基于局部主分量分析的胎儿心电信号提取算法

Fetal ECG extraction based on local principal component analysis

为了从孕妇腹部一个导联混合信号中提取微弱的胎儿心电信号,提出了一种基于局部主分量分析的非线性特征信号提取算法。该算法首先利用重构相空间理论从观测时间序列中重构系统的动力学特征,然后在每个相点的局部范围内利用主分量分析的算法提取母亲和胎儿心电信号。由于提取的胎儿心电信号中还含有噪声,再一次利用局部主分量分析算法就能分别提取纯净的胎儿心电信号和噪声。两组实测信号的仿真结果表明,算法具有较好的提取效果和鲁棒性。

In order to extract fetal electrocardiogram （FECG） from one single cutaneous potential recording of a pregnant woman, we propose a nonlinear local principal component analysis （LPCA） algorithm to extract FECG. Firstly, based on the theory of phase space reconstruction, a scalar sequence of measurements is converted into state vectors. Secondly, for every state vector in local area, PCA is adopted to extract mother ECG （MECG） and FECG. Because of the existence of noise, it is better to adopt PCA for the second time to extract more pure FECG. Simulation results show the effectiveness and robustness of the algorithm.