Heartbeat Recognition from ECG Signals Using Hidden Markov Model with Adaptive Features

Abstract

A heartbeats recognition system for recognizing four different cardiac diseases was developed based on electrocardiogram (ECG) in this paper. The Hidden Markov model (HMM) was applied to the recognition of heartbeats from electrocardiogram (ECG). The ECG features developed by existing papers are used to train the HMM model. However, since different set of features are suitable to recognize different cardiac diseases, this paper proposed a strategy of using adaptive features to recognize different set of cardiac disease. The four abnormal heartbeats including the left bundle branch block (LBBB), the right bundle branch block (RBBB), the ventricular premature contractions (VPC), and the atrial premature contractions (APC) are recognized from the ECG data in the MIT-BIH Arrhythmia Database. Experimental results in this paper shown that the proposed strategy performed well and had very excellent recognition rate for some heartbeat cases.