Rahul Kher and Shivang Gohel
In this work, arrhythmia detection and classification from ECG signals has been performed using a digital signal processor- TMS320C6713. Two of the predominant ECG arrhythmias- premature ventricular contraction (PVC) and atrial fibrillation (AF) have been addressed in this work. In order to distinguish the PVC and AF beats from normal ECG beats, algorithms based on the morphological characteristics of arrhythmias have been applied. The PVC and AF beats present in ECG signals have been classified using correlation-based algorithm, in which a PVC or AF beats are compared/correlated with a normal ECG beat. The correlation coefficient value for normal ECG beats for a particular ECG signal is above 0.9 (highly correlated) whereas for a PVC or AF beats its value is in the range of 0.09 to 0.3 (highly uncorrelated). Another algorithm, based on slope/amplitude, has been implemented for detecting the PVC beats from ECG signals. The slope/ amplitude-based algorithm detects the PVC beats with 98.94% accuracy as compared to 65.20% accuracy by correlation-based algorithm. Thus, slope/amplitude-based algorithm outperforms the correlation-based algorithm as two parameters -the slope of QRS complex and R wave amplitude- are considered for detecting the abnormal beats. This work presents a DSP processor-based system, ideal for use in real time applications, for detecting PVC and AF beats from ECG signals.
