A two-channel training algorithm for hidden Markov model to identify visual speech elements

Author

Foo, Say Wei ; Lian, Yong ; Dong, Liang

Author_Institution

Sch. of Electr. & Electron. Eng., Nanyang Technol. Univ., Singapore

Volume

2

fYear

2003

Abstract

A novel two-channel algorithm is proposed in this paper for discriminative training of Hidden Markov Models (HMMs). It adjusts the symbol emission coefficients of an existing HMM to maximize the separable distance between a pair of confusable training samples. The method is applied to identify the visemes of visual speech. The results indicate that the two-channel training method provides better accuracy on separating similar visemes than the conventional Baum-Welch estimation.

Keywords

hidden Markov models; maximum likelihood estimation; speaker recognition; speech recognition; confusable training samples; discriminative training; hidden Markov model; maximum likelihood HMM; separable distance; sequence recognition problems; speaker identification; speech recognition; symbol emission coefficients; two-channel training algorithm; viseme recognition; visemes identification; visual speech elements identification; Convergence; Handwriting recognition; Hidden Markov models; Management training; Maximum likelihood estimation; Signal processing; Speech recognition; Training data;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 2003. ISCAS '03. Proceedings of the 2003 International Symposium on

Print_ISBN

0-7803-7761-3

Type

conf

DOI

10.1109/ISCAS.2003.1206038

Filename

1206038