Non-parametric ICA Algorithm for Hybrid Sources Based on GKNN Estimation

摘要

Novel independent component analysis(ICA) algorithm based on non-parametric density estimation-generalized k-nearest neighbor(GKNN) estimation is proposed using a linear ICA neural network. The proposed GKNN density estimation is directly evaluated from the original data samples, so it solves the important problem in ICA: how to choose nonlinear functions as the probability density function(PDF) estimation of the sources. Moreover the GKNN-ICA algorithm is able to separate the hybrid mixtures of source signals using only a flexible model and it is completely blind to the sources. It provides the way to wider applications of ICA methods to real world signal processing. Simulations confirm the effectiveness of the proposed algorithm.