A Principal Curves-Based Method for Electronic Tongue Data Analysis

Abstract

Electronic tongues are a type of sensor inspired in the biological recognition system, where sensorial and instrumental techniques are used to determine flavors or substances present in the samples analyzed. The sensor arrays used in electronic tongues are generally non-specific and are designed to provide global information about the solution's response. As a consequence, it usually becomes necessary to use data processing techniques that are capable of providing specific information from the samples under investigation. In this context, here we employed a Principal Curves based method to evaluate the experimental data obtained from an impedimetric electronic tongue used to evaluated distinct flavors enhancers of similar compostions. For the classification of concentrations, the best results were in the range of 88.02 to 91.15%, using electrode architectures E1 (Poly allylamine hydrochloride (PAH), reduced graphene oxide (rGO), polyaniline (PANI) and copper tetrasulfonated phthalocyanine (CuTsPc)) and E6 (PAH and CuTsPc). In the classification of substances, again the E1 architecture was highlighted, with 90.15% accuracy. These results shown that principal curves may be a promising technique for electronic tongues data analysis.