Automatic detection of premature ventricular contraction beat using morphological transformation and cross-correlation

This paper presents a premature ventricular contraction beat (PVC) detection algorithm based on morphological transformation and cross-correlation technique. A modified morphological filtering (MMF) technique is used for signal preprocessing and multiscale morphological derivative (MMD) is performed on the MMF conditioned signal to detect each ECG beat present in the signal. A template beat is chosen and compared with the rest ECG beats using cross-correlation technique. PVC beats are then detected using a decision parameter which is a linear function of two equally weighted indices. One of the indices is linearly dependent on inter-beat duration and the other is an exponential function of the cross-correlation coefficient between template beat and the ECG beat. Potential of this proposed method was examined using MIT-BIH arrhythmia database. Results show high sensitivity (96.67%) and specitivity (95.2%) on premature beat recognition.