Discriminative training for Bayesian sensing hidden Markov models

George Saon*, Jen-Tzung Chien

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

7 Scopus citations

Abstract

We describe feature space and model space discriminative training for a new class of acoustic models called Bayesian sensing hidden Markov models (BS-HMMs). In BS-HMMs, speech data is represented by a set of state-dependent basis vectors. The relevance of a feature vector to different bases is determined by the precision matrices of the sensing weights. The basis vectors and the precision matrices of the reconstruction errors are jointly estimated by optimizing a maximum mutual information (MMI) criterion. Additionally, we discuss the training of an fMPE-style discriminative feature transformation under the same criterion given these models. Experimental results on an LVCSR task show that the proposed models outperform discriminatively trained conventional HMMs with Gaussian mixture models (GMMs). Cross-adapting the baseline GMM-HMMs to the BS-HMM output yields a 6% relative gain which indicates that the two systems make different errors.

Original languageEnglish
Title of host publication2011 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2011 - Proceedings
Pages5316-5319
Number of pages4
DOIs
StatePublished - 18 Aug 2011
Event36th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2011 - Prague, Czech Republic
Duration: 22 May 201127 May 2011

Publication series

NameICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
ISSN (Print)1520-6149

Conference

Conference36th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2011
CountryCzech Republic
CityPrague
Period22/05/1127/05/11

Keywords

  • basis representation
  • Bayesian learning
  • discriminative training

Fingerprint Dive into the research topics of 'Discriminative training for Bayesian sensing hidden Markov models'. Together they form a unique fingerprint.

Cite this