Maximum Likelihood Blind Source Separation: A Context-Sensitive Generalization of ICA


In the square linear blind source separation problem, one must nd a linear unmixing operator which can detangle the result xi(t) of mixing n unknown independent sources si(t) through an unknown n n mixing matrix A(t) of causal linear lters: xi = P j aij sj . We cast the problem as one of maximum likelihood density estimation, and in that framework introduce an algorithm that searches for independent components using both temporal and spatial cues. We call the resulting algorithm \Contextual ICA," after the (Bell and Sejnowski 1995) Infomax algorithm, which we show to be a special case of cICA. Because cICA can make use of the temporal structure of its input, it is able separate in a number of situations where standard methods cannot, including sources with low kurtosis, colored Gaussian sources, and sources which have Gaussian histograms. 1 The Blind Source Separation Problem Consider a set of n indepent sources s1(t); : : : ; sn(t). We are given n linearly distorted sensor reading which combine these sources, xi = P j aijsj , where aij is a lter between source j and sensor i, as shown in gure 1a. This can be expressed as xi(t) = X

Extracted Key Phrases

3 Figures and Tables


Citations per Year

214 Citations

Semantic Scholar estimates that this publication has 214 citations based on the available data.

See our FAQ for additional information.

Cite this paper

@inproceedings{Pearlmutter1996MaximumLB, title={Maximum Likelihood Blind Source Separation: A Context-Sensitive Generalization of ICA}, author={Barak A. Pearlmutter and Lucas C. Parra}, booktitle={NIPS}, year={1996} }