Advanced search
1 file | 363.24 KB

Improving large vocabulary continuous speech recognition by combining GMM-based and reservoir-based acoustic modeling

Fabian Triefenbach (UGent) , Kris Demuynck (UGent) and Jean-Pierre Martens (UGent)
Author
Organization
Abstract
In earlier work we have shown that good phoneme recognition is possible with a so-called reservoir, a special type of recurrent neural network. In this paper, different architectures based on Reservoir Computing (RC) for large vocabulary continuous speech recognition are investigated. Besides experiments with HMM hybrids, it is shown that a RC-HMM tandem can achieve the same recognition accuracy as a classical HMM, which is a promising result for such a fairly new paradigm. It is also demonstrated that a state-level combination of the scores of the tandem and the baseline HMM leads to a significant improvement over the baseline. A word error rate reduction of the order of 20\% relative is possible.
Keywords
reservoir computing, tandem acoustic modeling, continuous speech recognition

Downloads

  • ftriefen SLT2012.pdf
    • full text
    • |
    • open access
    • |
    • PDF
    • |
    • 363.24 KB

Citation

Please use this url to cite or link to this publication:

Chicago
Triefenbach, Fabian, Kris Demuynck, and Jean-Pierre Martens. 2012. “Improving Large Vocabulary Continuous Speech Recognition by Combining GMM-based and Reservoir-based Acoustic Modeling.” In IEEE Workshop on Spoken Language Technology, Proceedings, 107–112. Piscataway, NJ, USA: IEEE.
APA
Triefenbach, F., Demuynck, K., & Martens, J.-P. (2012). Improving large vocabulary continuous speech recognition by combining GMM-based and reservoir-based acoustic modeling. IEEE Workshop on Spoken Language Technology, Proceedings (pp. 107–112). Presented at the IEEE Workshop on Spoken Language Technology, Piscataway, NJ, USA: IEEE.
Vancouver
1.
Triefenbach F, Demuynck K, Martens J-P. Improving large vocabulary continuous speech recognition by combining GMM-based and reservoir-based acoustic modeling. IEEE Workshop on Spoken Language Technology, Proceedings. Piscataway, NJ, USA: IEEE; 2012. p. 107–12.
MLA
Triefenbach, Fabian, Kris Demuynck, and Jean-Pierre Martens. “Improving Large Vocabulary Continuous Speech Recognition by Combining GMM-based and Reservoir-based Acoustic Modeling.” IEEE Workshop on Spoken Language Technology, Proceedings. Piscataway, NJ, USA: IEEE, 2012. 107–112. Print.
@inproceedings{3000500,
  abstract     = {In earlier work we have shown that good phoneme recognition is possible with a so-called reservoir, a special type of recurrent neural network. In this paper, different architectures based on Reservoir Computing (RC) for large vocabulary continuous speech recognition are investigated. Besides experiments with HMM hybrids, it is shown that a RC-HMM tandem can achieve the same recognition accuracy as a classical HMM, which is a promising result for such a fairly new paradigm. It is also demonstrated that a state-level combination of the scores of the tandem and the baseline HMM leads to a significant improvement over the baseline. A word error rate reduction of the order of 20{\textbackslash}\% relative is possible.},
  author       = {Triefenbach, Fabian and Demuynck, Kris and Martens, Jean-Pierre},
  booktitle    = {IEEE Workshop on Spoken Language Technology, Proceedings},
  isbn         = {9781467351256},
  keyword      = {reservoir computing,tandem acoustic modeling,continuous speech recognition},
  language     = {eng},
  location     = {Miami, FL, USA},
  pages        = {107--112},
  publisher    = {IEEE},
  title        = {Improving large vocabulary continuous speech recognition by combining GMM-based and reservoir-based acoustic modeling},
  year         = {2012},
}

Web of Science
Times cited: