Discrimination of arabica natural, arabica washed and robusta coffee through NIR spectroscopy and artificial tongue analysis

Food authentication is a still challenging issue regarding food quality control and safety. Identification of coffee variety has gained increasing attention as a means to control and avoid coffee adulteration, mainly considering the great variability of the final sale price depending on coffee varietal or geographic origin. Most commercially available coffees are produced from Arabica and Robusta beans or blends of these two species. Both varieties differ not only in relation to their botanical, chemical and organoleptic characteristics, but also in terms of commercial value, with Arabica coffees usually achieving market prices 20–25% higher than Robusta. Therefore, there is a clear need for suitable analytical methods to differentiate between these two coffee types. Near infrared spectroscopy (NIRS) combined with chemometrics, can reply to this demand representing a fast, clean and inexpensive technology. This methodology has emerged in the last years as a very promising non-destructive alternative method for constructing, on the basis of spectral features and in combination with pattern recognition methods, reliable classification models for assessing the quality of a given product in many food applications, including coffee varietal authentication [1,2]. Electronic tongue is a technological attempt to mimic human taste. This device consists of chemical sensor arrays, coupled with an appropriate pattern recognition system able to interpret complex signals from such sensor and produce a fingerprint of the product. Different approaches can be recognizing tastes and giving information about groups of substances present in complex liquid systems [3, 4]. In this study, near infrared spectroscopy (NIRS) and electronic tongue analysis, combined with technical computing, has been used to discriminate between Arabica and Robusta types of coffee and between two categories of Arabica samples. Suitable pre-processing methods were applied aiming at correcting spectral data by minimizing the contribution of physical effects to NIR signals and thus enhancing the relevant chemical information contained. The multivariate analysis applied on both kind of data dealt with PCA, used to get an overview of multivariate data, and then with linear discriminant analysis (LDA), to achieve a classification model of coffee samples on the basis of the species and of the different treatments for Arabica samples. Both these techniques represent a valid alternative to usual expensive analytical methods for their rapidity and simplicity; furthermore, they are advantageous in terms of environment protection and safety, since no chemicals are needed and no hazards are associated with this kind of determinations.