Research On Classification Algorithm Of Arrhythmia

According to the report of the World Health Organization(WHO),the number of annual deaths due to cardiovascular diseases is increasing year by year.In 2019,more than 20 million people around the world died from cardiovascular diseases,accounting for 33% of the annual global deaths.At the same time,cardiovascular diseases are characterized by a wide variety of diseases,high concealment and high fatality rate,so it can be seen that cardiovascular diseases have become the primary threat to human life and health.The most effective test for cardiovascular disease is an electrocardiogram.In recent years,the artificial intelligence technology has obtained the unprecedented development,at the same time,because the diagnosis of telepathogram mainly relies on the doctor’s artificial diagnosis.The researchers combined arrhythmia diagnosis with artificial intelligence technology to develop an automated ECG analysis system to help doctors make the diagnosis.This technique has been widely used in clinical practice and has important applications in the fields of ECG monitoring and auxiliary diagnosis.Automatic classification of arrhythmias includes four steps: ECG signal preprocessing,feature wave detection,feature extraction and classification diagnosis.Aiming at the above problems,a new algorithm for classification of arrhythmias is designed in this paper.The main work is as follows:ECG signal preprocessing.The preprocessing in this paper includes denoising and heartbeat segmentation.The main noise of ECG signal is baseline drift,EMG interference and power frequency interference.In this paper,a multiscale wavelet threshold filter is designed to filter and denoise the ECG signal.For heartbeat segmentation,the Pan-Tompkins QRS detection algorithm was used to detect the R ware position,then,149,150 sample points were intercepted before and after the R ware position,and total of 300 sample points were taken as a complete heartbeat.Feature wave detection.For the detection of P wave and T wave,the fixed range search method is adopted in this paper,and the search range is determined by the interval between the QRS wave and R-R interval.Feature extraction.This paper presents a method for extracting the timefrequency feature of ECG signal by Non-negative Matrix Factorization(NMF).One dimensional ECG signals were analyzed by non-negative smoothing wignerville time-frequency analysis,then,the two dimensional matrix was obtained.In this paper,four R-R intervals,P amplitude and T amplitude are defined.Classification.In this paper,a multi-layer SVM multi-classifier is designed for multi-classification of arrhythmias.Finally,MIT-BIH arrhythmia dataset was used to verify the algorithm in this paper,and the accuracy of the test set reached 93.3%.