Research On R-peak Detection And Premature Beats Recognition Algorithm For Dynamic ECG

The intelligent real-time diagnosis of arrhythmia is of great significance to the out-ofhospital monitoring,early prevention and timely treatment of cardiovascular diseases.Premature beats are the most common type of arrhythmia.Early detection of premature beats and timely corresponding treatment measures can reduce the risk of cardiovascular disease and prevent the development and deterioration of heart disease.Dynamic ECG signals are characterized by a large amount of data,large noise,and strong individual and environmental variability.The real-time and sensitivity indicators of intelligent diagnosis of premature beats still fall short of clinical standards.In this paper,aiming at the above problems,on the basis of analyzing the biomedical mechanism and morphological characteristics of arrhythmia,and the noise distribution characteristics of the dynamic ECG signals,the real-time positioning algorithm of the main wave of the dynamic ECG signals is studied.A real-time diagnosis strategy for ambulatory electrocardiogram arrhythmia based on extrusion-excitation residual network(SE-Res Net)and composite deep learning model is used to identify premature beats,aiming to improve real-time complex wave detection algorithm and arrhythmia diagnosis accuracy.The main contents are as follows:(1)A real-time R-peak detection technology of ECG signals based on Shannon energy envelope combined with Hilbert transform is proposed to solve the problem of low real-time detection accuracy of R-peak for dynamic ECG signals.On this basis,the performance and characteristics of current R-peak detection algorithm recognition are evaluated,and a decision-making diagnosis model for self-learning multi-Rpeak detection algorithm fusion is proposed.After testing with multiple public dynamic ECG data sets,the model effectively improves the accuracy of R-peak positioning.(2)A classification model for premature ventricular contraction and supraventricular premature beats of single-lead ECG was proposed based on the improved LSTM-Res Net composite model.Experimental tests show that the model has high real-time performance,can fully mine the time characteristics and morphological characteristics of ECG,and effectively improves the accuracy and sensitivity of arrhythmia recognition compared with a single model.(3)The optimized model structure of the multi-lead ECG premature ventricular contraction feature extraction network based on SE-Res Net is proposed.The model can fully mine multi-channel ECG information,and the experimental results show that the model improves the classification accuracy and sensitivity.