I nt J r of Advanced Computer Engineering and Architecture

Vol. 1, No. 1, J une 2011

Copyright Mind Reader Publications
www.ijacea.yolasite.com
The Applications and Simulation of
Adaptive Filter in Speech Enhancement
Soni Changlani, Dr. M.K. Gupta
Assistant Professor Lakshmi Narain College of Technology, Bhopal
Department of Electronics and Communication Engineering,
e-mail:sonibpl@rediffmail.com
Professor Maulana Azad National Institute of Technology, Bhopal
Department of Electronics and Communication Engineering,
e-mail: gupta4@yahoo.com
Abstract
In many applications of noise cancellation the changes in signal
characteristics could be quite fast. This requires the utilization of
adaptive algorithms, which converge rapidly. Least mean square
(LMS) and Normalized LMS (NLMS) adaptive filters have been
used in a wide range of signal processing applications because of
its simplicity in computation and implementation. The Recursive
Least Squares (RLS) algorithm has established itself as the
ultimate adaptive filtering algorithm in the sense that it is the
adaptive filter exhibiting the best convergence behavior.
Unfortunately, practical implementations of this algorithm are
often associated with high computational complexity and/or poor
numerical properties. In this paper we have performed and
compared these classical adaptive filters for attenuating noise in
speech signals. In each algorithm, the optimum orders of filter of
adaptive algorithms have also been found through experiments...
The objective of noise cancellation is to produce the estimate of the
noise signal and to subtract it from the noisy signal and hence to
obtain noise free signal. We simulate the adaptive noise canceller
with MATLAB.
Keywords- Noise canceller; adaptive algorithm; LMS; NLMS RLS;
Soni Changlani 96
1 Introduction
Liner filtering is required in a variety of applications. A filter will be optimal only
if it is designed with some knowledge about input data. If this information is not
known, then adaptive filters [1] are used .An adaptive filter is essentially a digital
filter with self-adjusting characterstics.It adapts, automatically, to changes in its
input signals. In adaptive filtering, the adjustable filter parameters are to be
optimized. The criteria arrived at for optimization should consider the filter
performance and realisability. Adaptive algorithms are used to adjust the
coefficient of the digital filter. Common algorithms that have found widespread
applications are the least mean square (LMS), the recursive least square (RLS),
and the kalman filter [3] algorithms.RLS algorithm is one of the most popular
adaptive algorithm. As compared to the least-mean-square (LMS) algorithm, the
RLS offers a superior convergence rate, especially for highly correlated input
signals. The price to pay for this is an increase in the computational complexity.
2 RLS Adaptive Algorithm
The RLS (recursive least squares) algorithm is another algorithm for determining
the coefficients of an adaptive filter. In contrast to the LMS algorithm, the RLS
algorithm uses information from all past input samples (and not only from the
current tap-input samples) to estimate the (inverse of the) autocorrelation matrix
of the input vector. The RLS algorithm is a recursive form of the Least Squares
(LS) algorithm. It is recursive because the coefficients at time n are found by
updating the coefficients at time n-1 using the new input data. Whereas the LS
algorithm is a block-update algorithm where the coefficients are computed from
scratch at each sample time. To decrease the influence of input samples from the
far past, a weighting factor for the influence of each sample is used. This
weighting factor is introduced in the cost function.
( ) ( )
2
1
i e n
n
i
i n

= (1)
where ( ) i e is the difference between the desired response and output produced by
a filter. is special form of weighting factor, known as forgetting factor, the
performance of RLS-type algorithms in terms of convergence rate, tracking,
midadjustment, and stability depends on the forgetting factor. The classical RLS
algorithm uses a constant forgetting factor and needs to compromise between the
previous performance criteria. When the forgetting factor is very close to one, the
algorithm achieves low misadjustment and good stability, but its tracking
capabilities are reduced. A smaller value of the forgetting factor improves the
tracking but increases the misadjustment, and it could affect the stability of the
97 The Applications and Simulation
of

algorithm. The use of the forgetting factor is intended to ensure that data in the
distant past are forgotten in order to afford the possibility of following the
statistical variations of the observable data when the filter operates in a
nonstationary environment. is a positive constant close to, but less than, 1. The
rate of convergence of the RLS also depends on .
3 Adaptive Noise Cancellation.
The basic idea of an adaptive noise cancellation algorithm is to pass the noisy
signal through a filter that tends to suppress the noise while leaving the signal
unchanged. This process is an adaptive process, which means it can not require a
priori knowledge of signal or noise characteristics. The technique adaptively
adjusts a set of filter coefficients so as to remove the noise from the noisy signal.
To realize the adaptive noise cancellation, we use two inputs and an adaptive filter.
One input is the signal corrupted by noise (Primary Input, which can be expressed
as (s(n) +N
o
(n)) ,the other input contains noise related in some way to that in the
main input but does not contain anything related to the signal (Noise Reference
Input expressed as N
1
(n) .the noise reference input pass through the adaptive
filter and output y(n) is produced as close a replica as possible of N
o
(n) . The
filter readjusts itself continuously to minimize the error between N
o
(n) and y (n)
during this process. Then the output y (n) is subtracted from the primary input to
produce the system output e= S+N
o
-y, which is the denoised signal. Assume that
S, N
0
, N
1
and y are statistically stationary and have zero means. Suppose that S is
uncorrelated with N
0
and N
1
, but N
1
is correlated with N
0
. We can get the
following equation of expectations:
| | | | ( ) | |
2 2 2
y N E s E e E
o
+ = (2)
When the filter is adjusted so that E [e
2
] is minimized, E[(N
o
-y)
2
] is also
minimized. So the system output can serve as the error signal for the adaptive
filter.
Adaptive
+
+
?
REFERENCE
+
y(n)
-
e(n) OUTPUT
d(n)
S(n)+N
o
(n)
x(n) N
1
(n)
Fig. 1 : Adaptive noise canceller.
PRIMARY INPUT
Soni Changlani 98
4 Simulation And Results:
This section describes parts of the sample MATLAB program and results. The
Program can be divided into four parts: First employ the randn(N,1) function
within MATLAB to generate x(k) which will have unity power and zero mean,
where N is number of system point. The information bearing signal is a sine wave
of 0.055 cycles/sample the noise picked up by the secondary microphone is the
input for the RLS adaptive filter. The noise that corrupts the sine wave is a low
pass filtered version of (correlated to) this noise. The sum of the filtered noise and
the information bearing signal is the desired signal for the adaptive filter. Set and
initialize RLS adaptive filter parameters and values Filter order N=32 Exponential
weighting factor lam=1;
0 20 40 60 80 100 120 140 160 180 200
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
The information bearing signal
Fig 2: The information bearing signal
99 The Applications and Simulation
of

Fig 3: Noise picked up by the secondary microphone
0 20 40 60 80 100 120 140 160 180 200
-4
-3
-2
-1
0
1
2
3
4
Desired input to the Adaptive Filter =Signal +Filtered Noise
Fig 4: Input to the Adaptive filter
0 100 200 300 400 500 600 700 800 900 1000
-4
-3
-2
-1
0
1
2
3
4
Soni Changlani 100
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Normalized Frequency (t rad/sample)
M
a
g
n
i
t
u
d
e
Adaptive Filter Response
Required Filter Response
Fig 5: Response of the Adaptive filter
0 50 100 150 200 250 300 350 400 450 500
-4
-3
-2
-1
0
1
2
3
4
Original information bearing signal and the error signal
Original Signal
Error Signal
Fig 6: Information Signal and Error signal
5 Conclusion
Simulation results for filter length N=32 ,and forgetting factor=1shows that if
Speech enhancement is done by using RLS algorithm and Filter order should be
taken high ,in this The RLS algorithm converges quickly, but its complexity
grows with the square of the number of weights, roughly speaking. This algorithm
can also be unstable when the number of weights is large
101 The Applications and Simulation
of

References
.
[1] Emmanuel C.ifehor, Barrie W.jervis Digital Signal Processing a Pactical
Approach" II edition, Pearson Education .ISBN81-7808-609-3.
[2] Simon Haykin, AdaptiveFilterTheory, fourth edition, Pearson
Education .ISBN978-81-317-0869-9.
[3] Bernard widrow, Samuel D.Stearns. Adaptive Signal Processing third
Impression, (2009), Pearson Education. ISBN 978-81-317-0532-2
[4] Ying He, Hong He, Li Li, Yi Wu The Applications and Simulation of
Adaptive Filter in Noise Canceling (2008) International Conference on
computer Science and Software Engineering pp-1-4.
[5] Gnitecki1, Z. Moussavi, H. Pasterkamp, Recursive Least Squares Adaptive
Noise Cancellation Filtering for Heart Sound Reduction in Lung Sounds
Recordings (2003) IEEE proc pp 2416-2419,
[6] Ondracka J., Oravec R., Kadlec J., Cocherov E. Simulation of RLS and
LMS Algorithms for Adaptive Noise Cancellation in MATLAB.
[7] Honing. M., Messerschmitt. D, 1984, Adaptive Filters: Structures,
Algorithms and Applications, ISBN-13: 978-0898381634, Kluwer
Academic Publishers, pp. 1061- 1064.