Speaker adaptations in sparse training data for improved speaker verification

Sungjoo Ahn, Hanseok Ko

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

The over-training problem in speaker verification occurs when modelling a speaker with sparse training data. The authors propose to solve this problem by employing effective speaker adaptations using a hybrid version of the maximum a posteriori (MAP) and maximum likelihood linear regression (MLLR) methods. Experimental results show that the speaker verification system using the proposed hybrid adaptation scheme outperforms systems based on speaker models without adaptation by a factor of up to 5.

Original languageEnglish
Pages (from-to)371-373
Number of pages3
JournalElectronics Letters
Volume36
Issue number4
DOIs
Publication statusPublished - 2000 Feb 17

Fingerprint

Linear regression
Maximum likelihood

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Cite this

Speaker adaptations in sparse training data for improved speaker verification. / Ahn, Sungjoo; Ko, Hanseok.

In: Electronics Letters, Vol. 36, No. 4, 17.02.2000, p. 371-373.

Research output: Contribution to journalArticle

@article{3de1093349714cf2998f5294951c1d4a,
title = "Speaker adaptations in sparse training data for improved speaker verification",
abstract = "The over-training problem in speaker verification occurs when modelling a speaker with sparse training data. The authors propose to solve this problem by employing effective speaker adaptations using a hybrid version of the maximum a posteriori (MAP) and maximum likelihood linear regression (MLLR) methods. Experimental results show that the speaker verification system using the proposed hybrid adaptation scheme outperforms systems based on speaker models without adaptation by a factor of up to 5.",
author = "Sungjoo Ahn and Hanseok Ko",
year = "2000",
month = "2",
day = "17",
doi = "10.1049/el:20000330",
language = "English",
volume = "36",
pages = "371--373",
journal = "Electronics Letters",
issn = "0013-5194",
publisher = "Institution of Engineering and Technology",
number = "4",

}

TY - JOUR

T1 - Speaker adaptations in sparse training data for improved speaker verification

AU - Ahn, Sungjoo

AU - Ko, Hanseok

PY - 2000/2/17

Y1 - 2000/2/17

N2 - The over-training problem in speaker verification occurs when modelling a speaker with sparse training data. The authors propose to solve this problem by employing effective speaker adaptations using a hybrid version of the maximum a posteriori (MAP) and maximum likelihood linear regression (MLLR) methods. Experimental results show that the speaker verification system using the proposed hybrid adaptation scheme outperforms systems based on speaker models without adaptation by a factor of up to 5.

AB - The over-training problem in speaker verification occurs when modelling a speaker with sparse training data. The authors propose to solve this problem by employing effective speaker adaptations using a hybrid version of the maximum a posteriori (MAP) and maximum likelihood linear regression (MLLR) methods. Experimental results show that the speaker verification system using the proposed hybrid adaptation scheme outperforms systems based on speaker models without adaptation by a factor of up to 5.

UR - http://www.scopus.com/inward/record.url?scp=0033882589&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0033882589&partnerID=8YFLogxK

U2 - 10.1049/el:20000330

DO - 10.1049/el:20000330

M3 - Article

AN - SCOPUS:0033882589

VL - 36

SP - 371

EP - 373

JO - Electronics Letters

JF - Electronics Letters

SN - 0013-5194

IS - 4

ER -