Blind Source Separation Based on random mixing coefficient

Blind source separation (BSS) is the extraction (estimation) of the original signals from mixed signals, without knowledge (or very little information) on the source signals or the mixture process. Many applications are used blind source separation like cocktail parties and conferences. Therefore, source separation is considered a vital and important subject in signal processing. The separation problem is based on the mixing criteria, while the gap is for considering the sources is fixed. This paper introduces a random propagation mixing coefficient of an indirect path (moving sources) for Quasi of cocktail parties and press conference. The research also presents new evaluation criteria for the separation process and separation algorithms based on measuring the similarity of source and estimated signals. Where this evaluation criterion is based on signals features similarity. Three separation algorithm SVD, PCA, and ICA are tested with different noise amount with many source signals. The advancement of ICA as a good separation algorithm is clear with gaining a separation ration of 96.1%.