基于二次特征函数高阶导数联合对角化的盲源分离算法

Second Characteristic Function High Order Derivative Joint Diagonalization Based on BSS Algorithm

针对线性混叠信号,提出一种瞬时混叠盲源分离的批处理算法.利用观测信号的二次特征函数高阶导数联合对角化得到盲分离准则和实验公式.采用非正交对角化方法,不需要白化,实现盲源分离.仿真实验结果表明此算法不仅对非高斯信源具有良好的分离性能,而且对概率密度为非对称分布信源,具有很好的性能.

For linear mixing signals, a temporal mixing Blind Source Separation batch algorithm is proposed. The principle and experiment formula of BSS is established based on the second characteristic function high order derivative joint diagonalization. Non-orthogonal diagonal method is used without the need for BSS. Simulation shows that the proposed algorithm has good separation performance for nongaussian sources, and have perfect outcome for asymmetry sources.