Abstract: In order to explore the complexity of pulse wave,combined with the concept of “atrial fibrillation pulse” in traditional Chinese medicine,a complex network method to detect atrial fibrillation was presented.The photoplethysmograph pulse wave is thereby transformed to a network topology using visibility graph method.A binary classification support vector machine (SVM) based on Gausssian kernel function is designed to distinguish between normal sinus rhythm and atrial fibrillation.The degree distribution of the network and the average heart rate are extracted as the input features of the SVM.According to the experimental results of patients with paroxysmal atrial fibrillation,this methodcan effectively identify the patient’s disease status and normal status.

Key words: Pulse wave,Paroxysmal atrial fibrillation,Complex network,Visibility graph,Support vector machine