一种基于免疫遗传神经网络的盲信号分离方法

提出一种基于免疫遗传优化神经网络的盲信号分离算法。该算法用网络的第一层先对观测矢量作预处理，将其降维与白化，再用网络的第二层对信号进行分离：分离层的权矩阵设计成正交矩阵，并采用免疫遗传优化与独立分量分析相结合的算法，对网络分离层的权值进行训练，其中，取高阶统计量峭度的变形作为训练的代价函数。实验表明，该算法对于盲信号分离是有效的。

Blind source separation algorithm based on immune genetic neural network

A neuralnetwork based on immune genetic optimization forblind source separation is proposed. The first layer performs pretreatment, which finishes pre-whitening and dimensionality reduction. Thesecond layerperforms separation of sources, where the column vectors of the weight matrix are orthogonal. The weights of separate network are updated by immune genetic algorithm (IGA) and Independent Component Analysis (ICA), where cost function is based on the transformation of high order statistics of Kurtosis. The applicability of proposed method for blind source separation is demonstrated by simulation.