A Study of Minimum Classification Error (MCE) Linear Regression for Supervised Adaptation of MCE-Trained Continuous-Density Hidden Markov Models

Author

Wu, Jian ; Huo, Qiang

Author_Institution

Dept. of Comput. Sci., Hong Kong Univ.

Volume

15

Issue

2

fYear

2007

Firstpage

478

Lastpage

488

Abstract

In this paper, we present a formulation of minimum classification error linear regression (MCELR) for the adaptation of Gaussian mixture continuous-density hidden Markov model (CDHMM) parameters. Two optimization approaches, namely generalized probabilistic descent (GPD) and Quickprop are studied and compared for the optimization of the MCELR objective function. The effectiveness of the proposed MCELR technique is confirmed via a series of supervised speaker adaptation experiments on a task of continuous Putonghua (Mandarin Chinese) speech recognition

Keywords

Gaussian processes; hidden Markov models; regression analysis; speech processing; speech recognition; Gaussian mixture; MCE-trained continuous-density hidden Markov models; Mandarin Chinese; Quickprop; continuous Putonghua; generalized probabilistic descent; minimum classification error linear regression; speech recognition; Adaptation model; Automatic speech recognition; Computer science; Hidden Markov models; Linear regression; Maximum likelihood estimation; Mutual information; Parameter estimation; Speech recognition; Training data; HMM adaptation; Hidden Markov model (HMM); minimum classification error linear regression (MCELR); speaker adaptation;

fLanguage

English

Journal_Title

Audio, Speech, and Language Processing, IEEE Transactions on

Publisher

ieee

ISSN

1558-7916

Type

jour

DOI

10.1109/TASL.2006.881692

Filename

4067055