An Algorithm of Nonlinear Blind Source Separation Based on Score Function Estimation for Spaceborne Radar Signals
Nonlinear blind source separation (NBSS) has attracted increasing attention in recent years because of its application and importance. In this paper, we propose a method of separating spaceborne radar signals based on blind source separation. The BSS algorithm is optimized by minimizing the output mutual information based on specific nonlinear mixing model known as post-nonlinear-linear (PNLL). In this paper, we present a new method for the estimation of score function derived from the Pearson model. The proposed algorithm can efficiently approximate the sub-Gaussian and super-Gaussian signal, overcomes the defects that Pearson gets the same score function for estimation the same kind signals (e.g. sub-Gaussian signals) and improves the precision of the estimation of the score function. The separating system incorporates one group of multilayer perceptrons with linear outputs and two linear matrixes. The multilayer perceptrons network is used for approximation of nonlinear function. We perform simulations based on our method with MATLAB. The simulation results demonstrate that the proposed method is effective and efficient.
Blind Source Separation Nonlinear Mixture Score Function Spaceborne Radar Signals Multilayer Perceptrons
Jing An Li-dong Zhu
National Key Laboratory of Science and Technology on Communications, UESTC, Chengdu, China
国际会议
海口
英文
146-150
2011-07-15（万方平台首次上网日期，不代表论文的发表时间）