On the generation and use of a parallel-branch subunit model in continuous HMM

Author

Park, Y.K. ; Un, C.K.

Author_Institution

Dept. of Electr. Eng., Korea Adv. Inst. of Sci. & Technol., Seoul, South Korea

Volume

4

Issue

4

fYear

1996

fDate

7/1/1996 12:00:00 AM

Firstpage

290

Lastpage

293

Abstract

In this paper, we propose two methods of obtaining a parallel-branch subunit model for improved recognition performance. In one method, the model is obtained by adding a new subunit branch based on misrecognized data in training to the previous parallel branches for that submit. In the other method, it is obtained by splitting off each subunit branch based on mixture components in continuous hidden Markov model. We propose to use a good initialization point obtained by error corrective estimation rather than by random or probabilistic statistics in the parallel-branch model when the number of mixtures and parallel branches increases

Keywords

error correction; hidden Markov models; maximum likelihood estimation; speech recognition; continuous HMM; continuous hidden Markov model; error corrective estimation; misrecognized data; mixture components; parallel-branch subunit model; speech recognition performance; training; Approximation algorithms; Convergence; Error analysis; Error correction; Estimation error; Helium; Hidden Markov models; Maximum likelihood estimation; Speech recognition; Training data;

fLanguage

English

Journal_Title

Speech and Audio Processing, IEEE Transactions on

Publisher

ieee

ISSN

1063-6676

Type

jour

DOI

10.1109/89.506932

Filename

506932