Presentation 2004/12/13
Aggregate a Posteriori Linear Regression Adaptation of Hidden Markov Models
Jen-Tzung Chien, Chih-Hsien Huang,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We present a rapid and discriminative speaker adaptation algorithm for hidden Markov model (HMM) based speech recognition. The adaptation is based on the linear regression framework. Attractively, we estimate the regression matrices from speaker-specific adaptation data according to the aggregate a posteriori criterion, which is expressed in a form of classification error function. The aggregate a posteriori linear regression (AAPLR) is proposed to achieve discriminative adaptation so that the classification errors of adaptation data are minimized. The superiority of AAPLR to maximum a posteriori linear regression (MAPLR) is demonstrated. Different from minimum classification error linear regression (MCELR), AAPLR has closed-form solution to fulfill rapid adaptation. Experimental results reveal that AAPLR speaker adaptation does improve speech recognition performance with moderate computational cost compared to the maximum likelihood linear regression (MLLR), MAPLR and MCELR.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Hidden Markov model / MLLR / MAPLR / MCELR / Discriminative training / Aggregate a posteriori criterion / Speaker adaptation / Speech recognition
Paper # NLC2004-53,SP2004-93
Date of Issue

Conference Information
Committee NLC
Conference Date 2004/12/13(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Natural Language Understanding and Models of Communication (NLC)
Language ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Aggregate a Posteriori Linear Regression Adaptation of Hidden Markov Models
Sub Title (in English)
Keyword(1) Hidden Markov model
Keyword(2) MLLR
Keyword(3) MAPLR
Keyword(4) MCELR
Keyword(5) Discriminative training
Keyword(6) Aggregate a posteriori criterion
Keyword(7) Speaker adaptation
Keyword(8) Speech recognition
1st Author's Name Jen-Tzung Chien
1st Author's Affiliation Department of Computer Science and Information Engineering National Cheng Kung University()
2nd Author's Name Chih-Hsien Huang
2nd Author's Affiliation Department of Computer Science and Information Engineering National Cheng Kung University
Date 2004/12/13
Paper # NLC2004-53,SP2004-93
Volume (vol) vol.104
Number (no) 538
Page pp.pp.-
#Pages 6
Date of Issue