Abstract
This paper describes a new method for generating the recombination weights in sub-band based speaker verification. The approach, which is based on the use of background speaker models, attempts to reduce the effect of any mismatch between the band-limited segments of the test utterance and the corresponding sections in the target speaker model. The discussion also includes an analysis of other possible methods for determining these weights. Moreover, a problem generally associated with the sub-band cepstral features is pointed out and a possible solution is presented.
Original language | English |
---|---|
Title of host publication | Proc. International Conference on Spoken Language Processing (ICSLP '98) |
Pages | 77-80 |
Volume | 2 |
Publication status | Published - 1998 |