An Effective Algorithm for Blind Separation of Arbitrary Source Signals
Zhang Hong · Dianzi xuebao · 2001
A new blind source separation algorithm called DBBSS (Density Based Blind Source Separation) is proposed.Instead of using nonlinear functions,the DBBSS algorithm use nonparametric kernel density estimation to directly estimate the score functions of the signals.The key advantage of the proposed method over many existing blind source separation algorithms is its ability to separate hybrid mixtures that contain both super Gaussian and sub Gaussian sources.The DBBSS algorithm is also very simple in implementation.Simulations show good performance of the proposed algorithm.