基于独立分量分析的盲源分离方法研究

Blind Source Separation Based on Independent Component Analysis

盲源分离是指在没有任何先验知识的前提下,从观测到的多个混合信号中提取或分离出未知源信号的过程。本文主要探讨了基于独立分量分析的盲源分离自然梯度算法及活动函数的选取,并利用该算法实现了5路混合信号和3路自然语音信号的分离,最后在Matlab2008下进行了仿真验证。结果表明该算法能够有效实现语音信号的分离。

Blind Source Separation means extracting or separating the statistically independent source signals from their mixture Under the premise of the unknown signal source. An important algo- rithm- natural gradient algorithm is given, which is based on Independent Component Analysis. This passage use different active functions to simulate the blind source separation of speech signal and communication signal. Both simulation of speech signal and communication signal have come out to prove the validity of the natural gradient algorithm.