Publication Details
Discriminatively Trained i-vector Extractor for Speaker Verification
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Brümmer Niko
Plchot Oldřich, Ing., Ph.D. (DCGM)
Matějka Pavel, Ing., Ph.D. (DCGM)
speaker verification, i-vectors, PLDA, discriminative training
We have proposed a technique for discriminative training of the i-vector extractor parameters using cross-entropy as the error function. We have applied the technique both to the original i-vector extractor and to its simplified version. In both cases, the discriminative training was effective, giving higher relative improvement in the simplified case.
We propose a strategy for discriminative training of the ivector extractor in speaker recognition. The original i-vector extractor training was based on the maximum-likelihood generative modeling, where the EM algorithm was used. In our approach, the i-vector extractor parameters are numerically optimized to minimize the discriminative cross-entropy error function. Two versions of the i-vector extraction are studied-the original approach as defined for Joint Factor Analysis, and the simplified version, where orthogonalization of the i-vector extractor matrix is performed.
@inproceedings{BUT76447,
author="Ondřej {Glembek} and Lukáš {Burget} and Niko {Brümmer} and Oldřich {Plchot} and Pavel {Matějka}",
title="Discriminatively Trained i-vector Extractor for Speaker Verification",
booktitle="Proceedings of Interspeech 2011",
year="2011",
journal="Proceedings of Interspeech",
volume="2011",
number="8",
pages="137--140",
publisher="International Speech Communication Association",
address="Florence",
isbn="978-1-61839-270-1",
issn="1990-9772",
url="http://www.fit.vutbr.cz/research/groups/speech/publi/2011/glembek_interspeech2011_386.pdf"
}