基于分块扩展Infomax算法的胎儿心电分离

Block-extend Infomax algorithms in fetal electrocardiogram extraction

背景:胎儿心电是围产期对胎儿发育状况监测的重要参数,准确将胎儿心电从母体心电中分离出来是目前研究的重点。目的:采用分块扩展Infomax算法能够快速、有效的将胎儿心电分离出来。方法:利用分块扩展Infomax算法结合四阶统计去相关学习规则,使信号加权协方差阵的非对角元素最小化,改善算法收敛速度。结果与结论:扩展Infomax能实现超高斯和亚高斯信号的同时分离,文章提出的算法在收敛速度上要快于扩展Infomax的收敛速度。扩展Infomax算法能对胎儿心电实现有效的提取,由于分块扩展Infomax算法每次处理的数据量远少于扩展Infomax算法处理的数据量,且分块Infomax算法的迭代次数也远小于扩展Infomax的迭代次数,从而实现了胎儿心电的快速分离。

BACKGROUND： Fetal electrocardiogram is an important parameter in perinatal period for fetus growing. How to extract the fetal electrocardiogram from maternal electrocardiogram quickly and effectively has become an important research topic. OBJECTIVE： To use the block-extend Infomax algorithm for the separation of the fetal electrocardiogram quickly and effectively. METHODS： Block-extened Infomax and statistics into four ranks were combined to study the relate rules which make the non-cross element of weight covariance matrix minimum, and improve the astringency of arithmetic. RESULTS AND CONCLUSION： Extended infomax algorithm could achieve the separation of super-Gaussian and sub-Gaussian signal. The convergence rate of the algorithm presented in the paper was faster than that of the extended infomax algorithm. Extended Infomax algorithm could effectively extract the fetal electrocardiogram. The amount of the data treated by the block-extend Infomax algorithm was less than that treated by the extended infomax algorithm, and the iteration number of the block-extend Infomax algorithm was smaller than that of the extended infomax algorithm, so the fetal electrocardiogram could be separated quickly and effectively.