Near-infrared region of optical spectrum extends from 0.7 to 2.5μm and can be used for quantitative measurement of organic functional groups, especially C-H, O-H, N-H and C=O. Analyzing sample concentration by near infrared spectroscopy focuses on the design of the optical sensor and the spectral data processing. Noninvasive blood glucose measurement methods using near-infrared spectroscopy usually apply a beam of light to irradiate the blood region of human, and then extract the information of blood glucose from the spectrum. The key is to improve the signal to noise ratio so that very low glucose absorption can be detected. In this paper, according to the absorption of glucose, a noninvasive blood glucose measurement system based on three wavelengths in the near-infrared region was designed. The system included several important parts such as the light source, the optical chopper, the detector and the lock-in amplifier. The three wavelengths were respectively chosen at the signal wavelength 1610nm by glucose absorption peak in the overtone band and the reference wavelengths 1200/1350nm to eliminate the interference effect. The optical probe used an annular light bundle to greatly increase the intensity of incidence light and improve the signal-noise ratio. Two group experiments of glucose aqueous solutions with different concentration interval in the normal human physiological blood glucose range (0-500mg/dL) have been done to evaluate the predictive performance of the system. In these group experiments, the partial least square algorithm was used to predict the glucose concentration. The preliminary results showed when the interval was 100mg/dL, the correlation coefficient (R) was 0.998 and the root mean square error of prediction (RMSEP) was 17.08mg/dL; and when the interval was 20mg/dL, the values of R and RMSEP were 0.959 and 23.22mg/dL, respectively.