Research On Non-contact Blood Pressure And Blood Glucose Detection Method Based On IPPG Technology

Blood pressure has always been a potential risk factor for cardiovascular disease.Blood pressure measurement is one of the most useful methods for early diagnosis,prevention and treatment of cardiovascular diseases.Blood sugar is also linked to health.Hypertension and diabetes are serious diseases that endanger human health.Therefore,portable,reliable and sustainable blood pressure and glucose monitoring equipment is particularly important for the prevention and treatment of diseases.At present,the mainstream blood pressure measuring equipment is mainly divided into mercury sphygmomanometer and electronic sphygmomanometer.Mercury sphygmomanometer although high accuracy,but complex operation,high professional requirements.Electronic sphygmomanometers are usually used for home monitoring,but the use of electronic sphygmomanometers has many limitations,such as accuracy affected by sensor position,can not carry out continuous measurement and other shortcomings.At present,the mainstream blood glucose detection methods are mainly determined by blood test or glucose meter.These methods are all measured in an invasive way,which will cause damage to the human body.Frequent and invasive blood collection can cause severe discomfort and.Therefore,it is urgent to seek an accurate,convenient,safe and comfortable blood glucose measurement method for diabetes patients.Image photoelectric volume pulsegraphy(IPPG)is a new non-contact optical detection technology.The technology collects rich and reliable blood flow information(pulse wave signals)from human skin tissue,which is closely related to our physiological condition.In addition,the technology can be implemented on any device that can capture and process video(e.g.,webcams,smartphones,etc.).The method has the advantages of simple operation,comfortable use and continuous measurement,which is convenient for daily monitoring.This paper aims to achieve the non-contact measurement of blood pressure and blood sugar by IPPG technology.The specific research work of this paper is as follows:1.This paper designs a NCBP system based on PTT.The system uses a high-speed industrial camera to capture video of the human face and hand,and extracts IPPG signals from the face and hand respectively.PTT was obtained from two signals to predict blood pressure.The system uses a single camera to take pictures of two parts of the human body,and experiments can be carried out under ambient light without the need for external light sources.The results of this study were in line with the international standards.The method is simple in equipment and convenient in experiment,providing a new solution for portable blood pressure detection and assisted diagnosis of hypertension.2.In the research work based on PTT blood pressure detection,IPPG signal acquisition needs to obtain signals from two parts of the human body,which leads to inconvenience in measurement and high cost of industrial cameras with high frame speed.Therefore,in the second project,a non-contact blood pressure detection system based on machine learning method is designed.The system captures a video of the face via a webcam under ambient light and extracts pulse wave signals from the video.In this work,we extracted 26 pulse wave features from IPPG signals and screened out features highly correlated with blood pressure.Different machine learning algorithms were used for feature training and blood pressure prediction model was established.Meanwhile,a region of interest(ROI)selection method based on facial pseudo-color image and human appearance is proposed.The method visually observed the light intensity distribution of volunteers’ faces in the form of pseudo-color images,combined with male and female physical features,and found the most suitable area for analysis.By comparing the four models with 191 volunteers,support vector regression(SVR)was the best model for predicting blood pressure,and the experimental results fully matched the two international indicators mentioned above.The system has the potential to replace the traditional cuff sphygmomanometer and has guiding significance for the future development of blood pressure measuring equipment.3.In the BP detection work based on ML,the PTT needs to be defined and calculated manually,which will lead to a large workload.Moreover,in the definition of some features,such as the pulse wave shape area feature,because the pulse wave valley is not strictly on the same horizontal line,there will be ambiguity in the definition of some features,which will lead to calculation error.Based on this,in the third work,we propose a multi-feature fusion(MTFF)deep neural network model for blood pressure prediction.The model consists of two modules: one is the convolutional neural network(CNN)module,which is used to train the morphological and time-frequency characteristics of pulse wave signals;the other is the bidirectional long and short-term memory(BLSTM)network,which is used to train the time characteristics of pulse wave signals.After training,these features were fused by specific fusion modules,and the relationship model between fusion features and blood pressure was established.The results of this model fully comply with the international standards of the AAMI and BHS.The difference between the method and the machine learn-based blood pressure prediction method is that the method automatically extracts pulse wave characteristics through deep learning model,which can easily deal with complicated calculation.Meanwhile,the training integrates three different characteristics,and further improves the accuracy of blood pressure prediction.4.The current mainstream BG detection methods are invasive,which will cause damage to the human body.NCBG system can avoid this problem.Therefore,in the fourth work,a NCBG detection method based on near-infrared camera was proposed.Because blood glucose has strong absorbability in near-infrared band,and other components in blood(water,hemoglobin,etc.)in this band have very low absorbability.Therefore,in this method,we realized blood glucose detection by receiving light reflected back after blood glucose absorption by near-infrared camera.At the same time,the pulse wave characteristics highly correlated with blood glucose were analyzed and discussed,and the blood glucose prediction model was established by using multiple regression algorithms.Finally,the experimental results are analyzed by Clark error curve,which shows that the proposed method is in good agreement with commercial continuous glucose monitor.Compared with the traditional invasive blood glucose detection method,the non-contact blood glucose detection method has more application prospects.