A new method for model selection in speech recognition

Author

Wu, Yahui ; Liu, Gang ; Guo, Jun

Author_Institution

Pattern Recognition & Intell. Syst. Lab., Beijing Univ. of Posts & Telecommun., Beijing

fYear

2008

fDate

19-22 Oct. 2008

Firstpage

1

Lastpage

4

Abstract

A new method based on model selection for acoustic model training is proposed .The MPE trained model and the MLE trained model is used for model selection for the following training. The selection criteria is based on the ratio of the inter-variance to the intra-variance of each model. Besides we also propose a cluster method for the model in order to get the accuracy information for the weight calculation. The experiments demonstrate that the new model can get better performance than any of the directly trained models.

Keywords

maximum likelihood estimation; speech recognition; accuracy information; acoustic model training; inter-variance; intra-variance; maximum likelihood estimation; minimum phone error; model selection; speech recognition; weight calculation; Computer errors; Hidden Markov models; Intelligent systems; Laboratories; Maximum likelihood estimation; Pattern recognition; Speech analysis; Speech recognition; Statistical analysis; Training data; MLE; MPE; Model selection; speech recognition;

fLanguage

English

Publisher

ieee

Conference_Titel

Natural Language Processing and Knowledge Engineering, 2008. NLP-KE '08. International Conference on

Conference_Location

Beijing

Print_ISBN

978-1-4244-4515-8

Electronic_ISBN

978-1-4244-2780-2

Type

conf

DOI

10.1109/NLPKE.2008.4906801

Filename

4906801