Extraction of Fetal ECG Signals Based on PCA and Improved FastICA Algorithm

Abstract

In order to solve the problem that the fast independent component analysis (FastICA) algorithm is sensitive to the initial value and the separated signal contains interference signals, this paper proposes a method combining principal component analysis (PCA) and improved FastICA algorithm, and applies it to extract fetal Electrocardiogram (FECG). Firstly, a number of major sources signals are extracted from the parent abdominal mixed signal using PCA. Secondly, the randomly generated initial values are processed by the steepest descent method, and then the source signals are separated by the double convergence factor FastICA algorithm to drive them into the Newtonian convergence region, and finally, the channel where the FECG is located is determined according to the sample entropy and processed by the wavelet thresholding method. The simulation results show that the algorithm not only alleviates the problem of initial value sensitivity, but also improves the convergence speed, extracts clear FECG signals, and almost does not lose the separation accuracy.