(29) Two-Dimensional Correlation Analysis of NIR Fractional Derivative Spectra: Correlation with Constituent Concentrations
Jun Uozumi and Yoshika Kawakami
Abstract Book of the 14th International Conference on Near Infrared Spectroscopy, p.72, 2009-11
Derivatives of spectra sometimes play a key role in improving calibration performance of NIR prediction. Although the order of ordinary derivatives is practically limited to one and two, it can be extended to any positive real number by means of the Fourier transform framework, leading to a fractional derivative (FD) and a fractional absolute derivative (FAD), the latter being a peak-shift suppressed version of FD. As the order increases, FD and FAD spectra in the NIR region exhibit continuous and complicated change in their forms. In this paper, generalized two-dimensional (2D) correlation analysis is investigated to analyze interrelation of emerging peaks of FD and/or FAD. Furthermore, a new approach for extracting 2D correlation peaks for individual constituents is proposed and employed.
>>
NEXT