一种新的基于极大似然估计的盲分离算法

盲源分离（Blind Source Separation，BSS）或独立成分分析（Independent Component Analysis，ICA）是一种矩阵处理或数据分析技术，主要目的就是在仅在源信号相互独立的假设下从混合的观测信号中恢复出源信号。由于它在生物医学信号分析、语音识别、无线通信等领域应用的不断拓广，它已成为一个热门的研究领域。本文对于如何分离混合信号模型的问题，从观测信号与分离信号的概率密度函数（Probability Densitv Function PDF）之间的关系推导出了一种新的基于极大似然估计的盲分离算法，通过选择一个带参数的非线性函数近似超高斯与亚高斯的PDF，以此来分离源信号。并通过模拟实验验证了此算法的有效性。

NEW MAXIMUM LIKELIHOOD ESTIMATION ALGORITHM FOR BLIND SEPARATION

Blind Source Separation (BSS) or Independent Component Analysis (ICA) are emerging techniques of array processing and data analysis, aiming at recovering unobserved signals from observed mixtures, exploiting only the assumption of mutual independence between the signals. It has become an increasing important research field due to its rapid growing applications in various areas, such as biomedical signal analysis, speech recognition, tele communication and so on. Based on the question about separating the arbitrary source signals, according to the Probability Density Function (PDF) of the observed signals and the separated signals, this paper chooses a nonlinear function with a parameter to be approximate to the PDF of super-Gaussian and sub-Gaussian sources, and then we can separate the sig nals. The effectiveness and performance of the algorithm are verified by the computer simu lations.