Spices are added in order to enhance the organoleptic characteristics of food and culinary dishes, making them more attractive for consumers. The use of illicit cheap colourants might be profitable along the food supply chain, posing undue risks to human health. This work evaluates the feasibility of NIR spectroscopy with chemometrics as a rapid, simple, non-destructive and affordable screening tool to determine the presence of Sudan I, II, III, IV and Para-red dyes in paprika. The dataset comprised unadulterated and adulterated samples with the five studied dyes at different concentration levels. Several multivariate classification models were built with Linear Discriminant Analysis (LDA) and different machine learning techniques. Preliminary results show that a classifier based on only six wavenumbers is able to determine the presence of some of these dyes in food samples in levels that may represent risk to human health. Sensitivities and specificities above 90% were obtained in almost all cases. These results show the feasibility of inexpensive and portable devices that can be useful for screening out adulterated stock along the food chain supply.
Keywords: NIR; Paprika adulteration; Sudan dyes; few-variables sensing; machine learning analysis; screening methods.