An improved MMIE training algorithm for speaker-independent, small vocabulary, continuous speech recognition

Author

Normandin, Yves ; Morgera, Salvatore D.

Author_Institution

Centre de Recherche Inf. de Montreal, Que., Canada

fYear

1991

fDate

14-17 Apr 1991

Firstpage

537

Abstract

Recently, Gopalakrishnan et al. (1989) introduced a reestimation formula for discrete HMMs (hidden Markov models) which applies to rational objective functions like the MMIE (maximum mutual information estimation) criterion. The authors analyze the formula and show how its convergence rate can be substantially improved. They introduce a corrective MMIE training algorithm, which, when applied to the TI/NIST connected digit database, has made it possible to reduce the string error rate by close to 50%. Gopalakrishnan´s result is extended to the continuous case by proposing a new formula for estimating the mean and variance parameters of diagonal Gaussian densities

Keywords

Markov processes; estimation theory; information theory; speech recognition; MMIE; MMIE training algorithm; TI/NIST connected digit database; continuous speech recognition; convergence rate; diagonal Gaussian densities; hidden Markov models; maximum mutual information estimation; mean; rational objective functions; reestimation formula; small vocabulary; speaker independent recognition; string error rate; variance; Communication systems; Convergence; Error analysis; Hidden Markov models; Maximum likelihood decoding; Maximum likelihood estimation; Mutual information; NIST; Speech recognition; Vocabulary;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech, and Signal Processing, 1991. ICASSP-91., 1991 International Conference on

Conference_Location

Toronto, Ont.

ISSN

1520-6149

Print_ISBN

0-7803-0003-3

Type

conf

DOI

10.1109/ICASSP.1991.150395

Filename

150395