Inter-patient Heartbeat Classification Using Higher Order Statistics and Symbolic Dynamics of ECG Signal

Abstract

Computer assisted heartbeat classification techniques have gained significant importance in the recent years as they aid the cardiologists in identifying abnormal heartbeats in ECG recordings. These techniques are of vital importance especially for the long term ECG recordings where it is not feasible to manually analyze the numerous heartbeats. Electrocardiogram (ECG) is an electrical representation of human heart activities and change of these activities is reflected in the morphology of ECG signal. Anomalous activities of the heart are characterized by significant variation in the ECG heartbeat morphology and the time related aspects. In the present paper, the morphological and temporal features are explored for ECG heartbeat classification. A novel feature set is formulated by selecting symbolic dynamics, RR intervals and the higher order statistics (HOS). Subsequently the feature set is fed to the K-nearest neighbor (KNN) classifier that follows the inter-patient classification scheme for beat categorization. The proposed method is validated on MIT-BIH arrhythmia database and heartbeat samples of this database are grouped into five classes as Normal beats (N), Supraventricular beats (S), Ventricular beats (V), Fusion beats (F), and Unclassified beats (Q). The proposed method achieves an overall accuracy of 94.35%. The obtained results outperform existing state-of-art works reported in the literature for inter-patient heartbeat classification.

Keywords: Symbolic dynamics, heartbeat classification, K-nearest neighbors, higher order statistics

Cite this Article
Krishnakant Chaubey, Barjinder Singh Saini, Atul Kumar Verma. Inter-patient Heartbeat Classification Using Higher Order Statistics and Symbolic Dynamics of ECG Signal. Current Trends in Signal Processing. 2017; 7(2): 37–45p.