Publication Details

Full-covariance UBM and Heavy-tailed PLDA in I-Vector Speaker Verification

MATĚJKA Pavel, GLEMBEK Ondřej, CASTALDO Fabio, ALAM Jahangir, PLCHOT Oldřich, KENNY Patrick, BURGET Lukáš and ČERNOCKÝ Jan. Full-covariance UBM and Heavy-tailed PLDA in I-Vector Speaker Verification. In: Proceedings of the 2011 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2011. Praha: IEEE Signal Processing Society, 2011, pp. 4828-4831. ISBN 978-1-4577-0537-3.
Czech title
Plné kovarianční matice a Heavy-tailed PLDA v ověřování mluvčího pomocí i-vektorů
Type
conference paper
Language
english
Authors
Matějka Pavel, Ing., Ph.D. (DCGM FIT BUT)
Glembek Ondřej, Ing., Ph.D. (DCGM FIT BUT)
Castaldo Fabio (POLITO)
Alam Jahangir (CRIM)
Plchot Oldřich, Ing., Ph.D. (DCGM FIT BUT)
Kenny Patrick (CRIM)
Burget Lukáš, doc. Ing., Ph.D. (DCGM FIT BUT)
Černocký Jan, prof. Dr. Ing. (DCGM FIT BUT)
URL
Keywords

GMM, speaker recognition, PLDA, heavytailed PLDA, full-covariance UBM, i-vectors

Abstract

The work we presented aims at the best performance of the single stand alone system. We have presented full-covariance UBM and i-vector extraction with different kind of modeling. Our analysis shows that for the best performance it is necessary to have fullcovariance i-vector without any approximation.

Annotation

In this paper, we describe recent progress in i-vector based speaker verification. The use of universal background models (UBM) with full-covariance matrices is suggested and thoroughly experimentally tested. The i-vectors are scored using a simple cosine distance and advanced techniques such as Probabilistic Linear Discriminant Analysis (PLDA) and heavy-tailed variant of PLDA (PLDA-HT). Finally, we investigate into dimensionality reduction of i-vectors before entering the PLDA-HT modeling. The results are very competitive: on NIST 2010 SRE task, the results of a single full-covariance LDA-PLDA-HT system approach those of complex fused system.

Published
2011
Pages
4828-4831
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
DOI
10.1109/ICASSP.2011.5947436
EID Scopus
2-s2.0-80051621424
BibTeX 
@INPROCEEDINGS{FITPUB9657,
   author = "Pavel Mat\v{e}jka and Ond\v{r}ej Glembek and Fabio Castaldo and Jahangir Alam and Old\v{r}ich Plchot and Patrick Kenny and Luk\'{a}\v{s} Burget and Jan \v{C}ernock\'{y}",
   title = "Full-covariance UBM and Heavy-tailed PLDA in I-Vector Speaker Verification",
   pages = "4828--4831",
   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",
   doi = "10.1109/ICASSP.2011.5947436",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/9657"
}
Back to top