Abstract: Based on the near infrared absorption spectrum analysis of noninvasive blood glucose, this paper puts forward a new technique for t he infrared absorption spectrum analysis of blood. By means of the measurement of in frared absorption spectra of the normal people’s whole blood, serum and high  glucose’s whole blood, application of the artificial neural network BP arithmetic, and with the absorption values of sixteen characteristic wavelengths as the network characteristic parameters, the network training was carried out. Only repeating 11 times, can the network training reach the error precision required. The error between -0.01~0.03 mmol/L fulfils the requirement of national metrological verification regulation on biochemical analysis. This research result shows valuable prospects for blood infrared absorption spectrum analysis and disease diagnoses.

Key words: infrared absorption spectrum, artificial neural network, noninvasive blood glucose measurement