Research On Arrhythmia Classification Based On Feature Extraction Of ECG Signal

In recent years,cardiovascular and cerebrovascular diseases have become one of the diseases with high incidence and high risk,and prevention and treatment of cardiovascular and cerebrovascular diseases has become a key point in medicine.Arrhythmia is a cardiac activity or conduction of the origin of abnormal disease,long-term arrhythmia will cause a variety of heart diseases.Therefore,based on ECG signal analysis,auxiliary medical diagnosis of arrhythmia has important research significance.In this paper,the improved weighted linear discriminant analysis is used to classify the features so as to realize the diagnosis of arrhythmia.First of all,preprocessing and data set construction based on data.This paper uses digital filter to preprocess the data.Avoid the situation that the sample number is too small and the classification result is inaccurate when directly processing a piece of data.In this paper,the ECG signal is segmented and the data set is constructed.In order to ensure the accuracy of the experiment,the number selected each time is random.Secondly,feature extraction and dimension reduction based on difference threshold method.In this paper,the difference threshold method is used to extract R waves.According to the information of R wave,this paper selects the corresponding time-domain feature and combines the morphological feature as a general feature of classification.Too high data dimension will affect the classification effect.This paper chooses principal component analysis to reduce the dimension of high-dimensional data.Finally,the improved linear discriminant analysis was used as adjuvant therapy for arrhythmias.When traditional linear discriminant analysis classifies data,it often happens that some samples have little difference from other samples,and the difference between them and similar samples is very large,which makes the data borderline.The weights of the samples that are well distinguished are reduced,and the weights of the edge samples are increased,so that they can be well distinguished.Numerical experiments show that the accuracy of the traditional linear discriminant analysis classification model is about 84% and that of the improved linear discriminant analysis classification model is about 91%.