Linearization of R-R Peak in Abdominal ECG Signals for Fetal ECG Separation Using Adaptive Filter

Abstract

This paper proposes a fetal electrocardiogram (ECG) extraction algorithm from the abdominal maternal ECG signals. The method uses the signal collected from the thorax as the reference signal and the signal collected from the abdomen as the primary input for the adaptive filter. To reduce the noise in the fetal ECG signal, this method initially detects the maternal R peaks of the thorax ECG signals using the Pan–Tompkins algorithm. Using the estimated R-R peaks, the abdominal and thorax signal is framed from which the slopes of the abdominal and thorax signal are estimated. The slope of the abdominal signal between the R-R peaks is made equal to the slope of the R-R thorax signal frame. The slope of the abdominal is adjusted such that the first R peak remains unchanged, while the second R peak highly changes. This also creates a linear change in the samples between the two peaks. The framed thorax signal is used as the reference input, and the framed linearized signal of the abdomen is used as the primary input of the adaptive filter. The evaluation of the proposed fetal ECG extraction was done using the metrics such as correlation and quality of signal to noise ratio (qSNR) for various values of fetal to maternal signal to noise ratio (fmSNR). The evaluation results using the fetal ECG synthetic dataset show that the method outperforms the traditional fetal ECG extraction algorithms.