t + ϕ 1 (t)] + mA c [cos ω c t + ϕ 2 (t)] = r (t) cos [φ 1 (t)] (6)

where

(7)

ω d t = ϕ 2 (t) ϕ 1 (t)

B. Co-channel Interference under AWGN Channel. Typical methods for reducing

co-channel FM interference, it estimates the stronger signal’s instantaneous frequency amplitude and phase. CCI is the interference due to the mixture of signals with similar carrier frequencies. It is necessary to find efficient techniques to reduce the harmful effects of CCI in FM analogue or digital communication system. The mobile users often operate in the presence of cumbersome interference along with multi-path, Rayleigh fading channel and AWGN channel that leads to signal distortion

1004

G.-J. JONG AND G.-J. HORNG

and signal fading at the receiver. The interference between signals from these cells is called CCI. The co-channel signals are transmitted by the same carrier frequency f c in the AWGN channel. Accordingly, the complex representation of the received signal is defined as

y(t) = y 1 (t) + my 2 (t)

=

A c

= A

c

cos [ 2πf c t + 2πk f

cos [2πf c t + θ 1 (t)]

t m 1 (τ )] + mA c cos [ 2πf c t + 2πk f t m 2 (τ )]

0

0

+ mA c cos [2πf c t + θ 2 (t)]

and

θ i (t) = 2πk f t m i (τ )dτ,

0

i = 1, 2

(8)

(9)

The parameter m is defined as the interference to carrier ratio (ICR). It is denoted in Equation (8) and the vector diagram is shown in Figure 3.

y(t) = A c {cos [2πf c t + θ 1 (t)] + mA c cos [2πf c t + θ 2 (t)]} = r(t) cos φ 1 (t) (10)

where

(11)

ω d t = θ 2 (t) θ 1 (t)

(11) ω d t = θ 2 ( t ) − θ 1 ( t )

Figure 3. Phase diagram of phase modulation with noise

From Figure 3, it can be found a(t) = mA c cos ω d t, b(t) = mA c sin ω d t for the following equations:.

r(t) = [A c + a(t)] 2 + b 2 (t)

= 2A c a(t) + a 2 (t) + b 2 (t)

A

2

c

+

 

A

2

c

+

2

cos ω d t + A

2

m 2 cos 2 ω d t + A

2

m 2 sin 2 ω d t

c

c

c

= 2mA

= A c 1 + 2m cos ω d t + m 2

(12)

THE IMPROVEMENT OF ADALL SEPARATION ANALYSIS

φ(t) = tan 1

b(t)

A c + a(t) = tan 1

mA c sin ω d t ω d t = tan 1 A c + mA c cos

φ 1 (t) = 2πf c t + θ 1 (t) +

φ(t)

m sin ω d t

1 + m cos ω d t

1005

(13)

(14)

In this paper, the FM transmitted signals are operated at the same carrier frequency, the transmission channel results in the co-channel interference (CCI). For the convenient reason, we adopt the ADPLL FM model for the FM scheme to analyze the FM separation system. In wireless transmission channel system, it is included many extrinsic noise interference factors such as flat-fading channel, AWGN channel, multi-path channel, etc. This com- posite signals go through ideal band-pass filter (BPF) completely and the wideband noise naturally because the narrownband noise. The band-limited noise is obtained in terms of in-phase and quadrature components and represented by

n(t) = n I (t) cos(2πf c t) n Q (t) sin(2πf c t) (15)

where n I (t) and n Q (t) are the in-phase and quadrature noise components of n(t), respec- tively. Equivalently, we may express n(t) in terms of its envelope and phase as

(16)

where

n(t) = r n (t) cos [2πf c t + ϕ(t)]

and

r n (t) = n I (t) + n Q (t)

2

2

ϕ n (t) = tan 1 n n Q I (t) (t)

n ( t ) = tan − 1 n n Q I ( t ) (

Figure 4. ADPLL FM model with noise

The phase diagram of phase modulation with noise signal is shown as Figure 4. The mathematical representation of this two modulated co-channel FM signals under AWGN channel can be described as:

y n (t)

= s(t) + i(t) + n(t)

= A c cos [2πf c t + θ 1 (t)] + mA c cos [2πf c t + θ 2 (t)] + r n (t) cos [2πf c t + ϕ n (t)]

= r(t) cos φ(t) + r n (t) cos

= R(t) cos φ 2 (t)

[2πf c t + ϕ n (t)]

φ 2 (t) = 2πf c t + θ 1 (t) + φ(t) + φ n (t)

(17)

(18)

1006

where

and

G.-J. JONG AND G.-J. HORNG

ω d t = θ 2 (t) θ 1 (t)

˙

˙

ω d = θ 2 (t) θ 1 (t)

ω n t = ϕ n (t) θ 1 (t) φ(t)

˙ ˙

ω n = ϕ n (t) θ 1 (t) φ˙ (t)

(19)

(20)

n ( t ) − θ 1 ( t ) − φ ˙ ( t )

Figure 5. System-level depiction of ALL application

From Figure 5, it can be found c(t) = r n (t) cos ω n t, d(t) = r n (t) sin ω n t

R(t) = [r(t) + c(t)] 2 + d 2 (t)

= r 2 (t) + 2r(t)r n (t) cos ω n t + r n (t) cos 2 ω n t + r n (t) sin 2 ω n t

2

2

= r 2 (t) + 2r(t)r n (t) cos ω n t + r n (t)

2

φ n (t) = tan 1

d(t)

r n (t) sin ω n t

r(t) + c(t) = tan 1 r(t) + r n (t) cos ω n t

(21)

(22)

The envelope and phase of the receiver are represented (21) and (22), respectively. The phase mathematical representation of two modulated co-channel FM signals under AWGN channel for ADPLL model is shown as Figure 5.

C. ADALL Analysis in the Discrete-Time. The conventional amplitude-locked loop (ALL) configuration is a feedback-based control system, the dual of the familiar phase- locked loop (PLL). The ALL functions have been provided for applications such as tuning the Q-factor in integrated bandpass filters, noise cancellation, suppression of co-channel FM interference and demodulation of suppressed carrier signals [1]. More recently, ALLs have begun to adopt in high data rate wireless communication systems. Many modulation schemes used to achieve high data rates have non-constant envelopes, creating stringent linearity requirements for power amplifiers (PAs) [13]. To proceed with the ADALL analysis, we replace each blocks of Figure 5. It is demon- strated the system-level depiction of ALL application for the frequency-domain equivalent, individually. It is shown and represented as Figure 6. Here A DES is the desired amplitude input to the comparator and A OUT is represented by a summing junction and the gain block Kc. The loop filter is assumed to have a transfer function F (s), the variable gain amplitude (VGA) is assumed to have a linear gain characteristic with a slope of K V GA , and the peak detector is approximated as having a one-pole response with a low-frequency gain of K PD . In this ADALL analysis, we assume that the amplitude of the input signals to the K V GA and A IN , remains constant and thus is incorporated into the gain of the

THE IMPROVEMENT OF ADALL SEPARATION ANALYSIS

1007

THE IMPROVEMENT OF ADALL SEPARATION ANALYSIS 1007 Figure 6. Frequency-domain analysis model of ADALL The transfer

Figure 6. Frequency-domain analysis model of ADALL

The transfer function of Figure 6 for the frequency-domain for of the system is defined as:

A OUT (s)

(s) = H(s) =

A DES

KF (s)(1 +

ω

s

PD

)

s

1 + ω PD + KF(s)

(23)

where it is assumed K PD = 1 and K = K C A IN K V GA . It is also expressed the loop gain of the system as:

(24)

KF(s)

s

G(s) =

1 +

ω PD

We substitute in the transfer function for various choices of loop filters, and examine their effects on the stability and transient response of the ALL. Consider the case for the loop filter is a simple gain, which can be incorporated into the constant K. From (24), we consider the loop is unconditionally stable and from (23) we express the closed-loop transfer function as:

H(s) =

K

1 + K

1 +

s

ω

PD

s

1 + ω PD (1 + K)

(25)

From this, we see that the setting time will be determined by the pole of the peak detector, ω PD , in conjunction with the gain factor K. It is also clear that the output will have a steady-state amplitude error, determined by the size of K. As K increases, the steady-state amplitude error reduced. To eliminate the steady-state amplitude error, we replace the simple gain with an integrator. From (24), we see that the loop remains unconditionally stable; however, it may be necessary to add a zero to the loop filter to maintain a desired phase margin. From (23), it is replaced the new closed-loop transfer function as:

PD (1 +

s

ω

PD

)

(26)

s 2 + PD + where we have again incorporated the loop filter gain into the constant K. By equating the denominator to the standard form of the second-order characteristic equation, we can solve the damping factor and the natural frequency of the loop:

H(s) =

PD

ξ

= 1

2 ω PD ,

K

ω n = PD

(27)

Using (27), we design the loop to achieve the desired loop dynamics. More complex loop filter choices are also possible, and can be analyzed for stability and transient behavior in a manner to that carried out above.

1008

G.-J. JONG AND G.-J. HORNG

D. The Separation Signal for the ADPLL and ADALL Analysis for Adaptive FIR Filter. The separation system of transmission communication system block model is shown in Figure 7. The adaptive FIR filter is proposed for detecting the variation of the dominant and subdominant signals. Both of dominant and subdominant signals are through the ADPLL and ADALL separation model. We developed and derived new algorithm for the discrete and digital mathematical representations of the ADPLL and ADALL discrete output signals.

of the ADPLL and ADALL discrete output signals. Figure 7. The separation system for the ADPLL

Figure 7. The separation system for the ADPLL and ADALL analysis for adaptive FIR filter

It is demodulated by an ideal ADPLL demodulator for the output signal f ADP LL (τ ). The detail derivation mathematical functions are shown as the following equations and Appendix 1 [35].

f ADALL (τ) =

· τ ) 2 ) × 1 + 4m · cos (

1

· τ ) , 2 ·

m (m 1) 2 ) ×

2ω d · τ ) 2

1

×

×

2 · ( 1 cos (

1

2m + m 2

2ω d

2ω d

(m 1) 2

1 2m + m 2

sin (

2 · ω d · τ )

1

EllipticE ( cos (

1 + 4m · cos (

2ω d · τ ) 2

1

2m + m 2

τ

The dominant signal is derivative as follows:

1

f ADP LL (τ) × f ADALL2 (τ) =

2

ω d · di log(m · e iω d τ + 1) · s 1 (τ)

+

1

2

) · f ADALL1 (τ) · s 2 (τ)

ω d · di log (

e d τ

m

(28)

(29)

THE IMPROVEMENT OF ADALL SEPARATION ANALYSIS

1009

The subdominant signal is derivative as follows:

f ADP LL (τ) × f (ADALL1) (τ)

1 2 f ADALL1

(τ)

= ω d · di log(m · e iω d τ + 1) ·

f ADALL2 (τ) · s 1 (τ)

+

1 2 ) · f (ADALL1) (τ) · f ADALL1 (τ)

e d τ

m

f ADALL2 (τ)

ω d · di log (

·

s 2 (τ)

(30)

4. Multi-Channel FM Signals with Multi-Input Multi-Output (MIMO) Model for Adaptive FIR Filter ll-Digital Amplitude-Locked Loop of CCI. We present the theory and application of optimum linear filters and predictors. It is concentrated on linear filters for the optimum in the sense of minimizing the mean square error (MSE). The minimum MSE (MMSE) criterion leads to a theory of linear filtering that is elegant and simple, involves only second-order statistics and is useful in many practical applications. The optimum filter is designed for a given set of second-order moments can be used for any realizations of stochastic processes with the same moments [25]. The MIMO communication techniques have been an important area of focus for next- generation wireless systems because of their potential for high capacity, increased diversity and interference suppression. For applications such as wireless and mobile communica- tions, MIMO systems will likely be deployed in environments where a single base be communicated with many users simultaneously [30]. Wireless systems with transmit and receive diversity have become a very attractive area of research due to the potential of achieving extraordinarily high data bit rates and high spectral efficiency. In this section, we suggest a systematic design procedure to the adaptive equalization problem for time-varying channels [26]. For given adaptive finite impulse response (FIR) MIMO channels with maximum al- lowable average input powers, the transmitter and receiver adaptive FIR MIMO filters are jointly optimized such that the MSE between the desired and reconstructed signals is minimized [29]. It is well known that in a MIMO system with n T transmit and n R receive antennas. The communication channel capacity grows linearly with min (n T , n R ). It is also desirable to transmit the same information-bearing signal over several channels. The mode of transmission is used primarily in situations where there is high probability that one or more of the channels will be unreliable from time to time [32]. The MIMO channel with n T transmitters and n R receivers is typically represented as a matrix H of dimensions n R × n T , where each of the coefficients [H] i,j represents the transfer function from the j th transmitter to the i th receiver. We denote the signal or symbol transmitted from the j th transmitter y j , and collect all such symbols into an n T dimensional vector y. With this notation, the matrix model of the channel is

(31)

s = Hy + w

where w is a vector of additive noise with a variance of σ n and s is the vector of received data, with an element in w and s for each receive antenna. In a point-to-point MIMO link (single-user case), all outputs are available to the user for processing. In the multi-user case, the n R receivers are distributed among different users; for example, if each user has only one antenna, each user has access to only one element of s. H is the narrowband transmission matrix of the MIMO channel, the complex correlation coefficient between antenna elements i and j at the transmitter or at the receiver side is

2

1010

given in accordance with as:

G.-J. JONG AND G.-J. HORNG

ρ T x ,R x i,j

=

R

T x ,R x

i,j

σ

T x ,R

i

2

x

T x ,R

j

σ

2

x

(32)

between the interacting

antenna elements i and j at the transmitter or at the receiver and the respective variances

σ

matrices R Tx and R Rx as follows:

. Antenna correlations at the T x and R x sides are combined in correlation

Antenna correlation ρ T x R x depends on antenna covariance R

i,j

T x R

i

2

x

and σ

T x R

j

2

x

T x R x i,j

R T x ,R x =

ρ T x R x

11

ρ T x R x

21

.

.

.

ρ T x R x

n1

···

···

.

···

···

···

.

.

.

···

ρ T x R x

1n

ρ T x R x

2n

.

ρ T x R x nn

n T x ,R x ×n T x ,R x

(33)

For this reason, R T x and R R x explicitly determine the correlations within the regarded channel model. Subsequently, the eigen-analysis method is used to compute the eigen-

values λ k with k = 1,

gains of the respective subchannels, where [·] H denotes Hermitian transpose. Each chan- nel matrix is given by n T x transmit and n R x receive antennas, it offers K min ( n T x , n R x ) parallel subchannels. It is also provided a uniform power allocation at the T x antenna array, the information theoretic MIMO channel capacity may be evaluated as a sum of the singular capacity values for each of the K subchannels:

, K of the instantaneous matrix HH H that represent the power

C =

K

k=1

log 2 ( 1 + λ k P k )

σ

2

n

(34)

In (34), P k denotes the amount of power assigned to the k th subchannel and σ n is the total noise power at the receiver. The signal-to-noise ratio (SNR) amount to:

2

SNR = E [P T x ]

σ

2

n

(35)

where E [·] is represented the expectation operator to compute the expectation of the total signal power P T x at the transmitter. An important tool for characterizing any communication channel is capacity. In a single- user channel, capacity is the maximum amount of information that can be transmitted as a function of available bandwidth given a constraint on transmitted power. In single-user MIMO channels, it is common to assume that there is a constraint on the total power broadcast by all transmitted antennas. For the multi-user MIMO channel, the separation signal model for the ADPLL and ADALL analysis with MIMO system.

5. Simulated Performance. The CCI problem could be become a serious impairment for any mobile communication system. This simulation results demonstrate the per- formance of the ADALL combined with the ADPLL separation model using the novel adaptive FIR filter. This simulations present to verify how co-channel FM signal can be separated. The simulations of sinusoidal and mobile signals are indicated to improve the separation performance capability comparing the ADALL and novel separation system models [1-3,13]. The proposed co-channel FM signal demodulation and separation systems with noise interference environment consists the following subsystems: the FM signal generator, the ADPLL combined ADALL model and the adaptive FIR filter with noise canceller. Figure

THE IMPROVEMENT OF ADALL SEPARATION ANALYSIS

1011

8 is shown the dominant and subdominant signals are separated using proposed model at m = 0.1. The simulations demonstrate the performance of the proposed co-channel signal separation model design with AWGN channel based on the adaptive FIR filter algorithm. The MMSE criteria using the step size and the optimum weight coefficient values are defined by Wiener-Hopf with steepest descent method. The demodulated output signals are approximately approached to the desired signals. The impairments of the channel noise affect the MMSE value. By assuming the input SNR ratio ranged between 20dB and 20dB step 10dB, it is operated at the random statistic process in this paper.

is operated at the random statistic process in this paper. Figure 8. The separation dominant and

Figure 8. The separation dominant and subdominant signals for adaptive FIR filter at m = 0.1

Two important improvement performance cases are discussed in simulation results. The proposed algorithm of this paper can be suited to implement the perfect separation performance index for both of the low SNR and the large ICR conditions for the proposed algorithm. The simulated performances can be presented the ADPLL with ADALL model for the adaptive FIR filter better then ADPLL with ADALL for the butterworth filter. The performance of MMSE via SNR ratio is shown as Figure 9 by varying different m values. In Table 1, we summarize the MMSE of proposed and conventional methods with dominant signal by varying ICR (i.e. m) and the SNR ratios.

with dominant signal by varying ICR (i.e. m ) and the SNR ratios. Figure 9. The

Figure 9. The performance of MMSE via SNR ratio

1012

G.-J. JONG AND G.-J. HORNG

Table 1. The MMSE with dominant signal for SNR via ICR

(a) Proposed method for ADPLL and ADALL

❳ ❳ ❳ ❳ ❳ ❳ ❳❳

ICR

SNR (dB)

20dB

10dB

 

0dB

10dB

20dB

 

0.1

0.0017

0.0009

0.00015

0.000098

0.000011

 

0.5

0.0029

0.0012

0.00023

0.000105

0.000023

 

0.9

0.0047

0.0014

0.00030

0.000160

0.000037

 

(b) Traditional method for PLL and ALL

 

❳ ❳ ❳ ❳ ❳ ❳ ❳❳

ICR

SNR (dB)

20dB

10dB

0dB

10dB

20dB

 

0.1

0.3705

0.1855

0.0391

0.0108

0.0015

0.5

0.4294

0.2421

0.0502

0.0115

0.0032

0.9

0.5053

0.2999

0.0546

0.0125

0.0061

6. Conclusions. In this paper, the proposed algorithm combined the ADPLL with the ADALL system for all-digital analysis that is more perfect to separate the CCI or multi- channel FM signals in the same carrier frequency environment. It is also found to suppress the distortion that caused by the AWGN channel transmission. The new combined all-digital method can achieve the expectative goal by modifying the conventional ALL system (i.e. analog ALL) with the adaptive filter algorithm. For CCI FM signal with AWGN interference, the dominant and subdominant signals have been separated completely. The proposed algorithms recover the multiple original or modulating signals and suppressed the distortion caused by conventional all analog ALL or PLL systems [1]. It also solve the co-channel or multi-channel transmission problems for suppressing the AWGN interference in the noisy transmission channel. The major contribution of paper adopts to separate co-channel and multi-channel transmissions by using MIMO system. In this paper, the performance of the fully digital separation system can be obtained more effectively. The advantage of the adaptive FIR filter is proved to obtain the better tracking ca- pability and all of the weights are updated to MMSE value. We compare ADPLL and ADALL with all analog PLL and ALL for channel SNR ratio between 20dB and 20dB, the performance index difference is approximately 260 times. The proposed method can be adopted to the monitor and security systems for the signal intercepted applications for the future. The proposed system of this paper is the virtue and advantage to implement the security field. In general words, it is substituted and replaced the encryption and decryption systems but operated at no multi-access communication systems to save the cost and complex circuits.

Appendix 1.

y [n] =

τ

K d

1

n

τ

n1

φ(τ )+ e L [n] =

τ

K d

1

n

τ

n1

2πf c τ + θ 1 (τ) + ϕ(τ) + φ n (τ )

f ADP LL (τ) = [2πf c τ + θ 1 (τ) + ϕ (τ) + φ n (τ )] ·

= πf c τ 2 + θ 1 (τ )+ [ 2 log ( e iw d τ ) +

1

1

2

w d · di log(m · e iw d τ + 1)

(A.1)

(A.2)

THE IMPROVEMENT OF ADALL SEPARATION ANALYSIS

1

2

1

2

1

4

1

2

w d τ 2 ]

+

w d · di log (

e iw d τ

m

) +

w d · log (e iw d τ + m) · log (

1

2

w n · di log ( r n (τ)·e iw n τ

r(τ)

) +

e iw d τ

m

)

+ [ 1 2 log ( e iw n τ )

w n · di log ( r(τ)·e iw n τ

r n (τ)

)

1

2

+ w n · log (r (τ ) · e iw n τ + r n (τ )) · log ( r(τ)·e iw n τ

r n (τ)

)

1

4

w n τ 2 ]

f ADP LL (τ) =

2 1 log ( e iw d τ ) + πf c τ 2 + θ 1 (τ )

1

4

w n · log (e iw n τ ) 2

1013

1

1

2

2

+

+

+

+

w d · di log (

m·e iw n τ

r(τ)

)

1

2

w n · di log ( r n (τ)·e iw n τ

r(τ)

)

1

2

w n · log (m · e iw n τ + r (τ )) · log (

m·e iw n τ

r(τ)

)

w n · log (r (τ ) · e iw n τ + r n (τ )) · log ( r(τ)·e iw n τ

m

)

1

2

w d · di log (m · e iw d τ + 1) +

1

2

w d · di log (

e iw d τ

m

)

1

2

w d · log (e iw d τ + m) · log (

e iw d τ

m

)

1

4

w d τ 2 ]

f ADP LL1 (τ) =

f ADP LL2 (τ) =

1

2

ω d · di log(m · e iω d τ + 1) · s 1

1

2

ω d · di log (

e d τ

m

) · s 2

REFERENCES

(A.3)

(A.4)

(A.5)

[1] A. M. Pettigrew, The Amplitude-Locked Loop the Theory and Operation of the FM 201/5 High Performance FM Demodulator, Ampsys Company Document, 1996. [2] G. J. Jong, T. J. Su and T. J. Moir, The performance of the amplitude-locked loop with co-channel interference, Electronics Letters, vol.34, no.8, pp.719-720, 1998. [3] G. J. Jong, T. J. Moir, A. M. Peettigrew and T. J. Su, Improvement of FM demodulator with co-channel FM interference, Electronics Letters, vol.35, no.20, pp.1758-1759, 1999. [4] J. E. Ohlson, Phased-locked loop operation in the presence of impulsive and gaussian noise, IEEE Trans. on Communications, vol.Com21, no.9, pp.991-996, 1973. [5] T. J. Moir and A. M. Pettigrew, Inband noise canceling in FM systems: The white noise case, Electronics Letters, vol.28, no.9, pp.814-815, 1992. [6] B. C. Sarkar, Phase error dynamics of a first-order phase locked loop in the presence of co-channel tone interference additive noise, IEEE Trans. on Communications, vol.38, no.7, pp.962-965, 1990. [7] Y. L. Dong and G. Y. He, On spurious phase modulation caused by second-harmonic terms for digital phase-locked loop, Proc. of the 6th International Conference on ITS Telecommunications, pp.1180-1183, 2006. [8] M.-T. Liu, A Fuzzy Rule-based DPLL for Maximum Likelihood Phase Estimation in Signal Trans- mission, Master Thesis, National Tsing Hua University, 1995.

1014

G.-J. JONG AND G.-J. HORNG

[9] T. Banerjee and B. C. Sarkar, Phase error dynamics of a class of DPLLs in presence of cochannel interference, Signal Processing, vol.85, no.6, pp.1139-1147, 2005. [10] W. Li and J. Meiners, Introduction to phase-locked loop system modeling, Analog Applications Journal, pp.5-10, 2000. [11] S. Haykin, Communication System, 4th Edition, Wiley, 2001. [12] G. J. Jong, T. J. Moir and A. M. Pettigrew, Co-channel FM interference suppressed by using the amplitude-locked loop for digital communication, Electronics Letters, vol.34, no.8, pp.719-720, 1998. [13] C. T. Charles and D. J. Allstot, Amplitude-locked loop analysis for calibration applications, Elec- tronics Letters, vol.42, no.11, pp.614-615, 2006. [14] Y.-J. Chen, G.-J. Horng and G.-J. Jong, The separated speech signals combined the hybrid adaptive algorithms by using power spectral density and total harmonic distortion, Multimedia and Ubiquitous Engineering, pp.825-830, 2007. [15] Y.-J. Chen, G.-J. Horng, G.-J. Jong and T.-J. Su, The amplitude-locked loop system of co-channel interference in AWGN channel, Proc. of the 6th Annual Conference on Wireless Telecommunications Symposium, Pomona, CA, 2007. [16] G.-J. Jong, P.-J. Liao, C.-Y. Jung and T.-J. Su, Multi-channel interference separation for the AWGN channel, Intelligent Signal Processing and Communication Systems, pp.581-584, 2005. [17] J. Gaspar, S. F. Chen, A. Gordillo, M. Hepp, P. Ferreyra and C. Marqu´es, Digital lock in amplifier:

Study, design and development with a digital signal processor, Microprocessors and Microsystems, vol.28, no.4, pp.157-162, 2004. [18] Y. Iwanami, A DSP DPLL demodulator with sequence estimator for CP-FSK signals in the presence of large Doppler shift, Universal Personal Communications, vol.1, pp.351-356, 1998. [19] Y. Iwanami, A DSP DPLL demodulator with sequence estimation for MSK signal in the presence of Doppler shift on Rician fading channel, Personal, Indoor and Mobile Radio Communications, Wireless Networks-Catching the Mobile Future, vol.1, pp.332-337, 1994. [20] B. C. Sarkar and S. Chattopadhyay, Symmetric lock-range multilevel quantized digital phase locked FM demodulator, IEEE Trans. on Communications, vol.38, no.12, pp.2114-2116, 1990. [21] A. Jameel, M. Y. Siyal and N. Ahmed, Transform-domain and DSP based secure speech communi- cation system, Microprocessors and Microsystems, vol.31, no.5, pp.335-346, 2007. [22] I. Galton, Analog-input digital phase-locked loops for precise frequency and phase demodulation, IEEE Trans. on Circuits and Systems II: Analog and Digital Signal Processing, vol.42, no.10, pp.621- 630, 1995. [23] T. Banerjee and B. C. Sarkar, Phase error dynamics of a class of modified second-order digital phase-locked loops in the background of cochannel interference, Signal Processing, vol.85, no.8, pp.1611-1622, 2005. [24] J. G. Proakis, Digital Signal Processing, John Wiely & Sons, pp.258-272, 2000. [25] D. G. Manolakis and V. K. Ingle, Statistical and Adaptive Signal Processing: Spectral Estimation, Signal Modeling, Adaptive Filtering, and Array Processing, Artech House Inc., 2005. [26] H. M. Karkhanechi, E. A. Fain and B. C. Levy, An investigation of carrier recovery techniques for space-time coded MIMO wireless systems, Wireless Communications and Networking Conference, vol.1, pp.454-458, 2002. [27] T. K. Yeung and S. F. Yau, A multi-input multi-output adaptive FIR filter with guaranteed conver- gence for feedforward ANC system, Proc. of the 1999 IEEE International Symposium on Circuits and Systems, vol.3, pp.17-20, 1999. [28] A. Maleki-Tehrani, B. Sayyarrodsari, B. Hassibi, J. How and J. Cioffi, Estimation-based synthesis of H -optimal adaptive equalizers over wireless channels, Proc. of GLOBECOM, vol.1A, pp.457-461,

1999.

[29] A. Hjorungnes, Minimum MSE transmitter and receiver FIR MIMO filters for multi-user uplink com- munications, Proc. of IEEE International Conference on Acoustics, Speech, and Signal Processing, vol.4, pp.497-500, 2004. [30] Q. H. Spencer, C. B. Peel, A. L. Swindlehurst and M. Haardt, An introduction to the multi-user MIMO downlink, IEEE Communications Magazine, vol.42, no.10, pp.60-67, 2004. [31] O. Klemp and H. Eul, Analytical approach for MIMO performance and electromagnetic coupling in linear dipole arrays, Wireless Communication Systems, pp.586-590, 2005. [32] J. G. Proakis, Digital Signal Processing, John Wiely & Sons, pp.709, 2000. [33] M. Hassani, Approximation of the dilogarithm function, Journal of Inequalities in Pure and Applied Mathematics, vol.8, no.1, pp.3-10, 2007.

THE IMPROVEMENT OF ADALL SEPARATION ANALYSIS

1015

[34] E. S. Ginsberg and D. Zaborowski, The dilogarithm function of a real argument, Communications of the ACM, vol.18, no.4, pp.200-202, 1975. [35] G.-J. Horng, All-Digital Amplitude-Locked Loop Separation System for Adaptive FIR Filter Anal- ysis, Master Thesis, Institute of Electronic Engineering National Kaohsiung University of Applied Sciences, 1997. [36] C.-H. Hsu, Downlink MIMO-SDMA optimization of smart antennas by phase-amplitude perturba- tions based on memetic algorithms for wireless and mobile communication systems, International Journal of Innovative Computing, Information and Control, vol.5, no.2, pp.443-460, 2009. [37] H.-F. Huang, C.-W. Chan, C.-H. Lin and H.-W. Wang, A low-computation conference key system for mobile communications, International Journal of Innovative Computing, Information and Control, vol.5, no.2, pp.461-466, 2009. [38] K. Yamada, K. Kimura, H. Yuki and K. Yoshida, The home network system by mutual complement of wireless and wired communications, ICIC Express Letters, vol.2, no.1, pp.73-79, 2008. [39] H. Chen, Q. Ding, L. Ding and X. Dong, Experimental study on secure communication of different scroll chaotic systems with identical structure, ICIC Express Letters, vol.2, no.2, pp.201-206, 2008.

International Journal of Innovative Computing, Information and Control Volume 7, Number 3, March 2011

ICIC International c 2011 ISSN 1349-4198

pp. 1017–1038

VALUE BASED INTELLIGENT REQUIREMENT PRIORITIZATION (VIRP): EXPERT DRIVEN FUZZY LOGIC BASED PRIORITIZATION TECHNIQUE

Muhammad Ramzan, M. Arfan Jaffar and Arshad Ali Shahid

Department of Computer Science National University of Computer and Emerging Sciences (FAST-NU) A. K. Brohi Road, H11/4 Islamabad, Pakistan { muhammad.ramzan; arfan.jaffar; arshad.a.shahid }@nu.edu.pk

Received July 2009; revised November 2009

Abstract. Requirement Prioritization is a very critical but often neglected area of re- quirement engineering. Experience has shown that without proper prioritization of re- quirements presented by various stakeholders, the end product usually fails to meet its objectives optimally. In fact in many instances, the product is considered a failure be- cause it fails to meet its core objectives. Several requirement prioritization techniques have been presented by various researchers over the past years. Working with these tech- niques has exposed several limitations when applied in software projects. In this paper, we have presented a novel multi-level value based intelligent requirement prioritization technique using fuzzy logic. We have introduced and applied the concept of requirement value to prioritize requirements. We have performed extensive experimentation using our proposed technique along with existing techniques. Results have shown that our technique has achieved superior prioritization results and consistency. The experiments have also shown that proposed technique is capable of delivering impressive prioritization under various circumstances. Keywords: Requirement engineering, Requirements prioritization, Fuzzy systems, In- telligent requirements prioritization

1. Introduction. Software Engineering is one of the youngest engineering domains which

emerged as recently as somewhere in the middle of 1980 as an accepted engineering disci- pline. The aim of SE is to create software products, services or their artifacts in order to meet the requirements posed by stakeholders while meeting quality constraints imposed on them. In order to meet both these objectives, any software development derives its pur- pose and meaning from the requirements posed by various stakeholders. In this context, better elicitation, modeling and analysis of requirements plays a very critical role towards development of a quality software. Requirement Engineering is an established domain of knowledge within software engineering which establishes practices and principles for effective requirement elicitation, modeling, specification, documentation, etc. One very important but often neglected practice of software requirement engineering is requirement prioritization. Requirement prioritization is the process of establishing worth and value of various requirements posed by multiple stakeholders based on certain established cri- teria of their utility for the ultimate software product. Several requirement prioritization techniques have been presented by authors. These techniques are both quantitative and qualitative in their nature. Some well known requirement prioritization techniques include Analytical Hierarchy Process (AHP), Cumulative Voting, Numerical Assignment, Rank- ing, Theory W, Requirement Triage and Wieger’s Method, etc. And there are several other techniques which we shall discuss in this paper.

1017

1018

M. RAMZAN, M. A. JAFFAR AND A. A. SHAHID

Requirement prioritization enables us to understand the significance of requirements vis-`a-vis the system to be developed and among requirements as well. With requirement prioritization, we can identify the focus areas which need most of our attention in order to develop a product which optimally meets the requirements of the stakeholders. In most of the situations, due to budget and time constraints, it becomes impossible to implement all the requirements posed by stakeholders. Also the nature of many projects is such that requirements are implemented in a staged environment. In both of these scenarios, we need requirement prioritization. We can prioritize requirement to realize which requirements can be delayed or altered so that other urgent requirements can be implemented and to what degree. We can also use requirement prioritization to determine which requirements to be implemented in earlier stages or later stages. We have been working with several funded projects during our research. These projects are faced with both of the above mentioned situations. We have found it very important to prioritize requirements in their true sense in order to develop a meaningful and successful product. Requirement prioritization was a new practice in our specific development environment. So, our project needs required us to study further into various requirements prioritiza- tion techniques so that we can select one which can best suit our peculiar development environment. Our finding was that there is severe deficiency of any experimental results to determine which technique to prefer. Consequently, during this period of research and development, we studied various requirement prioritization techniques and tried to imple- ment them on experimental level at various projects. We soon realized that all of these techniques worked well within certain situations but had some inherent problems attached with them which made it impossible to implement any one of these across the organization for all different kinds of projects. The main hindrances faced by us while implementing these techniques were related to cost, time and handling of evolving and creeping require- ments. It was very important for us to work with a time and cost efficient prioritization technique. Since we were testing these techniques for small to medium software projects which needed a very efficient prioritization mechanism, our ultimate aim was to decide upon a technique which yielded better accuracy in requirement prioritization while at the same being efficient in terms of cost and time. We also were in search of a technique which could reduce the dependence on human factor due to certain obvious reasons which we shall elaborate in later sections. Unfortunately any of existing techniques could not satis- factorily answer all these problems. If some techniques like cumulative voting or ranking were efficient in terms of cost and time, their results were not very impressive. On the other hand, better techniques like AHP required very skilled human resource and a lot of capital to be applied in a proper environment. In order to overcome these problems, one solution before us was to develop an artificially intelligent expert driven requirement prioritization technique. In one of our previous works, we had presented the initial sketch of a “value based requirements prioritization” technique [52]. This technique was very much similar to Theory W. In this approach, the end users and experts were asked to prioritize their requirements based upon the value that accomplishment of this requirement may have for the system. The salient feature of this technique was a combination of end users and experts in the process of requirement prioritization process. However, while implementing this technique, we encountered two major problems.

The technique generated a lot of conflicts at the end of requirement prioritization process. Conflict resolution was a very long and time consuming process which needed to be followed at the end of every prioritization session.

VALUE BASED INTELLIGENT REQUIREMENT PRIORITIZATION

1019

The technique was completely manual. The prioritization was done through human endeavor and element of human bias was noticeable. While applying the technique proposed in [48], we realized the need for an automated requirement prioritization technique in order to overcome both the above mentioned lim- itations. In this paper, we have proposed and elaborated upon a fuzzy logic based intel- ligent requirement prioritization technique. This technique uses fuzzy logic to prioritize requirements presented by various stakeholders. This modified scheme is basically a mul- tilevel prioritization where end users, experts and intelligent system perform requirement prioritization at various levels. By using this technique, we achieved multiple benefits. On one hand, we achieved certain degree of automation which ultimately resulted in a more efficient technique. On the other hand, we were also able to reduce the emphasis of our approach on skilled requirement engineer solely as now intelligent component as well as end user was also contributing towards ultimate prioritization. At the same time developing this technique also offered us with the opportunity of eliminating the role of requirement engineer altogether since we are currently working on incorporating a neural network based component to replace engineer during second stage of prioritization. In order to establish the utility of intelligent requirement prioritization technique, we applied this as well as a representative group of other techniques on several projects and determined the degree of success. We have also presented these findings and observations in this paper. Based on these, we have established the case for an artificially intelligent hybridized technique on which we are currently working. We have also presented a frame- work for future evaluations of results acquired by application of our proposed mechanism with existing requirements prioritization techniques. We have presented preliminary re- sults of application of this framework on various techniques. The paper is arranged as follows: After a brief introduction in Section 1, we have described a comprehensive literature review in Section 2. In Section 3, we have presented an elaborate overview of existing requirement prioritization techniques. In Section 4, we have described the concept and implementation of fuzzy logic based intelligent requirement prioritization technique. These techniques have been applied on several projects to observe their performance. Findings and evaluations of these experiments have been presented in Section 5. Conclusion and future work is described in Section 6.

1.1.

Major contributions. Following are the major contributions:

In this paper, we have proposed and elaborated upon a fuzzy logic based intelligent requirement prioritization technique. This technique uses fuzzy logic to prioritize requirements presented by various stakeholders.

This modified scheme is basically a multilevel prioritization where end users, experts and intelligent system perform requirement prioritization at various levels.

In order to establish the utility of intelligent requirement prioritization technique, we applied this as well as a representative group of other techniques on several projects and determined the degree of success. We have also presented these findings and observations in this paper. Based on these, we have established the case for an artificially intelligent hybridized technique on which we are currently working.

We have also presented a framework for future evaluations of results acquired by ap- plication of our proposed mechanism with existing requirements prioritization tech- niques. We have presented preliminary results of application of this framework on various techniques.

1.2.

Paper organization. The paper is arranged as follows: After a brief introduction

in Section 1, we have described a comprehensive literature review in Section 2. In Section

1020

M. RAMZAN, M. A. JAFFAR AND A. A. SHAHID

3, we have presented an elaborate overview of existing requirement prioritization tech-

niques. In Section 4, we have described the concept and implementation of fuzzy logic based intelligent requirement prioritization technique. These techniques have been ap-

plied on several projects to observe their performance. Findings and evaluations of these experiments have been presented in Section 5. Conclusion and future work is described

in Section 6.

2. Literature Review. Requirement Engineering (RE) is one of the earliest and very critical phases of software engineering. RE as a knowledge stream is basically aimed

at acquisition, modeling and documentation of requirements for the software product.

Requirement Engineering is a unique discipline in the sense that it not only incorporates

the concepts of engineering but also of human and social sciences. Sometimes, referred

to as requirements analysis, RE is treated as a sub discipline of system engineering and

software engineering. Requirement engineering aims to define precisely the requirements that need to be met. This is not an ordinary task. According to Fred Brooks, deciding what needs to be built is the most difficult part of software development [1]. We can visualize one software requirement as one documented need that software product should accomplish. Usually requirements are classified as either as process based and product based or functional and non functional requirements. Software requirement can best be defined as the description of system functionality along with its quality concerns. Requirement prioritization is the next logical task once requirements have been elic- itated and properly analyzed. In most cases, it is really difficult to meet all the re- quirements that have been given by various stakeholders. Most of the times, elicitated requirements are vague, conflicting or out rightly false. Over period of time, as our un- derstanding of the system becomes more and more clear, the requirements start attaining their actual or specific shape. Similarly, in many cases, requirements are implemented in a staggered fashion. In such circumstances, it becomes important to arrange the require-

ments in a prioritized order to develop the system in more realistic way. This task becomes even more difficult when performed early in the lifecycle. According to Karlsson et al. [2], one of the greatest problems of software engineers is development of such a product which doesn’t satisfy the needs and expectations of stakeholders. To overcome this problem, same authors came up with the idea of prioritizing the requirements according to their value in the paper [3] titled “A Cost-Value Approach for Prioritizing Requirements”. Sub- sequently, many other researchers [4,5] emphasized upon the significance of requirement prioritization. According to Ed Yourdon, prioritization of requirements is an extremely important issue [6] where as Lubars et al. stated that prioritization of requirements was one major topic of discussion during the survey that they undertook[7]. There are various methods for prioritizing requirements [8]. Some major techniques are Analytical Hierarchy Process (AHP) [9-11], Binary Search Trees [12-15], 100 points method [16,17], planning game [18,19], numerical assignment technique [20,21] and theory

W 20 [22], etc. The consensus of all of these studies is that the project’s success or failure

is strongly determined by how effectively, we can prioritize the requirements. Computational Intelligence and soft computing are established techniques which have helped resolve many real world problems. These methodologies include Artificial Neural

Networks, Fuzzy Logic, and Evolutionary Computing etc. Fuzzy logic is a technique centered on fuzzy set theory. Thus it is considered as an extension of classical set theory [23,24]. The concept of fuzzy sets as introduced by Lotfi Zadeh [25] can be considered

as generalization of the classical sets which are crisp in their nature. The purpose of

fuzzy logic is to both reduce the complexity of existing solutions as well as increasing the accessibility of control theory [26]. Computational intelligence based techniques including

VALUE BASED INTELLIGENT REQUIREMENT PRIORITIZATION

1021

fuzzy logic have been widely used recently to tackle many real world problems. Some of the recent applications of computational intelligence can be found in [27,28]. Fuzzy logic has also found its way in software engineering where it has most recently been used in effort estimation [29], software project similarity [30], software development [31], project evaluation [32], software evolution [33] etc. Fuzzy logic has been used in requirement engineering as well for various tasks [34,35]. Chengdong Li et al. [36] presented a novel approach for using prior knowledge and sampled data in fuzzy systems. Some other recent developments in the domain of fuzzy logic [37-39] have also presented new vistas of research in software engineering. What is evident after studying extensively is the fact that researchers in software engineering need to apply artificial intelligence in various domains of knowledge of SE to propose such techniques which are evolvable and can intelligently generate efficient results. In this paper, we have proposed another application of fuzzy logic in software engineering. We have proposed to use fuzzy logic in the domain of requirement engineering. We suggest introducing fuzzy logic to determine the priority of requirements. The next section is devoted to elaboration of this proposed technique. In this paper, we proposed another application of fuzzy logic in the domain of re- quirement engineering. We suggest introducing fuzzy logic to determine the priority of requirements. The next section is devoted to elaboration of this proposed technique.

3. Requirement Prioritization Techniques: An Overview. As mentioned in the literature review, there are various requirement prioritization techniques. However, no evaluation of these techniques has been made so far so that their utility and relevance can be determined. In this section, we give a comprehensive overview of various requirement prioritization techniques.

3.1. Analytical hierarchy process (AHP). AHP is a relative assessment based sta- tistical technique to prioritize requirements for software products. If we have n number of requirements, AHP makes nx(n 1)/2 comparisons at each hierarchy level. In real life, we are usually working with requirements which have multiple objectives. AHP works as an efficient technique in these kinds of situations by making pair wise comparison to calculate relative value and cost of each requirement against the other one. This sig- nificantly large number of comparisons makes the technique less effective as increase in number of comparisons always takes place at the rate of O(n 2 ). AHP is regarded as a five step method.

1. Establish completeness of requirements.

2. Apply the pair-wise comparison method to assess the relative value.

3. Apply the pair-wise comparison method to assess the relative cost.

4. Calculate each candidate requirement’s relative value and implementations cost, and plot each on a cost-value diagram.

5. Use the cost-value diagram as a map for analyzing the candidate requirement.

A sample cost value diagram is shown in Figure 1. A large number of studies have been made in recent past to determine the effectiveness of AHP for requirements prioritization. Karlsson [3,40] has made a number of studies which have shown the effectiveness of this technique in industrial settings. At the same time, some other studies [19,41] have talked about AHP as being difficult, less efficient and time consuming. AHP can be considered as a highly sophisticated and complex technique which can establish prioritization at the level of individual requirements. Efforts have been made to reduce the number of comparisons. However, this has al- ways enhanced the margin of error. In our opinion, this tradeoff is necessary since some comparisons may actually never bee needed.

1022

M. RAMZAN, M. A. JAFFAR AND A. A. SHAHID