Research Seminars

Batch and online underdetermined source separation using Laplacian Mixture Models

Dr. Nick Mitianoudis, Imperial College London

Wednesday 8 February 2006, **4.15pm**, Room 105
(Note unusual time)

Abstract

In this work, the authors explore the problem of sound source separation and identification from a two-sensor instantaneous mixture. The estimation of the mixing and the sources is performed using Laplacian Mixture Models (LMM). The proposed algorithm fits the model using batch processing of the observed data and performs separation using either a hard or a soft decision scheme. An extension of the algorithm to online source separation, where the samples are arriving in a real-time fashion, is also presented. The proposed scheme exhibits good performance as far as separation quality and convergence speed are concerned. The online version also demonstrates several promising source separation possibilities in the case of nonstationary mixing.