Abstract
This paper introduces a novel pre-processing method based on optimizing the window size for the Savitzky-Golay smoothing coupled with bandpass filtering to further enhance the prediction performance of the of glucose concentration from both Near Infrared and Mid Infrared spectra. The proposed method is compared to the bandpass filtering with Savitzky-Golay using fixed window size and RReliefF pre-processing technique for further evaluation. The developed prediction models have been validated to predict the concentration of the glucose from both Near and Mid Infrared spectra of a mixture of glucose and human serum albumin in a phosphate buffer solution. The results confirm that the proposed technique enhance prediction performance of the linear calibration models the Principal Component Regression and the Partial Least Squares Regression models and achieve better results than the bandpass filtering with Savitzky-Golay with fixed window size technique.