Intelligent ECG Signal Noise Removal with Moving Median Filter Using Neural Network

Abstract

In this paper, the electrocardiogram (ECG) signal is susceptible to noise and artifact and it is essential to remove noise using neural network. The noise in order to support first 3600 of noisy heart signal is collected from MIT-BIH data base. In this paper, the use of average median filter and artificial neural network is analyzed. The available filters for power-line interference need a reference channel or regard that the frequency is fixed 50/60 Hz. In the literature of the last twenty-five years, several solutions for noise removal on electrocardiogram (ECG) signal can be found. The spectrum of the ECG signal is extracted from the two databases - arrhythmia and supraventricular. Baseline wander is removed using the average median filter. The results show that the intelligent artificial neural network system successfully de-noised ECG signal. This study mainly focuses on cutoff frequency calculating best performance MSE.

Keywords: Finite impulse response (FIR), low-pass filter, artificial neural network, cutoff frequency, average median filter