Non Linear Blind Source Separation Using Different Optimization Techniques

Abstract

The Independent Component Analysis technique has been used in Blind Source separation of non linear mixtures. The project involves the blind source separation of a non linear mixture of signals based on their mutual independence as the evaluation criteria. The linear mixer is modeled by the Fast ICA algorithm while the Non linear mixer is modeled by an odd polynomial function whose parameters are updated by four separate optimization techniques which are Particle Swarm Optimization, Real coded Genetic Algorithm, Binary Genetic Algorithm and Bacterial Foraging Optimization. The separated mixture outputs of each case was studied and the mean square error in each case was compared giving an idea of the effectiveness of each optimization technique.