Comparision of Several Preprocessing Algorithms based on Near Infrared Spectroscopic Measurement of Glucose in Aqueous Glucose Solutions

Abstract

Glucose concentration measurement is the basis of noninvasive detection of blood glucose concentration. It is significant in scientific research. In this study, Near Infrared Spectrum (NIRS) and regression analysis methodology were combined to measure the glucose concentration. The spectrum of glucose solutions was obtained with the Fourier Transformed Infrared Spectrometer, and then the data was used for regression analysis. In addition, the method of Paritial Least Squares (PLS) was used to achieve principle components and various spectral preprocessing methods were discussed. During PLS modeling, the Savitzky-Golay could improve the Prediction Residual Error Sum of Squares (PRESS) within 6%. The experiment results demonstrate that NIRS has the potential for the measurement of glucose solution.

DOI : http://dx.doi.org/10.11591/telkomnika.v12i4.4060