Classification of ECG Signals Using Wigner-Ville Distribution

Abstract

ECG signals by the Wigner-Ville Distribution (WVD) in order to classify related pathologies. The major problem of the WVD is the interferences (cross-terms). The Smoothed Pseudo-Wigner-Ville Distribution (SPWVD) makes it possible to reduce these interferences by a suitable choice of two windows H and G and their corresponding sizes Lg and Lh. We propose here a classification method based on the SPWVD and the nearest representative decision rule, the latter requires the choice of a distance and the definition of a representative for each training class. Two beat classes were considered: normal beats (NOR) and premature ventricular beats (PVC). By using the MIT-BIH data-base, the method was tested on a set composed of more than 80.000 beats. The best classification rate that we obtain is 90,66%, this one was carried out by using only the first most discriminant coefficient.