Selection of shared-state Hidden Markov model structure using Bayesian criterion

Shinji Watanabe, Yasuhiro Minami, Atsushi Nakamura, Naonori Ueda

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

A Shared-State Hidden Markov Model (SS-HMM) has been widely used as an acoustic model in speech recognition. In this paper, we propose a method for constructing SS-HMMs within a practical Bayesian framework. Our method derives the Bayesian model selection criterion for the SS-HMM based on the variational Bayesian approach. The appropriate phonetic decision tree structure of the SS-HMM is found by using the Bayesian criterion. Unlike the conventional asymptotic criteria, this criterion is applicable even in the case of an insufficient amount of training data. The experimental results on isolated word recognition demonstrate that the proposed method does not require the tuning parameter that must be tuned according to the amount of training data, and is useful for selecting the appropriate SS-HMM structure for practical use.

Original languageEnglish
Pages (from-to)1-9
Number of pages9
JournalIEICE Transactions on Information and Systems
VolumeE88-D
Issue number1
DOIs
Publication statusPublished - 2005 Jan
Externally publishedYes

Fingerprint

Hidden Markov models
Model structures
Speech analysis
Decision trees
Speech recognition
Tuning
Acoustics

Keywords

  • Bayesian criterion
  • Model structure selection
  • Shared-state HMM
  • Speech recognition
  • Variational bayes

ASJC Scopus subject areas

  • Information Systems
  • Computer Graphics and Computer-Aided Design
  • Software

Cite this

Selection of shared-state Hidden Markov model structure using Bayesian criterion. / Watanabe, Shinji; Minami, Yasuhiro; Nakamura, Atsushi; Ueda, Naonori.

In: IEICE Transactions on Information and Systems, Vol. E88-D, No. 1, 01.2005, p. 1-9.

Research output: Contribution to journalArticle

Watanabe, Shinji ; Minami, Yasuhiro ; Nakamura, Atsushi ; Ueda, Naonori. / Selection of shared-state Hidden Markov model structure using Bayesian criterion. In: IEICE Transactions on Information and Systems. 2005 ; Vol. E88-D, No. 1. pp. 1-9.
@article{a07bdd54e4e14cfe8df1e34a79d389a3,
title = "Selection of shared-state Hidden Markov model structure using Bayesian criterion",
abstract = "A Shared-State Hidden Markov Model (SS-HMM) has been widely used as an acoustic model in speech recognition. In this paper, we propose a method for constructing SS-HMMs within a practical Bayesian framework. Our method derives the Bayesian model selection criterion for the SS-HMM based on the variational Bayesian approach. The appropriate phonetic decision tree structure of the SS-HMM is found by using the Bayesian criterion. Unlike the conventional asymptotic criteria, this criterion is applicable even in the case of an insufficient amount of training data. The experimental results on isolated word recognition demonstrate that the proposed method does not require the tuning parameter that must be tuned according to the amount of training data, and is useful for selecting the appropriate SS-HMM structure for practical use.",
keywords = "Bayesian criterion, Model structure selection, Shared-state HMM, Speech recognition, Variational bayes",
author = "Shinji Watanabe and Yasuhiro Minami and Atsushi Nakamura and Naonori Ueda",
year = "2005",
month = "1",
doi = "10.1093/ietisy/E88-D.1.1",
language = "English",
volume = "E88-D",
pages = "1--9",
journal = "IEICE Transactions on Information and Systems",
issn = "0916-8532",
publisher = "Maruzen Co., Ltd/Maruzen Kabushikikaisha",
number = "1",

}

TY - JOUR

T1 - Selection of shared-state Hidden Markov model structure using Bayesian criterion

AU - Watanabe, Shinji

AU - Minami, Yasuhiro

AU - Nakamura, Atsushi

AU - Ueda, Naonori

PY - 2005/1

Y1 - 2005/1

N2 - A Shared-State Hidden Markov Model (SS-HMM) has been widely used as an acoustic model in speech recognition. In this paper, we propose a method for constructing SS-HMMs within a practical Bayesian framework. Our method derives the Bayesian model selection criterion for the SS-HMM based on the variational Bayesian approach. The appropriate phonetic decision tree structure of the SS-HMM is found by using the Bayesian criterion. Unlike the conventional asymptotic criteria, this criterion is applicable even in the case of an insufficient amount of training data. The experimental results on isolated word recognition demonstrate that the proposed method does not require the tuning parameter that must be tuned according to the amount of training data, and is useful for selecting the appropriate SS-HMM structure for practical use.

AB - A Shared-State Hidden Markov Model (SS-HMM) has been widely used as an acoustic model in speech recognition. In this paper, we propose a method for constructing SS-HMMs within a practical Bayesian framework. Our method derives the Bayesian model selection criterion for the SS-HMM based on the variational Bayesian approach. The appropriate phonetic decision tree structure of the SS-HMM is found by using the Bayesian criterion. Unlike the conventional asymptotic criteria, this criterion is applicable even in the case of an insufficient amount of training data. The experimental results on isolated word recognition demonstrate that the proposed method does not require the tuning parameter that must be tuned according to the amount of training data, and is useful for selecting the appropriate SS-HMM structure for practical use.

KW - Bayesian criterion

KW - Model structure selection

KW - Shared-state HMM

KW - Speech recognition

KW - Variational bayes

UR - http://www.scopus.com/inward/record.url?scp=27544496222&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=27544496222&partnerID=8YFLogxK

U2 - 10.1093/ietisy/E88-D.1.1

DO - 10.1093/ietisy/E88-D.1.1

M3 - Article

AN - SCOPUS:27544496222

VL - E88-D

SP - 1

EP - 9

JO - IEICE Transactions on Information and Systems

JF - IEICE Transactions on Information and Systems

SN - 0916-8532

IS - 1

ER -