Estimation of Prior Probabilities in Speaker Recognition

Dat Tran

doi:10.1109/ISIMP.2004.1434020

Estimation of Prior Probabilities in Speaker Recognition

Dat Tran

Information Systems

Research output: A Conference proceeding or a Chapter in Book › Conference contribution › peer-review

4 Citations (Scopus)

29 Downloads (Pure)

Abstract

According to Bayesian decision theory, the maximum a posteriori (MAP) decision rule is used to minimize the speaker recognition error rate. The a posteriori probability is determined if the a priori probability and the likelihood function are known. However, there has been no method to determine the a priori probability, therefore the maximum likelihood (ML) decision rule is used instead. The paper proposes a method to estimate the a priori probability for speakers based on a training data set and speaker models. Speaker identification experiments performed on 138 Gaussian mixture speaker models in the YOHO database using the MAP rule showed lower error rates than using the ML rule.

Original language	English
Title of host publication	Proceedings of the 2004 International Symposium on Intelligent Multimedia, Video and Speech Processing
Editors	J Kwok, LM Po
Place of Publication	Hong Kong
Publisher	IEEE, Institute of Electrical and Electronics Engineers
Pages	141-144
Number of pages	4
ISBN (Print)	0-7803-8688-4
DOIs	https://doi.org/10.1109/ISIMP.2004.1434020
Publication status	Published - 2004
Event	2004 International Symposium on Intelligent Multimedia, Video and Speech Processing - , Hong Kong Duration: 19 Oct 2004 → 21 Oct 2004

Conference

Conference	2004 International Symposium on Intelligent Multimedia, Video and Speech Processing
Country/Territory	Hong Kong
Period	19/10/04 → 21/10/04

Access to Document

10.1109/ISIMP.2004.1434020

PaperFinal published version, 765 KB

Cite this

@inproceedings{b21ef1c7f5604fb585f511c8b32a6170,

title = "Estimation of Prior Probabilities in Speaker Recognition",

abstract = "According to Bayesian decision theory, the maximum a posteriori (MAP) decision rule is used to minimize the speaker recognition error rate. The a posteriori probability is determined if the a priori probability and the likelihood function are known. However, there has been no method to determine the a priori probability, therefore the maximum likelihood (ML) decision rule is used instead. The paper proposes a method to estimate the a priori probability for speakers based on a training data set and speaker models. Speaker identification experiments performed on 138 Gaussian mixture speaker models in the YOHO database using the MAP rule showed lower error rates than using the ML rule.",

author = "Dat Tran",

year = "2004",

doi = "10.1109/ISIMP.2004.1434020",

language = "English",

isbn = "0-7803-8688-4",

pages = "141--144",

editor = "J Kwok and LM Po",

booktitle = "Proceedings of the 2004 International Symposium on Intelligent Multimedia, Video and Speech Processing",

publisher = "IEEE, Institute of Electrical and Electronics Engineers",

address = "United States",

note = "2004 International Symposium on Intelligent Multimedia, Video and Speech Processing ; Conference date: 19-10-2004 Through 21-10-2004",

}

Tran, D 2004, Estimation of Prior Probabilities in Speaker Recognition. in J Kwok & LM Po (eds), Proceedings of the 2004 International Symposium on Intelligent Multimedia, Video and Speech Processing. IEEE, Institute of Electrical and Electronics Engineers, Hong Kong, pp. 141-144, 2004 International Symposium on Intelligent Multimedia, Video and Speech Processing, Hong Kong, 19/10/04. https://doi.org/10.1109/ISIMP.2004.1434020

Estimation of Prior Probabilities in Speaker Recognition. / Tran, Dat.
Proceedings of the 2004 International Symposium on Intelligent Multimedia, Video and Speech Processing. ed. / J Kwok; LM Po. Hong Kong: IEEE, Institute of Electrical and Electronics Engineers, 2004. p. 141-144.

Research output: A Conference proceeding or a Chapter in Book › Conference contribution › peer-review

TY - GEN

T1 - Estimation of Prior Probabilities in Speaker Recognition

AU - Tran, Dat

PY - 2004

Y1 - 2004

N2 - According to Bayesian decision theory, the maximum a posteriori (MAP) decision rule is used to minimize the speaker recognition error rate. The a posteriori probability is determined if the a priori probability and the likelihood function are known. However, there has been no method to determine the a priori probability, therefore the maximum likelihood (ML) decision rule is used instead. The paper proposes a method to estimate the a priori probability for speakers based on a training data set and speaker models. Speaker identification experiments performed on 138 Gaussian mixture speaker models in the YOHO database using the MAP rule showed lower error rates than using the ML rule.

AB - According to Bayesian decision theory, the maximum a posteriori (MAP) decision rule is used to minimize the speaker recognition error rate. The a posteriori probability is determined if the a priori probability and the likelihood function are known. However, there has been no method to determine the a priori probability, therefore the maximum likelihood (ML) decision rule is used instead. The paper proposes a method to estimate the a priori probability for speakers based on a training data set and speaker models. Speaker identification experiments performed on 138 Gaussian mixture speaker models in the YOHO database using the MAP rule showed lower error rates than using the ML rule.

U2 - 10.1109/ISIMP.2004.1434020

DO - 10.1109/ISIMP.2004.1434020

M3 - Conference contribution

SN - 0-7803-8688-4

SP - 141

EP - 144

BT - Proceedings of the 2004 International Symposium on Intelligent Multimedia, Video and Speech Processing

A2 - Kwok, J

A2 - Po, LM

PB - IEEE, Institute of Electrical and Electronics Engineers

CY - Hong Kong

T2 - 2004 International Symposium on Intelligent Multimedia, Video and Speech Processing

Y2 - 19 October 2004 through 21 October 2004

ER -

Estimation of Prior Probabilities in Speaker Recognition

Abstract

Conference

Access to Document

Fingerprint

Cite this