Beyond the Narrowband Approximation: Wideband Convex Methods for Under-Determined Reverberant Audio Source Separation - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Article Dans Une Revue IEEE Transactions on Audio, Speech and Language Processing Année : 2010

Beyond the Narrowband Approximation: Wideband Convex Methods for Under-Determined Reverberant Audio Source Separation

Résumé

We consider the problem of extracting the source signals from an under-determined convolutive mixture assuming known mixing filters. State-of-the-art methods operate in the time-frequency domain and rely on narrowband approximation of the convolutive mixing process by complex-valued multiplication in each frequency bin. The source signals are then estimated by minimizing either a mixture fitting cost or a l1 source sparsity cost, under possible constraints on the number of active sources. In this article, we define a wideband l2 mixture fitting cost circumventing the above approximation and investigate the use of a l12 mixed-norm cost promoting disjointness of the source timefrequency representations. We design a family of convex functionals combining these costs and derive suitable optimization algorithms. Experiments indicate that the proposed wideband methods result in a signal-to-distortion ratio improvement of 2 to 5 dB compared to the state-of-the-art on reverberant speech mixtures.
Fichier principal
Vignette du fichier
kvg_taslp.pdf (292.38 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-00435897 , version 1 (25-11-2009)
hal-00435897 , version 2 (24-10-2010)
hal-00435897 , version 3 (20-11-2010)

Identifiants

Citer

Matthieu Kowalski, Emmanuel Vincent, Rémi Gribonval. Beyond the Narrowband Approximation: Wideband Convex Methods for Under-Determined Reverberant Audio Source Separation. IEEE Transactions on Audio, Speech and Language Processing, 2010, 18 (7), pp.1818 - 1829. ⟨10.1109/TASL.2010.2050089⟩. ⟨hal-00435897v2⟩
1042 Consultations
1088 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More