基于微分进化算法的盲源分离  

Blind source separation based on differential evolution algorithm

作　　者：柯维[1] 张永祥[1] 吕博[1] 

机构地区：[1]海军工程大学动力工程学院,武汉430033

出　　处：《海军工程大学学报》2012年第5期12-17,共6页Journal of Naval University of Engineering

基　　金：国家自然科学基金资助项目(50979109)

摘　　要：将微分进化算法的优点引入到盲源分离中,提出了基于微分进化的盲源分离算法。该算法以混合信号的峰度为代价函数,采用独立分量分析的方法对瞬时混合的信号进行盲分离。盲源分离中常用的自然梯度算法是一种局部寻优算法且收敛速度较慢,而微分进化算法是一种全局寻优算法且具有并行性、易实现等优点。分别用无噪仿真信号和有噪仿真信号对提出的算法进行仿真实验,比较了基于微分进化算法的盲源分离、基于粒子群优化算法的盲源分离和基于自然梯度算法的盲源分离的分离结果。结果表明:基于微分进化的盲分离算法收敛速度快,分离效果也比较好。According to the advantage of differential evolution algorithm, a new algorithm is obtained from blind source separation based on differential evolution （DE） optimization. This algorithm uses the kurtosis of the signals as the cost function of blind source separation. The independent component analysis method is used for the blind source separation of the unknown source signals from a set of col- lection signals without prior information. The algorithm of natural gradient, a commonly used blind source separation method, is a local optimization method which is low in convergence but DE algo- rithm is a global optimization method which has such advantages as parallelism, and easy operation. This paper compares the proposed algorithm with the blind source separation method based on the particle swarm algorithm （PSO） and that based on nature gradient algorithm （NG） by simulating noiseless and noisy signals respectively. The results show the algorithm is fast in convergence and sta- ble in performance.

关 键 词：盲源分离 微分进化算法 代价函数 全局优化 

分 类 号：TN911.7[电子电信—通信与信息系统]