ABSTRACT
Spectroscopic prediction of quantitative mixture composition is based on comparing a mixture spectrum with a set of standard spectra. The simple linear model can also be used to analyze multicomponent mixtures if spectral areas of individual components are sufficiently distinct and if spectroscopic signals of components do not overlap. The multicomponent linear model is a generalization of the single-component model. It is supposed that the spectroscopic responses of individual components are additionally summed into a spectrum of a multicomponent system. The method of partial least squares also known as partial least squares regression, also involves the matrix of calibration standards in the factor decomposition of a calibration spectra matrix. Kalman filtering can be used to determine the number of components present in a mixture, as was shown by Didden and Poulisse in their examples of numerically simulated spectra.
