A novel evolutionary approach for blind source separation based on stone's method

In recent years, much research emerged to modify Stone's BSS method for solving a blind source separation problem; Stone's method used to recover original signals from the mixture. In this work, new direction has been opened to use an intelligent soft computing technique (Fast Genetic Algorithm) with the temporal predictability of signals based on Stone's BSS method. Proposed algorithm has been compared with wellknown BSS algorithms (JADE, FICA, and Stone's BSS) over super-Gaussian, sub-Gaussian, and Gaussian signals with linear mixture combination. Then eight voices have mixed randomly; and the proposed approach has successfully recovered the voices with high efficiency. Interpretation based on the responses of two different linear scalar filters to the same set of signals, which indicate to Short-term and Long-term linear predictors with tuned Half-life values (h_L, h_S) genetically is a powerful new technique for solving BSS problem. In addition to the benefits of the Stone's method, the proposed algorithm overcomes the local minima problem by successfully jump out of the potential local minimum. Usually recovery methods depend on the difference between signals and mixture proprieties; generally, there are three famous properties for any signal: (1) Gaussian probability density function based on the central limit theorem (2) Degree of statistical independence (3) Temporal predictability. So (1&2) proprieties, have previously been used as a base for separation but in this work only 3rd property has been used. In order to check the effectiveness of the proposed algorithm two performance indexes are used: Interference to signal ratio (ISR), and Integral square error (ISE).