Design And Research Of Electronic Tongue System Based On Microcontroller And Machine Learning

The electronic tongue is electronic detection equipment coupled with sensor array,pattern recognition methods and multivariate statistical methods of qualitative analysis of liquid. It can be used to detect different liquid foods’tastes, and made the different tastes information "visualize". So the electronic tongue is currently new hot spot in the field of food testing.The main purpose of this study is to explore designing a set of cheap and stable electronic tongue system, and at the same time through a series of experiments and data analysis to verify the feasibility of this system. The subject has a great significance for the rapid non-destructive testing of food research.In this subject we independently developed a set of electronic tongue system, mainly consists of (1) ion selective electrode sensor array,(2) the data acquisition module,(3) the data acquisition module,(4) the data analysis module, the above total of four parts. And we the developed a set of software, in order to control the electronic tongue system, and analysis the data collected by electronic tongue or pattern recognition.This paper mainly studied the following contents:(1.) Introduced some common electronic tongue sensor, and the general structure of it.(2) Designed a small signal amplification circuit,and the electronic tongue sensor signal filter and amplification circuit, and a piece of modulus signal conversion interface card which can collected sensor signal to the PC(3) Compiled a windows program, achieved data transmission and human-computer interaction based on VC++platform.(4) In the MATLAB7.11environment, writed a set of data processing program, including PCA, LDA, and BP, PNN, RBF neural network and its parameters optimization tools, based on Lib-SVM toolbox of pattern recognition software, based on the two dimensional search parameters optimization tools and SVM based on GAOT genetic algorithm toolbox of SVM parameters optimization tools also included an pattern recognition software based on LS-SVM. We Writted BP neural network electronic tongue data processing program by ANSI C and verified it. Finally I integrated the above software as original ET GUI TEST electronic tongue data processing software.I applicated the electronic tongue system testind four different brands of mineral water and five different kinds of apple juice, processed the data with data analysis and pattern recognition, the results showed that the electronic tongue system can effectively distinguish the type of apple juice and mineral water.Due to these tests, it has been proved that the electronic tongue prototype is feasible and effective.And we make the outlook for the future development of the hardware design of the electronic tongue and a more advanced design of a new electronic tongue.