基于累积量展开的神经网络盲源分离方法

A Cumulant-Expansion Based Neural Networks Method for Blind Separation of Sources

本文基于源信号统计独立性假设,推导了表示信号独立性的准则函数,将它用于独立分量分析(ICA)神经网络得出了一种新的信号盲源分离的方法,计算机仿真实验结果表明该方法可较好地进行瞬时混迭信号的分离,并具有良好的稳定性能。

Based on the statistically independent assumption of source signals,this paper formu lates an objective function which is a measure of statistically independent. We then propose a neu ral network algorithm for blind separation of source signals. The computer simulations show that the proposed algorithm can well separate sources from their instantaneous mixtures.