28622 articles – 22134 references  [version française]

hal-00482287, version 1

Blind Spectral-GMM Estimation for Underdetermined Instantaneous Audio Source Separation

Simon Arberet 1, Alexey Ozerov () 2, Rémi Gribonval () a, 1, Frédéric Bimbot () 1

International Conference on Independent Component Analysis and Blind Source Separation (ICA) (2009) pp. 751 - 758

Abstract: The underdetermined blind audio source separation problem is often addressed in the time-frequency domain by assuming that each time-frequency point is an independently distributed random variable. Other approaches which are not blind assume a more structured model, like the Spectral Gaussian Mixture Models (Spectral-GMMs), thus exploiting statistical diversity of audio sources in the separation process. However, in this last approach, Spectral-GMMs are supposed to be learned from some training signals. In this paper, we propose a new approach for learning Spectral-GMMs of the sources without the need of using training signals. The proposed blind method significantly outperforms state-of-the-art approaches on stereophonic instantaneous music mixtures.

  • a –  INRIA
  • 1:  METISS (INRIA - IRISA)
  • CNRS : UMR6074 – INRIA – Institut National des Sciences Appliquées (INSA) - Rennes – Université de Rennes 1
  • 2:  Laboratoire Traitement et Communication de l'Information [Paris] (LTCI)
  • Télécom ParisTech – CNRS : UMR5141
  • Domain : Engineering Sciences/Signal and Image processing
    Computer Science/Signal and Image Processing
  • Keywords : blind audio source separation – Spectral Gaussian Mixture Models – training signals
 
  • hal-00482287, version 1
  • http://hal.archives-ouvertes.fr/hal-00482287
  • oai:hal.archives-ouvertes.fr:hal-00482287
  • From: Alexey Ozerov <>
  • Submitted on: Monday, 10 May 2010 10:50:27
  • Updated on: Friday, 7 January 2011 14:12:28