Bayesian approaches to acoustic modeling

A review

Shinji Watanabe, Atsushi Nakamura

Research output: Contribution to journalReview article

1 Citation (Scopus)

Abstract

This paper focuses on applications of Bayesian approaches to acoustic modeling for speech recognition and related speechprocessing applications. Bayesian approaches have been widely studied in the fields of statistics and machine learning, and one of their advantages is that their generalization capability is better than that of conventional approaches (e.g., maximum likelihood). On the other hand, since inference in Bayesian approaches involves integrals and expectations that are mathematically intractable in most cases and require heavy numerical computations, it is generally difficult to apply them to practical speech recognition problems.However, there have beenmany such attempts, and this paper aims to summarize these attempts to encourage further progress on Bayesian approaches in the speech-processing field. This paper describes various applications of Bayesian approaches to speech processing in terms of the four typical ways of approximating Bayesian inferences, i.e., maximum a posteriori approximation, model complexity control using a Bayesian information criterion based on asymptotic approximation, variational approximation, and Markov chain Monte Carlo-based sampling techniques.

Original languageEnglish
Article numbere5
JournalAPSIPA Transactions on Signal and Information Processing
Volume1
DOIs
Publication statusPublished - 2012
Externally publishedYes

Fingerprint

Speech processing
Acoustics
Speech recognition
Markov processes
Maximum likelihood
Learning systems
Statistics
Sampling

Keywords

  • Approximate bayesian inference
  • Bayesian approach
  • Machine learning
  • Speech processing

ASJC Scopus subject areas

  • Information Systems
  • Signal Processing

Cite this

Bayesian approaches to acoustic modeling : A review. / Watanabe, Shinji; Nakamura, Atsushi.

In: APSIPA Transactions on Signal and Information Processing, Vol. 1, e5, 2012.

Research output: Contribution to journalReview article

@article{eb399bfa999f4ccebf87941dd2ab6084,
title = "Bayesian approaches to acoustic modeling: A review",
abstract = "This paper focuses on applications of Bayesian approaches to acoustic modeling for speech recognition and related speechprocessing applications. Bayesian approaches have been widely studied in the fields of statistics and machine learning, and one of their advantages is that their generalization capability is better than that of conventional approaches (e.g., maximum likelihood). On the other hand, since inference in Bayesian approaches involves integrals and expectations that are mathematically intractable in most cases and require heavy numerical computations, it is generally difficult to apply them to practical speech recognition problems.However, there have beenmany such attempts, and this paper aims to summarize these attempts to encourage further progress on Bayesian approaches in the speech-processing field. This paper describes various applications of Bayesian approaches to speech processing in terms of the four typical ways of approximating Bayesian inferences, i.e., maximum a posteriori approximation, model complexity control using a Bayesian information criterion based on asymptotic approximation, variational approximation, and Markov chain Monte Carlo-based sampling techniques.",
keywords = "Approximate bayesian inference, Bayesian approach, Machine learning, Speech processing",
author = "Shinji Watanabe and Atsushi Nakamura",
year = "2012",
doi = "10.1017/ATSIP.2012.6",
language = "English",
volume = "1",
journal = "APSIPA Transactions on Signal and Information Processing",
issn = "2048-7703",
publisher = "Cambridge University Press",

}

TY - JOUR

T1 - Bayesian approaches to acoustic modeling

T2 - A review

AU - Watanabe, Shinji

AU - Nakamura, Atsushi

PY - 2012

Y1 - 2012

N2 - This paper focuses on applications of Bayesian approaches to acoustic modeling for speech recognition and related speechprocessing applications. Bayesian approaches have been widely studied in the fields of statistics and machine learning, and one of their advantages is that their generalization capability is better than that of conventional approaches (e.g., maximum likelihood). On the other hand, since inference in Bayesian approaches involves integrals and expectations that are mathematically intractable in most cases and require heavy numerical computations, it is generally difficult to apply them to practical speech recognition problems.However, there have beenmany such attempts, and this paper aims to summarize these attempts to encourage further progress on Bayesian approaches in the speech-processing field. This paper describes various applications of Bayesian approaches to speech processing in terms of the four typical ways of approximating Bayesian inferences, i.e., maximum a posteriori approximation, model complexity control using a Bayesian information criterion based on asymptotic approximation, variational approximation, and Markov chain Monte Carlo-based sampling techniques.

AB - This paper focuses on applications of Bayesian approaches to acoustic modeling for speech recognition and related speechprocessing applications. Bayesian approaches have been widely studied in the fields of statistics and machine learning, and one of their advantages is that their generalization capability is better than that of conventional approaches (e.g., maximum likelihood). On the other hand, since inference in Bayesian approaches involves integrals and expectations that are mathematically intractable in most cases and require heavy numerical computations, it is generally difficult to apply them to practical speech recognition problems.However, there have beenmany such attempts, and this paper aims to summarize these attempts to encourage further progress on Bayesian approaches in the speech-processing field. This paper describes various applications of Bayesian approaches to speech processing in terms of the four typical ways of approximating Bayesian inferences, i.e., maximum a posteriori approximation, model complexity control using a Bayesian information criterion based on asymptotic approximation, variational approximation, and Markov chain Monte Carlo-based sampling techniques.

KW - Approximate bayesian inference

KW - Bayesian approach

KW - Machine learning

KW - Speech processing

UR - http://www.scopus.com/inward/record.url?scp=84887091716&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84887091716&partnerID=8YFLogxK

U2 - 10.1017/ATSIP.2012.6

DO - 10.1017/ATSIP.2012.6

M3 - Review article

VL - 1

JO - APSIPA Transactions on Signal and Information Processing

JF - APSIPA Transactions on Signal and Information Processing

SN - 2048-7703

M1 - e5

ER -