Research On Non-invasive Blood Glucose Detection Technology Based On PPG Signal

With the change of people’s lifestyle and eating habits,the occurrence and prevalence of diabetes has undergone new changes.The latest report from the International Diabetes Federation shows that the number and proportion of people with diabetes are on the rise worldwide.The potential threat of diabetes to health and people’s gradual attention to health make the traditional invasive and invasive blood glucose detection method unable to meet people’s needs,and people are in urgent need of a simple,real-time non-invasive blood glucose detection method.In this paper,a non-invasive blood glucose detection system based on photoplethysmographic pulse wave(PPG)signal is designed.Firstly,the red and infrared PPG signals after the transmitted fingertip were collected by the transmission photoelectric acquisition equipment.The collected PPG signal line is filtered,and the median filter is used in this paper to remove the noise interference in PPG signal.Aiming at the problem that the median filter will lead to smooth waveform,an improved median filter is proposed in this paper.On the basis of median filtering,an adaptive threshold value is designed according to the window,and the median value is used to determine whether the original signal value needs to be replaced by the threshold value.The improved median filter can not only remove noise interference well,but also retain the integrity of PPG signal waveform.In this paper,the baseline drift of PPG signal is solved based on cubic spline interpolation,and the desired effect is achieved.The main wave peak and trough of pretreated PPG signal were extracted by differential threshold method,and then the comprehensive characteristic parameters were calculated.The fitting formula between the comprehensive characteristic parameters and the blood glucose value was established by the least square method,and the initial blood glucose value could be calculated by the fitting formula.Secondly,due to the difference of human body,the calculated initial blood glucose value has a large error,and the simple fitting formula is difficult to meet the requirements of blood glucose detection.On this basis,physiological parameters affecting blood glucose detection were extracted,including heart rate,age,height,weight and BMI.The preliminary blood glucose calculated by combining the above physiological parameters constituted the input parameters of the subsequent blood glucose detection model.In this paper,a blood glucose detection model based on Particle Swarm optimization-radial Basis Function(PSO-RBF)is proposed.Subclustering algorithm and K-means clustering algorithm were used to calculate the number of hidden layer neuron nodes and basis function center of RBF neural network,and PSO algorithm was used to optimize the weight of hidden layer to output layer of RBF neural network,which optimized the network model and improved the prediction ability.Through experimental analysis,compared with multiple linear regression model,partial least squares regression model and traditional RBF neural network model,the PSO-RBF neural network model proposed in this paper has better blood glucose detection ability,with the mean absolute error of 0.56mmol/L and the relative error of 10%,which meets the relevant national standards.Bland-altman analysis showed that the blood glucose values obtained by PSO-RBF neural network model were in good agreement with the standard blood glucose values.Finally,based on the non-invasive blood glucose detection model based on PSO-RBF neural network,this paper proposes a networked non-invasive blood glucose detection system.The hardware part of the system includes a PPG signal acquisition module and a network communication module.The software part includes data transmission,non-invasive blood glucose algorithm and visual interface design.The system allows the tester to check the personal blood sugar level in real time simply and in real time through the smart terminal.Finally,the practical feasibility of the system is verified by the experimental comparison with the invasive blood glucose meter.