Blind dereverberation of speech signals using independence transform matrix

Jong-Hwan Lee, Soo Young Lee

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

In the real room environment, sound source is distorted with delayed versions of itself reflected from walls. This room reverberation severely degrades the intelligibility of speech and performance of automatic speech recognition system. Blind deconvolution is to find the inverse of reverberation channel when only convolved versions of the sources are available at the receiver. However existing blind deconvolution algorithms assume that a source signal has an independent identically-distributed (IED) non-Gaussian probability density function (PDF). In this research, colored nonstationary non-IID speech signals were transformed into an IID-like signal as possible by ICA-based independence transform and the resulting signals were processed using infomax blind deconvolution algorithm for the simulated minimum-phase finite impulse response (FIR) channels. Compared to the pre-whitening method by Torkkola, the proposed method demonstrated much better performance of about 30dB signal-to-reverberant components ratio (SRR).

Original languageEnglish
Title of host publicationProceedings of the International Joint Conference on Neural Networks
Pages1453-1457
Number of pages5
Volume2
Publication statusPublished - 2003 Sep 24
Externally publishedYes
EventInternational Joint Conference on Neural Networks 2003 - Portland, OR, United States
Duration: 2003 Jul 202003 Jul 24

Other

OtherInternational Joint Conference on Neural Networks 2003
CountryUnited States
CityPortland, OR
Period03/7/2003/7/24

Fingerprint

Deconvolution
Reverberation
Speech intelligibility
Independent component analysis
Impulse response
Speech recognition
Probability density function
Acoustic waves

ASJC Scopus subject areas

  • Software

Cite this

Lee, J-H., & Lee, S. Y. (2003). Blind dereverberation of speech signals using independence transform matrix. In Proceedings of the International Joint Conference on Neural Networks (Vol. 2, pp. 1453-1457)

Blind dereverberation of speech signals using independence transform matrix. / Lee, Jong-Hwan; Lee, Soo Young.

Proceedings of the International Joint Conference on Neural Networks. Vol. 2 2003. p. 1453-1457.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Lee, J-H & Lee, SY 2003, Blind dereverberation of speech signals using independence transform matrix. in Proceedings of the International Joint Conference on Neural Networks. vol. 2, pp. 1453-1457, International Joint Conference on Neural Networks 2003, Portland, OR, United States, 03/7/20.
Lee J-H, Lee SY. Blind dereverberation of speech signals using independence transform matrix. In Proceedings of the International Joint Conference on Neural Networks. Vol. 2. 2003. p. 1453-1457
Lee, Jong-Hwan ; Lee, Soo Young. / Blind dereverberation of speech signals using independence transform matrix. Proceedings of the International Joint Conference on Neural Networks. Vol. 2 2003. pp. 1453-1457
@inproceedings{3fb73825d1504ba5b60885de67ccc59d,
title = "Blind dereverberation of speech signals using independence transform matrix",
abstract = "In the real room environment, sound source is distorted with delayed versions of itself reflected from walls. This room reverberation severely degrades the intelligibility of speech and performance of automatic speech recognition system. Blind deconvolution is to find the inverse of reverberation channel when only convolved versions of the sources are available at the receiver. However existing blind deconvolution algorithms assume that a source signal has an independent identically-distributed (IED) non-Gaussian probability density function (PDF). In this research, colored nonstationary non-IID speech signals were transformed into an IID-like signal as possible by ICA-based independence transform and the resulting signals were processed using infomax blind deconvolution algorithm for the simulated minimum-phase finite impulse response (FIR) channels. Compared to the pre-whitening method by Torkkola, the proposed method demonstrated much better performance of about 30dB signal-to-reverberant components ratio (SRR).",
author = "Jong-Hwan Lee and Lee, {Soo Young}",
year = "2003",
month = "9",
day = "24",
language = "English",
volume = "2",
pages = "1453--1457",
booktitle = "Proceedings of the International Joint Conference on Neural Networks",

}

TY - GEN

T1 - Blind dereverberation of speech signals using independence transform matrix

AU - Lee, Jong-Hwan

AU - Lee, Soo Young

PY - 2003/9/24

Y1 - 2003/9/24

N2 - In the real room environment, sound source is distorted with delayed versions of itself reflected from walls. This room reverberation severely degrades the intelligibility of speech and performance of automatic speech recognition system. Blind deconvolution is to find the inverse of reverberation channel when only convolved versions of the sources are available at the receiver. However existing blind deconvolution algorithms assume that a source signal has an independent identically-distributed (IED) non-Gaussian probability density function (PDF). In this research, colored nonstationary non-IID speech signals were transformed into an IID-like signal as possible by ICA-based independence transform and the resulting signals were processed using infomax blind deconvolution algorithm for the simulated minimum-phase finite impulse response (FIR) channels. Compared to the pre-whitening method by Torkkola, the proposed method demonstrated much better performance of about 30dB signal-to-reverberant components ratio (SRR).

AB - In the real room environment, sound source is distorted with delayed versions of itself reflected from walls. This room reverberation severely degrades the intelligibility of speech and performance of automatic speech recognition system. Blind deconvolution is to find the inverse of reverberation channel when only convolved versions of the sources are available at the receiver. However existing blind deconvolution algorithms assume that a source signal has an independent identically-distributed (IED) non-Gaussian probability density function (PDF). In this research, colored nonstationary non-IID speech signals were transformed into an IID-like signal as possible by ICA-based independence transform and the resulting signals were processed using infomax blind deconvolution algorithm for the simulated minimum-phase finite impulse response (FIR) channels. Compared to the pre-whitening method by Torkkola, the proposed method demonstrated much better performance of about 30dB signal-to-reverberant components ratio (SRR).

UR - http://www.scopus.com/inward/record.url?scp=0141459932&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0141459932&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0141459932

VL - 2

SP - 1453

EP - 1457

BT - Proceedings of the International Joint Conference on Neural Networks

ER -