Skip to main content

Recurrent neural networks for speech recognition

  • Applications
  • Conference paper
  • First Online:

Part of the book series: Lecture Notes in Computer Science ((LNCS,volume 540))

Abstract

In this paper we present some results from a net-like structure for Hidden Markov Models, applied to speech recognition. Net topology is a Recurrent Neural Network in which each temporary step is identified as a layer. Backpropagation techniques are used to train the RNN-HMM. Two types of training estimations are used: Maximum Likelihood and Competitive Training. Maximum Likelihood estimation algorithm using backpropagation provides the same updating equations as Baum-Welch algorithm used in HMM. Competitive Training is based on the probability of correct labelling the sequences from the Maximum Likelihood measures. Our results have shown that the best procedure is to train first with Maximum Likelihood estimation and then with Competitive Training reestimation.

This work has been supported by CICYT under project TIC 88-0774

This is a preview of subscription content, log in via an institution.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

VIII. References

  1. Bourlard, H; Wellekens, C.J. "Speech dynamics and Recurrent Neural Networks". Proc. ICASSP-89 pp. 33–36.

    Google Scholar 

  2. Demichelis, P; et als. "On the use of Neural Networks for Speaker Independent Isolated Word Recognition". Proc. ICASSP-89 pp. 314–317.

    Google Scholar 

  3. Sakoe, H.: et als. "Speaker Independent Word Recognition Using Dynamic Programming Neural Networks" in Proc. ICASSP-89 pp. 29–32.

    Google Scholar 

  4. Hwang, J.N.; Vlontzos, J.; Kung, S. "A Systolic Neural Network Architecture for Hidden Markov Models", in IEEE Trans. Acoust., Speech, Signal Processing, vol. 37, pp. 1967–1979, Dec. 1989.

    Google Scholar 

  5. Kung, S.; Hwang, J. "A Unifying Algorithm/Architecture for Artificial Neural Networks" in Proc. ICASSP-89, pp. 2505–2508.

    Google Scholar 

  6. Bahl,L.R; et als. "Maximum Mutual Information Estimation of Hidden Markov Model Parameters for Speech Recognition" in Proc. ICASSP-86, pp. 49–52. Tokio.

    Google Scholar 

  7. Ephrain, Y.; Rabiner, L. "On the Relations Between Modelling Approaches for Speech Recognition", in IEEE Trans. Acoust., Speech, Signal Processing, vol. 36, pp. 372–379. March, 1990.

    Google Scholar 

  8. Bridle, J.S. "Alpha-nets: A Recurrent Neural Network Architecture with a Hidden Markov Model Interpretation" in Speech Communication, vol. 9, 1990.

    Google Scholar 

  9. Rabiner, L. "A Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition". in Proc. of the IEEE, vol. 77, n. 2, Feb. 1989.

    Google Scholar 

  10. Sadaoki, Furui. "Speaker-Independent Isolated Word Recognition Using Dynamic Features of Speech Spectrum" in "IEEE Trans. Acoust., Speech, Signal Processing, vol. ASSP-34, pp. 52–59, Feb. 1986.

    Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Editor information

Alberto Prieto

Rights and permissions

Reprints and permissions

Copyright information

© 1991 Springer-Verlag Berlin Heidelberg

About this paper

Cite this paper

Díaz Verdejo, J.E., Herreros, A.P., Segura Luna, J.C., Benitez Ortúzar, M.C., Ayuso, A.R. (1991). Recurrent neural networks for speech recognition. In: Prieto, A. (eds) Artificial Neural Networks. IWANN 1991. Lecture Notes in Computer Science, vol 540. Springer, Berlin, Heidelberg. https://doi.org/10.1007/BFb0035915

Download citation

  • DOI: https://doi.org/10.1007/BFb0035915

  • Published:

  • Publisher Name: Springer, Berlin, Heidelberg

  • Print ISBN: 978-3-540-54537-8

  • Online ISBN: 978-3-540-38460-1

  • eBook Packages: Springer Book Archive

Publish with us

Policies and ethics