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.
出处
《计算机仿真》
CSCD
北大核心
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)
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
Kurtosis
Blind source separation （ BSS ）
Blind signal extraction （ BSE ）
Threshold parameter
Inde- pendent component analysis with reference
