Research On Arrhythmia Feature Extraction And Classification Based On ECG

In recent years,with the improvement of people's living standard and consumption ability,the incidence of cardiovascular disease is increasing year by year,and has a higher mortality rate.The early onset of cardiovascular disease is often accompanied by arrhythmia symptoms,so early detection of arrhythmia is of vital significance for early prevention of cardiovascular disease and early intervention.The classification of arrhythmia types has always been an important research topic in electrocardiogram(ECG)automatic analysis.However,because ECG signals belong to weak bioelectrical signals,it is very vulnerable to noise.There are still some challenges to realize accurate feature extraction and classification of arrhythmia.Based on this,this paper focuses on ECG feature extraction and arrhythmia type classification.In this paper,a mathematical morphological filter is designed for the low frequency noise and high frequency noise,which often appears in the ECG signal.By selecting the appropriate structure elements of different types and lengths,the waveform characteristics of the original signal are retained on the premise of effectively filtering the noise,so as to avoid the signal distortion and lead to the subsequent features extraction of errors.On the premise of accurate detection of each characteristic waveform of ECG signal,the most intuitionistic feature of waveform shape,that is interval and amplitude,is extracted from ECG signal,and the waveform information of ECG signal is obtained in an all-round way.Secondly,dynamic time normalization(DTW)is used to calculate the shape between single heart beat and standard heartbeat.The correlation between the heart beat and the distance of the state is described.Finally,the principal component analysis(PCA)is carried out on the QRS wave group with the highest information content in the ECG signal,and the main component of the QRS wave group.The three parts feature fusion is used as a classification feature vector set,which lays the foundation for subsequent classification work.In this paper,several common classification algorithms are compared,and the support vector machine is selected for the classification of normal ECG types and three types of arrhythmia types with strong nonlinear random signal processing performance.The type of kernel function used in support vector machine is radial basis kernel function.In view of the disadvantage that the standard particle swarm optimization algorithm is easy to fall into local optimal,this paper proposes an improved particle swarm optimization algorithm to quickly and effectively find the kernel parameters of radial basis kernel function and the penalty coefficient C.Then the feature vectors of the training samples are input into the support vector machine and the optimal classification model is trained.Then the characteristic vectors of the test samples are input into the classification model for classification.The classification results are evaluated by four indexes,sensitivity,specificity,positive predictive value and accuracy rate,and the classification results are divided into other related literature.The results were compared and analyzed.Experimental results show that the algorithm proposed in this paper can accurately classify heart beat types.The accuracy of classification can reach 97.32%.