A New Algorithm of Blind Source Separation Based on ICA

Blind source separation has been one of the hottest areas in the signal processing fields, and it has application in the telecommunication system, speech enhancement, remote sensing and medical imaging. To improve the fast fixed-point algorithm based on independent component analysis (ICA) with only one activated function, we propose a new algorithm which includes three activated functions. In this paper, the nonlinear functions are hyperbolic cosine function, Beta distribution function and Pearson system function. Results from experiments show that it will not only maintain the characteristics of the original algorithm, but also can separate some signals which can´t be separated by the original algorithm.