Non-contact Blood Pressure Measurement Based On Skin Color Images

Compared to traditional contact blood pressure measurement,non-contact blood pressure measurement enables people to feel more comfortable,and makes remote blood pressure monitoring possible,which has important influence on early diagnosis and effective prevention of hypertension.These days,most of non-contact blood pressure measurements have highly strict requirement on the measuring equipment,and still rely on contact sensor.Hence,this thesis has constructed the blood pressure prediction model of BP neural network to measure blood pressure by following research:based on the Imaging Photoplethysmography theory,the research has collected different parts of human body through color CCD camera on the first step;then it has differentiated and collected the skin color of human body through modified Adaboost algorithm;after that,it has obtained the human blood pulse signal through modified Euler amplification technology to deal with the image of skin color.Here are some details about the conclusion of this thesis.Firstly,in order to get two-way blood pulse signal,the experiment has designed a set of human skin color image acquisition hardware platform,and in order to extract the skin color part of the original image,based on the study of Adaboost algorithm,an improved Adaboost algorithm is proposed to recognize and extract the face and palm images.Secondly,the model of blood pressure can be constructed only if the IPPG signal extracted from skin color image is accurate.This thesis has extracted and analyzed the blood pulse signal contained in the skin color image through modified Euler amplification technology to get the pulse wave eigenvalue and heart rate signal.Thirdly,at the blood pressure measurement period,the two-aspect blood pressure prediction model has been constructed.The research has built the linear relation between pause transmit time with blood pressure to vaguely estimated the blood pressure,then accurately predicted the blood pressure by using the improved BP neural network with a trained parameter library.Above all,the experiment proves that the accuracy of recognition rate for human face and palm has achieved 92.5%.The pulse wave signal obtained based on Euler's amplified skin color enhancement algorithm has very good agreement with the standard pulse wave signal,and the accuracy of the heart rate signal calculated by it is 97.25%,while the accuracy of blood pressure prediction model has achieved 93.6%in general,which enables acccurate analysis and conclusion of unusual blood pressure data?...