A Tutorial on Blind Source Separation using Independent Component Analysis and Related Methods

Blind Source Separation (BSS) is needed to recoverseveral source signals from several mixture-signals. The mixture-signals are linear combinations of the sources signals. Such asetup is encountered for example when it is desired to recoverthe speech ofNspeakers, speaking simultaneously fromNmicrophone signals placed at various positions in the sameroom with the speakers. Conversely the Independent ComponentAnalysis (ICA) is a term covering a methods that aim to representa set of observations from several random variables in terms ofa linear combination of observations from several other randomvariables that are independent from one another. The solutionsto the BSS problem usually imply some weak assumptions onthe source signals, like for example independence. Thus, the ICArepresents a possible solution to the BSS problem.