Classification of honeys of different floral origins by artificial neural networks

In this study, the data obtained from a potentiometric electronic tongue system are analyzed using a Fuzzy-Artmap artificial neural network. The analysis is focused on determining the optimal configuration of the neural network in order to achieve the proper classification of the honey analysed, as well as to adjust its size for further implementation in a microcontroller system. The application submitted has been employed to discriminate several varieties of honey according to the types of floral origin (citrus, rosemary, polyfloral and honeydew) and considering three different physical treatment: raw, liquefied and pasteurized. The measures have been carried out using a set of metal electrodes of different materials. We have used eight electrodes of various metals (pure and chemically treated). In order to implement the neural networks, a Graphic User Interface (GUI) in Matlab has been developed, fixing thereby the parameters of neural networks and obtaining the classification of the samples. Moreover, the information needed to implement neural networks in microcontroller systems is also generated to ensure an autonomous system of electronic tongue applied to the analysis of honey..