Multimedia Signal Processing Projects

Subband Acoustic Echo Cancellation for Mobile and Hands-Free Phones in Car

Gobu Rasalingham, Saeed Vaseghi

Objective

To simulate and Implement a real time Acoustic Echo Cancellation (AEC) system for mobile phones and hands free phones in moving noise environments. This can be breaked down into a number of sub-objectives.

Introduction

Acoustic echo is created by a sound wave originating in the receiving side of the telephones. When a person is on the phone, a sound wave received on the earpiece enters through mouthpiece of the handset via reflections off objects in the sound path and circulate the feedback loop. Figure 1 shows the core system of one direction acoustic echo canceller in a mobile phone.


Figure1 one direction AEC in mobile phones

Technical Issues

In AEC systems Multi-path echo, time varying echo, combined echo and noise reduction and performance of the echo canceller are some of the issues that have to be thoroughly scrutinized when designing the AEC system. The European Telecommunication Standards Institute (ETSI) has specified the AEC system.


Figure 2

Sub-band adaptive filtering is one of the common solutions for AEC systems. Figure 2 shows the sun-band echo canceller with two sub-bands. There are few known adaptive algorithms such as LMS, NLMS and RLS. As far as computational and performance concerns, NLMS offers a good solutions. The main challenge in AEC is that of echo path variation, that leads to the echo delay. In a typical office room environment echo delay vary from 100ms to 250ms. For a sampling rate of 8 KHz, this means about 800 to 2000 taps. This requires long adaptive filter length. Setting adapting the filter length according to the echo signal is complex task. Double Talk (DT) is one other issue related to AEC. If DT is not detected during the echo cancellation process, it can cause divergence of the adaptive algorithm. Convergence Rate (CR) is a performance specification of acoustic echo cancellers. Factors like multi path echo and performance of adaptive filter can affect the CR. The ITU G.167 recommendation for acoustic echo controllers recommends that an acoustic echo canceller have a total convergence of 30db without non-linear processing and convergence rate of 20db per second. Now days, in real time processing an average DSP processor can handle 20db/sc CR and up to 50db (typically above 20db) Echo Return Loss Enhancement (ERLE).

AEC Applications

Main applications of AEC systems are in


Figure 3 complexity car environment

Increasingly car environments have electronic appliances which performance is dependent on issues like echo and noise. The latest electronic gadgets in motor industries are mobile phones with speaker and hand free set, and multimedia appliances such as small screen TV with DVD, highly digitized Hi-Fi audio system and play stations. Therefore reducing the acoustic echo and noise inside a car is an important issue. Our expectations see motor vehicles as more intelligent and robotic customized system in near by future. From engineering point of view achieving this goal is a challenging task and the high performance AEC system in an integral part of it. Figure 3 shows complexity at how a modern car gadgets connections.

Figure 4 shows the basic PC telephone block diagram and how AEC system fits into it.


Figure 4 PC telephony and AEC system

AEC Demonstration on PC Telephone and Audio Conferencing Systems

This demonstration is based on Acoustic Echo Cancellation in PC telephones and in principal is similar to audio conferencing. In this demonstration an AEC system offering the best performance, in echo cancellation is applied to echo containing signal. More technical details of this demonstration as well as other AEC systems practical results are laid out in the simulation analysis section.


Figure 5 shows the full echo signal

 Press the button to listen the echo signal.


Figure 6 signal after echo cancellation

 Press the button to listen the echo signal.

Press the button to listen the reduced echo signal.

AEC Systems simulation Analysis

Various methods of AEC systems been designed and simulated in Matlab. In all simulations, real signal and its real echo were used to monitor the performance of the echo canceller. Table 1 shows the performances of individual system. As the work progresses the table of results will updated. All AEC systems measurement based on following units

Click on each system to find out more of its details.

  AEC system ERLE in db TCL in db ABR in db CR in db/sc
01 Full band -5.25 -10.44 - 300
02 2 Sub-band -11.99 -17.18 -4.59
03 4 Sub-band -16.20 -21.40 -3.35
04 8 Sub-band -21.75 -26.94 -2.93
05 12 Sub-band -20.25 -25.44 -3.95

Table 1

Comments

As the work progress all the achievements will be updated.

If anyone has any suggestions or comments regarding this project please do not hesitate to contact me. I am also open to any discussion in related field.

Email: gobu.rasalingam@brunel.ac.uk

Thank you for visiting this web page.

Related Links

News

List of MPhil/PhD Projects in MultiMedia Mobile Digtial Signal Processing

Advanced Signal processing and Noise Reduction 3rd Ed, S. Vaseghi, John Wiley 2006

Multimedia Signal Processing: Theory and Applications in Speech, Music and Communications, S. Vaseghi, John Wiley 2007

DeNoise Toolkit

This is a collection of established and new speech denoising methods developed at Brunel in collaboration with Southampton and UEA (Project sponsored by EPSRC).

Voice Morph

VoiceMorph is a software tool developed in our lab for analysis, modelling modification and of voice profile parameters including speaker correlates and Accent correlates of voice.

MSc Digital Signal Processing

Multimedia, communication and inteligent systems.