Blind source separation by independent component analysis applied to electroencephalographic signals

Abstract

Independent Component Analysis (ICA) is a statistical based method, which goal is to find a linear transformation to apply to an observed multidimensional random vector such that its components become as statistically independent from each other as possible. Usually the Electroencephalographic (EEG) signal is hard to interpret and analyse since it is corrupted by some artifacts which originates the rejection of contaminated segments and perhaps in an unacceptable loss of data. The ICA filters trained on data collected during EEG sessions can identify statistically independent source channels which could then be further processed by using event-related potential (ERP), event-related spectral perturbation (ERSP) or other signal processing techniques. This paper describes, as a preliminary work, the application of ICA to EEG recordings of the human brain activity, showing its applicability.