Fast discrimination of chocolate varieties using near infrared spectroscopy

Abstract

In this paper we present the use of non-contact near infrared spectroscopy (NIRS) technology employing a diffuse reflection fiber optic probe for discrimination of chocolate varieties. 120 samples of 8 typical varieties of chocolate are selected randomly, and the samples are scanned in diffuse reflectance mode by a cooled InGaAs array spectrometer (950-1700 nm). Partial least squares (PLS) method and support vector machine (SVM) method are used for calibration models development. The calibration models are built according to full spectrum and three separate spectral regions, respectively. The results show that the model of the whole spectral region performs better than those separate spectral region models. The combination of PLS and SVM yields better predictive accuracy than PLS method, and greatly reduces the modeling time compared with the SVM method. So the PLS-SVM method is an effective approach of pattern recognition for mass spectra data. And the NIR spectroscopy with a fiber optic reflection probe has the substantial potential for non-destructive discriminating chocolate varieties.