Quantitative Calibration of Near-Infrared Spectra by Wavelet Packet Transform, Orthogonal Signal Correction and Information Entropy Theory

摘要

A new hybrid algorithm (MIE-WPTOSC), which is the combination of wavelet packet transform (WPT), orthogonal signal correction (OSC) and maximum information extraction (MIE), is proposed for interferences elimination in near-infrared (NIR) spectra. In MIE-WPTOSC algorithm, WPT is firstly employed for de-noising by threshold method, and then MIE is applied to remove the baseline of the spectra based on information theory. At last, the information uncorrelated to the concentrations of analyte is eliminated by the OSC in each frequency band of spectra. To validate the effectiveness of the MIE-WPTOSC algorithm, a real NIR spectral dataset of milk was analyzed by different methods for the concentration determination of fat and protein. Experimental results show that the prediction ability and robustness of calibration models developed by MIE-WPTOSC are superior to those developed by either WPT or OSC individually. The root mean square errors of the calibration models for fat and protein can reach up to 0.0832% and 0.0846%, which indicates that the MIE-WPTOSC algorithm is efficient to eliminate the interference information in NIR spectra.