Study On The Diagnosis Method Of Ventricular Premature Beat Based On Fuzzy C-means

At present,the incidence of cardiovascular disease is getting higher and higher,and more and more people die from cardiovascular disease.Among all cardiovascular diseases,premature ventricular contraction(PVC)is the most common,with a large number of cases,a wide range,and large variability,which seriously affects people's normal lives.Electrocardiogram,as a non-invasive diagnostic method,is an important basis for judging whether a patient has cardiovascular disease,but it is affected by the environment and the subject's movement,psychology and other factors.The detected electrocardiogram(ECG)contains a variety of Interference cannot be used directly for medical diagnosis.In addition,in order to more comprehensively grasp the patient's condition,long-term monitoring is required,which generates a large amount of ECG data,which is not suitable for manual classification and judgment.Therefore,it is particularly important to improve the efficiency of the denoising preprocessing and automatic classification algorithms of ECG signals.Based on the characteristics of ECG signals,this paper proposes a fuzzy C-mean deep belief network structure to diagnose ventricular premature beats.The specific research content is as follows:First,a filtering and interception filtering algorithm for ventricular premature beats is proposed.This algorithm uses sym4 wavelet to decompose on 8 scales,and uses adaptive threshold method to denoise ECG signals.Comprehensively consider the energy and frequency of different waveforms,reconstruct the wavelet signal on the 3,4,and 5 scales to highlight the QRS wave,mark the R wave,and then calculate the RR interval;based on the non-uniformity of the RR interval,screen out the possible inclusion Premature beats;the screened beats are automatically cut into candidate bands containing R and T waves.The preprocessing process greatly reduces the amount of ECG data,which can reduce the intensity of subsequent classification work.Second,use deep belief networks to generate feature vectors to avoid human subjective factors.Using a Gaussian-Bernoulli Restricted Boltzmann machine(GBRBM)and a multilayer Bernoulli-Bernoulli Restricted Boltzmann machine(BBRBM)The deep beliefnetwork(DBN)is composed of feature vectors through layer-by-layer training.Third,a new fuzzy C-means clustering algorithm(FCM)with a degree of membership with confirmation degree is used to classify feature vectors.Adding the degree of confirmation to FCM can reduce the impact of outliers and get more accurate cluster centers.The objective function of the FCM was modified,and the degree of confirmation related to membership and the iterative formula of the cluster center were derived to realize the diagnosis of ventricular premature beats.This paper uses 20 sets of different types of ECG data verification algorithms in the MIT-BIH ECG database.The results show that the proposed fuzzy C-mean deep belief network algorithm has high accuracy and strong stability.