Study On Noninvasive Continuous Blood Pressure Prediction Model Based On PPG

As we all know,the heart is the center of the blood circulation of the human body,and it bears the important task of making the organs of the human body run normally.Blood pressure is one of the body’s most important physical signals produced by the heart.The blood pressure in normal person remains stable under a variety of factors;for the elderly or patients,the fluctuations of blood pressure tend to exceed normal range,while the abnormal fluctuations in blood pressure are helpful in determining a person’s physical state.Therefore,How to measure blood pressure effectively is of great significance in medical treatment and daily life.In daily life,the most commonly used blood pressure measurement equipment is electronic sphygmomanometer,which the theoretical basis is korotkoff sounds Oscillographic method.In the use of an electronic sphygmomanometer,the instrument needs to exert pressure on the subject.This method is cumbersome and can not be continuously monitored,and it is easy to cause discomfort to those who have been tested,so it is necessary to seek a better method for continuous noninvasive blood pressure monitoring.Thanks to the development of sensor technology,people can easily obtain Photoplethysmogram(PPG)signals of human pulse,and many studies have also made prediction of blood pressure based on PPG signals.since the PPG signal is a set of continuously fluctuating waveform data,A regression model was developed by extracting features that were significantly related to blood pressure from the waveform of the PPG signal,then the blood pressure can be predicted.At present,most researchers establish the relationship between PPG features and blood pressure by linear regression.Through the experiment,we can see that the model obtained in the linear way has some shortcomings in the prediction accuracy,so this paper will model the data by multiple regression methods.On the basis of the regression model,in this paper,linear and nonlinear regression models of PPG signals and blood pressure are modeled by means of linear regression,neural networks and other machine learning methods,and then the prediction results of each method are compared.In view of the characteristics of blood pressure,this paper presents two optimization methods to improve the accuracy of blood pressure prediction.First of all,according to the differences among different groups of people,the original data are classified by clustering method,and then the regression model is established for each category,which can reduce the error of the model to a certain extent.In addition,through the analysis of data we can find that there is a certain correlation between systolic pressure and diastolic pressure: the difference between them is generally stable in a certain range,Therefore,this paper uses this feature to propose an improved gradient boosting algorithm to optimize the basic regression model.The final experimental results show that the proposed method can quickly and effectively improve the effect of blood pressure prediction,and the prediction results have lower errors.