Skip to main content
ARS Home » Research » Publications at this Location » Publication #182640
Title: NIR-FT/RAMAN SPECTROSCOPY AND NIR SPECTROSCOPY FOR NUTRITIONAL CLASSIFICATION OF CEREAL FOODS

Author
item SOHN, MI RYEONG
item KAYS, SANDRA
item HIMMELSBACH, DAVID
item BARTON II, FRANKLIN

Submitted to: Near Infrared Spectroscopy International Conference Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 7/30/2005
Publication Date: 3/1/2007
Citation: Sohn, M., Kays, S.E., Himmelsbach, D.S., Barton Ii, F.E. 2007. Nir-ft/raman spectroscopy and nir spectroscopy for nutritional classification of cereal foods, pp. 674-678. Near Infrared Spectroscopy: Proceedings of the 12th International Conference, Auckland, New Zealand 9-15 April 2005, New Zealand Near Infrared spectroscopy Society Inc., Hamilton, New Zealand.

Interpretive Summary: Two spectroscopic methods of FT-Raman and near-infrared (NIR) were used to investigate the possibility of classification of cereal samples into two classes, high level and low level in protein and fat, respectively. The spectroscopic methods provide, within a five-minute total sample preparation and analysis time, a rapid means of classification for additional (or new) products based on their spectra as a substitute for time-consuming laboratory assays, about 1~2 days, for protein and fat. The information obtained in this study could be useful to food processing quality control and to groups and agencies concerned with product labeling of the final products.

Technical Abstract: In this study, we investigated a potential of FT-Raman spectroscopy as a method for cereal food classification and the result was compared to the NIR. Particularly, we focused on the problem of improving the performance of linear calibration algorithms in the presence of high non linearity. To improve the classification accuracy, we selected the frequency region that is related to only target component and used various preprocessing of the spectral data. FT-Raman and NIR spectroscopy have the potential to classify cereal foods into two classes according to their percent level. For Raman spectroscopy, the use of the selected x-variables for each component improved the classification result, and PLS-based classification was slightly better than SIMCA. For NIR spectroscopy, the classification result was affected by preprocessing of the spectra, and also PLS was better than SIMCA for both components. FT-Raman and NIR have the same accuracy for fat, but Raman was slightly better than NIR for protein.