基于统计估计的盲信号分离算法

最大熵法（ＭａｘｉｍｕｍＥｎｔｒｏｐｙ，ＭＥ）和最小互信息量法（ＭｉｎｉｍｕｍＭｕｔｕａｌＩｎｆｏｒｍａｔｉｏｎ，ＭＭＩ）是两种目前最常用的盲信号分离算法．在分析ＭＥ与ＭＭＩ算法的基础上，提出了一种利用反馈结构的输出信号概率密度函数（ｐｄｆ）估计的增强ＭＥ算法．与传统ＭＥ算法相比较，新算法无需给出传统ＭＥ算法中神经元非线性函数的具体表达形式，而是直接利用输出信号ｐｄｆ估计来推导算法的迭代核，进行算法自适应．分析了应用几种不同ｐｄｆ估计方法的新算法迭代公式．通过计算机模拟表明，新算法比传统ＭＥ算法对于解决卷积混合输入的盲信号分离问题时，具有更好的算法性能．

Blind Signal Separation Based on Statistical Estimation

Maximum Entropy (ME) and Minimum Mutual Information (MMI) algorithms are mainly two kinds of the Blind Signal Separation (BSS) algorithms. Based on the recursive architecture and the relationship between ME and MMI algorithms, an extended ME(EME) algorithm was proposed by using a special probability density function (pdf) estimation of outputs instead of searching the appropriate neurons' nonlinear function in the traditional ME algorithm. Several kinds of pdf estimation were studied. Based on the simulation, it can be concluded that the proposed algorithm has better performance than the traditional ME algorithm in convoluted mixture BSS problems.