Application of Gaussian Mixture Models for Blind Separation of Independent Sources
In this paper, we consider the problem of blind separation of an instantaneous mixture of independent sources by exploiting their non-stationarity and/or non-Gaussianity. We show that non-stationarity and non-Gaussianity can be exploited by modeling the distribution of the sources using Gaussian Mixture Model (GMM). The Maximum Likelihood (ML) estimator is utilized in order to derive a new source separation technique. The method is based on estimation of the it sensors distribution parameters via the Expectation Maximization (EM) algorithm for GMM parameter estimation. The separation matrix is estimated by applying simultaneous joint diagonalization of the estimated GMM covariance matrices. The performance of the proposed method is evaluated and compared to existing blind source separation methods. The results show superior performance.
KeywordsGaussian Mixture Model Independent Component Analysis Independent Source Source Distribution Separation Matrix
Unable to display preview. Download preview PDF.
- 8.Bilmes, J.: A Gentle Tutorial on the EM Algorithm and its Application to Parameter Estimation for Gaussian Mixture and Hidden Markov Models, Technical Report, University of Berkely, ICSI-TR-97-021 (1997)Google Scholar