Publication Details

The subspace Gaussian mixture model-A structured model for speech recognition

POVEY, D.; BURGET, L.; AGARWAL, M.; AKYAZI, P.; GHOSHAL, A.; GLEMBEK, O.; GOEL, N.; KARAFIÁT, M.; RASTROW, A.; ROSE, R.; SCHWARZ, P.; THOMAS, S. The subspace Gaussian mixture model-A structured model for speech recognition. COMPUTER SPEECH AND LANGUAGE, 2011, vol. 25, no. 2, p. 404-439. ISSN: 0885-2308.
Czech title
Sub-space gaussovský model - strukturovaný model pro rozpoznávání řeči
Type
journal article
Language
English
Authors
Povey Daniel
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Agarwal Mohit
Akyazi Pinar
Ghoshal Arnab
Glembek Ondřej, Ing., Ph.D.
Goel Nagendra
Karafiát Martin, Ing., Ph.D. (DCGM)
Rastrow Ariya
Rose Richard
Schwarz Petr, Ing., Ph.D. (DCGM)
Thomas Samuel
and others
URL
Keywords

Speech recognition; Gaussian Mixture Model; Subspace Gaussian Mixture Model

Abstract

Speech recognition based on the Hidden Markov Model-Gaussian Mixture Model (HMM-GMM) framework generally involves training a completely separate GMM in each HMM state.We introduce a model in which the HMM states share a common structure but the means and mixture weights are allowed to vary in a subspace of the full parameter space, controlled by a global mapping from a vector space to the space of GMM parameters.

Annotation

We describe a new approach to speech recognition, in which all Hidden Markov Model (HMM) states share the same Gaussian Mixture Model (GMM) structure with the same number of Gaussians in each state. The model is defined by vectors associated with each state with a dimension of, say, 50, together with a global mapping from this vector space to the space of parameters of the GMM. This model appears to give better results than a conventional model, and the extra structure offers many new opportunities for modeling innovations while maintaining compatibility with most standard techniques.

Published
2011
Pages
404–439
Journal
COMPUTER SPEECH AND LANGUAGE, vol. 25, no. 2, ISSN 0885-2308
Book
Computer Speech & Language, Volume 25, Issue 2, April 2011
Publisher
Elsevier Science
UT WoS
000284670200017
EID Scopus
2-s2.0-78049502526
BibTeX 
@article{BUT76383,
  author="Daniel {Povey} and Lukáš {Burget} and Mohit {Agarwal} and Pinar {Akyazi} and Arnab {Ghoshal} and Ondřej {Glembek} and Nagendra {Goel} and Martin {Karafiát} and Ariya {Rastrow} and Richard {Rose} and Petr {Schwarz} and Samuel {Thomas}",
  title="The subspace Gaussian mixture model-A structured model for speech recognition",
  journal="COMPUTER SPEECH AND LANGUAGE",
  year="2011",
  volume="25",
  number="2",
  pages="404--439",
  issn="0885-2308",
  url="https://www.fit.vut.cz/research/publication/9670/"
}
Back to top