A new EM algorithm for underdetermined convolutive blind source separation

Zaher El Chami (1) , Dinh-Tuan Pham (2) , Christine Serviere (3) , Alexandre Guérin (1)

1 Orange Labs [Lannion]
2 SMS - Statistique et Modélisation Stochatisque
3 GIPSA-SA-IGA - GIPSA - Signal et Automatique pour le Diagnostic et la Surveillance

This paper presents a new statistical method for separating more than two sound sources from a two-channel record­ing. It is based on a probabilistic model of the Interchannel Level/Phase Difference and the model para­meters are estimated using the maximum likelihood criterion and an Expectation-Maximization algorithm. The source separation task is achieved by soft time-frequency masking of the observation. These masks are derived from the es­timated source position model. Algorithm performance is evaluated on the real and synthetic convolutive mixtures data of the first audio source evaluation campaign as well as the Signal Separation Evaluation campaign (SiSEC). Promising results are obtained when comparing to the other methods presented in these two campaigns.

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

Zaher El Chami, Dinh-Tuan Pham, Christine Serviere, Alexandre Guérin. A new EM algorithm for underdetermined convolutive blind source separation. EUSIPCO 2009 - 17th European Signal Processing Conference, Aug 2009, Glasgow, United Kingdom. pp.1457-1461. ⟨hal-00435933⟩