Combining of confidence measures under Bayesian framework for speech recognition

Tae Yoon Kim, Hanseok Ko

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

Abstract

Bayesian combining of confidence measures is proposed for speech recognition. Centralized and distributed schemes are considered for confidence measure combining under Bayesian framework. Centralized combining is feature level fusion which combines the values of individual confidence scores and makes a final decision. In contrast, distributed combining is decision level fusion which combines the individual decision makings made by each individual confidence scoring method. Both methods are basically based on the statistical modeling of confidence features. In addition, adaptation of confidence score using the statistical models also presented. The proposed methods reduced the classification error rate by 17% from conventional single feature based confidence scoring method in isolated word Out-of-Vocabulary rejection test.

Original languageEnglish
Title of host publicationProceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, ISPACS 2004
EditorsS.J. Ko
Pages164-168
Number of pages5
Publication statusPublished - 2004 Dec 1
EventProceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, ISPACS 2004 - Seoul, Korea, Republic of
Duration: 2004 Nov 182004 Nov 19

Other

OtherProceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, ISPACS 2004
CountryKorea, Republic of
CitySeoul
Period04/11/1804/11/19

Fingerprint

Speech recognition
Fusion reactions
Decision making
Statistical Models

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Kim, T. Y., & Ko, H. (2004). Combining of confidence measures under Bayesian framework for speech recognition. In S. J. Ko (Ed.), Proceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, ISPACS 2004 (pp. 164-168)

Combining of confidence measures under Bayesian framework for speech recognition. / Kim, Tae Yoon; Ko, Hanseok.

Proceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, ISPACS 2004. ed. / S.J. Ko. 2004. p. 164-168.

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

Kim, TY & Ko, H 2004, Combining of confidence measures under Bayesian framework for speech recognition. in SJ Ko (ed.), Proceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, ISPACS 2004. pp. 164-168, Proceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, ISPACS 2004, Seoul, Korea, Republic of, 04/11/18.
Kim TY, Ko H. Combining of confidence measures under Bayesian framework for speech recognition. In Ko SJ, editor, Proceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, ISPACS 2004. 2004. p. 164-168
Kim, Tae Yoon ; Ko, Hanseok. / Combining of confidence measures under Bayesian framework for speech recognition. Proceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, ISPACS 2004. editor / S.J. Ko. 2004. pp. 164-168
@inproceedings{39d24b3345864079aaced17a0bb96135,
title = "Combining of confidence measures under Bayesian framework for speech recognition",
abstract = "Bayesian combining of confidence measures is proposed for speech recognition. Centralized and distributed schemes are considered for confidence measure combining under Bayesian framework. Centralized combining is feature level fusion which combines the values of individual confidence scores and makes a final decision. In contrast, distributed combining is decision level fusion which combines the individual decision makings made by each individual confidence scoring method. Both methods are basically based on the statistical modeling of confidence features. In addition, adaptation of confidence score using the statistical models also presented. The proposed methods reduced the classification error rate by 17{\%} from conventional single feature based confidence scoring method in isolated word Out-of-Vocabulary rejection test.",
author = "Kim, {Tae Yoon} and Hanseok Ko",
year = "2004",
month = "12",
day = "1",
language = "English",
isbn = "0780386396",
pages = "164--168",
editor = "S.J. Ko",
booktitle = "Proceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, ISPACS 2004",

}

TY - GEN

T1 - Combining of confidence measures under Bayesian framework for speech recognition

AU - Kim, Tae Yoon

AU - Ko, Hanseok

PY - 2004/12/1

Y1 - 2004/12/1

N2 - Bayesian combining of confidence measures is proposed for speech recognition. Centralized and distributed schemes are considered for confidence measure combining under Bayesian framework. Centralized combining is feature level fusion which combines the values of individual confidence scores and makes a final decision. In contrast, distributed combining is decision level fusion which combines the individual decision makings made by each individual confidence scoring method. Both methods are basically based on the statistical modeling of confidence features. In addition, adaptation of confidence score using the statistical models also presented. The proposed methods reduced the classification error rate by 17% from conventional single feature based confidence scoring method in isolated word Out-of-Vocabulary rejection test.

AB - Bayesian combining of confidence measures is proposed for speech recognition. Centralized and distributed schemes are considered for confidence measure combining under Bayesian framework. Centralized combining is feature level fusion which combines the values of individual confidence scores and makes a final decision. In contrast, distributed combining is decision level fusion which combines the individual decision makings made by each individual confidence scoring method. Both methods are basically based on the statistical modeling of confidence features. In addition, adaptation of confidence score using the statistical models also presented. The proposed methods reduced the classification error rate by 17% from conventional single feature based confidence scoring method in isolated word Out-of-Vocabulary rejection test.

UR - http://www.scopus.com/inward/record.url?scp=21444459920&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=21444459920&partnerID=8YFLogxK

M3 - Conference contribution

SN - 0780386396

SP - 164

EP - 168

BT - Proceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, ISPACS 2004

A2 - Ko, S.J.

ER -