Independent Component Analysis: A Review, with Emphasis on Commonly used Algorithms and Contrast Function

Rasmikanta Pati, Arun K. Pujari, Padmavati Gahan, Vikas Kumar


Independent Component Analysis ( ICA) is a powerful tool for blind sources to be separated from their determined or signals obtains at that moment in the fields of signal processing, machine learning , data mining, finance, bio-medical, communications, artificial intelligence etc., ICA focuses primarily on finding an Objective Function (Contrast Function) and an appropriate optimization method to solve the problem. Different methods of ICA are differs in how one model the contrast functions between themselves. ICA focuses mainly on finding components that are as independent as possible and as non-Gaussian as possible of an observed unexplained non-Gussian Signal Mixture. ICA is an extremely important subject of great interest in numerous technological and scientific applications. In this article we review the few different contrast function in addition to the much earlier survey of Aapo Hyvarinen and widely used existing ICA algorithms in different scenarios for source separation. This article presents basic ideas on ICA, ICA algorithms and contrast functions.


Independent component analysis, unsupervised learning, particle swarm optimization, higher order statistics, blind source separation

Full Text: PDF