源信号数目大于观察信号数目情况下的盲源分离

独立分量分析(ICA)作为一种有效的盲源分离技术(BSS）是信号处理领域的热点。传统的独立分量分析都要求观察信号数目大于或者等于源信号数目，然而对于脑电图(EEG)等的一些信号处理中存在的源信号数目大于观察信号数目的情况，传统的独立分量分析算法不能有效分离。该文针对源信号数目大于观察信号数目的情况，在传统的独立分量分析技术的基础上，给出了一个新的学习算法，并将新算法与传统的独立分量算法进行了比较。实验仿真结果证明该算法在给定2个混合信号的情况下能够较好地分离3个未知语音信号源，成功实现了源信号数目大于观察信号数目情况下的盲源分离。

Blind Source Separation of More Sources Than Mixtures

The Independent Component Analysis (ICA) as a method used in blind source separation is a hotspot in signal processing.The standard ICA algorithm require that there must be more mixtures than sources.While in some signal processing of EEG and other sinals,there are more sources than mixtures,so the standard ICA algorithm fail to separate the mixtures successfully.Based on analysis of existing ICA algorithm,this paper presents a new algorithm for the blind source separation of more sources than mixtures and also compares it with the standard ICA algorithm.At last,simulation experiments prove that three speech signals can be separated with good fidelity in only two mixtures of the three signals,thus achieving the blind source separation of more sources than mixtures.