Advertisement

On the use of Expert Systems for Speaker Verification

  • D. A. Gaganelis
  • E. D. Frangoulis
Chapter
Part of the Microprocessor-Based and Intelligent Systems Engineering book series (ISCA, volume 9)

Abstract

In this paper we report on the use of Expert System shells for automatic speaker verification over the public telephone network. The operation of the automatic speaker verification system, (ASV), is based on LPC normalized and differenced cepstrum coefficients. An Expert System has been introduced to assist in the construction of the reference templates. This Expert System is based on rules that relate to outlier detection through K-Means clustering among the training tokens for each verification word. The Expert System has additional rules referring to the combination of independent statistical classifiers for the classification of unlabelled utterances. The classifiers considered in this paper are the standard Gaussian classifier, the Quadratic classifier and the Fourier-Bessel classifier.

The effect of the introduction of the Expert System into the Verification process was evaluated in terms of the average verification error rates using three verification words and a population of sixty speakers.

Keywords

Expert System Speaker Verification Training Token Reference Template Speaker Verification System 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    D. A. Gaganelis, E. Frangoulis, “A novel approach to speaker verification”, ICASSP 1991, Toronto, CANADA.Google Scholar
  2. 2.
    S. Furui, “Cepstral analysis technique for automatic speaker verification”, IEEE Trans. ASSP-29, April 1981.Google Scholar
  3. 3.
    J. B. Attili, M. Savic, J. P. Campbell, Jr., “A TMS32020-based real time, text independent, automatic speaker verification system”, ICASSP 1988, New York, USA.Google Scholar
  4. 4.
    H. Gish, “Robust discrimination in automatic speaker identification”, ICASSP 1990, Albuquerque, New Mexico, USA.Google Scholar
  5. 5.
    J. S. Bird and D. A. George, “The use of the Fourier-Bessel series in calculating error probabilities for digital communication systems”, IEEE Trans. Commun., vol. Com-29, September 1981.Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht 1991

Authors and Affiliations

  • D. A. Gaganelis
    • 1
  • E. D. Frangoulis
    • 1
  1. 1.Speech & Language DivisionLogica Cambridge Ltd.CambridgeUK

Personalised recommendations