The classification of electrocardiographic signal (ECG) perturbed by noise using the wavelet theory [articol]

Abstract

ECG signal is a non-stationary signal meaning that it changes its statistical proprieties over time. Therefore, the most powerful tool for analyzing this type of signals is the wavelet theory. The aim of this paper is to classify the ECG signals belonging to a given database in four classes: one class for ECGs without noise and three classes corresponding to ECGs perturbed by three types of noise. The Discrete Wavelet Transform and Daubechies wavelets were used to filter and analyze the four types of signals. The ECG data is taken from the standard MIT-BIH Arrhythmia database, while the signals of noise belong to MIT-BIH Noise Stress Test Database. Various tests were elaborated in this sense and the method used was the descriptive statistics.