University of South Florida

Digital Commons @ University of

South Florida

USF Tampa Graduate Theses and Dissertations USF Graduate Theses and Dissertations

10-15-2003

Implementation of an Acoustic Echo Canceller Using Matlab

Srinivasaprasath Raghavendran
University of South Florida

Follow this and additional works at: https://digitalcommons.usf.edu/etd

Part of the American Studies Commons

Scholar Commons Citation

Raghavendran, Srinivasaprasath, "Implementation of an Acoustic Echo Canceller Using Matlab" (2003).
USF Tampa Graduate Theses and Dissertations.
https://digitalcommons.usf.edu/etd/1453

This Thesis is brought to you for free and open access by the USF Graduate Theses and Dissertations at Digital
Commons @ University of South Florida. It has been accepted for inclusion in USF Tampa Graduate Theses and
Dissertations by an authorized administrator of Digital Commons @ University of South Florida. For more
information, please contact digitalcommons@usf.edu.
 Implementation of an Acoustic Echo Canceller

Using Matlab

by

Srinivasaprasath Raghavendran

A thesis submitted in partial fulfillment

of the requirements for the degree of
Master of Science in Electrical Engineering
Department of Electrical Engineering
College of Engineering
University of South Florida

Major Professor: Wilfrido A. Moreno, Ph.D.

James T. Leffew, Ph.D.
Wei Qian, Ph.D.

Date of Approval:
October 15, 2003

Keywords: aec, nlms, dtd, nlp, matlab

© Copyright 2003, Srinivasaprasath Raghavendran

 ACKNOWLEDGEMENTS

I would like to express my sincere gratitude to my Major Professor Dr. Wilfrido

A. Moreno, for being a constant source of help and inspiration throughout my work. His

timely advice and guidelines have assisted me to get through a lot of difficult situations.

My other committee members, Dr. James T. Leffew and Dr. Wei Qian have been very

considerate and cooperative with me. I would like to thank them for their prompt

feedback and being approachable and available whenever I needed any assistance. I

would also like to thank the IEC forum for the help and valuable suggestions.

At this juncture, I thank my parents, my sister and my friends for their total

support and encouragement. This Master’s thesis would not have been possible without

their support.
 TABLE OF CONTENTS

LIST OF TABLES iii

LIST OF FIGURES iv

ABSTRACT vi

CHAPTER 1 INTRODUCTION 1
1.1 Need for Echo Cancellation 1
1.2 Basics of Echo 2
1.3 Types of Echo 3
1.4 The Process of Echo Cancellation 3
1.4.1 Adaptive Filter 4
1.4.2 Doubletalk Detector 5
1.4.3 Nonlinear Processor 5
1.5 Echo Cancellation Challenges 5
1.5.1 Avoiding Divergence 6
1.5.2 Handling Doubletalk 6
1.5.3 Preventing Clipping 7
1.6 Research Motivation and Thesis Outline 7

CHAPTER 2 ECHOES IN TELECOMMUNICATION NETWORKS 9

2.1 Hybrid / Electrical Echo 9
2.2 Acoustic Echo 11
2.3 Long Distance Calls between Fixed Telephones 12
2.3.1 Echo Suppressors 14
2.4 Full-duplex Data Transmission between Voice-band Modems 15
2.5 Short Distance Connections between Fixed and Cellular Lines 17
2.6 Teleconference/Videoconference Communication Systems 18

CHAPTER 3 THE ECHO CANCELLATION ALGORITHM 21

3.1 Basic Echo Canceller 21
3.2 Components of Acoustic Echo Canceller (AEC) 24
3.3 Adaptive Filtering 25
3.3.1 Least Mean Square (LMS) Algorithm 26
3.3.1.1 Generic LMS Algorithm 27
3.3.2 Normalized Least Mean Square (NLMS) Algorithm 29

i
 3.4 Doubletalk Detector (DTD) 31
3.4.1 The Generic Doubletalk Detection Schemes 33
3.4.2 The Geigel Algorithm 34
3.4.3 Cross Correlation Method 35
3.4.4 Normalized Cross Correlation Method 36
3.5 Nonlinear Processor (NLP) 37
3.5.1 Noise Gate as NLP 38
3.5.2 A Generic Expander 38
3.5.3 Noise Gate 40

CHAPTER 4 SIMULATION AND RESULTS 43

4.1 Why MATLAB? 43
4.2 Simulation Flowchart 44
4.3 Description of the Simulation Setup 46
4.4 Results 46
4.5 Evaluation of the Echo Cancellation Algorithm 52
4.5.1 Convergence Test 53
4.5.2 Echo Return Loss Enhancement (ERLE) 53
4.5.3 Auditory Test 54

CHAPTER 5 CONCLUSION AND FURTHER WORK 55

5.1 Conclusion 55
5.2 Further Work 56

REFERENCES 57

ii
 LIST OF TABLES

Table 3.1: LMS Algorithm 29

Table 3.2: NLMS Algorithm 31

iii
 LIST OF FIGURES

Figure 1.1: Block Diagram of a Generic Echo Canceller 4

Figure 2.1: Hybrid Echo 10

Figure 2.2: Sources of Acoustic Echo in a Room 12

Figure 2.3: Simplified Long Distance Connections 13

Figure 2.4: Echo Suppressor at Near-end Talker B Path 15

Figure 2.5: Echo Canceller at Modem Locations for Full-Duplex Voice-band Modems 16

Figure 2.6: Cellular to Fixed Telephone Connection 18

Figure 2.7: Adaptive Acoustic Echo Cancellation in an Enclosed Environment 19

Figure 3.1: A Basic Echo Canceller 21

Figure 3.2: A Generic Adaptive Echo Canceller 22

Figure 3.3: Echo Canceller with Doubletalk Detector and Nonlinear Processor 24

Figure 3.4: LMS Algorithm 26

Figure 3.5: Generic LMS Algorithm 27

Figure 3.6: Basic Block Diagram of an Expander 38

Figure 3.7: Input / Output Characteristics of an Expander 40

Figure 3.8: The Effect of an Expander on a Signal 42

Figure 4.1: Flowchart of the MATLAB Simulation 45

Figure 4.2: Plot of the Far-end Signal, x(n) 47

iv
Figure 4.3: Plot of the Echo Signal, r(n) 48

Figure 4.4: Plot of the Near-end Signal, v(n) 49

Figure 4.5: Plot of the Desired Signal, d(n) 49

Figure 4.6: Plot of the Error Signal e(n) 51

Figure 4.7: Plot of the Error Signal after Nonlinear Processing 52

Figure 4.8: Plot of ERLE Vs the Number of Samples 54

v
 IMPLEMENTATION OF AN ACOUSTIC ECHO CANCELLER

USING MATLAB

Srinivasaprasath Raghavendran

ABSTRACT

The rapid growth of technology in recent decades has changed the whole dimension of

communications. Today people are more interested in hands-free communication. In

such a situation, the use a regular loudspeaker and a high-gain microphone, in place of a

telephone receiver, might seem more appropriate. This would allow more than one

person to participate in a conversation at the same time such as a teleconference

environment. Another advantage is that it would allow the person to have both hands

free and to move freely in the room. However, the presence of a large acoustic coupling

between the loudspeaker and microphone would produce a loud echo that would make

conversation difficult. Furthermore, the acoustic system could become instable, which

would produce a loud howling noise to occur.

The solution to these problems is the elimination of the echo with an echo

suppression or echo cancellation algorithm. The echo suppressor offers a simple but

effective method to counter the echo problem. However, the echo suppressor possesses a

main disadvantage since it supports only half-duplex communication. Half-duplex

communication permits only one speaker to talk at a time. This drawback led to the

vi
invention of echo cancellers. An important aspect of echo cancellers is that full-duplex

communication can be maintained, which allows both speakers to talk at the same time.

This objective of this research was to produce an improved echo cancellation

algorithm, which is capable of providing convincing results. The three basic components

of an echo canceller are an adaptive filter, a doubletalk detector and a nonlinear

processor. The adaptive filter creates a replica of the echo and subtracts it from the

combination of the actual echo and the near-end signal. The doubletalk detector senses

the doubletalk. Doubletalk occurs when both ends are talking, which stops the adaptive

filter in order to avoid divergence. Finally, the nonlinear processor removes the residual

echo from the error signal. Usually, a certain amount of speech is clipped in the final

stage of nonlinear processing. In order to avoid clipping, a noise gate was used as a

nonlinear processor in this research. The noise gate allowed a threshold value to be set

and all signals below the threshold were removed. This action ensured that only residual

echoes were removed in the final stage. To date, the real time implementation of echo an

cancellation algorithm was performed by utilizing both a VLSI processor and a DSP

processor. Since there has been a revolution in the field of personal computers, in recent

years, this research attempted to implement the acoustic echo canceller algorithm on a

natively running PC with the help of the MATLAB software.

vii
 CHAPTER 1

INTRODUCTION

1.1 Need for Echo Cancellation

In this new age of global communications, wireless phones are regarded as

essential communications tools and have a direct impact on people’s day-to-day personal

and business communications. As new network infrastructures are implemented and

competition between wireless carriers increases, digital wireless subscribers are

becoming ever more critical of the service and voice quality they receive from network

providers. Subscriber demand for enhanced voice quality over wireless networks has

driven a new and key technology termed echo cancellation, which can provide near wire

line voice quality across a wireless network.

Today’s subscribers use speech quality as a standard for assessing the overall

quality of a network. Regardless of whether or not the subscribers’ opinion is subjective,

it is the key to maintaining subscriber loyalty. For this reason, the effective removal of

hybrid and acoustic echoes, which are inherent within the telecommunications network

infrastructure, is the key to maintaining and improving the perceived voice quality of a

call. Ultimately, the search for improved voice quality has led to intensive research into

the area of echo cancellation. Such research is conducted with the aim of providing

solutions that can reduce background noise and remove hybrid and acoustic echoes

1
before any transcoder processing occurs. By employing echo cancellation technology,

the quality of speech can be improved significantly. This chapter discusses the overall

echo problem. A definition of echo precedes the discussion of the fundamentals of echo

cancellation and the voice quality challenges encountered in today’s networks.

1.2 Basics of Echo

Echo is a phenomenon where a delayed and distorted version of an original sound

or electrical signal is reflected back to the source. With rare exceptions, conversations

take place in the presence of echoes. Echoes of our speech are heard as they are reflected

from the floor, walls and other neighboring objects. If a reflected wave arrives after a

very short time of direct sound, it is considered as a spectral distortion or reverberation.

However, when the leading edge of the reflected wave arrives a few tens of milliseconds

after the direct sound, it is heard as a distinct echo [1].

Since the advent of telephony echoes have been a problem in communication

networks. In particular, echoes can be generated electrically due to impedance

mismatches at various points along the transmission medium. The most important factor

in echoes is called end-to-end delay, which is also known as latency. Latency is the time

between the generation of the sound at one end of the call and its reception at the other

end. Round trip delay, which is the time taken to reflect an echo, is approximately twice

the end-to-end delay.

Echoes become annoying when the round trip delay exceeds 30 ms. Such an echo

is typically heard as a hollow sound. Echoes must be loud enough to be heard. Those

less than thirty (30) decibels (dB) are unlikely to be noticed. However, when round trip

2
delay exceeds 30 ms and echo strength exceeds 30 dB, echoes become steadily more

disruptive. However, not all echoes reduce voice quality. In order for telephone

conversations to sound natural, callers must be able to hear themselves speaking. For this

reason, a short instantaneous echo, termed side tone, is deliberately inserted. The side

tone is coupled with the caller’s speech from the telephone mouthpiece to the earpiece so

that the line sounds connected.

1.3 Types of Echo

In telecommunications networks there are two types of echo. One source for an

echo is electrical and the other echo source is acoustic [1]. The electrical echo is due to

the impedance mismatch at the hybrids of a Public Switched Telephony Network,

(PSTN), exchange where the subscriber two-wire lines are connected to four-wire lines.

If a communication is simply between two fixed telephones, then only the electrical echo

occurs. However, the development of hands-free teleconferencing systems gave rise to

another kind of echo known as an acoustic echo. The acoustic echo is due to the coupling

between the loudspeaker and microphone. These electrical and acoustic echoes are

discussed in greater detail in chapter 2.

1.4 The Process of Echo Cancellation

An echo canceller is basically a device that detects and removes the echo of the

signal from the far end after it has echoed on the local end’s equipment. In the case of

circuit switched long distance networks, echo cancellers reside in the metropolitan

3
Central Offices that connect to the long distance network. These echo cancellers remove

electrical echoes made noticeable by delay in the long distance network.

An echo canceller consists of three main functional components:

• Adaptive filter

• Doubletalk detector

• Non-linear processor

A brief overview of these components is presented in this chapter. However, a

detailed sketch that involves mathematical illustrations is provided in chapter 3.

Input signal x(n)

Doubletalk
Filtered signal Clear signal e(n)
Doubletalk decision Non-Linear
Adaptive Filter
detector Processor

Reference signal
y(n)

Figure 1.1: Block Diagram of a Generic Echo Canceller

1.4.1 Adaptive Filter

The adaptive filter is made up of an echo estimator and a subtractor. The echo

estimator monitors the received path and dynamically builds a mathematical model of the

line that creates the returning echo. The model of the line is convolved with the voice

stream on the receive path. This yields an estimate of the echo, which is applied to the

4
subtractor. The subtractor eliminates the linear part of the echo from the line in the send

path. The echo canceller is said to converge on the echo as an estimate of the line is built

through the adaptive filter.

1.4.2 Doubletalk Detector

A doubletalk detector is used with an echo canceller to sense when far-end speech

is corrupted by near-end speech. The role of this important function is to freeze

adaptation of the model filter when near-end speech is present. This action prevents

divergence of the adaptive algorithm.

1.4.3 Nonlinear Processor

The non-linear processor evaluates the residual echo, which is nothing but the

amount of echo left over after the signal has passed through the adaptive filter. The

nonlinear processor removes all signals below a certain threshold and replaces them with

simulated background noise which sounds like the original background noise without the

echo.

1.5 Echo Cancellation Challenges

An echo canceller has to deal with a number of challenges in order to perform

robust echo cancellation.

5
1.5.1 Avoiding Divergence

The process of divergence is an adaptive filter problem that arises when a suitable

solution for the line model is not found through the use of a mathematical algorithm.

Under specific conditions, certain algorithms are bound to diverge and corrupt the signal

or even add echo to the line. Good echo cancellers are tuned to avoid divergence

situations in nearly all conditions.

1.5.2 Handling Doubletalk

In an active conversation, both talkers often speak at the same time or interrupt

each other. Those situations are called “doubletalk”. Doubletalk presents a special

processing challenge to echo cancellers. Taken step-by-step, doubletalk proceeds as

follows:

1. A speaks. The echo canceller must compare the received speech from

Speaker A to what would be transmitted back to A in order to approximate

an echo point.

2. B speaks over the echo signal. B speaking constitutes doubletalk. The

echo canceller must detect the doubletalk and cancel the echo without

affecting what is heard locally, which is speaker B’s words.

3. The echo canceller must send B’s speech, as well as the echo-cancelled

version of A’s own speech, back to A.

Handling doubletalk so that it sounds natural is technically challenging. A good

echo canceller must be able to do the following:

6
 • It must detect doubletalk and distinguish it from background noise.

• The echo canceller must be capable of choosing not to update the line

model in order to avoid divergence if divergence could result.

• It needs to make a smooth transition between doubletalk detection,

processing of doubletalk and return to the normal mode.

In summary, an important requirement for echo cancellation is the handling of

doubletalk in a natural manner that does not cause divergence.

1.5.3 Preventing Clipping

Clipping occurs during a telephone conversation when part of the speech is

erroneously removed. Clipping results due to the lack of a precise Non-Linear Processor,

(NLP). Specifically, the NLP fails to start and stop at the right time. Typically, an NLP

does not respond rapidly enough to the introduction of speech through the local end. It

replaces parts of words with background noise, which makes the conversation hard to

follow. The same can happen when the NLP confuses the fading of the voice level at the

end of a sentence with a residual echo.

1.6 Research Motivation and Thesis Outline

Since echo cancellation is a very demanding process, real-time implementation

has only been possible through the use of custom very large scale integration, (VLSI),

processors or digital signal processors (DSP). These processors are specially designed

for signal processing tasks. They provide parallel processing of commands and

optimized pipeline structures. However, since the computation power of regular home

7
personal computers, (PCs), has increased tremendously and powerful software has

evolved, it is now possible to perform real-time signal processing in the PC environment

as well. The advent of this growing capability was the motivation for this research. The

objective of the research was the implementation of a software echo canceller running

natively on a PC with the help of the MATLAB software.

This thesis provides an overview of an improved echo cancellation technique

using a noise gate for the NLP. Chapter 1 discusses the definition of echo, the necessity

of echo cancellers in telecommunications network, the basics of echo cancellation and the

challenges of echo cancellation. Chapter 2 gives an overview of the types of echo and

their sources. It also discusses, in great detail, the echo phenomena in four major

telecommunication systems. The proposed echo cancellation algorithm is explained step-

by-step in chapter 3. Chapter 4 discusses the simulation of the proposed algorithm,

details of the simulation environment and the results obtained. Finally Chapter 5

provides a summary and some ideas concerning further work in this field.

8
 CHAPTER 2

ECHOES IN TELECOMMUNICATION NETWORKS

This chapter deals with echoes that are generated in telecommunication systems.

As discussed in chapter one, there are two main types of echo, which are termed

electrical, or hybrid, and acoustic.

2.1 Hybrid/Electrical Echo

Hybrid echoes have been inherent within the telecommunications networks since

the advent of the telephone. This echo is the result of impedance mismatches in the

analog local loop. For example, this happens when mixed gauges of wires are used, or

where there are unused taps and loading coils. In the Public Switched Telephone

Network, (PSTN), by far the main source of electrical echo is the hybrid. This hybrid is a

transformer located at a juncture that connects the two-wire local loop coming from a

subscriber’s premise to the four-wire trunk at the local telephone exchange. The four-

wire trunks connect the local exchange to the long distance exchange. This situation is

illustrated in Figure 2.1.

9
 Hybrid
Echo

4W Trans Port

Hybrid Balance
2W Port Deivce Network

4W Recv Port

Figure 2.1: Hybrid Echo

The hybrid splits the two-wire local loop into two separate pairs of wires. One

pair is used for the transmission path and the other for the receiver path. The hybrid

passes on most of the signal. However, the impedance mismatch between the two-wire

loop and the four-wire facility causes a small part of the received signal to “leak” back

onto the transmission path. The speaker hears an echo because the far-end receives the

signal and sends part of it back again. Electrical echo is definitely not a problem on local

calls since the relatively short distances do not produce significant delays. However, the

electrical echo must be controlled on long distance calls.

In the early years, when the public network was entirely circuit switched, the

hybrid echo was the only significant source of echo. Since the locations of hybrids and

most other causes of impedance differences in circuit switched networks were known,

adequate echo control could be planned and provisioned. However, in today’s digital

networks the points where two wires split into four wires is typically also the point where

analog to digital conversion takes place. Regardless of whether the hybrid and analog to

10
digital conversion is implemented in the same device or in two devices, the two to four

wire conversions constitute an impedance mismatch and echoes are produced [1].

2.2 Acoustic Echo

The acoustic echo, which is also known as a “multipath echo”, is produced by

poor voice coupling between the earpiece and microphone in handsets and hands-free

devices. Further voice degradation is caused as voice-compressing and

encoding/decoding devices process the voice paths within the handsets and in wireless

networks. This results in returned echo signals with highly variable properties. When

compounded with inherent digital transmission delays, call quality is greatly diminished

for the wireline caller.

Acoustic coupling is due to the reflection of the loudspeaker’s sound waves from

walls, door, ceiling, windows and other objects back to the microphone. The result of the

reflections is the creation of a multipath echo and multiple harmonics of echoes, which

are transmitted back to the far-end and are heard by the talker as an echo unless

eliminated. Adaptive cancellation of such acoustic echoes has become very important in

hands-free communication systems such as teleconference or videoconference systems

[1]. The multipath echo phenomenon is illustrated in Figure 2.2.

11
 Loudspeaker

Reflections
Direct
coupling

Microphone

Figure 2.2: Sources of Acoustic Echo in a Room

In the following sections, the echo phenomena of four communication systems

will be described. The communication systems are:

• Long-distance connections between fixed telephones

• Full-duplex data transmission between voice-band modems

• Short-distance connections between fixed and cellular telephones

• Teleconference/videoconference systems

2.3 Long Distance Calls between Fixed Telephones

A simple long-distance telephone connection is presented in Figure 2.3. This

connection contains two-wire sections at the ends, the subscriber loops and possibly some

portion of the local network. It also contains a four-wire section in the center, which is a

carrier system for medium-range to long-range transmissions.

12
 Figure 2.3: Simplified Long Distance Connections

Every conventional telephone in a given geographical area is connected to the

local PSTN exchange by a two-wire line, called the subscriber loop, which carries a

connection for both directions of transmission. Simply connecting the two subscriber

loops at the local exchange sets up a local call. However, amplification of the speech

signal becomes necessary when the distance between the two telephones exceeds 35

miles. Therefore, a four-wire line is required, which segregates the two directions of

transmission. A hybrid is used to convert from the two-wire to four-wire line and vice

versa.

An echo can be decreased if the hybrid has a significant loss between its two four-

wire ports. To achieve this large loss the hybrid has to be perfectly balanced by

impedance located at its four-wire portion. Unfortunately, this is not possible in practice

since it requires knowledge of the two-wire impedance, which varies considerably over

the population of subscriber loops. When the bridge is not perfectly balanced, impedance

mismatch occurs. This causes some of the talker’s signal energy to be reflected back as

an echo. Adding an insertion loss to the four-wire portions of the connection can control

the effects of echo. Such action is effective since the echo signals experience this loss

two or three times while the talker’s speech suffers this loss only once. However, on

long-range connections the insertion loss can become very significant. Hence, it is not a
13
favorable solution and other echo control techniques such as echo suppression must be

used [1].

2.3.1 Echo Suppressors

Echo suppressors have been used since the introduction of long distance

communication. This device basically takes advantage of the fact that people seldom talk

simultaneously. The situation of two people talking simultaneously is termed “double

talking”. The echo suppressor is also helped by the fact that during such double talking

poor transmission quality is less noticeable. Figure 2.4 illustrates how the echo

suppressor dynamically controls the connection based on who is talking, which is decided

by the speech and double talking detector. Double talking is detected if the level of the

signal in path L1 is significantly lower than that in path L2. When the far-end talker A is

speaking, the path used to transmit the near-end speech is opened so that the echo is

prevented. Then, when the near-end talker B speaks, the same switch is closed and a

symmetric one at the far-end talker A’s path is opened. However, echo suppressors can

clip speech sounds and introduce impairing interruption. For example, if talker B is

initially listening to talker A but suddenly wants to talk, it is quite likely that the switch

preventing talker A’s echo from being transmitted will not close quickly enough. This

will cause the far-end talker A to not be able to receive all the messages from the near-

end talker B. This deletion is noticed by talker A, encouraging him/her to stop and wait

for talker B to finish. The resulting confusion may stop the conversation entirely while

each party waits for the other to say something [1]. Therefore the best solution for

removing echoes is to use echo cancellers. Echo cancellers are described in chapter 3.

14
 Arriving from Far-end
L1

Echo

Doubletalk detector to
Speech
overdrive echo Hybrid
Detector
suppessor
Near-end
Talker B

Echo
Suppessor
L2

Destined for
Far-end
Talker B Signal

Figure 2.4: Echo Suppressor at Near-end Talker B Path

2.4 Full-duplex Data Transmission between Voice-band Modems

The two-wire telephone line of a subscriber loop can be used for the transmission

of data through a modem. This can be accomplished either by using the entire bandwidth

of the wire or transmitting the data on a bandwidth that is slightly above the one used to

carry the speech signal. On an analog subscriber loop the speech signal occupies the

bandwidth between 300 to 3400 Hz. A higher bit rate of up to 16 kbps can be transmitted

by modulating the data signal onto a carrier signal at a band above 4000 Hz. Echo

cancellation is needed for full-duplex communication within the same bandwidth over the

subscriber loop as shown in Figure 2.5 where EC is the echo canceller, H is the hybrid,

RX is the receiver and TX is the transmitter.

15
 Figure 2.5: Echo Cancellers at Modem Locations for Full-Duplex Voice-band Modems

Typically the echo cancellers must be placed at the line interface where the

hybrids connect the modem to the two-wire subscriber loop. Several problems are

associated with this type of application and some of them are given below.

• It is not practical to freeze the adaptation algorithm during doubletalking in the

case of full-duplex operation since the echo path’s characteristic is likely to

change during a lengthy communication session.

• The far-end echo, which is returned from the far-end hybrid, must also be taken

into account. Therefore, the entire echo delay becomes very large, which is

unique to the echo cancellation at the station, or modem, location. If the circuit

includes a satellite communication network’s four-wire link, the far-end echo will

be delayed for more than 500ms. In such a case two cancellers will be required.

One for the near-end and one for the far-end echo at the modems.

• A significantly high level of echo cancellation is required. The data signal

coming from a far-end modem may be attenuated by 40 to 50dB. Therefore, the

near-end echo, which is returned from the first hybrid at the local station, can be
16
 40 to 50dB higher than the desired signal. For reliable communication the echo

canceller must be able to attenuate the near-end echo by 50 to 60dB in order to

maintain the signal power approximately 10dB above the echo [2].

2.5 Short Distance Connections between Fixed and Cellular Lines

In digital cellular communication, the combination of channel coding, speech

coding and signal processing involves considerable delays. In most cases, the delays are

increased further by time division multiple access framing. The total one-way delay can

be from 30 to 120 msec. Figure 2.6 illustrates that only one echo canceller, (EC), facing

the local PSTN exchange, (LE), is required in a digital cellular to fixed telephone

connection. This is only possible if the cellular telephone is assumed to behave in a

perfect four-wire fashion with no significant acoustic cross talk echo between the

microphone and the earpiece of the cellular phone. However, under certain conditions,

the cross talk echo in cellular handsets is still noticeable by users. Hence, the echo needs

to be removed by cellular cross talk control devices [2].

17
 Figure 2.6: Cellular to Fixed Telephone Connection

2.6 Teleconference/Videoconference Communication Systems

When the telephone connection is between hands-free telephones or between two

conference rooms, then an acoustic echo problem emerges that is due to the reflection of

the loudspeaker’s sound waves from the boundary surfaces and other objects back to the

microphone. This acoustic echo can be removed using an adaptive filter as illustrated in

Figure 2.7. The adaptive filter attempts to synthesize a model of the acoustic echo at its

output.

18
 Enclosed Environment
E.G., A Room or Vehicle

Loudspeaker
From Far-end
x(n)
Talker

Adaptive
Filter r(t)

ŷ(n)
To Far-end Talker
e(n) y(n) Near-end Talker
Σ

Microphone v(n)

Figure 2.7: Adaptive Acoustic Echo Cancellation in an Enclosed Environment

Adaptive acoustic echo cancellation is a more challenging problem than the

network echo cancellation for the following main reasons:

• The impulse response of the acoustic echo path is several times longer,

between 100 to 500 msec. than that of the network echo path.

• The characteristics of the acoustic echo path are more non-stationary due to

opening and closing of a door or movement of people inside the room while

the network echo path is almost stationary.

• The acoustic echo path has a mixture of linear and nonlinear characteristics.

The reflection of acoustic signals inside a room is almost linearly distorted.

However, the loudspeaker does introduce nonlinearity. The main causes of

this nonlinearity are the suspension nonlinearity that affects distortion at low
19
 frequency and the inhomogeneity of flux density that produces nonlinear

distortion at large input signal levels.

Due to the above mentioned reasons, the acoustic echo cancellers, (AECs), are

required to have more computing power in order to compensate for the longer impulse

response and to produce faster converging algorithms [2].

20
 CHAPTER 3

THE ECHO CANCELLATION ALGORITHM

This chapter discusses the echo cancellation algorithm for a VoIP environment.

The basic idea behind the algorithm, its terminology, modes of operation and the

problems addressed by the algorithm are discussed in detail.

3.1 Basic Echo Canceller

A basic echo canceller used to remove echo in telecommunication networks is presented

in Figure 3.1.

Far-end Talker
x(n)

Echo Echo Path

Echo Canceller

ŷ( n ) r(n)

_
e(n) d(n) = r(n) + v(n) + + v(n)
Σ Σ
+
Near-end
Talker

Figure 3.1: A Basic Echo canceller

21
 The echo canceller mimics the transfer function of the echo path in order to

synthesize a replica of the echo. Then the echo canceller subtracts the synthesized replica

from the combined echo and near-end speech or disturbance signal to obtain the near-end

signal. However, the transfer function is unknown in practice. Therefore, it must be

identified. This problem can be solved by using an adaptive filter that gradually matches

its estimated impulse response, ĥ , to that of the impulse response of the actual echo path,

h. This process is illustrated in Figure 3.2. The echo path is highly variable and can even

depend on such things as the movement of people in the room as well as other things.

These variations are accounted for by the adaptive control loop, which is built into the

canceller.

Far-end Talker
x(n)

Adaptive Filter Echo Echo Path

h
ĥ

ŷ( n ) r(n)

_
e(n) d(n) = r(n) + v(n) + + v(n)
Σ Σ
+
Near-end
Talker

Figure 3.2: A Generic Adaptive Echo Canceller

22
 The estimated echo, ŷ(n ) , is generated by passing the reference input signal, x(n),

through the adaptive filter, ĥ (n ) , that will ideally match the transfer function of the echo

path, h(n). The echo signal, r(n), is produced when x(n) passes through the echo path.

The echo r(n) plus the near-end talker or disturbance signal, v(n), constitute the desired

response,

d(n) = r(n)+v(n), (3.1)

for the adaptive canceller. The two signals x(n) and r(n) are correlated since the later is

obtained by passing x(n) through the echo path. The error signal e(n) is given by

e(n) = d(n) - ŷ(n ) . (3.2)

In the ideal case, e(n) = v(n), which represents the case when the adaptive echo canceller

is perfect.

Similar to the echo suppressors, adaptive echo cancellers also face the problem of

double talking when both near and far end speakers talk simultaneously. If double talk

occurs, the system may try to adjust the adaptive filter parameters to imperfectly cancel

the near-end talker signal. This will result in making large corrections to the estimated

echo path, ĥ , in an attempt to mimic h. In order to avoid this possibility the coefficients

in the adaptive filter must not be updated as soon as double talking is detected as

illustrated in Figure 3.3. The design of a good double talking detector is difficult. Even

with the assumption of a fast-acting detector, there is still a possibility of changes

occurring in the echo channel during the time that the echo canceller is not updated,

which leads to increasing amount of uncancelled echoes. Fortunately, the duration of

double talking is usually short. In addition to these problems, it sometimes occurs that a

well-working echo canceller leaves some residual uncancelled echo. In such a case, a
23
nonlinear processor is used to remove the residual echo. The goal of the nonlinear

processor is to block this small unwanted signal if the signal magnitude is lower than a

certain small threshold value during single talking. The nonlinear processor will only

distort and not block the near-end signal during double talking. The distortion is

generally unnoticeable and the processor does not have to be removed during double

talking [2].

Far-end Talker
x(n)

Adaptive Filter Echo Path

h
ĥ
Echo

Open
during
double talk
ŷ (n ) r(n)
Doubletalk
Detector
_
+ + v(n)
Nonlinear
Processor
Σ Σ
+ d(n) = r(n) + v(n) Near-end
e(n)
Talker

Figure 3.3: Echo Canceller with Doubletalk Detector and Nonlinear Processor

3.2 Components of an Acoustic Echo Canceller (AEC)

The previous section attempted to give some valuable first hand knowledge on the

functioning of a basic echo canceller. The following sections offer a detailed theoretical

24
and mathematical account of the three fundamental components of echo cancellers. The

three fundamental components that combine to form an echo canceller are:

1. Adaptive Filter

2. Doubletalk Detector

3. Nonlinear Processor

3.3 Adaptive Filtering

As previously demonstrated, the best solution for reducing the echo is to use some

form of adaptive algorithm. The theory behind such an algorithm and the reasons for

choosing that algorithm will be described in this section. Basically filtering is a signal

processing technique whose objective is to process a signal in order to manipulate the

information contained in the signal. In other words, a filter is a device that maps its input

signal into another output signal by extracting only the desired information contained in

the input signal. An adaptive filter is necessary when either the fixed specifications are

unknown or time-invariant filters cannot satisfy the specifications. Strictly speaking an

adaptive filter is a nonlinear filter since its characteristics are dependent on the input

signal and consequently the homogeneity and additivity conditions are not satisfied.

Additionally, adaptive filters are time varying since their parameters are continually

changing in order to meet a performance requirement. In a sense, an adaptive filter is a

filter that performs the approximation step on line.

25
3.3.1 Least Mean Square (LMS) Algorithm

The least mean square, (LMS), is a search algorithm that is widely used in various

applications of adaptive filtering. The main features that attracted the use of the LMS

algorithm are low computational complexity, proof of convergence in stationary

environments and stable behavior when implemented with finite precision arithmetic.

Figure 3.4 illustrates how such an algorithm works. A path that changes the signal x is

called h. Transfer function of this filter is not known in the beginning. The task of the

LMS algorithm is to estimate the transfer function of the filter. The result of the signal

distortion is calculated by convolution and is denoted by r. In this case r is the echo and

h is the transfer function of the hybrid. The near-end speech signal v is added to the echo.

The adaptive algorithm tries to create a filter w. The transfer function of the filter is an

estimate of the transfer function for the hybrid. This transfer function in turn is used for

calculating an estimate of the echo. The echo estimate is denoted by r̂ .

w h

r̂ r
+ v
Σ
+

_ d= v+r
Σ
+
d − r̂ = v + r − r̂ = v + e

Figure 3.4: LMS Algorithm

26
 The signals are added so that the output signal from the algorithm is

v + r – r̂ = v + e, (3.3)

where e denotes the error signal. The error signal and the input signal x are used for

estimation of the filter coefficient vector w. One of the main problems associated with

choosing the filter weight is that the path h is not stationary. Therefore, the filter weights

must be updated frequently so that the adjustment to the variations can be performed.

The filter is a FIR filter with the form

w = b0 +b1 z-1 + ··· +bL-1 z – (L – 1). (3.4)

A perfect FIR filter is linear, time-invariant and stable in a BIBO sense.

However, in a real-time environment, linearity is never a possibility and the first criterion

is not fulfilled so the filter can never be perfect. Updating of the filter weights is realized

in accordance with

w(k + 1) = w(k) - µ gw(k) (3.5)

for k = 0,1,2,··· where gw(k) represents an estimate of the gradient vector and µ is the

convergence factor or step size.

3.3.1.1 Generic LMS Algorithm [3]

The general case of the LMS algorithm is presented in Figure 3.5.

x w + e
ŷ _ Σ

∆w

LMS

Figure 3.5: Generic LMS Algorithm

27
 Figure 3.5 shows that

e(k) = d(k) – ŷ (k) = d(k) - xT(k) w(k), (3.6)

where w(k) is a vector containing the filter weights [b0, b1, b2, ···, b0] and x(k) represents

the vector [x(n), x(n-1), ···, x(n-L)]T. L is the length of the adaptive filter.

The derivation of the gradient estimate gw(k) is provided next.

The Wiener solution is given by

wo = R-1 p (3.7)

where

R = E [ x(k) xT(k)] (3.8)

and

p = E[d(k) x(k)], (3.9)

assuming d(k) and x(k) are jointly wide sense stationary. If good estimates of the matrix

R, denoted by R̂ (k), and of vector p, denoted by p̂ (k), are available, a steepest-descent

based algorithm can be used to search the Wiener solution is as follows

w(k + 1) = w(k) - µ gw(k)

= w(k) + 2 µ ( p̂ (k) – R̂ (k)w(k)). (3.10)

One possible solution is to estimate the gradient vector by employing instantaneous

estimates for R and p, which are given by:

R̂ (k) = x(k) xT(k), (3.11)

and

p̂ (k) = d(k) x(k). (3.12)

Then the gradient estimate gw(k) is given by

28
 gw(k ) = 2d(k )x(k) + 2x(k)xT(k)w(k)

= 2x(k)(d(k) + xT(k)w(k))

= 2e(k)x(k). (3.13)

The resulting gradient-based algorithm is known. It minimizes the mean of the squared

error, as the least-mean square (LMS) algorithm, whose updating equation is given by

w(k+1) = w(k) + 2 µ e(k)x(k). (3.14)

Table 3.1 presents the steps associated with the LMS algorithm in tabular form.

Table 3.1: LMS Algorithm

Initial Condition x(0) = w(0) = [0,···,0]T

For each instant of time, k = 1, 2, ···, compute

Filter output: y(k) =x(k)Tw(k)

Estimation Error: e(k ) = d (k ) − ŷ(k )

Tap-Weight Adaptation: w(k+1) = w(k) + 2 µ e(k)x(k)

3.3.2 Normalized Least Mean Square (NLMS) Algorithm [3]

There are a number of algorithms for adaptive filters, which are derived from the

conventional LMS algorithm. The objective of the alternative LMS-based algorithms is

either to reduce computational complexity or convergence time. The normalized LMS,

(NLMS), algorithm utilizes a variable convergence factor that minimizes the

instantaneous error. Such a convergence factor usually reduces the convergence time but

increases the misadjustment.

29
 The updating equation of the LMS algorithm can employ a variable convergence

factor µk in order to improve the convergence rate. In this case, the updating formula is

expressed as

w(k + 1) = w(k ) + 2 µk e(k )x (k ) = w (k ) + ∆ŵ (k ) , (3.15)

where µk must be chosen with the objective of achieving a faster convergence.

The value of µk is given by

1
µk = . (3.16)
2x̂ (k ) x (k )

Using the variable convergence factor the updating equation for the NLMS algorithm is

given by

e( k ) x ( k )
w(k+1) = w (k ) + . (3.17)
x T (k ) x (k )

Usually a fixed convergence factor µn is introduced in the updating formula in

order to control the misadjustment since all the derivations are based on instantaneous

values of the squared errors and not on the MSE. Also a parameter γ should be included

in order to avoid large steps when xT(k)x(k) becomes small. Then the coefficient updating

is by

2µ n
w(k+1) = w (k ) + e(k ) x (k ) (3.18)
γ + x T (k ) x (k )

30
Table 3.2 presents the steps associated with the NLMS algorithm in tabular form.

Table 3.2: NLMS Algorithm

Initial Condition 0 < µn ≤ 2

x(0) = w(0) = [0,···,0]T
γ = a small constant
For each instant of time, k = 1, 2, ···, compute

Filter output: y(k) = x(k)Tw(k)

Estimation Error: e(k ) = d (k ) − ŷ(k )

2µ n
Tap-Weight Adaptation: w(k+1) = w (k ) + e(k ) x (k )
γ + x T (k ) x (k )

3.4 Double Talk Detector (DTD)

An important characteristic of a good echo canceller is its performance during

double talk. The condition where both ends, the near-end and the far-end, are speaking is

referred to as double talk. If the echo canceller does not detect a double talk condition

properly the near end speech will cause the adaptive filter to diverge. Therefore, it is

important to have a reliable double-talk detector.

A DTD is used with an echo canceller to sense when the far-end speech is

corrupted by the near-end speech. The role of this important function is to freeze

adaptation of the model filter, ĥ , when the near-end speech, v, is present in order to avoid

divergence of the adaptive algorithm. The far-end talker signal, x, is filtered with the

impulse response, h, and the resulting signal. The echo is added to the near-end speech

signal, v, in order to obtain the corrupted signal

31
 d(n) = HTx(n) + v(n) (3.19)

where

H = [H0, H1 , ···, HL-1]T (3.20)

and

x (n) = [x(n), x(n-1),···,·x(n – L + 1)]T. (3.21)

L is the length of the echo path. The error signal at time n is defined by

e(n) = d(n) – Ĥ Tx(n). (3.22)

This error signal is used in the adaptive algorithm to adjust the L taps of the

filter, ĥ . For simplicity it is assumed that the length of the signal vector, x, is the same as

the effective length of the echo path, h. When v is not present, with any adaptive

algorithm, ĥ will quickly converge to an estimate of h, which is the best way to cancel

the echo. When x is not present, or very small, adaptation is halted by the nature of the

adaptive algorithm. When both x and v are present the near-end talker signal could

disrupt the adaptation of ĥ and cause divergence. Therefore, the goal of a double talk

detection algorithm is to stop the adaptation of ĥ when the level of v becomes significant

in relation to the level of x and to keep the adaptation going when the level of v is

negligible [4].

The basic double talk detection process starts with computing a detection statistic

and comparing it with a preset threshold. Different methods have been proposed to form

the detection statistic. The Geigel algorithm has proven successful in line echo

cancellers. However, it does not always provide reliable performance when used in

AEC’s. Cross-correlation based methods appear to be more suitable for AEC

32
applications. However, for the DTD algorithms only heuristic methods have been used to

select the threshold T with little justification for the choice. In addition, there has not

been an objective way to evaluate and compare these methods.

3.4.1 The Generic Doubletalk Detection Schemes

Almost all types of doubletalk detectors operate in the same manner. Therefore,

the general procedure for handling double talk is described by the following four steps.

1. A detection statistic, ξ , is formed using available signals such as x, d and e

and the estimated filter coefficients, ĥ .

2. The detection statistic, ξ , is compared to a preset threshold, T, (a constant),

and double talk is declared if ξ < T.

3. Once doubletalk is declared the detection is held for a minimum period of

time Thold. While the detection is held the filter adaptation is disabled.

4. If ξ ≥ T consecutively over a time Thold the filter resumes adaptation while

the comparison of ξ to T continues until ξ < T again.

The hold time, Thold, in steps 3 and 4 is essential to suppress detection dropouts due to the

noisy behavior of the detection statistic. Although there are some possible variations

most of the DTD algorithms keep this basic form and only differ in how they form the

detection statistic.

An optimum decision variable, ξ , for double talk detection should behave as

follows:

33
 • if v = 0 (doubletalk is not present), ξ ≥ T

• if v ≠ 0 (doubletalk is present ), ξ < T

The threshold T must be a constant, independent of data. Moreover ξ must be

insensitive to echo path variations when v = 0 [5].

In the following sections discussions of different DTD algorithms such as the

Geigel Algorithm, the Cross- correlation Method and the Normalized Cross-Correlation

Method are presented. The DTD algorithm used in this research was the Normalized

Cross-Correlation Method.

3.4.2 The Geigel Algorithm

One simple algorithm due to A. A. Giegel declares the presence of near-end

speech whenever

max{ x (k ) ,⋅ ⋅ ⋅, x (k − N + 1) }
ξ= <T (3.23)
d (k )

where N and T are suitably chosen constants. This detection scheme is based on a

waveform level comparison between the microphone signal, d, and the far-end speech, x,

assuming the near-end speech, v, in the microphone signal will be stronger than the echo.

The maximum, or norm, of the N most recent samples of x is chosen for the comparison

due to uncertain delay in the echo path. The threshold, T, is used to compensate for the

energy level of the echo path response, h, and is often set to ½ for line echo cancellers

since the hybrid loss is typically approximately 6dB. However, for an AEC, it is not easy

to set a universal threshold that will work reliably in all the various situations since the

34
loss through the acoustic echo path can vary greatly depending on many factors. For N,

one easy choice is to set it equal to the adaptive filter length L [5].

3.4.3 The Cross Correlation Method

This method uses the cross-correlation coefficient vector between x and d as a

means for double talk detection. The cross-correlation coefficient vector between x and d

is defined by

E{x (n )d (n )}
cxd = (3.24)
E{x 2 (n )}E{d 2 (n )

rxd
= (3.25)
σx σ d

= [cxd,0 cxd,1 ……cxd,L-1]T (3.26)

where E denotes the mathematical expectation and cxd,I is the cross-correlation coefficient

between x(n – I) and d(n). The idea is to compare

ξ = c xd (3.27)

= max c xd ,i , i = 0,1,···,L – 1 (3.28)

to a threshold level T. The decision rule is then very simple. If ξ ≥ T, double talk is not

present and if ξ < T, double talk is present.

The fundamental problem with this method is that the cross-correlation coefficient

vectors are not well normalized. In general, it is assumed that ξ ≤ 1. Therefore, if v =

0, it does not mean that ξ = 1 or any other known value. The value of ξ is not known in

general. The amount of correlation will depend greatly on the statistics of the signal and

35
of the echo path. As a result, the best value of T will vary from one experiment to

another. There is no natural threshold level associated with the variable ξ when v= 0.

These complexities lead to another DTD algorithm, which is termed the Normalized

Cross-Correlation method. This method is simply a modification of the existing Cross-

Correlation Method [4].

3.4.4 Normalized Cross Correlation Method

In this method a new normalized cross-correlation vector between a vector x and a

scalar d is derived. Suppose that v = 0. In this case

Rdd = E{d(n)dT(n)}

= HTRxxH (3.29)

where

Rxx = E{x(n xT(n)}. (3.31)

Since

d(n) = HTx(n), (3.32)

Rxd = RxxH, (3.33)

which allows Rdd to be rewritten as

Rdd = RTxdR-1xxRxd. (3.34)

In general, for v ≠ 0,

Rdd = RTxdR-1xxRxd + Rvv (3.35)

where

Rvv = E{v(n)vT(n)} (3.36)

36
is the covariance matrix of the near-end speech. The new decision variable is obtained by

dividing equation(3.35) by Rdd and extracting the square root, which yields

ξ = R T xd R −1 xx R xd R −1 dd (3.37)

= c xd (3.38)

where

cxd = R-1/2xxRxdR-1/2dd (3.39)

is the normalized cross-correlation vector between x and d. Substituting equation (3.33)

and equation (3.35) into equation (3.37) produces the decision variable, which is given by

H T R xx H
ξ = . (3.40)
H T R xx H + σ 2 v

Equation (3.30) shows that for v = 0; ξ = 1 and for v ≠ 0; ξ < 1. Therefore, the

threshold value can be set tone (1). It should also be noted that ξ is not sensitive to

changes of the echo path when v = 0 [4], [5].

3.5 Nonlinear Processor (NLP)

A nonlinear processor, (NLP), is a signal processing circuit or algorithm that is

placed in the speech path after echo cancellation in order to provide further attenuation or

removal of residual echo signals that cannot be removed completely by an echo canceller.

A non-linearity, a distortion, or an added noise signal are examples of signals that cannot

be fully cancelled by an echo canceller. Therefore, these signals are typically removed or

attenuated by a nonlinear processor.

37
3.5.1 Noise Gate as a NLP

In this research a noise gate was used as a NLP, which is a type of dynamic

processor. Noise gates belong to the family of expanders. As the name implies, it

increases the dynamic range of a signal such that low-level signals are attenuated while

the higher-level portions are neither attenuated nor amplified. The noise gate expansion

can be taken to the extreme where it will heavily attenuate the input or eliminate it

entirely leaving only silence.

While expanders are quite difficult to use effectively, noise gates are a very

common and effective way of reducing the apparent noise level in audio signals. The

noise gate offers a method of turning down the gain of an audio signal when the signal

level drops below some threshold value. The threshold value needs to be high enough

that only the background noise falls below but not so high that the audio signals are cut

off prematurely. Noise gates are most often used to eliminate noise or hiss that may

otherwise be amplified.

3.5.2 A Generic Expander

Figure 3.6 presents the basic structure of an expander.

Input Ouput
Σ

Level Detector Gain Control

Figure 3.6: Basic Block Diagram of an Expander

38
 An expander is essentially an amplifier with a variable gain control. The level of

the input signal is sensed by the level detector and applied to the gain control element.

The gain is never greater than one and is controlled by the level of the input signal.

When the input signal level is higher than a threshold value the expander has a unity gain

and acts as a normal unity gain amplifier. When the input signal level drops below the

threshold the gain decreases, which makes the signal even lower or the signal is

completely removed depending on the threshold value. This feature drove the choice of

using a noise gate as the NLP since the signal level of the echo is very much less than

that of the near-end signal.

The input/output relationship of the expander is represented in a simple graph,

which is presented in Figure 3.7. The level of the input signal is given by the horizontal

axis and the output level is given by the vertical axis. When the slope of the line is unity,

angled at 45 degrees, the gain of the expander is one (1). Therefore, the output level is

identical to the input level. A change in the line's slope means a change in the expander's

gain. For the expander, part of the line will have a larger slope. The point where the

slope of the line changes is called the threshold, which is adjustable in many expanders.

When the input signal level is above the threshold nothing happens. However, when the

input signal level drops below the threshold the gain reduction starts. The gain reduction

lowers the input level by increasing or expanding the dynamic range.

39
 Output Level
(dB)

No Expansion

2:1 Expansion
4:1 Expansion

10:1 Expansion
(Noise Gating)

Threshold Input Level (dB

Figure 3.7: Input / Output Characteristics of an Expander

The amount of expansion that is applied is usually expressed as a ratio such as 2:1

or 4:1. This implies that while the input is below the threshold a change in the input level

produces a change in the output that is two times or four times as large. Therefore, with a

4:1 expansion ratio and the input level below the threshold a dip of 3 dB in the input will

produce a drop of 12 dB in the output [6].

3.5.3 Noise Gate

When an expander is used with extreme settings where the input/output

characteristic becomes almost vertical below the threshold and when the expansion ratio

larger than 10:1, the expander is often termed a noise gate. In this case, the input signal

may be very heavily attenuated or removed entirely. Therefore, the expander acts like an

on/off switch for signals. When the signal is high enough, the switch is on and the input

appears at the output. However, when the signal drops below the threshold the switch is

off and there is no output. Hence, when the near-end signal passes through this on/off

40
switch or noise gate, because of the high signal level the switch is on and attenuation

does not occur. However, when the echo signal passes the switch is off and the echo is

completely removed or highly attenuated depending on the threshold. Hence the

important aspect of this device is the choice of a correct threshold value.

Since the level sensing function is a short time average it takes some time for a

change in the input level to be detected, which triggers a change in the gain. In general

an expander is characterized by its attack and release times. The attack time is the time

required for the expander to restore the gain to one once the input level rises above the

threshold. Likewise, the time taken for the expander to reduce its gain after the input

drops below the threshold is the release time. The attack and release times give the

expander a smoother change in the gain rather than abrupt changes that may produce

pops and/or other noise. Figure 3.8 illustrates how the attack and release times affect an

example input signal [7].

41
 Expander Input

Time

Expander Ouput

Time

Expander Gain
Attack Time Release Time

Time

Figure 3.8: The Effect of an Expander on a Signal

Only the middle portion of the input is above the expander's threshold value.

However, it takes some time for the expander to increase the gain when the input level

rises above the threshold. When the input level drops below the threshold the expander

gradually reduces its gain. Therefore, a noise gate fulfilled this research’s need for a

NLP. Another important aspect of the selection was that the noise gate does not facilitate

clipping of talker’s signal, which is very common in the with other NLP types.

42
 CHAPTER 4

SIMULATION AND RESULTS

The previous chapters provide a detailed sketch of an Acoustic Echo canceller,

(AEC). In this chapter the flowchart for the software simulation and the results of

simulation of the AEC algorithm, which was performed in MATLAB are discussed. The

idea that drove the simulation was to show that convincible results could be achieved in

the software environment.

4.1 Why MATLAB?

MATLAB is a powerful, general-purpose, mathematical software package.

MATLAB possesses excellent graphics and matrix handling capabilities. It integrates

mathematical computing in a powerful language to provide a flexible environment for

technical computing. The salient features of MATLAB are its in-built mathematical

toolboxes and graphic functions. Additionally, external routines that are written in other

languages such as C, C++, Fortran and Java, can be integrated with MATLAB

applications. MATLAB also supports importing data from files and other external

devices. Most of the functions in MATLAB are matrix-oriented and can act on arrays of

any appropriate dimension. MATLAB also has a separate toolbox for signal processing

43
applications, which provided simpler solutions for many of the problems encountered in

this research.

The MATLAB software environment suited the needs of this research for the

following reasons:

• The input signals (far-end and near-end talker signals) were voices. These

voices were stored as wav files and the wav files were easily imported into the

code.

• The intermediate signals (echo signals) and output signals (error signal and

signals obtained after echo cancellation) were obtained as wav files. Thus the

audio of the voice signals could be literally be heard, which aided immensely

judgments with respect to the results obtained.

• The signal processing toolbox has in-built functions for almost all signal

processing applications. The toolbox helped the efficiency of the code since

these functions could be called wherever necessary instead of writing separate

sub-routines.

• Since MATLAB supports graphics, the results of a simulation could be

presented in a graphical format with ease.

4.2 Simulation Flowchart

The flowchart for the simulation of the echo canceller algorithm is presented in

Figure 4.1.

44
 Start

Get far-end
signal, x(n)

Create echo
signal, r(n) from
the far-end signal

Get near-end
signal, v(n)

Combine r(n) and

v(n) to obtain the
desired signal,
d(n)

Yes
Does Filter Loop
Doubletalk Filter coefficients
Exist are frozen

No

NLMS Loop
Update Filter
coefficients

Subtract estimated Do Nonlinear

echo from d(n) to Processing to
produce residual remove residual
error signal, e(n) echo

Figure 4.1: Flowchart of the MATLAB Simulation

45
4.3 Description of the Simulation Setup

This section describes the simulation environment, its requirements and the

procedures adopted.

1. The input signals, both far-end and near-end signals, were simulated and

given to the AEC, which executed on a PC with the MATLAB environment.

2. The input signals seven seconds in duration.

3. A sampling rate of 8000 Hz was used for all the signals in the simulation.

4. The graphs plotted have x-axes denoting the time and y-axes denoting the

amplitude or magnitude of the signal.

4.4 Results

This section presents a graphical representation of the results obtained by

simulating the algorithm in MATLAB. The plot of the far-end signal x(n) is presented in

Figure 4.2.

46
 Figure 4.2: Plot of the Far-end Signal, x(n)

The far-end signal was delayed and scaled in order to produce the echo signal,

r(n), which is presented in Figure 4.3. The echo signal was produced when the far-end

signal, x(n), passed through the echo path, h.

47
 Figure 4.3: Plot of the Echo Signal, r(n)

The echo signal was added to the near-end signal, v(n), in order to produce the

desired signal, d(n), which became the input for the adaptive filter. The plot of the near-

end signal, v(n), is presented in Figure 4.4 and the plot of the desired signal, d(n). is

presented in Figure 4.5.

48
Figure 4.4: Plot of the Near-end Signal, v(n)

Figure 4.5: Plot of the Desired Signal, d(n)

49
 The desired signal, d(n),was passed through the adaptive filter and the double talk

detector. For the purpose of adaptive filtering the NLMS algorithm was used during the

simulation. The algorithm used the normalized cross correlation algorithm for double

talk detection.

Various parameters for the NLMS algorithm such as the convergence factor, µn,

and γ had to be set in order to avoid misadjustment. Additionally, the length of the filter

had to be established beforehand. The values of these parameters, which were used in the

simulation, are

• Length of the filter, N = 512

• Convergence factor, µn = 1.9. This value was found to produce faster

convergence of the NLMS algorithm.

• A small constant, γ = 0.9

For the purpose of the open simulation environment and faster convergence of the

algorithm, it was assumed that double talk did not take place during this simulation.

The output of this module is the error signal, e(n), which is presented in Figure

4.6. In the case of an ideal echo canceller the error signal should be the same as that of

the near-end signal, v(n). However, due to the presence of residual echo and

nonlinearities the error signal, e(n), was not a perfect copy of the near-end signal, v(n).

50
 Figure 4.6: Plot of the Error Signal e(n)

Since the error signal, e(n), contained a residual echo it was passed through a

NLP. As explained earlier, a noise gate was used for the NLP in this research. The

purpose of this device was to attenuate the residual echo and to pass on the speech signal

without any clipping. Figure 4.7 presents the plot of the error signal after nonlinear

processing.

51
 Figure 4.7: Plot of the Error Signal after Nonlinear Processing

Figure 4.7 clearly shows that the residual echo was completely removed and that s

no clipping occurred. Therefore, the signal output of the echo canceller was devoid of

any significant echoes.

4.5 Evaluation of the Echo Cancellation Algorithm

In order to evaluate the effective working of the algorithm, some basic tests were

conducted. This section provides a brief account of these tests.

52
4.5.1 Convergence Test

The first and paramount test of the algorithm was whether or not the algorithm

converged. If the filter coefficients used in the adaptive algorithm did not converge, the

code would be useless. Therefore, several tests were performed on the simulated data in

order to verify the convergence of the filter coefficients. These tests were conducted by

varying the convergence factor, µn, and examining the effect on the filter coefficients and

the plot of the error signal, e(n). Through careful observation it was determined that a

value of 1.9 produced faster convergence.

4.5.2 Echo Return Loss Enhancement (ERLE)

In order to evaluate the quality of the echo cancellation algorithm the measure of

ERLE was used. ERLE, measured in dB is defined as the ratio of the instantaneous

power of the signal, d(n), and the instantaneous power of the residual error signal, e(n),

immediately after cancellation. ERLE measures the amount of loss introduced by the

adaptive filter alone. Mathematically it can be expressed as

Pd (n ) E[d 2 (n )]
ERLE = 10log = 10log . (4.1)
Pe (n ) E[e 2 (n )]

For a good echo canceller circuit, an ERLE in the range of 30 dB – 40dB is

considered to be ideal. Figure 4.8 presents a plot of the ERLE with the ERLE plotted in

dB along the y-axis and the number of samples along the x-axis. The plot of ERLE

implies that the ERLE for this algorithm attained the required value.

53
 Figure 4.8: Plot of ERLE Vs Number of Samples

4.5.3 Auditory Test

The last test consisted of listening to the output for appropriate cancellation of

echoes. The audio of the output signals was presented to a panel of five members with no

technical expertise in this field. The panel was almost not able to distinguish the near-

end signal, v(n), and the output signal with the residual echo, e(n), removed. Some

discrepancies in the audio could be attributed to the fact that the real-time applications

cannot escape the factor called noise.

54
 CHAPTER 5

CONCLUTION AND FURTHER WORK

5.1 Conclusions

With the world shrinking into a global village because of superior

communications, telephones, both conventional and hands-free sets, occupy a prominent

position in solving people’s communication needs. One of the major problems in a

telecommunication application over a telephone system is echo. The Echo cancellation

algorithm presented in this thesis successfully attempted to find a software solution for

the problem of echoes in the telecommunications environment. The proposed algorithm

was completely a software approach without utilizing any DSP hardware components.

The algorithm was capable of running in any PC with MATLAB software installed.

Additionally, a new method, which utilized the noise gate device for nonlinear processing

was proposed. This new technique is faster and provides almost perfect results for

canceling residual echoes without clipping of the reference speech signals. In addition,

the results obtained were convincing. The audio of the output speech signals were highly

satisfactory and validated the goals of this research.

55
5.2 Further Work

The algorithm proposed in this thesis presents a solution for single channel

acoustic echoes. However, most often in real life situations, multichannel sound is the

norm for telecommunication. For example, when there is a group of people in a

teleconference environment and everybody is busy talking, laughing or just

communicating with each other multichannel sound abounds. Since there is just a single

microphone the other end will hear just a highly incoherent monographic sound. In order

to handle such situations in a better way the echo cancellation algorithm developed

during this research should be extended for the multichannel case.

56
 REFERENCES

[1] Sadaoki Furui and M. Mohan Sondhi, “Advances in Speech Signal Processing”,
Marcel Dekker, Inc, 1992

[2] Lester S.H Ngia, “System Modeling using Basis Functions and Application to Echo
Cancellation”, Ph. D. Dissertation, Chalmers University of Technology

[3] Paulo S.R. Diniz, “Adaptive Filtering Algorithms and Practical Implementation”,
Kluwer Academic Publishers, 1997

[4] J. Benesty, D.R. Morgan and J.H. Cho, “A New Class of Doubletalk Detectors
Based on Cross-correlation”, IEEE Trans. Speech Audio Processing, vol. 8, pp.
168-172, March 2000

[5] J. Benesty, T. Gansler, D.R. Morgan, M.M. Sondhi and S.L. Gay, “Advances in
Network and Acoustic Echo Cancellation”, Springer-Verlag, 2001

[6] Jerry C. Whitaker, “Master Handbook of Audio Production”, McGraw-Hill, 2003

[7] Eargle, John, “Handbook of Recording Engineering”, Van Nostrand Reinhold, 1996

[8] Mu Tian, P.K. Gupta, Marcus Harte and Danny Yip, “Improved Echo Canceller and
Implementation”, Dialogic Corporation

[9] J.G. Proakis and D.G. Manolakis, “Digital Signal Processing - Principles,
Algorithms and Applications”, Prentice Hall, 1996

[10] A.V. Oppenheim and R.W. Schafer, “Discrete Time Signal Processing”, Prentice
Hall, 1996

[11] Sophocles J. Orfanidis, “Optimum Signal Processing - An Introduction”, McGraw

Hill, 1998

[12] S.Haykin, “Adaptive Filter Theory”, Prentice Hall, 1996

[13] Andrzej Borys, “Nonlinear Aspects of Telecommunications”, CRC Press, 2001

57