SPECTRAL TRANSFORMATION AND WAVELENGTH SELECTION IN NEAR-INFRARED SPECTRA CLASSIFICATION

Near-infrared spectroscopy is used to discriminate between different d osage strengths of tablets in blister packs. Univariate and multivaria te feature selection and three data transformation methods are studied . The second derivative appears to be the most effective transformatio n. Univariate feature selection methods enable the selection of two se lective bands, the difference of which can directly separate all group s. The bipolar axis in the biplots of scores and loadings obtained wit h PCA or PLS also enable the selection of discriminating ratios of wav elengths.