A novel spectral subtraction scheme for robust speech recogniton: Spectral subtraction using spectral harmonics of speech

Jounghoon Beh, Hanseok Ko

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

19 Citations (Scopus)

Abstract

This paper addresses a novel noise-compensation scheme to solve the mismatch problem between training condition and testing condition for the automatic speech recognition (ASR) system, specifically in the car environments. The conventional spectral subtraction schemes rely on the signal to noise ratio (SNR) such that attenuation is imposed on that part of the spectrum that appears to have low SNR, and accentuation is made on that part of high SNR. However, since these schemes are based on the postulation that the power spectrum of noise is in general at the lower level in magnitude than that of speech. Therefore, while such postulation is adequate for high SNR environment, it is grossly inadequate for low SNR scenarios such as car environment. This paper proposes an efficient spectral subtraction scheme focused to specifically low SNR noisy environments by distinguishing the speech-dominant segment from the noise-dominant segment in speech spectrum. Representative experiments confirm the superior performance of the proposed method over conventional methods. The experiments are conducted using car noise-corrupted utterances of Aurora2 corpus.

Original languageEnglish
Title of host publicationICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Pages648-651
Number of pages4
Volume1
Publication statusPublished - 2003
Event2003 IEEE International Conference on Accoustics, Speech, and Signal Processing - Hong Kong, Hong Kong
Duration: 2003 Apr 62003 Apr 10

Other

Other2003 IEEE International Conference on Accoustics, Speech, and Signal Processing
CountryHong Kong
CityHong Kong
Period03/4/603/4/10

Fingerprint

subtraction
Signal to noise ratio
signal to noise ratios
harmonics
Railroad cars
speech recognition
Power spectrum
Speech recognition
Acoustic noise
power spectra
education
attenuation
Experiments
Testing

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Signal Processing
  • Acoustics and Ultrasonics

Cite this

Beh, J., & Ko, H. (2003). A novel spectral subtraction scheme for robust speech recogniton: Spectral subtraction using spectral harmonics of speech. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings (Vol. 1, pp. 648-651)

A novel spectral subtraction scheme for robust speech recogniton : Spectral subtraction using spectral harmonics of speech. / Beh, Jounghoon; Ko, Hanseok.

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 1 2003. p. 648-651.

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

Beh, J & Ko, H 2003, A novel spectral subtraction scheme for robust speech recogniton: Spectral subtraction using spectral harmonics of speech. in ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. vol. 1, pp. 648-651, 2003 IEEE International Conference on Accoustics, Speech, and Signal Processing, Hong Kong, Hong Kong, 03/4/6.
Beh J, Ko H. A novel spectral subtraction scheme for robust speech recogniton: Spectral subtraction using spectral harmonics of speech. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 1. 2003. p. 648-651
Beh, Jounghoon ; Ko, Hanseok. / A novel spectral subtraction scheme for robust speech recogniton : Spectral subtraction using spectral harmonics of speech. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 1 2003. pp. 648-651
@inproceedings{4d0f7e0fc3dc4f43a7e0fce4576c1c9a,
title = "A novel spectral subtraction scheme for robust speech recogniton: Spectral subtraction using spectral harmonics of speech",
abstract = "This paper addresses a novel noise-compensation scheme to solve the mismatch problem between training condition and testing condition for the automatic speech recognition (ASR) system, specifically in the car environments. The conventional spectral subtraction schemes rely on the signal to noise ratio (SNR) such that attenuation is imposed on that part of the spectrum that appears to have low SNR, and accentuation is made on that part of high SNR. However, since these schemes are based on the postulation that the power spectrum of noise is in general at the lower level in magnitude than that of speech. Therefore, while such postulation is adequate for high SNR environment, it is grossly inadequate for low SNR scenarios such as car environment. This paper proposes an efficient spectral subtraction scheme focused to specifically low SNR noisy environments by distinguishing the speech-dominant segment from the noise-dominant segment in speech spectrum. Representative experiments confirm the superior performance of the proposed method over conventional methods. The experiments are conducted using car noise-corrupted utterances of Aurora2 corpus.",
author = "Jounghoon Beh and Hanseok Ko",
year = "2003",
language = "English",
volume = "1",
pages = "648--651",
booktitle = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

}

TY - GEN

T1 - A novel spectral subtraction scheme for robust speech recogniton

T2 - Spectral subtraction using spectral harmonics of speech

AU - Beh, Jounghoon

AU - Ko, Hanseok

PY - 2003

Y1 - 2003

N2 - This paper addresses a novel noise-compensation scheme to solve the mismatch problem between training condition and testing condition for the automatic speech recognition (ASR) system, specifically in the car environments. The conventional spectral subtraction schemes rely on the signal to noise ratio (SNR) such that attenuation is imposed on that part of the spectrum that appears to have low SNR, and accentuation is made on that part of high SNR. However, since these schemes are based on the postulation that the power spectrum of noise is in general at the lower level in magnitude than that of speech. Therefore, while such postulation is adequate for high SNR environment, it is grossly inadequate for low SNR scenarios such as car environment. This paper proposes an efficient spectral subtraction scheme focused to specifically low SNR noisy environments by distinguishing the speech-dominant segment from the noise-dominant segment in speech spectrum. Representative experiments confirm the superior performance of the proposed method over conventional methods. The experiments are conducted using car noise-corrupted utterances of Aurora2 corpus.

AB - This paper addresses a novel noise-compensation scheme to solve the mismatch problem between training condition and testing condition for the automatic speech recognition (ASR) system, specifically in the car environments. The conventional spectral subtraction schemes rely on the signal to noise ratio (SNR) such that attenuation is imposed on that part of the spectrum that appears to have low SNR, and accentuation is made on that part of high SNR. However, since these schemes are based on the postulation that the power spectrum of noise is in general at the lower level in magnitude than that of speech. Therefore, while such postulation is adequate for high SNR environment, it is grossly inadequate for low SNR scenarios such as car environment. This paper proposes an efficient spectral subtraction scheme focused to specifically low SNR noisy environments by distinguishing the speech-dominant segment from the noise-dominant segment in speech spectrum. Representative experiments confirm the superior performance of the proposed method over conventional methods. The experiments are conducted using car noise-corrupted utterances of Aurora2 corpus.

UR - http://www.scopus.com/inward/record.url?scp=0141813765&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0141813765&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0141813765

VL - 1

SP - 648

EP - 651

BT - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

ER -