See discussions, stats, and author profiles for this publication at: [Link]

net/publication/294422511

Analysis on the adaptive filter based on LMS algorithm

Article in Optik · February 2016

DOI: 10.1016/[Link].2016.02.005

CITATIONS READS

31 3,611

6 authors, including:

Leilei Cao Yumeng Cai

Zhejiang University North China Electric Power University
40 PUBLICATIONS 540 CITATIONS 2 PUBLICATIONS 60 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Leilei Cao on 25 February 2018.

The user has requested enhancement of the downloaded file.

 Optik 127 (2016) 4698–4704

Contents lists available at ScienceDirect

Optik
journal homepage: [Link]/ijleo

Analysis on the adaptive filter based on LMS algorithm

Zhen Zhu a , Xiang Gao a,∗ , Leilei Cao a , Daoyuan Pan b , Yumeng Cai a , Yu Zhu c
a
School of Automotive and Traffic Engineering, Jiangsu University, Zhenjiang 212013, China
b
School of Mechanical and Automotive Engineering, Anhui polytechnic University, Wuhu 241000, China
c
Axle Branch Company, IVECO Co. Ltd , Nanjing, 212028, China

a r t i c l e i n f o a b s t r a c t

Article history: This article focuses on the application of adaptive filter based on the LMS algorithm. An adaptive filter
Received 1 January 2016 of the closed-loop system is introduced, including the elimination of interference signal, the prediction
Accepted 8 February 2016 of useful signal, and the approximation of expected signal. LMS (Least Mean Square) algorithm is used
to meet the optimum norm of error between estimated signal and expected signal. The structure of
Keywords: LMS algorithm is presented and the simulation of LMS algorithm is carried out. The results indicate that
Adaptive filter
the convergence performances of LMS algorithm are prefect, and the input signal can converge to the
LMS algorithm
expected signal. The application of adaptive filtering technology in this article includes the correction
Adaptive linear filter
Adaptive equalizer
of channel mismatch by an adaptive linear filter, the improvement of system performance by an adap-
Adaptive notch filter tive equalizer, and the filter of frequency signal by an adaptive notch filter. The analysis on adaptive
linear filter shows that the constant channel mismatch can be corrected quite well by the correction
algorithm. The analysis on adaptive equalizer shows that the error rate of system with an adaptive equal-
izer has significant improvement gains over that of system without an adaptive equalizer. The smaller
the error rate, the larger the SNR. The relationship between error rate and multi-path loss show that
the error rate is largest when the loss factor is 0.5. The analysis on adaptive notch filter shows that the
interference signal with two different known frequencies can be eliminated effectively by the adaptive
notch filter. The filtered signals accord with the corresponding useful signals very well.
© 2016 Published by Elsevier GmbH.

1. Introduction is suitable to the condition that the amplitude-frequency charac-

teristic is different from the typical filter characteristic or the linear
Commonly used filters can be divided into three kinds: FIR phase characteristic is required [1–4].
(Finite Impulse Response) filter, IIR (Infinite Impulse Response) fil- The precondition of FIR filter and IIR filter is that the statistical
ter, and adaptive filter. FIR filter can be designed as the system characteristics of input signals are known. That is to say, the per-
with arbitrary amplitude-frequency characteristic and linear phase formance indexes of filter, such as passband, stopband, tolerance
characteristic. The design target of FIR filter is that the require- of passband, and stopband, are definite. The design target of filter
ment of filter coefficients for performance indexes should be met is that the performance indexes of filter, which are not related to
and the hardware resources of system can be saved. Comparing time, should be met. The premise design condition of adaptive fil-
with FIR filter, IIR filter has higher efficiency and lower order under ter is that some of the statistical characteristics of input signals are
the condition of same amplitude-frequency response. However, unknown. The performance indexes of filter can be replaced by the
the disadvantage of IIR filter is the non-linear phase characteristic, estimated value of unknown signal [5,6]. In order to analyze the
which restricts the application range of filter. The theoretical basis adaptive filter based on LMS (Least Mean Square) algorithm, the
of FIR filter and IIR filter in this article is the Linear Time-Invariant principle and application of adaptive filter should be introduced,
System, which means the structure design and analysis method and the simulation results based on the statistical experimental
between FIR filter and IIR filter are similar. In a word, IIR filter is method are presented according to the principle and structure of
suitable to the condition that the amplitude-frequency character- LMS algorithm [7–9]. The applications of adaptive filtering technol-
istic is given and the phase characteristic is not considered; FIR filter ogy are shown by the introduction of three parts: an adaptive linear
filter for the correction of channel mismatch, an adaptive equalizer
for the improvement of system performance, and an adaptive notch
∗ Corresponding author. Tel.: +86 13775531530; fax: +86 51188722660. filter for the elimination of the interference signal with known
E-mail address: xianggaojs@[Link] (X. Gao). frequency.

[Link]
0030-4026/© 2016 Published by Elsevier GmbH.
 Z. Zhu et al. / Optik 127 (2016) 4698–4704 4699

Nomenclature

n sequence length
y (n) output signal
e (n) error signal
N2 (n) Interference signal 2
sN (n) narrow-band signal
 time-delay
kN sampling time of narrow-band signal
Fig. 2. Principle diagram for the elimination of interference signal.
X (n) Vector of input signal
e*(n) complex conjugate pair of e(n)
SNR signal-to-noise ratio to the optimal criterion for FIR adaptive filter and error signal.
Hk (j␻) frequency response of the kth channel The process of adjusting parameters of FIR adaptive filter is called
C (j␻) correction factor “learning process” when the statistical characteristics of input
GT (␻) transfer function of transmitting filter signals are tested; and the process of adjusting parameters of FIR
n (t) Gaussian white noise signal added in the channel adaptive filter is called “tracking process” when the statistical
G E (␻) transfer function of equalizer characteristics of input signals are unknown.
s (t) useful signal
␻1 /␻2 frequency of interference signal 2.1.2. Application of adaptive filter
␪1 /␪2 phase of interference signal With the development of the signal processing theory, the FIR
x (n) input signal adaptive filter has been widely used as a valuable signal-processing
t (n) excepted signal device to eliminate interference signal, predict useful signal and
N1 (n) Interference signal 1 approximate expected signal.
s (n) useful signal
sB (n) wide-band signal [Link]. Elimination of interference signal. The principle diagram for
kB sampling time of wide-band signal the elimination of interference signal is shown in Fig. 2. Expected
W (n) coefficient vector of filter signal t(n) is the sum of useful signal s(n) and noise signal N1 (n).
␮ step factor N2 (n) is another noise signal relevant to N1 (n). The parameters of
j (n) Gaussian white noise signal adaptive filter should be adjusted automatically to eliminate inter-
Dk (j␻) channel mismatch characteristic of the kth channel ference signal. Namely, output signal y(n) is approximate to N̂1 (n),
Hr (j␻) frequency response of the reference channel the optimal estimation of N1 (n), and error signal e(n) is approximate
H (␻) transfer function of signal system to s(n) [10,11].
C (␻) transfer function of channel There are two special conditions should be considered about the
GR (␻) transfer function of receiving filter elimination of interference signal:
x (t) input signal with the interference signal The system can realize the elimination of interference signal
r (t) reference signal when N2 (n) is correlated with N1 (n). However, it is difficult to elim-
A 1 /A 2 amplitude of interference signal inate the interference signal, which is uncorrelated with N2 (n) or
W1 ∼W4 weight value of single-frequency signal superimposed on s(n).
A part of useful signal will be eliminated when s(n) leaks into
the input terminal of adaptive filter, it is necessary to avoid this
situation.
2. Introduction of adaptive filter and LMS algorithm

2.1. Introduction of adaptive filter [Link]. Prediction of useful signal. The principle diagram for the
prediction of useful signal is shown in Fig. 3. The input signal of
2.1.1. Adaptive filter of the closed-loop system adaptive filter is the useful signal with time-delay, and the out-
The structure parameters of the adaptive filter studied in this put signal of predictive filter is the prediction of useful signal with
article can be adjusted automatically by the use of LMS algo- time-delay [12–14].
rithm, according to the statistical characteristics of input signal. One of the applications of adaptive prediction is the separation
The adaptive filter is composed of the digital filter with adjustable of narrow-band and wide-band signal. A mixed signal, which is
parameters and adaptive algorithm. This article focuses on the FIR added to the input terminal of predictive filter, can be expressed as
adaptive filter of the closed-loop system, which is shown in Fig. 1. follows:
Output signal y(n) is generated with the use of FIR digital filter, x(n) = sN (n) + sB (n) (1)
and is related to input signal x(n). Error signal e(n) is generated
with the comparison of output signal y(n) and target signal t(n).
The filter parameters can be modified automatically, according

Fig. 1. FIR adaptive filter of the closed-loop system. Fig. 3. Principle diagram for the prediction of useful signal.
4700 Z. Zhu et al. / Optik 127 (2016) 4698–4704

Fig. 4. Principle diagram for the approximation of expected signal. Fig. 5. Flow diagram of LMS algorithm structure.

The updating algorithms of whole weight values include 2N

The autocorrelation coefficient of narrow-band is shorter than
multiply operations, 2N addition operations, one subtraction oper-
that of wide-band. When the delay-time is set to kB <  < kN ,
ation and N shift operations. LMS algorithm is a closed-loop system,
sB (n − ) is uncorrelated with sB (n), and sN (n − ) is correlated with
so the updating algorithm of each weight value should be finished
sN (n), so the following equations can be obtained.
in the corresponding data cycle.
y(n) = ŝN (n) (2)
2.2.3. Simulation of LMS algorithm based on the statistical
ŝB (n) = sB (n) + sN (n) − ŝN (n) (3) experimental method
Statistical experimental method is a stochastic analysis method,
Where ŝN (n) and ŝB (n) are the best estimate values of sN (n) and
which tests the dynamic characteristics of system and concludes
sB (n), respectively. Therefore, sB (n) and sN (n) can be separated to
the statistical results by mass data of computer simulations. The
complete the prediction of useful signal.
basic idea of problem solving is that a similar probability model
should be built, then the statistical sampling should be carried out,
[Link]. Approximation of expected signal. The principle diagram
the estimated values of sample should be considered as the approx-
for the approximation of expected signal is shown in Fig. 4. The
imate solution of original problem finally. The theoretical basis of
weight values of adaptive filter are regulated automatically to
statistical experimental method is law of large numbers, and the
meet requirements. Namely, when the frequency components of
main method is the sampling analysis of random variables [20,21].
excitation source are unchanged, and the system noise is small
Simulation analysis should be carried out to visually display
enough, the adaptive filter will become an approximation model
the LMS algorithm. Input signal is the synthesis signal, which is
of unknown system [15,16].
synthesized by a sine signal and a white noise signal.
√ SNR
2.1.3. Adaptive algorithm x(n) = 2s(n) + sqrt(10− 10 )j(n) (6)
An adaptive filter is mainly composed of the digital filter with
Where s(n) is the sine signal, j(n) is the Gaussian white noise
the adjustable parameters and the adaptive algorithm. The final
signal (average value is zero, power is one). The Gaussian white
purpose of adaptive algorithm is to meet the optimal criterion
noise signal can form an interference cancellation system. It is
of error between estimated signal and expected signal, and the
noteworthy that there is not any time-delay in the LMS algorithm,
optimal algorithm is the decisive factor for the performance and
because the correlation function of Gaussian white noise signal
structure of adaptive criterion. This article adopts LMS algorithm
is the impulse function, the correlativity will disappear when the
as the optimal criterion to solve the optimal problem [17–19].
time-delay exists in the expected signal and input signal.
Simulation conditions are shown in Table 1.
2.2. Introduction of LMS algorithm
Simulation results are shown in Fig. 6. The results indicate that
the convergence performances of LMS algorithm are prefect, the
2.2.1. LMS algorithm principle
input signal can converge to the expected signal (sine signal or noise
By using the iteration operations, filter coefficients can converge
signal). Besides, the convergence speed and performances of LMS
to the optimal values. LMS algorithm, which is proposed by Widrow
algorithm have great relationship with the length of filter and step
and Hoff in 1960, is one of the iteration operations.
factor.
LMS algorithm can be expressed by the following recursion
equation.
3. Application of adaptive filtering technology
⎧
⎨ y(n) = W H (n)X(n)
The application of adaptive filtering technology in this article
e(n) = t(n) − y(n) (4)
⎩ includes the correction of channel mismatch by an adaptive lin-
W (n + 1) = W (n) + 2X(n)e∗ (n) ear filter, the improvement of system performance by an adaptive
equalizer, and the filter of frequency signal by an adaptive notch
The larger the ␮, the faster the convergence rate, and the larger
filter.
the error. So the performance of LMS algorithm can be adjusted
by controlling the step factor. Namely, larger  is used for the ini-
3.1. An adaptive linear filter for the correction of channel
tial stage to accelerate convergence, and smaller  is used for the
mismatch
algorithm convergence stage to improve stability.
FIR filter is one of the adaptive linear filters for its strict linear
2.2.2. LMS algorithm structure phase characteristic. Section 2.2.3 presents the simulation analysis
According to Equation (4), W(n) can be calculated by the cyclic of LMS algorithm, this section will carry out the channel mismatch
iteration method to realize the least mean square value of e(n). correction for an adaptive linear filter based on LMS algorithm.
∗ In the digital signal processing system, the compositions of ana-
W (n + 1) = W (n) + 2X(n)[t(n) − W H (n)X(n)] (5)
log processing units for each channel are not identical, which cause
The flow diagram of LMS algorithm structure is shown in Fig. 5. the inconsistency of amplitude and phase of frequency response
 Z. Zhu et al. / Optik 127 (2016) 4698–4704 4701

Time domain waveform of signal Output time domain signal(SNR=5dB,Expected signal is sine signal.)
2 2

x(n)

y(n)
0 0

-2 -2
0 20 40 60 80 100 120 0 20 40 60 80 100 120
n n
Time domain waveform of signal with additive noise(SNR=5dB) Output time domain signal(SNR=-5dB,Expected signal is sine signal.)
2 2

x(n)

y(n)
0 0

-2 -2
0 20 40 60 80 100 120 0 20 40 60 80 100 120
n n
Time domain waveform of signal with additive noise(SNR=-5dB) Output time domain signal(SNR=-5dB,Expected signal is noise signal.)
2
2
x(n)

y(n)
0 0
-2
-2
0 20 40 60 80 100 120 0 20 40 60 80 100 120
n n
(a) Input signals based on LMS algorithm (b) Output signals based on LMS algorithm

Error mean square signal(SNR=5dB,Expected signal is sine signal.)

0.02

e(n) 0.01

0
0 10 20 30 40 50 60 70 80 90 100
n
Error mean square signal(SNR=-5dB,Expected signal is sine signal.)
0.1
e(n)

0.05

0
0 10 20 30 40 50 60 70 80 90 100
n
Error mean square signal(SNR=-5dB,Expected signal is noise signal.)
0.1
e(n)

0.05

0
0 10 20 30 40 50 60 70 80 90 100
n
(c) Error signals based on LMS algorithm

Fig. 6. Signals based on LMS algorithm.

among channels. Therefore, it is necessary to correct the chan- reference signal and the corrected channel should trend to the min-
nel mismatch to acquire the higher performance of subsequent imum value, and the value of D(jw)C(jw) is approximate to one
processing [22–25]. finally.
The channel mismatch characteristic of the kth channel can be The time-delay  in the reference channel should make up for
defined as follows: the time difference, which is caused by D(jw)C(jw). In order to cor-
Hk (jw) rect the amplitude and phase factor at the same time, the operation
Dk (jw) = (7) of LMS algorithm is a complex operation, and the length of filter is
Hr (jw)
one. That is to say, a complex weight can realize the correction of
A channel, whose frequency response is H(jw), can be expressed amplitude and phase.
as follows: The simulation results can be shown in Fig. 8.
H(jw) The correction result is prefect when amplitude range is
D(jw) = = [1 + H(jw)]ejϕ(w) (8) 0∼1.5, and phase range is –75o ∼75o . Namely, the constant chan-
Hr (jw)
nel mismatch can be corrected quite well by the correction
H(jw) and ϕ(w) are constants because the channel mismatch algorithm.
is not related to the frequency. The input signal can be regarded
as stationary signal for the general signal system, and the main
problem of the correction of channel mismatch is attributed to the
determination of correction factor, which is shown in Equation (9). x 10
-5 Amplitude mismatch
 1
 1.5
SNR=-5dB
C(jw) = e−jϕ(w) (9)
After correction

SNR=0dB
1 + H(jw) 1 SNR=5db

This article analyzes the constant channel mismatch indepen- 0.5

dent from frequency response, and the correction model of channel
mismatch is shown in Fig. 7. 0
0 0.5 1 1.5
Input signal is transmitted to the reference channel and the pre- Before correction
Phase mismatch
dictive correction channel. The mean square deviation between the 100
SNR=-5dB
After correction

SNR=0dB
SNR=5db
50

0
-150 -100 -50 0 50 100 150
Before correction

Fig. 7. Correction model of channel mismatch. Fig. 8. Characteristic analysis of the correction algorithm of channel mismatch.
4702 Z. Zhu et al. / Optik 127 (2016) 4698–4704

Table 1
Simulation conditions of LMS algorithm.

Simulation parameter Data Simulation parameter Data

Iteration times 100 Sequence length of input signal 1024

Length of filter 128 Step factor 1/128
Useful signal sin (0.5*pi*t) SNR parameters of input signal –5 dB/5 dB

Relationship between error rate and SNR(p=0.1)

0
10

with equalizer
-1
10

Error rate
without equalizer
Fig. 9. Flow diagram of system model with an adaptive equalizer.
-2
10
3.2. An adaptive equalizer for the improvement of system
-3
performance 10
0 1 2 3 4 5 6 7 8 9 10
SNR
Because of the multi-path effect of communication channel, the 0
Relationship between error rate and multi-path loss(SNR=10dB)
10
limitation of channel bandwidth and the imperfection of chan-
nel characteristics in the wireless data communication system, the -1
with equalizer
10

Error rate
without equalizer
intersymbol interference are inevitably occurred when the data
pass through the channel. The adaptive equalizer can automatically -2
10
adjust its parameters to adapt to the change of channel character-
istics and eliminate interference to improve the communication -3
10
quality [26–29]. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
multi-path loss
Theoretical analysis and experimental results show that a tun-
able filter in the digital communication system can compensate Fig. 10. Error rate contrasting diagrams between the system with an adaptive equal-
system characteristics and eliminate intersymbol interference. The izer and the system without an adaptive equalizer.
tunable filter is called equalizer, and the main realization way of
adaptive filter is the transversal filter. The system model with an
adaptive equalizer is shown in Fig. 9. Table 2
The transfer function of signal system can be expressed as fol- Simulation conditions of LMS algorithm for an adaptive equalizer.
lows: Simulation parameter Data Simulation parameter Data

H(w) = GT (w)C(w)GR (w)GE (w) (10) Length of simulation data 20000 Length of training data 2000
Step factor 1/128 Length of filter 8
The transmitting filter and the receiving filter are matched to
each other, and the equalizer is used to compensate the channel dis-
tortion. The transfer function of equalizer should satisfy Equation
(11).
1 1
GE (w) = =   e−jϕ(ω) (11)
C(w) C(w)
The filter is used as the equalizer to compensate the distor-
tion of frequency characteristic. The demodulated output samples
obtained by the sampling decision device are the modified samples.
The adaptive filter adjusts the gain of input signal continuously
Fig. 11. The principle diagram of adaptive notch filter for the elimination of inter-
according to some algorithms, in order to adapt to the random
ference signal with two different known frequencies (ω1 , ω2 ).
variation of channels and keep the optimum operating condition
of equalizer. From parameters adjustment to convergence forma-
tion, the whole working process of system is the function on the
equalizer algorithm and communication change rate. The equalizer
should be trained periodically to eliminate intersymbol interfer- The simulation results show that the error rate of system with
ence, and be modified according to the same training sequence an adaptive equalizer has significant improvement gains over that
when the equalizer receives the new data. The training sequence is of system without an adaptive equalizer Table 1–3. The smaller the
the expected signal of some algorithm. error rate, the larger the SNR. The relationship between error rate
The simulation conditions of LMS algorithm for an adaptive and multi-path loss show that the error rate is largest when the loss
equalizer are shown in Table 2, and the error rate contrasting dia- factor is 0.5, which means the transmission paths of two signals are
grams between the system with an adaptive equalizer and the in the most interferential conditions. The simulation results accord
system without an adaptive equalizer are shown in Fig. 10. well with the theoretical analysis.

Table 3
Simulation conditions of LMS algorithm for an adaptive notch filter.

Simulation parameter Data Simulation parameter Data

Sampling frequency 4000 Hz Frequency of interference signal f1 = 50 Hz

f2 = 10 Hz
Frequency of useful signal (1)400 Hz ; (2)Random signal.
 Z. Zhu et al. / Optik 127 (2016) 4698–4704 4703

Useful signal Useful signal

2 2
1 1
0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Interference signal(10Hz) Interference signal(10Hz)
2 2
0 0
-2 -2
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Interference signal(50Hz) Interference signal(50Hz)
2 2
0 0
-2 -2
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Superposed signal Superposed signal
5 5
0 0
-5 -5
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Filtered signal Filtered signal
5 5
0 0
-5 -5
0 50 100 150 200 250 300 0 50 100 150 200 250 300

(a) signal frequency signal (400H z ) (b) randomsignal

Fig. 12. Simulation curves of different signals for an adaptive notch filter.

3.3. An adaptive notch filter for the elimination of the improvement gains over that of system without an adaptive
interference signal with known frequency equalizer. The smaller the error rate, the larger the SNR. The
relationship between error rate and multi-path loss show that
Notch filter is another kind of filter, which is used to eliminate the error rate is largest when the loss factor is 0.5. The analysis
the interference signal with known frequency, unknown phase, and on adaptive notch filter shows that the interference signal with
amplitude. The adaptive notch filter based on the LMS algorithm two different known frequencies can be eliminated effectively
can eliminate the interference signal with known frequency from by the adaptive notch filter. The filtered signals accord with the
the useful signal effectively. Fig. 11 presents the principle diagram corresponding useful signals very well.
of adaptive notch filter. A signal with two different known frequen-
cies can be eliminated at the same time by the adaptive notch filter
Acknowledgment
[30–32].
x(t) is the input signal with the interference signal, the adaptive
This article is financially supported by National Natural Science
notch filter is used to eliminate the interference signal with two
Foundation Project (51575001) and Scientific Research Innovation
different known frequencies (ω1 , ω2 ), s(t) is the useful signal. The
Project of Jiangsu University (KYXX-0008).
amplitudes of interference signal (A1 , A2 ) and the phases of interfer-
ence signal of ( 1 ,  2 ) should be estimated by the LMS algorithm.
In order to estimate the amplitude and phase of A1 cos(ω1 t +  1 ), References
two orthogonal single-frequency signals (sin(ω1 t), sin(ω2 t)) and
corresponding weight values (W1 , W2 ) are used to synthesis esti- [1] Faris Elasha, David Mba, Cristobal Ruiz-Carcel, A comparative study of adaptive
filters in detecting a naturally degraded bearing within a gearbox, Mech. Syst.
mated signal Â1 cos(ω1 t + ˆ 1 ). Therefore, the interference signal
Signal Pr. 3 (2016) 1–8.
with two different known frequencies can be eliminated effec- [2] Du Yong, MATLAB and FPGA implementation of digital filter, Publishing House
tively by four channel reference signals, which can be shown in of Electronics University, 2015.
[3] H. Olkkonen, P. Pesola, A. Valjakka, et al., Gain optimized cosine transform
Fig. 11. The interference signal with two different known frequen-
domain LMS algorithm for adaptive altering of EEG, Comput. Biol. Med. 29
cies can be measured by the estimated signal ŷ = Â1 cos(ω1 t + ˆ 1 ) + (1999) 129–136.
Â2 cos(ω2 t + ˆ 2 ), and the error signal based on the LMS algorithm [4] Hsiang-Feng Chi, X. Shawn, Gao, D. Sigfrid, Soli, et al., Band-limited feedback
is the useful signal without the interference signal. cancellation with a modified filtered-X LMS algorithm for hearing aids, Speech
Commun. 39 (2003) 147–161.
The simulation conditions of LMS algorithm for an adaptive [5] Zhong Bo, Jingli Yang, Chao Sun, et al., A filtered-x weighted accumulated LMS
notch filter are shown in Table 3. algorithm: stochastic analysis and simulations for narrowband active noise
Simulation curves of signal frequency signal for an adaptive control system, Signal Process. 104 (2014) 296–310.
[6] K. Mayyas, F. Momani, An LMS adaptive algorithm with a new step-size control
notch filter are shown in Fig. 12 (a); simulation curves of random equation, J. Franklin Inst. 348 (2011) 589–605.
signal for an adaptive notch filter are shown in Fig. 12 (b). [7] I. Dai, P. Kim, De Wilde, Performance analysis of the DCT-LMS adaptive altering
According to the simulation results in Fig. 12, the interference algorithm, Signal Process. 80 (2000) 1629–1654.
[8] D.S. Brito, Enio Aguiar, Fausto Lucena, et al., Influence of low frequency noise
signal with two different known frequencies can be eliminated in adaptive estimation using the LMS algorithm, Signal Process. 89 (2009)
effectively by the adaptive notch filter. The filtered signals accord 933–940.
with the corresponding useful signals very well. [9] Lingquan Zeng, Xin Li, Research on harmonic suppression in power system
based on improved adaptive filter, Energy Procedia 16 (2012) 1479–1486.
[10] Debi Prasad Das, J. Danielle, Moreau, S. Ben, Cazzolato, A computationally
4. Conclusions efficient frequency-domain filtered-X LMS algorithm for virtual microphone,
Mech. Syst. Signal Pr. 37 (2013) 440–454.
[11] X.L. Tang, C.-M. Lee, Time–frequency-domain filtered-x LMS algorithm for
(1) An adaptive filter of the closed-loop system is introduced, active noise control, J. Sound Vibration 331 (2012) 5002–5011.
including the elimination of interference signal, the prediction [12] Ho-Wuk Kim, Hong-Sug Park, Sang-Kwon Lee, et al., Modified-filtered-u LMS
of useful signal, and the approximation of expected signal. algorithm for active noise control and its application to a short acoustic duct,
Mech. Syst. Signal Pr. 25 (2011) 475–484.
(2) The application of adaptive filtering technology is presented, [13] Shiunn-Jang Chern, K. Jyh-Chau Horng, Max. Wong, The performance of the
including the correction of channel mismatch by an adaptive hybrid LMS adaptive algorithm, Signal Process. 44 (1995) 67–88.
linear filter, the improvement of system performance by an [14] Shyh-Neng Lin, Shiunn-Jang Chern, A new adaptive constrained LMS time delay
estimation algorithm, Signal Process. 71 (1998) 29-D 44.
adaptive equalizer, and the filter of frequency signal by an adap- [15] Eweda Eweda, A new approach for analyzing the limiting behavior of the nor-
tive notch filter. malized LMS algorithm under weak assumptions, Signal Process. 89 (2009)
(3) The analysis on adaptive linear filter shows that the constant 2143–2151.
[16] Ming [Link], Fan Wu, Structural system identification using Least Mean
channel mismatch can be corrected quite well by the correction Square(LMS) adaptive technique, Soil Dyn. Earthq. Eng. 14 (1995) 409–418.
algorithm. The analysis on adaptive equalizer shows that the [17] Reza Arablouei, Stefan Werner, Kutluyıl Doğançay, et al., Analysis of a reduced-
error rate of system with an adaptive equalizer has significant communication diffusion LMS algorithm, Signal Process. 117 (2015) 355–361.
4704 Z. Zhu et al. / Optik 127 (2016) 4698–4704

[18] Khaled Mayyas, Low complexity LMS-Type adaptive algorithm with selective [25] P.K. Dash, A.C. Liew, D.P. Swain, et al., Fast tracking of transient power system
coefficient update for stereophonic acoustic echo cancellation, Comput. Electr. signals using fuzzy LMS algorithm, Int. J. Elec. Power. 20 (1998) 555–561.
Eng. 35 (2009) 450–458. [26] W.M. Jan, Bergmans, Maria Yu Lin, David Modrie, et al., Asynchronous LMS
[19] Muhammad Tahir Akhtar, Binormalized data-reusing adaptive filtering algo- adaptive equalization, Signal Process. 85 (2005) 1301–1313.
rithm for active control of impulsive sources, Digit. Signal Process. 49 (2016) [27] Shengkui Zhao, Zhihong Man, Suiyang Khoo, et al., Variable step-size LMS algo-
56–64. rithm with a quotient form, Signal Process. 89 (2009) 67–76.
[20] Hongchang Zhao, Jian Zhou, Kaiyong Yang, Laser Doppler signal processing [28] Zhong Bo, Chao Sun, Yonghui Xu, et al., A variable momentum factor filtered-
based on least mean square adaptive filter, Optik 124 (2013) 2781–2783. x weighted accumulated LMS algorithm for narrowband active noise control
[21] Ibtissam Constantin, Regis Lengelle, Performance analysis of kernel adap- systems, Measurement 48 (2014) 282–291.
tive filters based on LMS algorithm, Procedia Comput. Sci. 20 (2013) [29] C.C. dos Santos, J.F. Galdino, E.L. Pinto, On the design of LMS-based channel
39–45. estimators using the Doppler spread parameter, Digit. Signal Process. 23 (2013)
[22] Ma rcio Holsbach Costa, Jose Carlos Moreira Bermudez, A noise resilient vari- 281–288.
able step-size LMS algorithm, Signal Process. 88 (2008) 733–748. [30] Yegui Xiao, Yoshiaki Tadokoro, LMS-based notch filter for the estimation of
[23] Michael J. Beerer, Hyungjoo Yoon, Brij N. Agrawal, Practical adaptive filter con- sinusoidal signals in noise, Signal Process. 46 (1995) 223–231.
trols for precision beam pointing and tracking with jitter attenuation, Control [31] K.C. Ho, P.C. Ching, Split filter structures for LMS adaptive filtering, Signal Pro-
Eng. Pract. 21 (2013) 122–133. cess. 46 (1995) 255–266.
[24] Lu Lu, Haiquan Zhao, Zhengyou He, et al., A novel sign adaptation scheme for [32] Hideaki Sakai, Yoichi Hinamoto, An exact analysis of the LMS algorithm with
convex combination of two adaptive filters, Int. J. Electron. Commun. (AEÜ) 69 tonal reference signals in the presence of frequency mismatch, Signal Process.
(2015) 1590–1598. 85 (2005) 1255–1262.

View publication stats