A Quantitative and Qualitative Research on Fetal ECG extraction using Wavelet based Adaptive Filtering

Abstract

Fetal electrocardiogram (ECG) extraction contributes a crucial part in monitoring the cardiac electrical activity of fetal heart. In this context, we propose a hybrid technique for fetal ECG extraction using recursive least square (RLS) based adaptive filtering algorithm and stationary wavelet transform (SWT). The accuracy of fetal beat detection is enhanced using improved spatially selective noise filtration (ISSNF) algorithm. The algorithm is designed for extracting the fetal ECG signal deriving out of the mother’s composite abdominal ECG signal and doctors can diagnose the health conditions of fetal heart during pregnancy period. The abdominal signal is decomposed into multi-resolution components using SWT. The scale of wavelet decomposition depends on different values of noise level. Fetal ECG extraction is performed using RLS algorithm. For denoising, ISSNF algorithm is used in wavelet domain. The proposed algorithm is experimentally evaluated using both synthetic dataset and clinical dataset. The performances of the proposed method are analyzed both qualitatively and quantitatively by observation, signal to noise ratio (SNR) calculation and R-peak detection. The results indicate superior performance when compared with the traditional methods of adaptive filtering. Hence the proposed system is well suited to fetal ECG extraction for generating clean fetal ECG signal with high SNRs and low distortions.