Scribd
Upload
83 views25 pages

My

This document provides an overview of a novel RLS-based adaptive filtering method for speech enhancement proposed by Rakesh Pogula. It introduces the goals of speech enhancement and common noise reduction techniques. Adaptive filtering and various algorithms are discussed, including LMS, NLMS, and RLS. The document proposes a new RLS algorithm for speech enhancement and provides simulation results demonstrating its improved performance over NLMS in terms of lower MSE, higher SNR, and lower SNR loss.

Uploaded by

Rakesh Pogula
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
83 views25 pages

My

This document provides an overview of a novel RLS-based adaptive filtering method for speech enhancement proposed by Rakesh Pogula. It introduces the goals of speech enhancement and common noise reduction techniques. Adaptive filtering and various algorithms are discussed, including LMS, NLMS, and RLS. The document proposes a new RLS algorithm for speech enhancement and provides simulation results demonstrating its improved performance over NLMS in terms of lower MSE, higher SNR, and lower SNR loss.

Uploaded by

Rakesh Pogula
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 25

A NOVEL RLS BASED ADAPTIVE

FILTERING
METHOD FOR SPEECH
ENHANCEMENT
By
Rakesh Pogula
Research scholar
Department of
ECE
NIT Warangal,
India

Contents

Introduction : Speech Enhancement


Motivation
Adaptive Filtering
Adaptive Noise Canceller
LMS, NLMS Algorithm
Proposed RLS Algorithm
Simulation Results
Conclusion
References

Introduction

The aim of speech enhancement is to improve


speech quality and intelligibility of degraded
speech
signal
using
speech
processing
techniques.
The approach to speech enhancement varies
considerably, depending on the type of
degradation.
Noise reduction, is the most important field of
speech enhancement, and used for many
applications such as mobile phones, VoIP,
teleconferencing systems, and hearing aids.

Motivation

Noise can severely reduce conversation quality.


Thus adaptive cancellation of such noises has
become
inherent
in
hands-free
telecommunication.
In order to enhance the speech quality, adaptive
filters are extensively used. Recursive algorithms
are applied in these types of filters, which adapt
dynamically to the random behavior of speech.

Adaptive Filters

An adaptive filter is a digital filter with


coefficients that are determined and updated by
an adaptive algorithm.
It is best suited when signal conditions are slowly
varying.

Continued

Applications of Adaptive
Filters
NOISE
SYSTEM
SIGNAL
CANCELLATIO
IDENTIFICATIO
PREDICTION :
N:
N:
Used to provide
Subtracts noise
Designs an
a prediction of
from received
adaptive filter
the present
signal
that provides an
value of a
adaptively to
approximation
random signal
improve SNR.
for an unknown
system.

ECHO
CANCELLATIO
N:
Used to cancel
unknown
interference
from a primary
signal

Adaptive Noise Canceller

Noise Cancellation is a variation of optimal


filtering that involves producing an estimate of
the noise by filtering the reference input and then
subtracting this noise estimate from the primary
input containing both signal and noise.

Various Adaptive
Algorithms for Noise
Properties of an ideal algorithm:
Cancellation

Practical to implement
Adapt to coefficient quickly to minimize error
Provide the Desired Performance

Different algorithms used are:


Least Mean Squares (LMS) algorithm
The Normalized Least Mean Squares(NLMS)
algorithm
The Recursive Least Squares (RLS) algorithm

LMS Algorithm

Adjusts the weight w(k) of the filter


Adaptively adjusts the filter tap weights according
to equation:
w(k+1)=w(k)+e(k)x(k)
Acts as negative feedback to minimize error
signal
It is robust in nature
Slow in convergence and sensitive to variations in
step size parameter

LMS Algorithm Steps

Initialization of filter weight coefficients vector,


(1)

where P is the filter order.

Calculate the output signal y(k) from the adaptive filter


(2)

Calculate the error signal e(k) by using the following equation


e(k) = d(k)-y(k)
(3)
Updates the filter coefficients
(4)by using the following equation
where is defined as the step-size parameter which affects the
convergence behavior of the filter weights.

Advantages and
Disadvantages
ADVANTAGES

It is simple in
implementation.
Stable and robust
performance.

DISADVANTAGE
S

Slow Convergence.

NLMS Algorithm

In structural terms NLMS filter is exactly same as


a standard LMS filter.
From one iteration to the next, the weight of an
adaptive filter should be changed in a minimal
manner.
One of the drawback of LMS is selection of step
size parameter .
In order to solve this difficulty, we can use the
NLMS (Normalized Least Mean Square) algorithm.
y[k ] xTk w k is normalised.
Here the step size, parameter
e[k ] d [k ] y[k ]
NLMS update equations:
w k 1 w k

x k e[k ]
T
xk xk c

NLMS Parameters

So the NLMS algorithm is a time varying step-size


algorithm, calculating the convergence factor as :
(k) =

2
c||x(k )||

Here is the adaptation constant, which optimizes


the convergence rate of the algorithm
Range of alpha is: 0<<2
c is the constant term for normalization and always
c< 1.
The updated filter weight is:

Advantages and
Disadvantages
As here is normalized this
ADVANTAGES

DISADVANTAGE
S

algorithm converges faster


than LMS.
Here estimated error value
between the desired signal
and filter output is less than
LMS.

But LMS is less complex than


NLMS and more stable.

Proposed RLS Algorithm

Steps involved in the RLS Algorithm:


1. Initialize the algorithm by setting
1
w(0) 0 and P(0) I
where = small positive constant for high
SNR and large positive constant for low SNR
2. calculate the output signal y(k)
3. calculate error signal e(k)=d(k)-y(k)
4. update filter coefficients w(k+1)=w(k)
+e(k)R(k)
Where R(k) is the gain vector.

Advantages and
Disadvantages
ADVANTAGES

DISADVANTAGES

RLS converges faster than LMS


and NLMS.
Its noise cancellation capacity is
the most.

This is the most complex


algorithm.

Speech Enhancement using


RLS Algorithm
Signal
Input
Output

Signal
Signal
Segmented

RLS
Filter

Adaptive Coefficient
Calculation

Block diagram of the proposed RLS


method

Simulation Results

Noise cancellation for pink noise at


10dB SNR using NLMS and proposed
RLS algorithm.

Spectrogram comparison for pink noise at 10dB


SNR using NLMS and proposed RLS algorithm.

Performance Measures

Signal-to-Noise Ratio (SNR): SNR is defined as


the ratio of signal power to the noise power
corrupting the signal. The higher the ratio the better
the performance in noise cancellation.
Mean Squared Error (MSE): The MSE is defined as
the average of the square of the difference between
the desired response and the actual system output.
SNR Loss: SNR loss is the increased signal-to-noise
ratio required by an individual to understand speech
in noise, as compared to normal performance. It is a
new
objective
measure
for
predicting
the
intelligibility of noise-suppressed speech.

Comparison of NLMS and proposed RLS


algorithm for Different Noise Types
based on MSE, SNR and SNR Loss
Values
Type of
SNR
NLMS
RLS
Noise

(dB)

AWGN
noise

-5dB

0.00233

7.7885

0.6216

3.1078x10-4

15.6976

0.4268

0dB

8.6265x10-4

11.6244

0.5833

2.8124x10-4

16.1172

0.4206

5dB

3.3883x10-4

15.4984

0.5631

2.6833x10-4

16.3762

0.4249

10dB

2.9533x10-4

15.9010

0.5330

2.6470x10-4

16.4280

0.4357

-5dB

0.0023

7.6840

0.6932

2.6419x10-4

16.3881

0.4475

0dB

9.7817x10-4

10.9904

0.6368

2.6378x10-4

16.3946

0.4535

5dB

5.4487x10-4

13.3827

0.5873

2.6350x10-4

16.3990

0.4533

10dB

3.4010x10-4

15.3756

0.5428

2.6310x10-4

16.4054

0.4522

-5dB

0.0013

9.8287

0.7125

3.5529x10-4

15.1027

0.4826

0dB

7.1655x10-4

12.2341

0.6601

3.5211x10-4

15.1402

0.4884

5dB

4.8776x10-4

13.8308

0.6174

3.5042x10-4

15.1598

0.4972

10dB

3.5229x10-4

15.0183

0.5789

3.4904x10-4

15.1754

0.5034

Pink
noise

Car
noise

MSE

SNR

SNR
Loss

MSE

SNR

SNR
Loss

Performance Comparison for


AWGN noise based on MSE,
SNR and SNR Loss

Conclusion

The performance of the proposed RLS algorithm is


compared with the NLMS algorithm. Based on the
simulation results, we can deduce that the
proposed method improves the performance in
quality of noisy speech signal.

References

Yang Liu, Mingli Xiao and Yong Tie, A Noise Reduction Method Based on LMS
Adaptive Filter of Audio Signals, 3rd International Conference on Multimedia
Technology (ICMT 2013), pp. 1001-1008, November 2013.
Kaur. H andTalwar. R, Performance comparison of adaptive filter algorithms
for noise cancellation, 2013 International Conference on Emerging Trends in
Communication, Control, Signal Processing & Computing Applications
(C2SPCA), pp. 1-5, October 2013.
Paulo S.R. Diniz, Adaptive Filtering: Algorithms and Practical Implementation,
Springer, 2013.
Jianfen Maa and Philipos C. Loizou, SNR loss: A new objective measure for
predicting
the
intelligibility
of
noise-suppressed
speech,
Speech
Communication, Vol. 53, No. 3, pp. 340354, March 2011.
Komal R. Borisagar and Dr. G.R.Kulkarni, Simulation and Comparative Analysis
of LMS and RLS Algorithms Using Real Time Speech Input Signal, Global
Journal of Researches in Engineering, Vol. 10, No. 5, pp. 44-47, October 2010.
Jingdong C., Jacob B. and Arden Huang., On the optimal linear filtering
techniques for noise reduction, Speech Communications, Vol. 49, No. 4, , pp.
305-316, April 2007.
J. H. Husoy and M. S. E. Abadi, A comparative study of some simplified RLS
type algorithm, Proceedings of International Symposium on control,
Communications and Signal Processing, pp.705-708, March 2004.

THANK YOU

You might also like