Sixth European Conference on Speech Communication and Technology
(EUROSPEECH'99)

Budapest, Hungary
September 5-9, 1999

Speaker Identification Using Subband HMMS

K. Yoshida, K. Takagi, K. Ozeki

The University of Electro-Communications, Chofu, Tokyo, Japan

This paper is concerned with optimum band splitting and optimum recombination weights in subband HMM-based speaker identification. In the first experiment, the full frequency band (8kHz) was split into two subbands, and speaker identification rate was measured for various splitting frequencies and recombination weights. It was found that subbands 0-2kHz and 2-8kHz with equal recombination weights gave the best identification rate, outperforming a baseline method without band-splitting. In the second experiment, the full-band was split into three subbands with various splitting frequencies. Splitting into 0-2kHz, 2-6kHz, and 6-8kHz gave the best result, slightly outperforming the two-subband case. Finally, four-subband experiment was conducted, the result of which suggests that the speaker information and the phonemic information are complementary to a considerable degree in the spectral domain.


Full Paper (PDF)   Gnu-Zipped Postscript

Bibliographic reference.  Yoshida, K. / Takagi, K. / Ozeki, K. (1999): "Speaker identification using subband HMMS", In EUROSPEECH'99, 1019-1022.