Publication Details

Fast Discriminative Speaker Verification in the I-Vector Space

CUMANI Sandro, BRÜMMER Niko, BURGET Lukáš and LAFACE Pietro. Fast Discriminative Speaker Verification in the I-Vector Space. In: Proceedings of the 2011 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2011. Praha: IEEE Signal Processing Society, 2011, pp. 4852-4855. ISBN 978-1-4577-0537-3.
Czech title
Rychlé diskriminativní ověřování mluvčího v i-vektorovém prostoru
Type
conference paper
Language
english
Authors
Cumani Sandro (POLITO)
Brümmer Niko (Agnitio)
Burget Lukáš, doc. Ing., Ph.D. (DCGM FIT BUT)
Laface Pietro, prof. (POLITO)
URL
Keywords

Discriminative Training,Two-covariance Kernel, Support Vector Machines, i-vectors

Abstract

A fast discriminative training approach for speaker verification based on i-vectors has been presented. On NIST telephone evaluation data, the resulting models perform better, without the need for normalization techniques, than the generative ones, even compared with heavy-tailed models.

Annotation

This work presents a new approach to discriminative speaker verification. Rather than estimating speaker models, or a model that discriminates between a speaker class and the class of all the other speakers, we directly solve the problem of classifying pairs of utterances as belonging to the same speaker or not. The paper illustrates the development of a suitable Support Vector Machine kernel from a state-of-the-art generative formulation, and proposes an efficient approach to train discriminative models. The results of the experiments performed on the tel-tel extended core condition of the NIST 2010 Speaker Recognition Evaluation are competitive or better, in terms of normalized Decision Cost Function and Equal Error Rate, compared to the more expensive generative models.

Published
2011
Pages
4852-4855
Proceedings
Proceedings of the 2011 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2011
Conference
International Conference on Acoustics, Speech and Signal Processing 2011, Praha, CZ
ISBN
978-1-4577-0537-3
Publisher
IEEE Signal Processing Society
Place
Praha, CZ
BibTeX
@INPROCEEDINGS{FITPUB9654,
   author = "Sandro Cumani and Niko Br{\"{u}}mmer and Luk\'{a}\v{s} Burget and Pietro Laface",
   title = "Fast Discriminative Speaker Verification in the I-Vector Space",
   pages = "4852--4855",
   booktitle = "Proceedings of the 2011 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2011",
   year = 2011,
   location = "Praha, CZ",
   publisher = "IEEE Signal Processing Society",
   ISBN = "978-1-4577-0537-3",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/9654"
}
Back to top