Premium
Extraction of near infra‐red spectral information by fast fourier transform and principal component analysis. Application to the discrimination of baking quality of wheat flours
Author(s) -
Devaux MarieFrançoise,
Bertrand Dominique,
Robert Paul,
Morat JeanLuc
Publication year - 1987
Publication title -
journal of chemometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.47
H-Index - 92
eISSN - 1099-128X
pISSN - 0886-9383
DOI - 10.1002/cem.1180010205
Subject(s) - principal component analysis , mathematics , fourier transform , standard deviation , linear discriminant analysis , calibration , statistics , pattern recognition (psychology) , near infrared spectroscopy , discriminant , analytical chemistry (journal) , computer science , artificial intelligence , chemistry , optics , chromatography , mathematical analysis , physics
Digitalized continuous near infra‐red reflectance (NIR) spectra are composed of a great number of data which must be reduced for microcomputer mathematical treatment. The sequence ‘fast Fourier transform preceding principal component analysis’ was tested to perform data size reduction without a large loss of information. The method was applied on a collection of wheat spectra composed of 351 data. Ten resulting data, which described 99.5% of the total variance, were kept. The relevance of the method was estimated by the ability of the resulting data (i) to regenerate the original signal, and (ii) to discriminate the baking quality of the wheat by stepwise multiple discriminant analysis. The average difference between initial and regenerated spectra was −2.4 × 10 −3 log (1/ R ) units and the standard deviation was 1.16 × 10 −3 log (1/ R ) units. The discrimination treatments gave 89.9% of well classified samples for the calibration test and 90.5% for the prediction test. The application of these mathematical treatments to other continuous signals is discussed.