An Improved Method for Discriminating ECG Signals using Typical Nonlinear Dynamic Parameters and Recurrence Quantification Analysis in Cardiac Disease Therapy

The discrimination of ECG signals using nonlinear dynamic parameters is of crucial importance in the cardiac disease therapy and chaos control for arrhythmia defibrillation in the cardiac system. However, the discrimination results of previous studies using features such as maximal Lyapunov exponent (lambda_max) and correlation dimension (D₂) alone are somewhat limited in recognition rate. In this paper, improved methods for computing lambda _max and D₂ are purposed. Another parameter from recurrence quantification analysis is incorporated to the new multi-feature Bayesian classifier with lambda_max and D₂ so as to improve the discrimination power. Experimental results have verified the prediction using Fisher discriminant that the maximal vertical line length (V_max) from recurrence quantification analysis is the best to distinguish different ECG classes. Experimental results using the MIT-BIH Arrhythmia Database show improved and excellent overall accuracy (96.3%), average sensitivity (96.3%) and average specificity (98.15%) for discriminating sinus, premature ventricular contraction and ventricular flutter signals