DocumentCode :
417258
Title :
Automatic generation of non-uniform HMM structures based on variational Bayesian approach
Author :
Jitsuhiro, Takatoshi ; Nakamura, Satoshi
Author_Institution :
ATR Spoken Language Translation Res. Labs., Kyoto, Japan
Volume :
1
fYear :
2004
fDate :
17-21 May 2004
Abstract :
We propose using the variational Bayesian (VB) approach for automatically creating nonuniform, context-dependent HMM topologies in speech recognition. The maximum likelihood (ML) criterion is generally used to create HMM topologies. However, it has an over-fitting problem. Information criteria have been used to overcome this problem, but theoretically they cannot be applied to complicated models like HMM. Recently, to avoid these problems, the VB approach has been developed in the machine-learning field. We introduce the VB approach to the successive state splitting (SSS) algorithm, which can create both contextual and temporal variations for HMM. We define the prior and posterior probability densities and free energy with latent variables as split and stop criteria. Experimental results show that the proposed method can automatically create a more efficient model and obtain better performance, especially for vowels, than the original method.
Keywords :
Bayes methods; hidden Markov models; speech recognition; topology; variational techniques; SSS algorithm; automatic generation; context-dependent HMM topologies; nonuniform HMM structures; speech recognition; successive state splitting algorithm; variational Bayesian approach; vowel performance; Bayesian methods; Clustering algorithms; Decision trees; Hidden Markov models; Laboratories; Maximum likelihood estimation; Natural languages; Speech recognition; Topology; Training data;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Acoustics, Speech, and Signal Processing, 2004. Proceedings. (ICASSP '04). IEEE International Conference on
ISSN :
1520-6149
Print_ISBN :
0-7803-8484-9
Type :
conf
DOI :
10.1109/ICASSP.2004.1326108
Filename :
1326108
Link To Document :
بازگشت