The Near Infrared Non-invasive Blood Glucose Detection System Based On Android

In recent years,with the rapid development of China's economy,a substantial change becomes in the social.Because of the upgrading of people's living and consumption patterns,people are seriously troubled by diabetes caused by high blood sugar and its complications.The proportion of patients being affected by diabetes and the proportion of patients dying due to the deterioration of diabetes is increasing year by year.Besides,people with certain diseases experience hypoglycemia that is a danger of syncope and sudden death.Therefore,real-time and convenient blood glucose detection is extremely important to improve people's life quality and health security today.Currently,most blood glucose detection methods on the market are anticipated by the electrochemical reaction between the finger blood and the paper.These methods require every measurement to damage the skin,causing physical and mental discomfort for patients.Especially,it would always want to increase the risk of wound infection in diabetic patients.Based on this situation,domestic and foreign experts and scholars have been working day and night in the field of non-invasive blood glucose detection so that divided the detection methods.Such as body fluid detection,energy metabolism conservation method,near-infrared spectroscopy,dynamic monitoring method with subcutaneous implanted sensors,microwave detection method.Because of the above situation,this study designed an android-based near-infrared blood glucose detection system,aiming to establish anear-infrared non-invasive blood glucose detection model through in vivo experiments of non-invasive blood glucose detection by near-infrared ray,and further improve the accuracy of the model through calibration.The work arrangement of this project is as follows:(1)Through the investigation of the phenomenon of existing non-invasive glucose detection technology,the research direction of this subject is determined as near-infrared non-invasive detection;The wavelength of the photometric data collector was determined by the comprehensive consideration of the absorption of the components of the near-infrared ray and the tissue fluid.The actual situation of collecting blood glucose data is helpful to the human body and the use of the instrument is in line with the ergonomic situation.Testing the validity and repeatability of photometric data ensure the validity of subsequent experiments.(2)This study established a near-infrared non-invasive glucose detection model based on xgb-lgb-catboost.Based on the original XGBoost model,this project adopted many improvements to enable it faster than other gradient enhancement algorithms.RMSE was tuned down to1.94,which the result was satisfactory.To further reduce the root-mean-square error coefficient and improve prediction accuracy,the model combines the prediction model based on LightGBM algorithm with the prediction model based on catBoost algorithm,and the RMSE is reduced to 1.80.The model performance is preferable to the first prior.Besides,the xgb-lgb-catboost model only needs to train 1 sample data to adjust the weighting coefficient to overcome individual differences due to skin color,gender,etc.(3)The near-infrared non-invasive detection system establishes a client on the Android Studio platform,connects the collector through low-power Bluetooth,and collects the photometric data of signals;Support to manually upload room temperature,humidity,high pressure,low pressure,heart rate,temperature and other relevantparameters collected in the experiment to the background database.Create a management side on the Visual Studio platform to manage users' personal information and health data uploaded to the background.The health data obtained through the management end were trained into the delicate blood glucose photometric data standard by a machine learning algorithm,and then real-time blood glucose reference value could be predicted by the real-time photometric data and related health data.In this paper,a non-invasive Blood Glucose detection model by near-infrared spectroscopy is studied,which increases the accuracy and robustness of the model,and promotes the progress of non-invasive Blood Glucose detection technology to a certain extent.According to the market situation,the software of the Blood Glucose health system based on Android is designed,which makes a series of exploration for the application of non-invasive Blood Glucose test in the market.