Automatic Generation of Non-uniform and Context-Dependent HMMs Based on the Variational Bayesian Approach
Summary :
We propose a new method both for automatically creating non-uniform, context-dependent HMM topologies, and selecting the number of mixture components based on the Variational Bayesian (VB) approach. Although the Maximum Likelihood (ML) criterion is generally used to create HMM topologies, it has an over-fitting problem. Recently, to avoid this problem, the VB approach has been applied to create acoustic models for speech recognition. We introduce the VB approach to the Successive State Splitting (SSS) algorithm, which can create both contextual and temporal variations for HMMs. Experimental results indicate that the proposed method can automatically create a more efficient model than the original method. We evaluated a method to increase the number of mixture components by using the VB approach and considering temporal structures. The VB approach obtained almost the same performance as the smaller number of mixture components in comparison with that obtained by using ML-based methods.
- Publication
- IEICE TRANSACTIONS on Information Vol.E88-D No.3 pp.391-400
- Publication Date
- 2005/03/01
- Publicized
- Online ISSN
- DOI
- 10.1093/ietisy/e88-d.3.391
- Type of Manuscript
- Special Section PAPER (Special Section on Corpus-Based Speech Technologies)
- Category
- Feature Extraction and Acoustic Medelings
Authors
Keyword
Latest Issue
Copyrights notice of machine-translated contents
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Cite this
Copy
Takatoshi JITSUHIRO, Satoshi NAKAMURA, "Automatic Generation of Non-uniform and Context-Dependent HMMs Based on the Variational Bayesian Approach" in IEICE TRANSACTIONS on Information,
vol. E88-D, no. 3, pp. 391-400, March 2005, doi: 10.1093/ietisy/e88-d.3.391.
Abstract: We propose a new method both for automatically creating non-uniform, context-dependent HMM topologies, and selecting the number of mixture components based on the Variational Bayesian (VB) approach. Although the Maximum Likelihood (ML) criterion is generally used to create HMM topologies, it has an over-fitting problem. Recently, to avoid this problem, the VB approach has been applied to create acoustic models for speech recognition. We introduce the VB approach to the Successive State Splitting (SSS) algorithm, which can create both contextual and temporal variations for HMMs. Experimental results indicate that the proposed method can automatically create a more efficient model than the original method. We evaluated a method to increase the number of mixture components by using the VB approach and considering temporal structures. The VB approach obtained almost the same performance as the smaller number of mixture components in comparison with that obtained by using ML-based methods.
URL: https://global.ieice.org/en_transactions/information/10.1093/ietisy/e88-d.3.391/_p
Copy
@ARTICLE{e88-d_3_391,
author={Takatoshi JITSUHIRO, Satoshi NAKAMURA, },
journal={IEICE TRANSACTIONS on Information},
title={Automatic Generation of Non-uniform and Context-Dependent HMMs Based on the Variational Bayesian Approach},
year={2005},
volume={E88-D},
number={3},
pages={391-400},
abstract={We propose a new method both for automatically creating non-uniform, context-dependent HMM topologies, and selecting the number of mixture components based on the Variational Bayesian (VB) approach. Although the Maximum Likelihood (ML) criterion is generally used to create HMM topologies, it has an over-fitting problem. Recently, to avoid this problem, the VB approach has been applied to create acoustic models for speech recognition. We introduce the VB approach to the Successive State Splitting (SSS) algorithm, which can create both contextual and temporal variations for HMMs. Experimental results indicate that the proposed method can automatically create a more efficient model than the original method. We evaluated a method to increase the number of mixture components by using the VB approach and considering temporal structures. The VB approach obtained almost the same performance as the smaller number of mixture components in comparison with that obtained by using ML-based methods.},
keywords={},
doi={10.1093/ietisy/e88-d.3.391},
ISSN={},
month={March},}
Copy
TY - JOUR
TI - Automatic Generation of Non-uniform and Context-Dependent HMMs Based on the Variational Bayesian Approach
T2 - IEICE TRANSACTIONS on Information
SP - 391
EP - 400
AU - Takatoshi JITSUHIRO
AU - Satoshi NAKAMURA
PY - 2005
DO - 10.1093/ietisy/e88-d.3.391
JO - IEICE TRANSACTIONS on Information
SN -
VL - E88-D
IS - 3
JA - IEICE TRANSACTIONS on Information
Y1 - March 2005
AB - We propose a new method both for automatically creating non-uniform, context-dependent HMM topologies, and selecting the number of mixture components based on the Variational Bayesian (VB) approach. Although the Maximum Likelihood (ML) criterion is generally used to create HMM topologies, it has an over-fitting problem. Recently, to avoid this problem, the VB approach has been applied to create acoustic models for speech recognition. We introduce the VB approach to the Successive State Splitting (SSS) algorithm, which can create both contextual and temporal variations for HMMs. Experimental results indicate that the proposed method can automatically create a more efficient model than the original method. We evaluated a method to increase the number of mixture components by using the VB approach and considering temporal structures. The VB approach obtained almost the same performance as the smaller number of mixture components in comparison with that obtained by using ML-based methods.
ER -
English
