5th European Conference on Speech Communication and Technology

Rhodes, Greece
September 22-25, 1997

A Multiresolutionally Oriented Approach for Determination of Cepstral Features In Speech Recognition

S. Dobrisek, F. Mihelic, N. Pavesic

University of Ljubljana, Faculty of Electrical Engineering, Ljubljana, Slovenia

This paper presents an effort to provide a more efficient speech signal representation, which aims to be incorporated into an automatic speech recognition system. Modified cepstral coefficients, derived from a multiresolution auditory spectrum are proposed. The multiresolution spectrum was obtained using sliding single point discrete Fourier transformations. It is shown that the obtained spectrum values are similar to the results of a nonuniform filtering operation. The presented cepstral features are evaluated by introducing them into a simple phone recognition system.

Full Paper

Bibliographic reference.  Dobrisek, S. / Mihelic, F. / Pavesic, N. (1997): "A multiresolutionally oriented approach for determination of cepstral features in speech recognition", In EUROSPEECH-1997, 1367-1370.