Research On Blood Pressure Monitoring With PPG Based On BP Neural Network

The blood pressure parameters are an important physiological state indicator for doctor's clinical diagnosis on their patients,which reflects the physiological state of the target organs such as the cardiovascular,kidney and brain of the human body.Hypertension is one of the most common chronic diseases,and it occurs mostly in the middle-aged and elderly people.Generally speaking,the main service subject for the community health service system are the elderly people living alone,so it is an indispensable task to achieve a non-invasive,continuous,real-time and convenient blood pressure monitoring for the elderly living alone,which has a positive effect on preventing sudden cardiovascular and cerebrovascular diseases in patients with chronic diseases.Photoplethysmograph(PPG)is one of the important physiological signals of the human body.It contains abundant physiological and pathological information,which intuitively reflects the fluctuation of blood in the human blood vessels caused by each heart beat.The collection technology of PPG signal has matured with the advantages of simple operation,low cost,safety,reliability,Similarly,continuous PPG signal monitoring and no discomfort to the test subject can be realized.This thesis focuses on the wave separation between PPG signal and noise,Characteristic parameter identification and blood pressure monitoring model for the photoelectric pulse wave signal of human fingertip,which can be summarized as follows:PPG signal was transmitted from the signal conditioning system to the host computer.The comparison of the three denoising effect on of the Infinite Impulse Response(IIR)digital filter,the Finite Impulse Response(FIR)digital filter,and the Wavelet Transform Modulus Maxima(WTMM)were conducted under the software.The denoising effect was analyzed by Peak signal-to-noise ratio(PSNR)and Normalized cross-correlation(NCC).The baseline drift of the signal was corrected by cubic spline interpolation.When the pulse wave Characteristic point recognition was performed,the initial position and the main peak position of the pulse wave were obtained by using the differential threshold method;the position of the tidal wave and repulsive wave were obtained by using the method of WTMM;For special waveforms,the tidal wave were obtained by fitting a partial pulse pattern with Gaussian bell curve.The relationships between the six Characteristic points are used to determine 20 groups Characteristic parameters related to blood pressure.Finally,the Back-Propagation Neural Network algorithm is used to establish the relationship model between pulse wave Characteristic parameters and blood pressure.The accuracy of the model is verified by the validation set.From the results of the validation set which includes 20 samples showed the mean error of systolic blood pressure was 1.08±2.95mmHg,and the mean error of diastolic blood pressure was 0.65±0.96mmHg,which met the error standard established by AAMI(Association for the Advancement of Medical Instrumentation).