ICA Blind Signal Separation Based on a New Probability Density Function
Zhang Juan-jua · Gongcheng shuxue xuebao · 2014
This paper is concerned with the blind source separation(BSS) problem of superGaussian and sub-Gaussian mixed signal by using the maximum likelihood method, which is based on independent component analysis(ICA) method. In this paper, we construct a new type of probability density function(PDF) which is different from the already existing PDF used to separate mixed signals in the previously published papers. Applying the new constructed PDF to estimate probability density of super-Gaussian and sub-Gaussian signals(assuming the source signals are independent of each other), it is not necessary to change the parameter values artificially, and the separation work may be performed adaptively. Numerical experiments verify the feasibility of the newly constructed PDF, and the convergence time and the separation effect are improved compared with the original algorithm.