Research On The Identification Of Atrial Fibrillation Based On Photoelectric Volume Pulse Wave

Atrial fibrillation(AF)occurs frequently in arrhythmia diseases.Although it is not a fatal disease,the patients are prone to palpitation,dizziness and other uncomfortable symptoms,and even cardiac failure or myocardial infarction in serious cases,thus threatening people’s life and health.At present,the research on AF monitoring and recognition at home and abroad is mainly based on electrocardiography(ECG)method.Considering the ECG signal acquisition instrument is difficult to operate,inconvenient to carry and costly,this paper proposes to monitor and identify AF by Photoplethysmography(PPG).There is a strong correlation between PPG signal and ECG signal,and the acquisition instrument of PPG signal is simple and portable,which is more suitable for long-term monitoring of human health status.The experimental data used in this paper is based on pulse wave data of AF patients(diagnosed through ECG)and non-AF patients in the open database “MIMIC-III ” of Massachusetts Institute Technology(MIT).A total of 9,681 pulse wave samples of 10 seconds long were collected.Based on the demonstration about the feature points of AF-related pulse waves in this paper,MATLAB program was used to extract AF-related characteristic parameters from the pulse wave data as the input of classical sorting algorithm model.The classical machine learning algorithms selected for model building experiments in this paper include: BP neural network,support vector machine and random forest algorithm.After the experimental optimization model,the AF recognition accuracy of the three classical machine learning algorithms reached 94.7%,93.8% and 97.1% respectively.This study also attempted to use deep learning recurrent neural network(RNN)to design an experiment on the automatic recognition of AF pulse waves,after the experimental optimization model,the accuracy rate reached96.8%.Through detailed model analysis and comparative evaluation,it is concluded that random forest algorithm has a better effect on the identification of AF pulse waves.This paper also designs a set of AF pulse wave monitoring and automatic recognition instrument.The instrument takes STM32F103 as the main control chip,and adopts HKG-07 B transmission-type infrared PPG sensor for acquisition of pulse wave signals.The collected pulse waves are transmitted into a PC through a SD card.The upper computer pre-processes the pulse waves.Then,the pre-possessed pulse waves are input into the algorithm model for sorting.The final result is displayed on the user interface made by using MATLAB.As the results of this study suggest,the method of identifying atrial fibrillation by PPG signals is feasible and effective,The six characteristic points related to the pulse wave of the atrial fibrillation state proposed in this paper can effectively distinguish the atrial fibrillation pulse wave,which provides a new idea for future research on the identification of AF.In addition,the AF pulse wave recognition algorithm model designed in this paper and the monitoring recognition instrument can realize long-term monitoring and automatic recognition of AF more easily.