改进的基于峭度的盲信号提取算法  

Improved Algorithm Based on Kurtosis for Blind Signal Extraction

作　　者：张守成[1] 张玉洁[2] 祁锐[2] 

机构地区：[1]华北科技学院基础部,河北廊坊065201 [2]中国地质大学数理学院,湖北武汉430074

出　　处：《计算机仿真》2014年第2期255-258,共4页Computer Simulation

基　　金：国家自然科学基金项目(61071188);中央高校基本科研业务费(JCB2013B11;JCB2013B10)

摘　　要：在盲信号优化提取问题的研究中,为解决基于峭度的参考独立分量分析算法中阈值参数难以确定的问题,通过对算法的理论分析,以输出信号峭度的绝对值和接近性量度函数的乘积作为目标优化函数,并在梯度算法的基础上根据经典的Kuhn-Tucker条件提出一种改进的定点算法,有效避免了人为选择阈值参数和步长,使算法收敛更快速稳定。针对合成数据和实际的心电图数据的计算机仿真,表明了所提改进算法的有效性和快速稳定性。To solve the problem that it is difficult to determine the threshold parameter of the algorithm based on kurtosis for independent component analysis with reference（ICA -R）. An improved fixed -point algorithm based on gradient algorithm with classical Kuhn - Tucker condition was proposed by taking the product of the absolute value of kurtosis and the closeness measure function as optimization function, and it can converge faster and better by avoiding to determine the threshold parameter and step size. Computer simulations with synthetic signals and real electrocardi- ogram（ECG） demonstrate its effectiveness and good separation.

关 键 词：峭度 盲源分离 盲信号提取 阈值参数 参考独立分量分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]