|Liu, Yongliang - UNIV OF GEORGIA|
|Barton Ii, Franklin|
Submitted to: Near Infrared Spectroscopy Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: December 15, 2003
Publication Date: January 15, 2004
Citation: Liu, Y., Barton II, F.E., Lyon, B.G., Chen, Y.R. 2004. Variations of large spectral set: two-dimensional correlation analysis of loadings spectra of principal component analysis. Journal of Near Infrared Spectroscopy. 2003. Vol. 11. Iss. 6. P. 457-466. Interpretive Summary: Generalized two-dimensional (2D) correlation spectroscopy has been established as a viable means to analyze and extract useful information from conventional one-dimensional spectral data. It has been considerably and successfully applied not only for a variety of optical spectroscopic techniques (IR, NIR, Raman, visible, fluorescence), but also for a number of different types of external perturbations (thermal, mechanical, chemical, spatial position, etc) and waveforms. However, most of its applications are limited to a small number of spectral data affected by simple and univariate static perturbations. Such an entire data set or smaller subset could accentuate the delicate feature and yield additional information of spectral intensity changes occurring around these particular ranges. On the other hand, one of specific challenges might be how to implement the 2D correlation analysis in much large spectral sets, which are common in chemometric model developments and actually include diverse fluctuations in chemical and physical components, and to acquire useful information from them. Here, we present an alternative method to explore the variations within the large spectral set by analyzing the loadings spectra of principal component analysis (PCA). Visible/near-infrared (NIR) spectra of chicken breast muscles under a variety of treatments were selected as examples, because muscles are one of the complicated agricultural commodities that vary greatly in color, chemical, physical and sensory attributes from one portion to another. Researchers working on visible/NIR spectroscopy and subsequent model developments will benefit from the findings of this research.
Technical Abstract: This work attempted to interpret the principle component loadings spectra of principle component analysis on large spectral data set with multi-variables by the use of two-dimensional (2D) correlation analysis. Three examples of visible/NIR spectra of chicken muscles under different conditions were given and discussed. 2D analysis indicated that characteristic bands from loadings spectra are in good agreement with those from a small number of spectra induced by simple external perturbations. Although some advantages of 2D correlation analysis (such as sequential changes in intensity) were not available, it might still be useful for the understanding of large and complex spectral data set with multi component variations.