基于互信息最小的非线性混合胎儿心电信号提取方法

FECG extraction from nonlinear mixture based on minimization of mutual information

胎儿心电信号提取具有重要的临床意义。目前大多数胎儿心电信号提取方法是基于线性瞬时混合模型,使用多路信号实现.本文基于更符合实际情况的非线性混合模型,使用两路心电信号,提出一种新颖的从母体腹部混合信号中提取胎儿心电信号方法。首先,在假定母体和胎儿两个心电信号相互独立的基础上,基于互信息最小原则,提出一种新的目标函数;然后,利用多层感知机拟合母体心电信号传导至腹部所经历的非线性变换,从而得到母体腹部混合信号中母体心电成分的最优估计;最后,从母体腹部混合信号中减去母体心电信号的最优估计,实现胎儿心电信号的成功提取。实验结果验证了该方法的正确性和有效性。

The extraction of fetal electrocardiogram（FECG） signal is of vital importance from the clinical point of view.Most of proposed methods for extracting FECG are based on the linear instantaneous mixing model and need to use multi-channel signals.In this paper,a new method which uses only two-channel ECG signals for extraction FECG from the more realistic nonlinear mixture is proposed.First,suppose that FECG and maternal electrocardiogram（MECG） are mutually independent and a new objective function is presented based on the minimization of the mutual information.Then,the distorted thoracic signal,which is mapped from the thorax to the abdomen,is approximated by multilayer perceptions network.Finally,the FECG is extracted by subtracting the mapped thoracic signal from the abdominal signal.The experiments on real-world ECG data confirm the correctness and the effectiveness of the proposed method.