What is Independent Component Analysis?

This article describes a relatively new research topic called independent component analysis (ICA), which is becoming very popular in the signal processing literature and amongst those working in machine learning and data mining. The primary focus of ICA is to resolve the classical problem of blind source separation (BSS), in which an unknown mixture of nonGaussian signals is decomposed into its independent component signals. The classical example of BSS is the so-called cocktail-party problem, where the mixture consists of simultaneous speech signals recorded by a number of microphones. Important applications include biomedical signal processing (usually brain wave activity in the form of EEG and MEG tracings), audio signal separation (mixed speech and music signals), telecommunications (a confusion of signals transmitted by multiple users of mobile phones), financial time series (portfolios of stocks), and data mining (text document analysis). The ICA methodology has much in common with that of projection pursuit (PP).