常用的盲分离算法有二阶统计量方法、高阶累积量方法、信息最大化（
Infomax
）以及独
立成分分析（
ICA
）等。这些方法取得最佳性能的条件总是与源信号的概率密度函数假设有关，
一旦假设的概率密度与实际信号的密度函数相差甚远，分离性能将大大降低。本文提出采用
核函数密度估计的方法进行任意信号源的盲分离，并通过典型算例与几种盲分离算法进行了
性能比较，验证了方法的可行性。(The commonly used blind separation algorithms include two order statistics, higher order cumulants, and maximization of information
Infomax) and independent component analysis (ICA). The conditions for obtaining the best performance of these methods are always related to the hypothesis of the probability density function of the source signal. The separation performance will be greatly reduced when the assumed probability density is very different from the density function of the actual signal. In this paper, a method of kernel function density estimation for blind separation of arbitrary signal sources is proposed. The performance of several blind separation algorithms is compared with some typical examples, and the feasibility of the method is verified.)