Comparative Study of Performance of Particle Swarm Optimization and Fast Independent Component Analysis method in Cocktail Party Problem

Abstract

     There are many methods used for solving the Blind Source Separation problem, such as Independent Component Analysis which became the most commonly used method. ICA methods depend on one of two properties: sample dependency or non-Gaussianity. In our study, the cocktail-party problem processed using ICA method.

In this work, we studied the performance of two techniques with the independent component analysis is standard FastICA, and PSO; and compare the results of each algorithm with others according to some evaluation metrics (objective such as SNR and SDR ) and (subjective such as signals plotting and playing). The implement of these algorithms was to be made with two source signals and three source signals. As in the evaluation process, the PSO gives more accurate results than FastICA.

Many input speech signals of 8 KHz sampling frequency, that achieve i.i.d. condition and well-condition were tested for different speeches for men and/or women, also music.