Research On Ecg Signal Denoising,Wave Group Detection And Arrhythmia Identification Algorithm

The prevalence of cardiovascular diseases is increasing gradually,posing a significant threat to human health.Accurate identification of cardiac arrhythmias is of great significance to the diagnosis and treatment of cardiovascular diseases.The use of computer technology to achieve automatic identification of ECG signals is divided into three main steps: pre-processing,waveform detection and classification identification.In practical applications,the classification of arrhythmias is not satisfactory due to different degrees of noise interference,incorrect positioning of feature waves and inadequate feature extraction.In this study,the following improvements were made around each of these three steps:(1)Traditional denoising methods are prone to the loss of useful information in high-frequency signals and are ineffective in denoising signals containing highintensity noise,which may lead to distorted waveforms.To tackle this problem,a WPDSVD-based denoising algorithm for ECG signals is proposed.The algorithm first uses wavelet packet decomposition to group the noisy signals,and then uses singular value decomposition to reconstruct the ECG signals.Compared with other denoising methods,this algorithm can effectively filter out the high-intensity noise in the signal and retain the waveform features of the original signal and the useful information in the highfrequency region to the maximum extent.(2)A waveform detection algorithm based on improved differential thresholding is proposed to address the problem that the traditional QRS waveform group detection algorithm has low localization accuracy under complex waveforms.To facilitate the localization of R peaks and highlight the amplitude and morphology of the characteristic waveform,the algorithm firstly enhances the signal using differential operations.The concept of pre-selected peaks is proposed,and they are relocated to the original signal by the score function,effectively avoiding the risk of difficult localization in complex waveforms.Compared with the classical Pan-Tompkins algorithm,this algorithm has a higher accuracy in identifying the characteristic waves,and the sensitivity can reach 99.8% and the positive prediction rate can reach 99.9%,which illustrates that this algorithm can accurately locate the characteristic waves in complex waveforms.(3)Through the feature extraction analysis of ECG signals,it is found that the methods used in the current study are relatively single and ignore the local features of heartbeats and the connection between heartbeats.To tackle this problem,an arrhythmia classification algorithm based on multi-feature fusion is proposed.The algorithm combines Lasso,t-SNE and DWT to extract three types of features in the ECG signal that are unrelated to each other to form a feature matrix and input it to the random forest classifier and this algorithm finally gets the classification accuracy of 99.9%.Compared with a single feature extraction method,this algorithm has better arrhythmia classification results.