Extended multiplicative signal correction and spectral interference subtraction: new preprocessing methods for near infrared spectroscopy

Abstract

Near infrared (NIR) spectroscopy spectra can be converted mathematically to precise quantitative information of chemical and physical nature by multivariate calibration. This makes NIR analysis useful for a variety of "difficult" sample types (powders, slurries), more or less without any sample preparation. The paper emphasizes the importance of using prior knowledge for spectral preprocessing of spectral data prior to the linear multivariate calibration modelling. Two new preprocessing methods are presented: extended multiplicative signal correction (EMSC) for elimination of uncontrollable path length or scattering effects, and spectral interference subtraction (SIS) for elimination of known spectral interferences. Determination of toluene in mixtures with benzene and xylene from NIR spectra with gross simulated light scattering effects is used for illustration.