Subspace independent component analysis using vector kurtosis

摘要

This discussion presents a new perspective of subspace independent component analysis (ICA). The notion of a function of cumulants (kurtosis) is generalized to vector kurtosis. This vector kurtosis is utilized in the subspace ICA algorithm to estimate subspace independent components. One of the main advantages of the presented approach is its computational simplicity. The experiments have shown promising results in estimating subspace independent components. (c) 2006 Pattern Recognition Society. Published by Elsevier Ltd. All rights reserved.