The non-invasive blood glucose measurement method based on near-infraredspectrum is one of the most promising non-invasive blood glucose detection methodat moment. In this paper, we constructed a near infrared glucose measurement systembased on Floating reference theory. By analyzing the characteristics of the floatingreference point existence position, we determined the detection position at allwavelengths. Then we evaluated every module of the system one by one andevaluated of the system as a whole. At last, we explored a set of possible options ofthe blood glucose prediction system which is based on the floating theory reference byexperiment. The paper includes the following four parts:Firstly, by studying the existence principle of the floating reference point viaMonte Carlo simulations, we determine the variation of the optical parameters. In thewavelength range from1200nm to1700nm, we simulated the5%intralipid solution atthe different glucose concentrations which are50mg/dL,100mg/dL and150mg/dL,to determine the existence position of the floating reference point changes in thissolution. Finally we designed the fiber probe according to this result.According to the theory of AOTF diffraction, we proposed a system-wide design.In view of AOTF spectral principle, we designed diffraction module. Finally, weremoved the effect of0level light through the orthogonal polarization method.According to near infrared spectral characteristic, we select the photoelectric detectorand operational amplifier to design circuit system. Finally, we complete dataacquisition part by LABVIEW data acquisition module.In evaluation stage, we verified the good sensitivity and stability of the AOTFcrystal for the input frequency firstly. Then we demarcated the relationship betweenthe driver frequency and the diffraction wavelength by using Fourier spectrometer andverify the calibration results by Ocean Optics spectrometer. The average deviation is6.1nm, which is less than the fluctuation range of the crystal diffraction wavelength.In dark background collecting4000data points to evaluate the data acquisition card’sstability, the coefficient of variation is7.0847E-06, who is meet the blood glucosemeter accuracy requirements. Also in dark background, we evaluated the stability ofphotoelectric conversion output voltage, the drift of the measurement path andreference path are2.3%and1.1%in120minutes. Then we analyzed the relationship between the change of the detector voltage output and the change of the driver’svoltage. To verify the system signal-to-noise ratio, the diffuse reflection of thestandard reflector was collected. The short-term coefficients of variation in referencepath measurement path in six minutes both are0.0004, and the long-term coefficientsof variation of the two paths in eight hour are about0.012.Take5%intralipid solution as a human tissue analog solution, concentration ofsugar from100mg/dL to2000mg/dL, take experiment to get spectral information.Use spectral correction method which is based on floating reference point to modifiedspectrum. Build mathematical model according to the Partial Least Squares to predictthe concentration of sugar and evaluate the system availability. The RMSEP andRMSEC in the model are95.4225and85.5487mg/dL, and the correlationcoefficient R is0.9889.The evaluation of the data shows that the system stability in the two hours afterbooting the system is in the best condition, and system signal-to-noise ratio is lower ineight hours, select a the data generated by specific time period has more importantsignificance for the future in non-invasive blood glucose. |