Research On Intelligent Discrimination Method For Human Physiological State Based On Heart Rate Variability

In recent years,cardiovascular disease has been widely concerned because of its high incidence rate,high mortality rate and low rate of complete cure.The number of deaths each year from cardiovascular disease is the highest of all causes of death.Timely and accurate detection and diagnosis of cardiovascular diseases is conducive to high-risk groups in advance to do a good job of cardiovascular disease prevention measures.The pathogenesis of cardiovascular disease is related to the fluctuation of the human autonomic nervous system,and heart rate variability(HRV)can reflect the fluctuation of the human autonomous nervous system,the physiological state of the human body can be determined by the HRV.Therefore,it has important theoretical and practical significance to carry out study on cardiovascular disease.This dissertation contains three main contents:The first content study obstructive sleep apnea(OSA),which is an independent factor for cardiovascular disease and a major factor in sudden night time death in patients with cardiovascular disease.In this dissertation,a novel method for detection OSA based on HRV signals is proposed.First,a Burst-AR model is used to estimate the power spectrum of HRV signals,and the time-frequency image of the power spectrum(TFSI)is obtained.Second,according to the physiological significance of HRV signals,the TFSI is divided into several sub-bands in the frequency domain.Finally,by studying the Shannon entropy and its interrelation of different sub-bands TFSIs,the fluctuating characteristics of HRV signals are measured to achieve screening of OSA patients.The validity and advantages of the MTFSE method are validated on a public data set in the PhysioNet database.The extracted HRV signal is first divided into short-term HRV segments of 5 minutes.Then,the features are extracted by the MTFSE method,and the HRV signals are classified by three classification methods:K-Neighbor algorithm(KNN),Support Vector Machine(SVM)and Decision Tree(DT).The results show that the SVM classification algorithm has the best classification effect,with an average accuracy of 91.89%,an average sensitivity of 88.01%,and an average specificity of 93.98%.The results show that the proposed MTFSE method improves the accuracy of OSA disease screening and provides a novel idea for the analysis and measurement of HRV signals.The second content studies Cardiovascular and Cerebrovascular Events(CCE),which are diseases caused by risk factors that cause vascular and cerebrovascular lesions in the heart and brain,including stroke,acute myocardial infarction,etc.In this dissertation,a novel method of predicting cardiovascular events HSE method is proposed.First,the instantaneous amplitude(IA),instantaneous frequency(IF)and instantaneous phase(IP)of the HRV signal are calculated based on the Hilbert transformation;Then the third-order accumulation and fourth-order accumulation of the HRV signals and IA sequence,IF sequence,IP sequence are calculated respectively;Finally,the singular entropy of the eight higher-order accumulations are calculated and used as characteristic vector to verify the validity and advantages of the HSE method through the public data set in PhysioNet database.Using KNN,SVM and DBN three classification methods to classify patients with cardiovascular events and without cardiovascular events,the DBN classification method shows the highest accuracy,sensitivity and specificity of the classification of cardiovascular events,90.77%,83.56%and 98.03%,respectively.The results show that the HSE method improves the accuracy of the classification of patients with cardiovascular events and provides a novel idea for the prediction of patients with cardiovascular events.The third content studies Atrial fibrillation(AF)and congestive heart failure(CHF),which are two typical cardiovascular diseases,and a typical method of determination of cardiovascular disease-ICBN is proposed.The method can distinguish between patients with AF and the healthy control group,and can also classify patients with CHF and health control group.This method analyzes HRV signals based on improved complete ensemble EMD with adaptive noise(ICEEMDAN)and complex networks,and mainly studies the coupling relationship between HRV signals in different frequency scale spaces.First,adopting the ICEEMDAN method on the HRV signals,the time-frequency scale decomposition of the HRV signals are obtained,the bubble entropy of each component is calculated to obtain the entropy vector,the vector is mapped into a complex network,and the parameters of the complex network are quantified by quantitative analysis.The validity and advantages of the ICBN method are validated through a publicly available data set in the PhysioNet database.The accuracy rate of recognition of patients with CHF is 89.66%,and the accuracy of recognition of patients with AF is 91.86%.As a result,this dissertation provides a novel idea for the research of typical cardiovascular AF and CHF.The physiological state of the human body is influenced by various factors.The analysis of the physiological state of human body by HRV is helpful to further explain the pathogenesis of cardiovascular disease.This dissertation focuses on the diagnosis of OSA,cardiovascular events,AF and CHF,which has some guiding significance for the prevention,diagnosis and treatment of cardiovascular disease.