International Journal of

Intelligent Engineering & Systems

http://www.inass.org/

Research on the Interference Cancellation Based on Adaptive Algorithms

Jing Dai∗ , Zhonghua Han, Feng Zhang

Faculty of Information and Control Engineering, Shenyang JianZhu University, Shenyang, China
∗ Corresponding author’s Email: [email protected]

Abstract: The adaptive interference cancellation algorithms(AICA) is proposed in this paper. The Matlab simulation
results of the performance of six different kinds of AICA are presented. According to the comparison of the simulation
results, especially the usage of resource, convergence rate and effects of cancellation, the NLMS adaptive interference
cancellation algorithm and frequency-domain block LMS algorithm are selected, which are suitable for the further
design on FPGA. By simulating the functions by Modelsim, and analyzing the PSD curve and the constellation of
output signals, the adaptive interference cancellation algorithm system based on frequency-domain block LMS have
achieved expected effect. The frequency-domain block LMS adaptive echo interference cancellation system has better
flatness on the in-band PSD curve and better convergence effect on the constellation. Therefore, frequency-domain
block LMS adaptive interference cancellation algorithm is chosen as the design with high performance-price ratio.
The more points there are, the more obvious advantages it will perform.
Keywords: Adaptive algorithms; Interference cancellation; NLMS; Frequency-domain block LMS; FPGA

1. Introduction ously. It is urgent that more complex, larger amount of

high-speed data signals be processed on FPGA chip [2].
In modern mobile communication systems, a mo-
In mobile communication system, repeater plays an
bile communication receiver should have a wider fre-
essential role in the coverage and optimization of net-
quency range and real time signal processing capa-
works. The weak received signals will first be am-
bilities in complex signal environment. An adaptive
plified by a repeater, and then are emitted to poten-
digital filter is the best option that can improve the re-
tial coverage area by the transmitting antenna, so as
ceiver sensitivity under the condition of low signal to
to realize the extension and coverage of communica-
noise ratio (SNR). Adaptive filters can process data
tion networks. However, due to the same-frequency
without any prior information about signal environ-
transmitting characteristics, without taking effective
ment and the received signals. The study of adaptive
measures, the signals emitted by the transmitting an-
digital filters with high-speed data processing capa-
tenna will be caught by the receiving antenna from the
bilities is currently an important issue in signal pro-
same repeater some time later, which may lead to the
cessing fields [1]. In recent years, the rapid advance-
strong interference to the effective signals, which is
ment in digital technologies has supported the imple-
echo interference. The interference will have a bad
mentation of sophisticated digital signal processing
influence on the communication. It may even cause
algorithms for real-time applications. Especially, with
the impossibility of communication in severe case. To
the development of FPGA (Field Program Gate Ar-
deal with the problem of echo interference, traditional
ray), many signal processing algorithms can be imple-
methods involve keeping some distance between the
mented, which are theoretically reasonable but could
receiving antenna and the transmitting antenna when
not be implemented for the hardware design previ-

International Journal of Intelligent Engineering and Systems, Vol.5, No.4, 2012 18

installing the repeater so as to effectively isolate sig- lation of echo interference is fulfilled. If only the
nals. With the development of digital signal process- transfer function of adaptive filter can fit well with the
ing technology and large scale programmable devices, transfer function of the echo channel, reflecting the
the adaptive echo interference cancellation technol- actual echo channel more truthfully, it will be possi-
ogy has become the effective method to cancel echo ble to realize the complete cancellation of echo inter-
interference in repeater systems. It can figure out the ference from receiving signals and get communication
problem of spatial isolation of a repeater; meanwhile, signals of high-quality. Therefore, the key of realizing
it can greatly improve the quality of communication. echo interference cancellation is adaptive filter.
This study focuses on the WCDMA repeater, under
the environment of MATLAB. It carries out the model
building and simulation research on interference can-
cellation system on the basis of LMS (Least Mean
Squares) adaptive filtering algorithm [1][3], NLMS
(normalized LMS) adaptive filtering algorithm[4], sym-
bolic LMS adaptive filtering algorithm, delay LMS
adaptive filtering algorithm, frequency-domain block
LMS adaptive filtering algorithm[5], and RLS (Recur-
sive Least Squares) adaptive filtering algorithm. In
the light of the easiness of hardware implementation
and high-speed processing, comparisons are made be-
tween six kinds of adaptive filtering algorithm [6][7]. Figure 1 The principle diagram of echo interference can-
Finally, the NLMS adaptive filtering algorithm and the cellation of adaptive filter
frequency-domain block LMS adaptive filtering algo-
rithm are chosen for further hardware architecture de- An adaptive digital filter can filter the time variant
sign and programming realization. And, the perfor- interference signals which exist in the communication
mance of the achieved system is verified. channels; thereby high quality communication signals
can be obtained. Many researchers focus on the devel-
2. Basic Principles of Adaptive Interference opment of adaptive algorithm in mobile communica-
Cancellation tion systems. Figure 2 shows the block diagram of an
adaptive digital filter, where d(n) is a desired signal,
Echo interference is the kind of time-varying, un-
y(n) filter’s output signal, and error e(n) adjusts the
known and multipath interference. This characteristic
adaptive digital filter coefficients in a weight vector
requires the filter can automatically track the changes
ω.
of signals, and respond to the changes by adjusting the
In order to determine the optimal weight vectors,
weight coefficient quickly. What is more, this just fits
some algorithms have been studied. Among all the al-
well with the functions of the adaptive filter [2]. Thus,
gorithms, Least Mean Square (LMS) is the most com-
the adaptive filter is what to be chosen for the echo in-
mon. The others are the improvement based on the
terference cancellation. The principles of adaptive fil-
LMS algorithm.
tering echo interference cancellation is shown in Fig-
ure 1.
In Figure 1, the signals emitted by the transmitting
antenna will produce echo interference after multipath
echo channel. The echo interference, superposed with
the weak signals from the base station, enters into the
receiving antenna and causes the echo interference to
the repeater. The adaptive filtering algorithm uses the
signals from the transmitting antenna as reference to
adjust the weight coefficient of the filter so that the
interference estimated value of echo channel is sim- Figure 2 An adaptive filter block diagram
ulated by the filtering processing unit. Then the esti-
mated value is subtracted from the receiving signals
superposed with echo interference, thus the cancel-

International Journal of Intelligent Engineering and Systems, Vol.5, No.4, 2012 19

3. Adaptive Algorithm Analysis ψ is used to modify the algorithm to avoid the possible
large step size which may result that the NLMS algo-
3.1 LMS algorithm
rithm will be diverged when xH (n)x(n) is very small.
In 1960 Widrow and Hoff presented the LMS al- As can be seen from equation (5), the step size µ (n)
gorithm. LMS algorithm was studied based on Min- varies adaptively by following the changes in the input
imum Mean Square Error (MMSE) and the steepest signal level. This prevents the update weights from
descent algorithm. From LMS algorithm we know diverging and makes the algorithm more stable and
that faster converging than when a fixed step size is used.
n Figure 3 shows the MSE (Mean square error) plot
y(n) = ∑ ω (i)x(n − i + 1) = ω T X(n). (1) or convergence for the LMS and the NLMS algorithm
i=1 respectively, which have the same filter orders. As
Where y(n) is the output at time n. The input sig- shown in Figure 3, for the same adaptation size or iter-
nal x(n) is multiplied with the coefficients in a weight ations, the NLMS algorithm can achieve much faster
vector ω . ω (n + 1) is estimated according to the fol- convergence than the LMS algorithm.
lowing recursion:

ω (n + 1) = ω (n) + µ x(n)e∗ (n), (2)

Where µ is a small positive constant, called the step

size, which controls system stability and convergence
rate. Because µ is a constant, convergence rate is
slower, and misadjustment coefficient δ is still big- Figure 3 MSE plot for the LMS and NLMS algorithm re-
ger when the system reaches steady state. δ could be spectively
expressed as:

δ = µ MPin . (3) 3.3 Recursive least squares algorithm

The Recursive Least Squares (RLS) algorithm is ba-
M is filter order, Pin is input signal power. As can be sed on the least square criterion; difference square
seen from Equation (3), when the step size µ , input sum is to be minimized between reference signals and
signal power Pin , or filter order M is large, the misad- output signals of filter as the ultimate objective. The
justment coefficient δ is large. LMS algorithm can not iterative equation of filter can be expressed as equa-
achieve both rapid convergence and small steady-state tion (6).
error simultaneously.
1 −1
ω (n) = ω (n − 1) + R (n − 1)e(n)x(n), (6)
3.2 Normalized LMS algorithm λ
Normalized LMS algorithm is an improved algo- where 0 < λ ≤ 1 is genetic factor, R(n −1) = E{x(n−
rithm of LMS. NLMS algorithm will be easy to im- 1)xT (n − 1) the autocorrelation matrix at time n − 1 of
plement with faster convergence speed and accurate input signals.
tracking capability [8]. Equation (3) shows misadjust-
ment coefficient δ is constant if µ and Pin are inversely 3.4 Delay LMS algorithm
proportional to the change. Thus NLMS algorithm
Considering the slow convergence process, and the
will be converged faster than LMS algorithm. (n + 1)
contradiction between step factor and convergence rate,
is estimated according to the following simple recur-
Long and Herzberg proposed the delay LMS (DLMS)
sion [8]:
algorithm which is derived from the traditional LMS
µ̂ algorithm. Comparing with LMS algorithm, the de-
ω (n + 1) = ω (n) + e∗ (n)x(n), (4)
xH (n)x(n) + ψ layed cycle number d in figure 4 used in updating the
coefficients of DLMS algorithm is to generate the co-
µ̂ efficient at the next clock. Thus, the throughput of
= µ (n). (5)
xH (n)x(n) + ψ DLMS algorithm data is twice as that of LMS, which
Where µ̂ is a small positive constant, called the step is more suitable for the high-speed signal processing [9].
size, whose value is between 0 and 1. A small constant

International Journal of Intelligent Engineering and Systems, Vol.5, No.4, 2012 20

 cancellation block, Wiener Baseband Power Ampli-
fier (PA), synchronization module and multipath chan-
nel. The signals from the WCDMA source are in-
putted to the system. Output signals from the echo
cancellation block (Point C) are amplified by the power
amplifier and then fed back via multipath channel to
the input terminal as interference signal. The inter-
Figure 4 Block diagram of DLMS algorithm
ference signals mix with the WCDMA source signals
and together enter the echo cancellation block. Mean-
The iterative equation of filter can be expressed as while, the output signals of the power amplifier serve
equation (7). as the reference signal of the cancellation block (Point
G). The cancellation of echo interference is processed
ω (n) = ω (n − 1) + 2µ e(n − d)x(n − d), (7) in the baseband. The original bandwidth of WCDMA
source is 3.84 MHz, and the stopband bandwidth be-
where the parameter d is the delayed cycle number,
comes 4.6848 MHz via a shaping filter with roll-off
which is introduced when the weight coefficients are
factor 0.22. The cancellation block adopts several al-
updated.
gorithms, including LMS adaptive filtering algorithm,
3.5 Sign LMS algorithm NLMS adaptive filtering algorithm, sign LMS adap-
tive filtering algorithm, DLMS adaptive filtering algo-
Stochastic gradient descent method is also used in
rithm, frequency-domain block LMS adaptive filter-
Sign LMS algorithm in order to obtain the optimal so-
ing algorithm, and RLS adaptive filtering algorithm.
lution. However, only optimal solution not the size of
In the simulation process, the multipath echo interfer-
gradient is given in this method. The iterative equa-
ence is three-path, and the order of the filter is 128,
tion of filter can be expressed as follows:
and the step size of the simulation 0.00008. For each
ω (n) = ω (n − 1) + 2µ · sign[e(n)]x(n), (8) kind of adaptive algorithm, echo interference cancel-
lation simulation is implemented to input signals with
ω (n) = ω (n − 1) + 2µ · e(n)sign[x(n)], (9) the signal to interference ratio -10, 0, 10, respectively.

where sign[] is sign function. Equation (8) is the Sign

LMS algorithm of error signals and equation (9) the
input signals.

3.6 Frequency-domain block LMS algorithm

The basic principle of the block LMS algorithm is
to divide the input sequences x(n) into the data blocks Figure 5 Adaptive echo interference cancellation system
at the length of L point. The adaptive filter updates the simulation model block diagram
weight coefficients block by block instead of one by
one. The block LMS algorithm mainly concerns the In analyzing the PSD curve and the constellation
liner convolution between the weight coefficients and of simulation results, it turns out that there is some
input signals, and the linear correlation between the improvement in the distribution on constellation for
input signals and the error signals. And fast fourier the sign LMS adaptive filtering algorithm. The output
transformation (FFT) is especially suitable for fast con- signals are not flat enough on the in-band PSD curve,
volution and relative calculation. Thus the block LMS which varies from the PSD curve of the source signals.
algorithm is conducted in frequency domain, or, Fre- Moreover, the attenuation of out-band does not reach
quency-domain Block LMS. the objective of 40dB when the signal to interference
ratio is -10. This shows that sign LMS adaptive fil-
4. Adaptive Interference Cancellation System tering algorithm does perform some inhibiting ability
Simulation Comparison to multipath echo interference. However, the effect is
not satisfying.
Figure 5 shows the simulation model block diagram, While the other five sorts of adaptive filtering al-
which indicates WCDMA source, echo interference gorithm perform excellent echo cancellation ability
judging from the PSD curve and the constellation of

International Journal of Intelligent Engineering and Systems, Vol.5, No.4, 2012 21

simulation results, and it can meet the design require- When the value range of equation xH (n)x(n) is judged,
ment of the echo cancellation system. In the view of the number of left shifting bit was decided for the
the easiness to implementation on FPGA chips, this error e(n), let η = 1/256, ψ = 1, then µ (n)max =
paper discusses the convergence rate and the usage 1/256 = 2−8 , µ (n)min = 1/2274877906945 ≈ 2−38 .
of resource of five adaptive algorithms-LMS, NLMS, Since the bit width of e(n) is 32, in segmentation op-
DLMS, frequency-domain block LMS and RLS, and eration, when xH (n)x(n) = 0, e(n) was shifted 8 bits
the one that best fit for high-speed signal processing to the left; when 1 ≤ xH (n)x(n) ≤ 2, e(n) was shifted
and implementation on FPGA is chosen. 9 bits to the left; when 3 ≤ xH (n)x(n) ≤ 7, e(n) was
As for the convergence rate, on the same condition shifted 10 bits to the left. And the rest may be known
of 128 order and 0.00008 step size, RLS has the high- by analogy. When xH (n)x(n) ≥ 4194303, e(n) will
est convergence rate, followed by frequency-domain be shifted 31 bits to the left. According to segmen-
block LMS, NLMS, DLMS, and LMS in sequence. tation and approximation method, shift operation was
When the tap number of filter M is 512, frequency- adopted instead of division. In addition, appropriate
domain block LMS is 8.8 times faster than LMS, and expansion and interception must be considered for ev-
16 times faster when M is 1024. ery calculating result to prevent data overflow. NLMS
And the usage of resource: if the length of the data algorithm adaptive filter structure is shown in Figure 6.
blocks to be processed is M, LMS will need to oper-
ate corresponsive 2M 2 times multiplication operation;
sign LMS will need M 2 times multiplication opera-
tion; frequency-domain block LMS will need 10M×
log2 M +26M times multiplication operation; RLS will
need (2M 2 +4M)M times multiplication operation and
twice division operation; and NLMS will need multi-
plier little more than LMS when adopts optimization
algorithm.
RLS has the highest convergence rate but huge us-
age of resource. Sign LMS has the least usage of re-
source, but it can not give satisfying filtering effect.
Figure 6 NLMS algorithm adaptive filter structure
LMS is too slow at the convergence to satisfy the re-
quirement of the high-speed signal processing, though In Figure 6, each block represents a computing node.
the usage of resource is relatively small and satisfying The adaptive algorithm parts consist of multiplier, de-
filtering effect. Therefore, this paper adopts NLMS cision device, and shifter. The multiplier performs the
adaptive algorithm and frequency-domain block LMS operation of equation xH (n)x(n). The decision device
adaptive algorithm to design the adaptive echo inter- which is based on case select structure chooses shift
ference cancellation system. factor according to the different range of xH (n)x(n).
The shifter implements shifting and interception for
5. NLMS Adaptive Algorithm Filter Imple- error e(n). The multiplier, decision device, and shifter
mentation on FPGA delays 1 clock period respectively, so x(n) must be
In this work, we considered a NLMS algorithm adap- delayed 3 clock periods to avoid time sequential con-
tive filter with 17 orders. And the filter has 16-bit- fusion. It reduces 17 clock periods and speeds up
wide input data and output data. The FPGA chips data processing that the multiplier, decision device,
supported by Xilinx Company are the Virtex5 series. and shifter replace a divider. The parallel adder will
The program was written by verilog HDL language. adds up the weight coefficient of each tap to the cor-
The Modelsim simulation software was used to run responding product of x(n). The function of Z −3 will
simulation for the designed system. The simulated re- produce 3 clock periods of delay to input signal x(n).
sults were carried out spectrum analysis under matlab
environment. Equation (5) shows that µ (n) can be ex-
6. Performance Analysis of the NLMS Adap-
pressed by a division formula. Because the value of η tive Echo Cancellation System
is between 0 and 1, the product of µ (n) and e(n) can With the help of verilog HDL hardware program-
be obtained by shift operation to reduce the computa- ming language, program for the NLMS adaptive filter
tional complexity. algorithm, synthesize in the environment of ISE, and

International Journal of Intelligent Engineering and Systems, Vol.5, No.4, 2012 22

 Figure 8 The comparison between the constellation before
Figure 7 The Comparison between the PSD curves before and after the interference cancellation when the signal to
and after the interference cancellation when the signal to interference ratio is -10. (a) Before the interference cancel-
interference ratio is -10. (a). Before the interference can- lation (b) After the interference cancellation
cellation (b). After the interference cancellation

number of the coefficients equals to the size of the

achieve function simulation using Modelsim. Based
data block, so as to get the highest computational ef-
on the adaptive algorithm above, echo interference can-
ficiency [11]. The frequency-domain block LMS al-
cellation was implemented to the signals with the sig-
gorithm based 1/2 overlapping reserving method cas-
nal to interference ratio -10, 0, 10 respectively. Then
cades the input signals and the expected signals, con-
the output data from Modelsim were imported to Mat-
ducts FFT, and then operates multiplication in frequency
lab. By analyzing the PSD curve and the constellation
domain, which needs to be operated with fast convolu-
of the imported data, estimate the performance of the
tion and relative calculation in time domain [12]. The
adaptive echo interference cancellation system [10].
basic block diagram of algorithm is shown in Figure 9.
Figure 7 and Figure 8 show the performance analy-
Figure 9 consists of 4 parts.
sis of NLMS adaptive echo interference cancellation
I ) Every M input signals x(n) in the time domain
system when the signal to interference ratio is -10.
make up a block; every two cascaded data blocks con-
Before interference cancellation, the in-band PSD curve
duct N point fast fourier transformation, and get trans-
has wider fluctuation, and the distribution of the con-
ferred X(k) in the frequency domain. X(k) is the input
stellation is dispersal. While after interference can-
of the adaptive filter. N is twice as much as M– the
cellation, the in-band PSD curve becomes more flat
number of the filter tap.
and the out-band attenuation exceeds 40dB. This indi-
II ) Update the weight coefficient W (k) according to
cates that the NLMS adaptive echo interference can-
the input signal X(k) and error signal E(k) and obtain
cellation system has remarkable inhibition ability to
the estimated value of interference Y (k).
multipath echo interference.
III ) Carry out inverse fast fourier transformation
(IFFT) for the output signal of adaptive filter Y (k), and
7. Frequency-Domain Block LMS Adaptive
get the time domain signals— interference estimated
Algorithm Filter Implementation on FPGA
value y(n).
Frequency-domain Block LMS adopts the 1/2 over- IV ) Calculate the differences between the interfered
lapping-reserving method in the operation process. The signal d(n) and y(n), and get the signal e(n) in which

International Journal of Intelligent Engineering and Systems, Vol.5, No.4, 2012 23

 Figure 9 The basic block diagram of frequency domain block LMS algorithm

the interference is removed. Transforming the signal

e(n) by fourier transformation will get the frequency
domain value E(k), which can serve for generation
filter tap coefficient of the next iteration.

8. Performance Analysis of the Frequency Do-

main Block LMS Adaptive Echo Cancel-
lation System
Figure 10 and Figure 11 show the performance anal-
ysis of the frequency-domain block LMS adaptive echo
interference cancellation system when the signal to
interference ratio is -10. Before interference cancel-
lation, the in-band PSD curve has wider fluctuation,
and the distribution of the constellation is dispersal.
While after interference cancellation, the in-band PSD
curve becomes more flat and the out-band attenua-
tion exceeds 40dB. This indicates that the frequency-
domain block LMS adaptive echo interference cancel-
lation system has remarkable inhibition ability to mul-
tipath echo interference.

9. Conclusion
After simulation research on six kinds of adaptive
echo interference cancellation system— on the basis
of LMS adaptive filtering algorithm, NLMS adaptive
filtering algorithm, DLMS adaptive filtering algorithm, Figure 10 The comparison between the PSD curve before
sign LMS adaptive filtering algorithm, frequency do- and after the interference cancellation when the signal to
main block LMS adaptive filtering algorithm, and RLS interference ratio is -10 (a). Before the interference cancel-
adaptive filtering algorithm, and the comparison be- lation. (b). After the interference cancellation.
tween the performance of the effects of cancellation,
convergence rate and the usage of resource, this pa-
per selected NLMS adaptive filtering algorithm and

International Journal of Intelligent Engineering and Systems, Vol.5, No.4, 2012 24

 References
[1] Y. H. Gong, Adaptive filtering (Second Edition) time-
domain adaptive filtering and smart antennas, Pub-
lishing House of Electronics Industry, BeiJing, 2003.
[2] Y. Tian, W. B. Xu, and Y. W. Zhang, Wireless FPGA
Design, Publishing House of Electronics Industry,
BeiJing, 2008.
[3] J.F. Meng and X.B. Hou, “The simulation and real-
ization of LMS adaptive filter,” Electronic Compo-
nent & Device Applications, Vol. 11(9), pp.67-73,
2009.
[4] M. T. Islam and Z. A. A. Rashid, “MI-NLMS adap-
tive beamforming algorithm for smart antenna sys-
tem applications,” Journal of Zhejiang University
SCIENCE A, Vol.7(10), pp.1709-1716, 2006.
[5] Liu Mingchun and Li Guojin, “Research on the appli-
cation of self-adaptive frequency-domain LMS algo-
rithm on interference cancellation,” World Technol-
ogy Research and Development, 1:pp57-58, 2009.
[6] C. So, “Modified LMS algorithm for unbiased im-
pulse response estimation in nonstationary noise,”
Electronics Letters, Vol.35(10), pp.791-792, 1999.
[7] J. Pesquet, G. Tziritas, and O. Macchi, “Modi-
Figure 11 The comparison between the constellation be- fied LMS algorithms for robust ADPCM,” Proc.
fore and after the interference cancellation when the signal IEEE Acoustics Speech and Signal Process-
to interference ratio is -10. (a) Before the interference can- ing(1990), IEEE Press, Mar. pp.1405-1408, doi:
cellation (b) After the interference cancellation 10.1109/ICASSP.1990.115652, 1990.
[8] H.C. So, “Modified LMS algorithm for unbiased im-
the frequency-domain block LMS algorithm for fur- pulse response estimation in nonstationary noise,”
Electronics Letters, Vol.35(10), pp.791-792, 1999.
ther hardware design on FPGA.
Based on the analysis and comparison of the PSD [9] W. B. Chen, X. M. Geng, and W. M. Ma, “High speed
curve and the constellation of NLMS and the frequency- adaptive DLMS algorithm and its hardware imple-
mentation,” Information Technology, Vol.10, pp.59-
domain block LMS adaptive echo interference cancel-
62, September, 2007.
lation system, it concludes that the frequency-domain
[10] CYLIX Communication Equipment Co., Ltd, The
block LMS adaptive echo interference cancellation sys-
Implementation of MATLAB and FPGA for Wireless
tem has better flatness on the in-band PSD curve and Communication, Posts and Telecom Press, 2009.
better convergence effect on the constellation. There-
[11] Paulo S. R. Diniz, Self-adaptive filter algorithm and
fore, frequency-domain block LMS adaptive interfer-
implementation: Second Edition, Liu Yulin Trans-
ence cancellation algorithm is chosen as the design lated. Beijing: Publishing House of Electronics In-
with high performance-price ratio. The more points dustry, 2004.
there exist, the more obvious advantages it will per- [12] Simon Haykin, Adaptive Filter Theory (Fourth Edi-
form. tion), Baoyu Zheng Translated, Beijing: Publishing
House of Electronics Industry, 2006.
Acknowledgments
This work was supported by the Dr. Start Fund of
Liaoning Province under project No. 20071003 and
the Discipline Construction Fund of College of Infor-
mation and Control Engineering of Shenyang Jianzhu
University.

International Journal of Intelligent Engineering and Systems, Vol.5, No.4, 2012 25