Research And Design Of Head-mounted Device Based On Pulse Wave Signal Processing

Pulse signal processingl is an important measure to make an analysis of human physiological and pathological.With the development of sensing technology and related theories,the monitoring based on the volumetric pulse wave of pulse oxygen saturation,pulse rate and pulse rate variability(PRV)has become an important index of modern medical detection.Because of the easy collection of pulse wave,portable monitoring of signs based on pulse wave and analysis device will play an important role in real-time monitoring of human health at the age of 5G and Internet of things.Based on the above background,this thesis proposes a design of a head-mounted physiological monitoring and analysis system based on signal processing of volume pulse wave.In this paper,we will carry out the analysis of the blood oxygen,pulse rate calculation and pulse rate variability according to the pulse wave signals collected by photoelectric reflection.The device can monitor the physical signs in real time,and can give early warning to the abnormal conditions in time.Through uploading them to the Cloud,the statistics and research on the physiological and mental state of the human body will be carried out,and it can provide the reference for the long-term health analysis of the human body.This thesis will do the research and analysis of monitoring system on the aspects of the accuracy of physiological monitoring,the feasibility of index analysis,the portability of the device,the optimization of signal processing.About the device design,this thesis chooses the modules of collecting and processing with better performance,high integration,and low power consumption,optimize collecting control software,designs wearing detecting procedure,improves collecting and processing performance and reduces overall power consumption.About the pulse rate variability analysis,the collecting methods of the pulse interval will be optimized and based on the EMD transform,a single component feature point extraction method is designed to analyze the PRV signal in time domain,and the EEMD decomposition and reconstruction of PRV signal is used to analyze the power spectral density.Experiments are designed to test the overall power consumption of the device,the accuracy of physical monitoring parameters,and the feasibility of PRV analysis.The duration of the device is about 9 hours.And the absolute value of blood oxygen error is within 3% when sitting and 4% when walking,and the average value of absolute value of pulse rate error is between 0.9% and 2.1%,which meet the expectant index.The correlation of frequency index between HRV and PRV is more than 0.95 and the correlation of time domain index between them is more than 0.98.In the Bland Altman consistency analysis chart of PRV and HRV,there are more than 95% samples in all index analysis which are within the range of consistency limit,and the PRV signal analysis index has a high consistency with HRV.