Echo cancellation in VoIP using digital adaptive filters

Kmita, Shane (2011) Echo cancellation in VoIP using digital adaptive filters. [USQ Project]

[img]
Preview
PDF
Kmita_2011.pdf

Download (3435Kb)

Abstract

VoIP calls frequently suffer from echoes which degrade the quality of voice transmissions. Traditional methods of using adaptive filters to cancel line echoes over the public
switched telephone network are not as effective when applied to VoIP channels. This is because VoIP echo paths are generally longer due to longer network delays and
non-stationary due to dynamic de-jitter buffering. Also, non-linearities introduced by dropped packets and lossy signal compression algorithms can reduce the performance
of the adaptive filters. This project researched the theory of digital adaptive filter algorithms and their application to the echo cancellation problem in VoIP networks. A VoIP, adaptive echo cancellation (AEC) simulation was designed and then implemented in MATLAB. The simulation modeled an echo path which incorporated VoIP channel characteristics and room acoustic effects. The simulation was used to test the echo cancelling effectiveness of three different adaptive algorithm schemes: a normalised least mean squares (NLMS) filter, a NLMS filter in Dual-H configuration and a recursive least squares (RLS) filter. Echo cancellation performance was primarily determined by measuring the loudness of echoes before and after they enter the system (represented by an ERLE value). The Telecommunication Standardization Sector of the International Telecommunications Union (ITU-T) G.131 echo objection rate gives the recommended echo attenuation levels required by an echo canceller. The Dual-H NLMS AEC system designed in this project achieved an ERLE of approximately 35 dB during simulations in the VoIP environment which is above the ITU-T recommended level. This showed that satisfactory echo cancellation performance in a VoIP environment could be achieved by the use of this relatively simple AEC system.


Statistics for USQ ePrint 22548
Statistics for this ePrint Item
Item Type: USQ Project
Refereed: No
Item Status: Live Archive
Faculty / Department / School: Historic - Faculty of Engineering and Surveying - Department of Electrical, Electronic and Computer Engineering
Supervisors: Leis, John
Date Deposited: 20 Dec 2012 01:31
Last Modified: 03 Jul 2013 01:36
Uncontrolled Keywords: digital adaptive filters, echo cancellation
Fields of Research : 08 Information and Computing Sciences > 0805 Distributed Computing > 080503 Networking and Communications
10 Technology > 1005 Communications Technologies > 100503 Computer Communications Networks
URI: http://eprints.usq.edu.au/id/eprint/22548

Actions (login required)

View Item Archive Repository Staff Only