Conference paperMATĚJKA Pavel, GLEMBEK Ondřej, CASTALDO Fabio, ALAM Jahangir, PLCHOT Oldřich, KENNY Patrick, BURGET Lukáš and ČERNOCKÝ Jan. Fullcovariance UBM and Heavytailed PLDA in IVector 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. 48284831. ISBN 9781457705373.  Publication language:  english 

Original title:  Fullcovariance UBM and Heavytailed PLDA in IVector Speaker Verification 

Title (cs):  Plné kovarianční matice a Heavytailed PLDA v ověřování mluvčího pomocí ivektorů 

Pages:  48284831 

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 

Place:  Praha, CZ 

Year:  2011 

ISBN:  9781457705373 

Publisher:  IEEE Signal Processing Society 

URL:  http://www.fit.vutbr.cz/research/groups/speech/publi/2011/matejka_icassp2011_4828.pdf [PDF] 

URL:  http://www.fit.vutbr.cz/research/groups/speech/publi/2011/matejka_icassp2011_poster4828.pdf [PDF] 

Keywords 

GMM, speaker recognition, PLDA, heavytailed PLDA, fullcovariance UBM, ivectors 
Annotation 

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

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

@INPROCEEDINGS{
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 = {Fullcovariance UBM and Heavytailed PLDA in IVector
Speaker Verification},
pages = {48284831},
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 = {9781457705373},
language = {english},
url = {http://www.fit.vutbr.cz/research/view_pub.php?id=9657}
} 
