摘要
利用基于随机变量概率密度函数的非参数密度估计的核密度估计法对评价函数进行直接估计,改进了盲分离算法的性能.理论推导和试验都证实了这种基于核密度估计的非参数密度估计盲分离算法能实现包含超高斯和亚高斯信号的杂系混合信号的盲分离,为盲分离问题在实际问题中的应用奠定了一定的基础.
Kernel density estimation (KDS) is a nonparametric density estimation based on random variables probability density functions, and it is not limited in any assumption about the density function model, and it is evaluated from the original data without the consideration of the above statistical features. Based on it, the score function was estimated by KDS, the performance of algorithms for BSS was improved. The theoretical derivation and experiments show that this algorithm succeeds in separating the hybrid mixing signals including the super-Gaussian and sub-Gaussian signals, and it paves the way to wider applications of BSS methods to real world signal processing.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2004年第2期203-206,共4页
Journal of Shanghai Jiaotong University
