一种基于组合核函数的非线性盲源分离方法研究

Nonlinear Blind Source Separation Based on Compound Kernel Function

核函数方法由于其有效性和简洁性在非线性盲源分离问题的探索中得到了应用,但其单一核的映射不能很好解决完全非线性问题。针对这一不足,提出了一种采用组合核函数分离非线性混合信号的方法。通过引入变化尺度因子将不同核函数纳入一个整体成为组合核函数,利用分离信号的互信息作为目标函数来反馈调节该组合核函数的尺度因子,以此寻找到对不同非线性的最佳映射。仿真结果证实了该算法的有效性,且在解决完全非线性问题时,组合核函数比单一核函数具有更好的性能。

Kernel Function method has been applied in exploring nonlinear blind source separation for its validity and simplicity. However, single kernel function cannot well solve the absolute nonlinear problem. So a novel separation algorithm based on compound kernel function was proposed. In the algorithm, different kernel function was combined as a whole through variable measure factors, mutual information of separated signal was used as target function to reflect and regulate the measure factors of the compound kernel function, so as to reach an optimum different nonlinear mapping. Simulation results verify the effectiveness of the proposed method; and in solving absolute nonlinear problem, compound kernel function unfolds a better performance than single kernel function.