Blind Source Separation: Application to biomedical signals

Blind Source Separation (BSS) is a prominent problem in signal processing. In the past few decades, it was applied to many fields, in which separation of compound signals, simultaneously observed by different sensors, is of interest. The problem can be considered as built-up of three physical elements: sources (also called transmitters), sensors (also called receivers) and communication channels which reflect the properties of the physical medium propagating the signals form the sources to the sensors. The signals detected by the sensors are commonly referred to as observations and are assumed to be algebraic combinations of the unknown sources signals. BSS approach assumes limited a priori information on the communication channels (linearity, memory properties...) and tries to reconstruct the source signals out of the detected signals only. Analysis of the communication channels is important mainly for selection of a proper processing technique. Namely, communication channels weight and/or filter the signals coming from the sources and, together with them, determine the temporal and spectral characteristics of the detected mixtures.