5 - Analyse en composantes indépendantes et identification aveugle

Independent Component Analysis (ICA) consists of searching a linear transformation that provides us with an exhaustive (or compressed) summary of the data in the form of statistically independent components . This concept may be seen as an extension of Principal Component Analysis, which only imposes independence up to the second order . ICA is first defined, and then algorithms for its calculation are proposed . It is shown how this new tool can be used in various applications, such as separation of signais, Bayesian detection, or blind deconvolution .