Multivariate Authentication of Herbs and Spices through UV-Vis and FT-IR Fingerprint

The aim of this study was to investigate the applicability of UV-Vis and FT-IR fingerprints combined with multivariate statistical tools to classify and authenticate Iranian standard herbs and spices, and their mislabeled and adulterated samples in single and fusion models. The proposed strategy is an alternative, rapid, easy, and economical approach for herbs and spices authentication. Sixty-three samples of different herbs and spices were collected across several cities in Iran. The potency of Savitzky-Golay(SG) smoothing in combination with autoscaling for improving the accuracy of clustering well studied and principal component analysis (PCA), PCA-linear discriminant analysis (PCA-LDA) and partial least squares-discriminant analysis (PLS-DA) were applied for classification. Additionally, data mining of spectral sets was performed using Kohonen self-organization maps (SOMs) of smoothed and unsmoothed individual data sets and classification results were compared. Also, the discriminant models using fusion matrix were built by concatenation of SG smoothed-autoscaled SOMs clusters of FTIR and UV-Vis (SG-autoscaled-SOMs) spectra. The results of different models showed that the accuracy of single SG-autoscaled-SOMs-FTIR data was better than SG-autoscaled-UV-Vis data and the accuracy of SG-autoscaled-SOMs-fusion technique was better than the other models. This method predicted the class of samples more accurately (more than 95%). The authentication and quality of fraud samples were identified more correctly with respect to raw data.