Lightweight arrhythmia detection using cross-correlation

Heart disease is a major health issue and the leading cause of death across the globe in recent times. Recent advanced technologies like IoT and AI can play a major role in remote monitoring and early detection of cardiovascular symptoms to control the effect of cardiovascular problems. Due to the crucial, diverse, and patient-specific characteristics of heart activity, the existing arrhythmia detection methods fail to identify the major heart abnormalities. Moreover, the existing Deep Learning based methods are computationally very expensive, and needs an expert knowledge to setup the architecture. In this paper, we have used a template-based cross-correlation measuring technique for the analysis of ECG signals. To make the system lightweight and more accurate, we have selected a limited number of most correlated reference signals from the normal and different abnormal categories of ECG signals. Performance of the proposed method is analyzed, and cross-validation is carried out among the proposed, three other correlation techniques, convolutional neural network, and handcrafted method using two different arrhythmia ECG datasets. Experimental results have proved that the proposed approach outperforms the state-of-the-art approaches in terms of identification accuracy (98.5%), and false detection rate (1.5%). Similarly, the experimental results also prove that the proposed technique is suitable for resource-constrained IoT applications with nearly 50% less computational resource usage and execution time compared to the other methods.