Source Separation: Principles, Current Advances and Applications

This paper is a survey on the source separation problem, and on methods used for solving this problem. In a blind context, i.e. without information about the sources but their mutual independence, methods are based on Independent Component Analysis (ICA). On the contrary, using priors on sources, one can developed semi-blind approaches which are very efficient and often much more simpler. Current advances aim to take into account various priors like positivity or sparsity. This paper will finish with sketches of source separation applications which will give practical examples.