Improved acoustic modeling based on selective data-driven PMC

Wooil Kim, Hanseok Ko

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

Abstract

This paper proposes an effective method to remedy the acoustic modeling problem inherent in the usual log-normal PMC intended for achieving robust speech recognition. In particular, the Gaussian kernels under the prescribed log-normal PMC cannot sufficiently express the corrupted speech distributions. The proposed scheme corrects this deficiency by judicially selecting the "fairly" corrupted component and by re-estimating it as a mixture of two distributions using data-driven PMC. As a result, some components become merged while equal number of components split. The determination for splitting or merging is achieved by means of measuring the similarity of corrupted speech model to those of clean model and noise model. The experimental results indicate that the suggested algorithm is effective in representing the corrupted speech distributions and attains consistent improvement over various SNR and noise cases.

Original languageEnglish
Title of host publicationICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume4
Publication statusPublished - 2002
Event2002 IEEE International Conference on Acoustic, Speech, and Signal Processing - Orlando, FL, United States
Duration: 2002 May 132002 May 17

Other

Other2002 IEEE International Conference on Acoustic, Speech, and Signal Processing
CountryUnited States
CityOrlando, FL
Period02/5/1302/5/17

Fingerprint

Acoustics
acoustics
speech recognition
Speech recognition
Merging
estimating

ASJC Scopus subject areas

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

Cite this

Kim, W., & Ko, H. (2002). Improved acoustic modeling based on selective data-driven PMC. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings (Vol. 4)

Improved acoustic modeling based on selective data-driven PMC. / Kim, Wooil; Ko, Hanseok.

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 4 2002.

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

Kim, W & Ko, H 2002, Improved acoustic modeling based on selective data-driven PMC. in ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. vol. 4, 2002 IEEE International Conference on Acoustic, Speech, and Signal Processing, Orlando, FL, United States, 02/5/13.
Kim W, Ko H. Improved acoustic modeling based on selective data-driven PMC. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 4. 2002
Kim, Wooil ; Ko, Hanseok. / Improved acoustic modeling based on selective data-driven PMC. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 4 2002.
@inproceedings{d991c07d275d43bebc93b3f95a9d28b0,
title = "Improved acoustic modeling based on selective data-driven PMC",
abstract = "This paper proposes an effective method to remedy the acoustic modeling problem inherent in the usual log-normal PMC intended for achieving robust speech recognition. In particular, the Gaussian kernels under the prescribed log-normal PMC cannot sufficiently express the corrupted speech distributions. The proposed scheme corrects this deficiency by judicially selecting the {"}fairly{"} corrupted component and by re-estimating it as a mixture of two distributions using data-driven PMC. As a result, some components become merged while equal number of components split. The determination for splitting or merging is achieved by means of measuring the similarity of corrupted speech model to those of clean model and noise model. The experimental results indicate that the suggested algorithm is effective in representing the corrupted speech distributions and attains consistent improvement over various SNR and noise cases.",
author = "Wooil Kim and Hanseok Ko",
year = "2002",
language = "English",
volume = "4",
booktitle = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

}

TY - GEN

T1 - Improved acoustic modeling based on selective data-driven PMC

AU - Kim, Wooil

AU - Ko, Hanseok

PY - 2002

Y1 - 2002

N2 - This paper proposes an effective method to remedy the acoustic modeling problem inherent in the usual log-normal PMC intended for achieving robust speech recognition. In particular, the Gaussian kernels under the prescribed log-normal PMC cannot sufficiently express the corrupted speech distributions. The proposed scheme corrects this deficiency by judicially selecting the "fairly" corrupted component and by re-estimating it as a mixture of two distributions using data-driven PMC. As a result, some components become merged while equal number of components split. The determination for splitting or merging is achieved by means of measuring the similarity of corrupted speech model to those of clean model and noise model. The experimental results indicate that the suggested algorithm is effective in representing the corrupted speech distributions and attains consistent improvement over various SNR and noise cases.

AB - This paper proposes an effective method to remedy the acoustic modeling problem inherent in the usual log-normal PMC intended for achieving robust speech recognition. In particular, the Gaussian kernels under the prescribed log-normal PMC cannot sufficiently express the corrupted speech distributions. The proposed scheme corrects this deficiency by judicially selecting the "fairly" corrupted component and by re-estimating it as a mixture of two distributions using data-driven PMC. As a result, some components become merged while equal number of components split. The determination for splitting or merging is achieved by means of measuring the similarity of corrupted speech model to those of clean model and noise model. The experimental results indicate that the suggested algorithm is effective in representing the corrupted speech distributions and attains consistent improvement over various SNR and noise cases.

UR - http://www.scopus.com/inward/record.url?scp=19244377591&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=19244377591&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:19244377591

VL - 4

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

ER -