Research On Wearable Noninvasive Continuous Blood Pressure Monitoring Technology Based On Signal Quality Classification

Hypertension is called "silent killer".It is often asymptomatic at an early stage,with a diagnosis rate as low as 46%,and less than 33%of patients can control hypertension after treatment.Therefore,Devices that can measure blood pressure timely,diagnose hypertension reliably,and monitor changes in blood pressure are key to preventing hypertension and its associated cardiovascular disease and for early treatment.At present,non-invasive continuous blood pressure monitoring devices on the market mainly use the volume compensation method,which are bulky and low in comfort,and are not suitable for continuous blood pressure monitoring for a long time in daily life.In view of this situation,this thesis has developed a set of portable,comfortable,scalable,and wearable non-invasive continuous blood pressure monitoring systems.The main research contents and results are as follows:1.Set up a wearable non-invasive continuous blood pressure monitoring system based on the mobile medical loT with wearability and scalability,achieving the synchronous collection of PPG and acceleration signals,the real-time communication between signal collection front-end and mobile relay,mobile relay and cloud platform.2.A multi-round signal quality evaluation algorithm for PPG based on exercise intensity and waveform characteristic indicators was proposed.The first round of signal quality evaluation based on exercise intensity removes abnormal signals that were severely disturbed by motion and improves the subsequent signal processing speed After the high-frequency noise and low-frequency noise were removed by the double-density wavelet transform and cubic spline interpolation,respectively,the pulse wave that was greatly affected by motion interference was removed from the remaining signal through the second round of signal quality evaluation based on the waveform characteristics.This ensured the quality of the PPG signal used for blood pressure estimation and provided a good data basis for subsequent blood pressure estimation.3.A blood pressure model construction algorithm based on signal quality classification and BP neural network was proposed.Signal quality was classified using triaxial acceleration.Pearson correlation coefficient method was used to select highly relevant pulse wave characteristic parameters for individuals.BP neural network was used for Construction of personalized BP blood pressure estimation model based on signal quality classification.Experiments showed that the average correlation coefficient of the method on systolic and diastolic blood pressure is increased by 0.12,the average absolute error is reduced by 0.4mmHg and 0.34mmHg,and the standard deviation of the average error is reduced by 0.5mmHg and 0.47mmHg,which is better than no signal quality classification,and the consistency analysis shows that it had good consistency with the Omron HEM-8622 wrist sphygmomanometer.