Algorithm of Nonlinear Blind Source Separation Based on Score Function Estimation
Lidong Zhu · Jisuanji fangzhen · 2012
In traditional methods of nonlinear source separation,the score function is chosen empirically.The performance of the present nonlinear source separation algorithm is degraded when the mixed signals contain super-Gaussian and sub-Gaussian signals and the nonlinear distortion is serious.In this paper,the proposed score function was derived from the Pearson model and can efficiently approximate the sub-Gaussian and super-Gaussian signal.The proposed algorithm overcome the defects that Pearson gets the same score functions through estimating the same kind of signals(e.g.sub-Gaussian signals),and improved the precision of the estimation of the score function.We performed simulations based on MATLAB and the experimental results demonstrate that the proposed method is effective and efficient.And the algorithm successfully estimates the score function and separates the nonlinear mixing signals.