Acoustic adaptation using nonlinear transformations of HMM parameters

Author

Abrash, Victor ; Sankar, Ananth ; Franco, Horacio ; Cohen, Michael

Author_Institution

Speech Res. & Technol. Lab., SRI Int., Menlo Park, CA, USA

Volume

2

fYear

1996

Firstpage

729

Abstract

Speech recognition performance degrades significantly when there is a mismatch between testing and training conditions. Linear transformation-based maximum-likelihood (ML) techniques have been proposed recently to tackle this problem. We extend this approach to use nonlinear transformations. These are implemented by multilayer perceptrons (MLPs) which transform the Gaussian means. We derive a generalized expectation-maximization (GEM) training algorithm to estimate the MLP weights. Some preliminary experimental results on nonnative speaker adaptation are presented

Keywords

acoustic signal processing; adaptive signal processing; learning (artificial intelligence); maximum likelihood estimation; multilayer perceptrons; speech processing; speech recognition; Gaussian means; HMM parameters; MLP; MLP weights estimation; acoustic adaptation; experimental results; generalized expectation-maximization training algorithm; maximum-likelihood techniques; multilayer perceptrons; nonlinear transformations; nonnative speaker adaptation; speech recognition performance; testing conditions; training conditions; Acoustic testing; Automatic speech recognition; Degradation; Hidden Markov models; Microphones; Multilayer perceptrons; Noise robustness; Nonlinear acoustics; Speech enhancement; Speech recognition;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech, and Signal Processing, 1996. ICASSP-96. Conference Proceedings., 1996 IEEE International Conference on

ISSN

1520-6149

Print_ISBN

0-7803-3192-3

Type

conf

DOI

10.1109/ICASSP.1996.543224

Filename

543224