Audio source separation using hierarchical phase-invariant models

Emmanuel Vincent 1
1 METISS - Speech and sound data modeling and processing
IRISA - Institut de Recherche en Informatique et Systèmes Aléatoires, Inria Rennes – Bretagne Atlantique
Abstract : Audio source separation consists of analyzing a given audio recording so as to estimate the signal produced by each sound source for listening or information retrieval purposes. In the last five years, algorithms based on hierarchical phase-invariant models such as single or multichannel hidden Markov models (HMMs) or nonnegative matrix factorization (NMF) have become popular. In this paper, we provide an overview of these models and discuss their advantages compared to established algorithms such as nongaussianity-based frequency-domain independent component analysis (FDICA) and sparse component analysis (SCA) for the separation of complex mixtures involving many sources or reverberation.We argue how hierarchical phase-invariant modeling could form the basis of future modular source separation systems.
Type de document :
Autre publication
2009 ISCA Tutorial and Research Workshop on Non-linear Speech Processing (NOLISP). 2009, pp.12--16
Liste complète des métadonnées

Littérature citée [19 références]  Voir  Masquer  Télécharger

https://hal.inria.fr/inria-00544170
Contributeur : Emmanuel Vincent <>
Soumis le : mardi 7 décembre 2010 - 14:17:52
Dernière modification le : mercredi 16 mai 2018 - 11:23:03
Document(s) archivé(s) le : mardi 8 mars 2011 - 04:22:21

Fichier

vincent_NOLISP09.pdf
Fichiers éditeurs autorisés sur une archive ouverte

Identifiants

  • HAL Id : inria-00544170, version 1

Citation

Emmanuel Vincent. Audio source separation using hierarchical phase-invariant models. 2009 ISCA Tutorial and Research Workshop on Non-linear Speech Processing (NOLISP). 2009, pp.12--16. 〈inria-00544170〉

Partager

Métriques

Consultations de la notice

424

Téléchargements de fichiers

313