Interval type-2 fuzzy system for ECG arrhythmic classification

Abstract

This chapter presents an Interval Type-2 fuzzy classifier and its application to ECG arrhythmic classification problem. The uncertainties associated with the membership functions are encapsulated by the footprint of uncertainty (FOU) and it is totally characterized by the upper membership function (UMF) and lower membership function (LMF). To enable designed membership functions (MFs) reflect the data, we proposed three types of FOU design strategies according to the dispersion of the data. The first and second designs comprise of Gaussian MFs with uncertain standard deviations and means respectively whereas the third design is the combination of both. The FOU is then further optimized through Genetic Algorithm. The proposed Type-2 fuzzy classifier has been applied to ECG arrhythmic classification problem to discriminate three types of ECG signals, namely the normal sinus rhythm (NSR), ventricular fibrillation (VF), and ventricular tachycardia (VT). The performance of the classifier is tested on MIT-BIH Arrhythmia database. The average period and pulse width of ECG data are extracted as the inputs to the classifier. Different sources of noises have been included to model the uncertainties associated with the vagueness in MFs and the unpredictability of the data. The results show that the proposed strategies to design the FOU are essential to achieve a high performance fuzzy rule-based classifier in face of the uncertainties. © 2009 Springer-Verlag Berlin Heidelberg.