Date:1992-08-24
Author:Pierre Comon
Pages:28
Summary:The paper introduces the concept of Independent Component Analysis (ICA), which aims to find a linear transformation that minimizes the statistical dependence between the components of a random vector. The authors utilize the expansion of mutual information as a function of cumulants of increasing orders to define suitable search criteria. An efficient algorithm is proposed to compute the ICA of a data matrix within polynomial time. ICA is seen as an extension of Principal Component Analysis (PCA), which can only enforce independence up to the second order and defines orthogonal directions. Potential applications of ICA include data analysis and compression, Bayesian detection, source localization, and blind identification and deconvolution. The paper also discusses related works, provides a detailed mathematical framework, and presents simulation results to validate the proposed methods.