Elimination of power line interference from ECG signals using recurrent neural networks

Abstract

This study proposed a method for the elimination of 50-Hz power line interference (PLI) in electrocardiogram (ECG) signals, using recurrent neural networks (RNN). The method is to preliminarily train an RNN-based model to obtain the signal pattern of PLI. The ECG signals are then filtered by subtracting the PLI signals extracted by the trained model. The results suggest that the proposed method could significantly relieve the distortion of QRS complex while filtering the ECG signals efficiently. The method was evaluated by comparing with a traditional IIR notch filter and an adaptive filtering method.