Simultaneous QRS detection and feature extraction using simple matched filter basis functions

A simple, easily implemented method is presented for extracting from an electrocardiogram (ECG) features describing QRS morphology. Tests using the American Heart Association (AHA) database indicate that these features are successful at enabling premature ventricular contractions (PVCs) to be associated with other PVCs on the basis of morphology 99% of the time. The good performance of this method suggests that basis functions for separating signal from noise in QRS complexes need not be highly detailed and that a concise description of QRS morphology in terms of 4 numbers is adequate for distinguishing normal QRSs from PVCs. The method of calculating features continually from ECG data makes clearer the role of window alignment in QRS feature extraction. The simplicity and effectiveness of the method should make it useful to researchers