Automatic Regularization of Cross-entropy Cost for Speaker Recognition Fusion

Ville Hautamäki 1 Kong Lee 2 David Van Leeuwen 3 Rahim Saeidi 4 Anthony Larcher 5 Tomi Kinnunen 1 Taufiq Hasan Seyed Sadjadi Gang Liu Hynek Boril John Hansen Benoît Fauve 6
1 University of Eastern Finland
2 Institute for Infocomm Research - I²R [Singapore]
3 Centre for Language Studies
4 Speech and Image Processing Unit, Department of Computer Science and Statistics
5 LIUM - Laboratoire d'Informatique de l'Université du Mans
6 Swansea University
Abstract : In this paper we study automatic regularization techniques for the fusion of automatic speaker recognition systems. Parameter regularization could dramatically reduce the fusion training time. In addition, there will not be any need for splitting the development set into different folds for cross-validation. We utilize majorization-minimization approach to automatic ridge regression learning and design a similar way to learn LASSO reg-ularization parameter automatically. By experiments we show improvement in using automatic regularization.
Type de document :
Communication dans un congrès
Annual Conference of the International Speech Communication Association (Interspeech), Aug 2013, Lyon, France
Liste complète des métadonnées
https://hal.archives-ouvertes.fr/hal-01927590
Contributeur : Anthony Larcher <>
Soumis le : lundi 19 novembre 2018 - 23:59:13
Dernière modification le : jeudi 22 novembre 2018 - 01:10:00

Fichier

I4UFusion_camera.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01927590, version 1

Collections

UNIV-LEMANS | LIUM

Citation

Ville Hautamäki, Kong Lee, David Van Leeuwen, Rahim Saeidi, Anthony Larcher, et al.. Automatic Regularization of Cross-entropy Cost for Speaker Recognition Fusion. Annual Conference of the International Speech Communication Association (Interspeech), Aug 2013, Lyon, France. 〈hal-01927590〉

Exporter

BibTeX TEI DC DCterms EndNote

Partager

Métriques

Consultations de la notice

11

Téléchargements de fichiers

5

Contact

support.ccsd.cnrs.fr
hal.support@ccsd.cnrs.fr
Logo CCSD