Analysis of Noninvasive Measurement of Human Blood Glucose with ANN-NIR Spectroscopy

The survey by diabetics themselves is the main means to decrease the combined illness caused by diabetes. The current measurement method is harmful or less harmful that brings pains to the patients and also has many insufficiencies in using it. Based on the task of noninvasive blood glucose measure, an outside body and inside body measure amphibious human blood glucose measuring system is designed by us in China which is used to measure and model analyze glucose liquid of different concentration, and in this paper we puts forward a new technology in analyzing the absorption spectrum of infrared ray by blood. After the measurement of the absorption spectrum of infrared by the total blood and normal human serum and higher glucose blood, this paper makes an artificial neural network training through the Levenberg-Marquardt BP neural network which depends on the character parameter in the value of the 16 special wavelength. Only by 11 times the training can match the accuracy of requirement that would meet error demands of the national measurement in biology and chemistry between -0.02~0.03 mmol/L. and the experiment is an advanced one. The research result has valuable promise to blood analysis of the infrared spectroscopy and diagnosing the disease. The experiment proves that this method, compared with the current method, is characterized by rapid training speed and accurate predication result