ECG arrhythmia classification using simple reconstructed phase space approach

ECG arrhythmias such as ventricular and atrial arrhythmias are one of the common causes of death. These abnormalities of the heart activity may cause an immediate death or cause a damage of the heart. In this paper, an arrhythmia classification algorithm is presented. The proposed method uses the nonlinear dynamical signal processing techniques to analyze the ECG signal in time domain. The classification algorithm is based upon the distribution of the attractor in the reconstructed phase space (RPS). The behavior of the ECG signal in the reconstructed phase space is used to determine the classification features of the whole classifier. To evaluate the performance of the presented classification algorithm, data sets are selected from the MIT database. Two groups of data, learning and testing datasets, are used to design and test the proposed algorithm. A classification sensitivity and specificity of 100% are used to fine tune the parameters of the selected features using the learning dataset. Forty five signals are used to test the proposed approach resulting in 85.7- 100% sensitivity and 86.7-100% specificity are obtained respectively.