Fast training of Large Margin diagonal Gaussian mixture models for speaker identification

Reda Jourani 1 Khalid Daoudi 2 Régine André-Obrecht 1 Driss Aboutajdine 3
1 IRIT - Institut de recherche en informatique de Toulouse
2 GeoStat - Geometry and Statistics in acquisition data
Inria Bordeaux - Sud-Ouest
3 GSCM-LRIT - Laboratoire de Recherche en Informatique et Télécommunications [Rabat]
Abstract : Gaussian mixture models (GMM) have been widely and successfully used in speaker recognition during the last decades. They are generally trained using the generative criterion of maximum likelihood estimation. In an earlier work, we proposed an algorithm for discriminative training of GMM with diagonal covariances under a large margin criterion. In this paper, we present a new version of this algorithm which has the major advantage of being computationally highly efficient. The resulting algorithm is thus well suited to handle large scale databases. We carry out experiments on a speaker identification task using NIST-SRE'2006 data and compare our new algorithm to the baseline generative GMM using different GMM sizes. The results show that our system significantly outperforms the baseline GMM in all configurations, and with high computational efficiency.
Type de document :
Communication dans un congrès
International Conference on Speech Technology and Human-Computer Dialogue (SpeD), May 2011, Brasov, Romania. 2011
Liste complète des métadonnées

Littérature citée [12 références]  Voir  Masquer  Télécharger
https://hal.inria.fr/hal-00647213
Contributeur : Khalid Daoudi <>
Soumis le : jeudi 1 décembre 2011 - 16:33:53
Dernière modification le : jeudi 11 janvier 2018 - 06:21:34
Document(s) archivé(s) le : vendredi 2 mars 2012 - 02:36:58

Fichier

SpeD-2011.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-00647213, version 1

Collections

UNIV-TLSE2 | UNIV-TLSE3 | INRIA | SMS

Citation

Reda Jourani, Khalid Daoudi, Régine André-Obrecht, Driss Aboutajdine. Fast training of Large Margin diagonal Gaussian mixture models for speaker identification. International Conference on Speech Technology and Human-Computer Dialogue (SpeD), May 2011, Brasov, Romania. 2011. 〈hal-00647213〉

Exporter

BibTeX TEI DC DCterms EndNote

Partager

Métriques

Consultations de la notice

376

Téléchargements de fichiers

208

Contact


archive-ouverte@inria.fr