Research And Realization Of Human Sub-health State Detection Method Based On Support Vector Machine

With severer social competition,the pace of life is increasing,which changes the people's diet and other factors.So the pressure of modern people is growing and more and more people is in sub-health state,which seriously affects people's quality of life and mental state.Therefore,the diagnosis of sub-health status,active intervention and prevention of sub-health,all have the scientific and social significance for the prevention of disease,and improve people's physical and mental health quality of life.This thesis explores the shortcomings of several common methods for the diagnosis of health status through the objective quantification of sub-health diagnostic parameters.It studies the health status recognition method based on multi-physiological signal fusion,and takes the physiological signals such as pulse,ECG and skin electrical signals as the research objects.It focuses on the design of the optimized feature extraction algorithm and the support vector machine classifier.And it finally puts forward a new sub-health diagnostic method.There are two main innovations in this thesis.First,in the current methods of health diagnosis based on quantification parameters,most of them only collected the physiological signals of volunteers as the evaluation parameters,while ignoring the influence of mental health.On the basis of the ECG signal and the pulse signal as the evaluation parameters,this thesis introduces the skin electrical signal which can represent the psychological state of the subject as the basis of the assessment,and more comprehensive analysis the health status of the subject,so that the test results are more of authenticity and representativeness.Second,in traditional linear discriminant method,too much emphasis on the impact of features which have a higher posterior classification rate,leads to the classification performance of edge samples with unobvious features reduced.In this thesis,the weighting factor based on posteriori classification rate of feature is introduced into the linear discriminant method to improve the problem.Compared with the traditional fisher linear discriminant method,the optimized algorithm not only reduces the intra-class dispersion,but also improves the support vector machine.In this thesis,three kinds of biological information of 67 volunteers were collected as experimental data.The linear discriminant method,traditional fisher linear discriminant method and principal component analysis were used to extract the37 kinds of features from three kinds of biological information,and diagnose the sub-health state of the 67 volunteers.The results show that the classification accuracy of the data processed by the optimized linear discriminant method is higher,which confirms the validity and feasibility of the algorithm.