改进的ICA胎儿心电信号提取法

Fetal Electrocardiogram Extraction Combining ICA and Highpass Filter

独立分量分析(ICA)模型中,有关独立性的假设对模型的可识别性至关重要,然而在实际应用中信号源之间往往有某种程度的联系。孕妇腹部信号是由胎儿心电信号(FECG)和母亲心电信号(MECG)线性混合而成的,但是FECG与MECG之间存在一定的相关性,不是完全统计独立的,针对信号源之间弱独立性的这一特点提出了一种新的非入侵式胎儿心电信号提取方法,试验表明该方法比直接用传统的ICA方法提取的胎儿心电信号噪声小、分离效果好。

In basic framework of ICA, the assumption of independence is crucial for the identifiability of the model, however, in practiceal application the source signals often have some relationship. The signals on the abdomens of pregnant women are made up of FEGG and MECG. They are linear mixture. The FEGG and MECG are correlated, not statistic independent. In this article, according to this opinion, a new method of noninvasive fetal electrocardiogram （FECG） extraction is introduced. Through experiments, FECG extracted with the algorithm has less yawp and the division effect is better than directly using traditional ICA method.