Sixth European Conference on Speech Communication and Technology
The front-end used in many speaker recognition systems extracts, from the input speech signal, a set of coeficients based on a mel-cepstrum technique. This paper addresses the problem of efficiency of melcepstrum coeficients in a speaker recognition system and suggests a technique permitting an appropriate choice of these coeficients. It is shown, by the results obtained, that this technique can significantly increase the performance of a speaker recognition system.
Full Paper (PDF) Gnu-Zipped Postscript
Bibliographic reference. Vergin, Rivarol / O'Shaughnessy, Douglas / Dumouchel, Pierre (1999): "Toward parametric representation of speech for speaker recognition systems", In EUROSPEECH'99, 795-798.