Classification of Branch Block Beats using Higher Order Spectral Analysis and Neural Networks

In this study, it is aimed to classify branch block beats. A total number of 6170 beats related to 3 types of classes are extracted from MIT/BIH arrhythmia database and their bispectrums are calculated using TOR method. The area defined by the frequency values where the value of the energy of bispectrum is 95% of the maximum value in both axes is calculated. This area information is used as a one dimensional feature vector to feed the neural network designed as a classifier. The overall performance of the system is calculated as 94.2%. This study shows that higher order spectral analysis is a promising tool for arrhythmia beat classification