An Improved FastICA Algorithm using a Sixth-order Newton's Method

Abstract

In this thesis, an improved fixed-point independent component analysis (FastICA) algorithm has been proposed by using the modified sixth-order Newton's Method. The FastICA algorithm is one of the most popular methods with better convergence speed to solve problems for ICA and blind source separation. In order to get the maximum of the negentropy and the minimizing mutual information between the independent components, an improved FastICA algorithm is proposed by using a sixth-order method for nonlinear equations, which is based on the modified Newton's Iteration Formula. Its good performance and remarkable convergence rate have been proved by the simulation results for the image separation.