ECG classification using morphological features derived from symbolic dynamics

The aim of this study is to estimate how far the symbolic analysis helps to characterize the nonlinear properties of Electrocardiogram (ECG) signals and thereby discriminate between normal and arrhythmia signals. Differences were found in the properties of the symbol sequence histogram of the resulting patterns and complexity measures of symbol sequences among the five different classes, namely normal, left bundle branch block (LBBB), right bundle branch block (RBBB), premature-ventricular contraction (PVC) and paced signals. The efficacy of the symbolic analysis features is also shown through classification achieved through the neural network which exhibits an average accuracy that exceeds 93.5%.