A unifying model for blind separation of independent sources

Many algorithms have been proposed for the blind separation of statistically independent sources. Most of the algorithms are based on one of the following properties: nongaussianity of the sources, their different autocorrelations, or their smoothly changing nonstationary variances. Each of the methods is able to separate sources if the respective assumptions are met. Here we propose a simple unifying model that is able to separate independent sources if any one of these three conditions is met. The model is a simple autoregressive model whose estimation can be performed by maximum likelihood estimation. We also propose a simple yet accurate approximation of the likelihood that gives a simple algorithm.