A multi-filter system for speech enhancement under low signal-to-noise ratios

In this paper, the problem of deteriorating performance of speech recognition under very low signal-to-noise ratios (SNR) is considered. In particular, for a given pre-trained speech recognizer and for a finite set of speech commands, we show that popular noise reduction methods have a mixed perform...

Full description

Bibliographic Details
Main Authors: Yiu, Ka Fai, Chan, Kit Yan, Low, Siow, Nordholm, Sven
Format: Journal Article
Published: American Institute of Mathematical Sciences 2009
Subjects:
Online Access:http://hdl.handle.net/20.500.11937/32975
_version_ 1848753815955701760
author Yiu, Ka Fai
Chan, Kit Yan
Low, Siow
Nordholm, Sven
author_facet Yiu, Ka Fai
Chan, Kit Yan
Low, Siow
Nordholm, Sven
author_sort Yiu, Ka Fai
building Curtin Institutional Repository
collection Online Access
description In this paper, the problem of deteriorating performance of speech recognition under very low signal-to-noise ratios (SNR) is considered. In particular, for a given pre-trained speech recognizer and for a finite set of speech commands, we show that popular noise reduction methods have a mixed performance in speech recognition accuracy under very low SNR. Although most noise reduction methods are attempting to reduce speech distortion or to increase noise suppression, it does not necessarily improve speech recognition accuracy very much due to the complexity of the recognizer. We propose a new hybrid algorithm to optimize on the speech recognition accuracy directly by mixing different noise reduction methods together. We show that this method can indeed improve the accuracy significantly.
first_indexed 2025-11-14T08:30:31Z
format Journal Article
id curtin-20.500.11937-32975
institution Curtin University Malaysia
institution_category Local University
last_indexed 2025-11-14T08:30:31Z
publishDate 2009
publisher American Institute of Mathematical Sciences
recordtype eprints
repository_type Digital Repository
spelling curtin-20.500.11937-329752017-09-13T16:07:33Z A multi-filter system for speech enhancement under low signal-to-noise ratios Yiu, Ka Fai Chan, Kit Yan Low, Siow Nordholm, Sven Speech enhancement Speech recognition Noise reduction Optimization In this paper, the problem of deteriorating performance of speech recognition under very low signal-to-noise ratios (SNR) is considered. In particular, for a given pre-trained speech recognizer and for a finite set of speech commands, we show that popular noise reduction methods have a mixed performance in speech recognition accuracy under very low SNR. Although most noise reduction methods are attempting to reduce speech distortion or to increase noise suppression, it does not necessarily improve speech recognition accuracy very much due to the complexity of the recognizer. We propose a new hybrid algorithm to optimize on the speech recognition accuracy directly by mixing different noise reduction methods together. We show that this method can indeed improve the accuracy significantly. 2009 Journal Article http://hdl.handle.net/20.500.11937/32975 10.3934/jimo.2009.5.671 American Institute of Mathematical Sciences unknown
spellingShingle Speech enhancement
Speech recognition
Noise reduction
Optimization
Yiu, Ka Fai
Chan, Kit Yan
Low, Siow
Nordholm, Sven
A multi-filter system for speech enhancement under low signal-to-noise ratios
title A multi-filter system for speech enhancement under low signal-to-noise ratios
title_full A multi-filter system for speech enhancement under low signal-to-noise ratios
title_fullStr A multi-filter system for speech enhancement under low signal-to-noise ratios
title_full_unstemmed A multi-filter system for speech enhancement under low signal-to-noise ratios
title_short A multi-filter system for speech enhancement under low signal-to-noise ratios
title_sort multi-filter system for speech enhancement under low signal-to-noise ratios
topic Speech enhancement
Speech recognition
Noise reduction
Optimization
url http://hdl.handle.net/20.500.11937/32975