New feedback method of hybrid HMM/ANN methods for continuous speech recognition

Tranzai Lee Tranzai Lee,Daowen Chen Daowen Chen

doi:10.1109/icassp.1998.674479

Tranzai Lee Tranzai Lee, Daowen Chen Daowen Chen

https://doi.org/10.1109/icassp.1998.674479

Copy DOI

Export

Save

Cite

Publication Date: May 12, 1998

Affiliation: Institute of Automation

Abstract
Full-Text
Similar Papers

Abstract

Listen

In continuous speech recognition, the co-pronunciation between two successive phonemes seriously disturbs the recognition effect. It is difficult for pure hidden Markov model (HMM) methods to cope with co-pronunciation, because HMM methods consider that two successive frames of speech are independant. The hybrid HMM and artificial neural network (ANN) methods with feedback multilayer perceptron (MLP) (Bourlard and Wellekens, 1990; Bourlard and Morgan 1994) provide the ability to cope with co-pronunciation by means of feedback input. In this paper, we propose a new feedback method for feedback hybrid HMM/ANN methods on the basis of the original methods. The new feedback method provides more information of co-pronunciation to the feedback ANN. From HMM/ANN with feedback double MLP structure, we discuss the method that reduces the computation of the feedback MLP during recognition.

Full Text