Estimating the number of sources in a noisy convolutive mixture using BIC 
Rasmus Kongsgaard Olsson, Lars Kai Hansen

Abstract  The number of source signals in a noisy convolutive mixture is determined based on the exact loglikelihoods of the candidate models. In (Olsson and Hansen, 2004), a novel probabilistic blind source separator was introduced that is based solely on the timevarying secondorder statistics of the sources. The algorithm, known as ‘KaBSS’, employs a Gaussian linear model for the mixture, i.e. AR models for the sources, linear mixing filters and a white Gaussian noise model. Using an EM algorithm, which invokes the Kalman smoother in the Estep, all model parameters are estimated and the exact posterior probability of the sources conditioned on the observations is obtained. The loglikelihood of the parameters is computed exactly in the process, which allows for model evidence comparison assisted by the BIC approximation. This is used to determine the activity pattern of two speakers in a convolutive mixture of speech signals. 
Keywords  Blind source separation, independent component analysis, speech processing, Bayes information criterion 
Type  Conference paper [With referee] 
Conference  5th International Conference on Independent Component Analysis and Blind Signal Separation 
Editors  C. G. Puntonet, A. Prieto 
Year  2004 Month September pp. 618625 
Publisher  Springer Berlin 
Address  Richard Petersens Plads, Building 321, DK2800 Kgs. Lyngby 
ISBN / ISSN  3540230564 
Electronic version(s)  [pdf] 
BibTeX data  [bibtex] 
IMM Group(s)  Intelligent Signal Processing 