多种概率分布源的盲源分离快速算法

Blind source separation for any source signal

基于目标函数迭代优化的方法在解决线性混合情况下,源信号存在多种概率密度分布的盲源分离问题时,需要对非线性函数以及迭代步长进行正确的选择,算法比较复杂;针对此问题,提出一种基于高阶统计的快速分离算法,该算法可以有效地避免上述问题.实验结果表明,该算法能够快速有效地分离出不同概率密度分布的混合信号.

The blind source separation based on optimizing target function could be used to separate sources of any probability density distribution in the linear mixing model. However, these algorithms are complex, and appropriate choice needs to be made on nonlinear function and steps. Herein, a new simpler and faster algorithm was introduced based on higher order statistics, which could overcome the problem of nonlinear function and step choice. Good performance in simulations supports the practicability and validity of the proposed algorithm.