Submitted to: International Conference on Fourier Transform Spectroscopy Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: June 4, 1999
Publication Date: N/A
Technical Abstract: Near-infrared Fourier transform Raman (NIR-FT/Raman) spectroscopy was investigated to provide a rigorous spectroscopic method by which to analyze rice flour for protein and apparent amylose content using a chemometric approach. Ninety rice samples from a 1996 collection of short, medium and long grain rices (grown in 4 states of the United States, Taiwan, Korea and dAustralia) were investigated. Milled rice flour samples were scanned in rotating cups with 1064 nm NIR laser at 500mW of power and Raman scatter was collected with a Ge (LN2) detector. Only Raman data over the shift range of 200-3600 cm-1 was utilized. This spectral data was preprocessed by either 1st derivative or baseline correction with normalization. Nearly equivalent results were obtained using either preprocessing method with partial least squares type-1(PLS1) chemometric models. However, models using baseline correction were slightly better and thus were selected for use. The model for protein (n=86, with 1 spectral and 3 chemical outliers removed) using 7 factors gave r2=0.988 with SECV=0.157% and bias= -0.0003%, over a range of 5.05-10.35 %, (Figure 1). The model for apparent amylose (n=86, with 4 chemical outliers removed) using 12 factors gave r2=0.982 with SECV=1.09% and bias=0.026%, over a range of 0.50-24.94% (Figure 2). The results obtained for prediction of protein and apparent amylose content by Raman spectroscopy were nearly statistically equivalent to those obtainable by NIR spectroscopy but were not affected by sample moisture content.