Dynamic ECG features for atrial fibrillation recognition

Highlights This study / paper highlights on several aspects as follows: • The characterization of atrial fibrillation using second order dynamic system. • The optimum windowing length of ECG signal processing during normal sinus rhythm (NSR) and during atrial fibrillation (AF) with normal sinus rhythm (N) according to pattern recognition machine learning using an artificial neural network (ANN) and a support vector machine (SVM) with k -fold cross validation ( k -CV) to develop an ECG recognition system. • The study proposed a method based on dynamic system, which achieved high sensitivity and specificity and able to describe the oscillatory behavior of heart according to the normal sinus rhythm of healthy people (NSR), normal (N) and abnormal sinus rhythms (AF) of atrial fibrillation patients ECG signals. • Our major results provide novel method in detection and classification of atrial fibrillation.