Improvement on the Identification and Discrimination Ability for Rice of Electronic Tongue Multi-Sensor Array Based on Information Entropy

An approach for improving the identification and discrimination ability of electronic tongue multi-sensor array was developed. The detail signal was obtained by decomposing the original voltammetric signal using wavelet packet decomposition, and the feature value was extracted by Fast Fourier transform in which the influence of collinearity was eliminated. Based on the principle of information entropy, the sensing entropy of single electrode and between electrodes in the multi-sensor array were defined, and the unit sensing vector and interactive sensing vector were constructed. The results showed that 6 unit sensing entropies could be effectively used for the identification of rice origin, and all interactive sensing vectors for the discrimination of rice type. SVM and KNN classifiers were employed. The results showed that the training and prediction accuracy of SVM with interactive sensing vector as the input for identifying rice origin were 89.0% and 82.9% respectively, and that for distinguishing rice type were 96.0% and 88.6% respectively. In conclusion, the SVM model with interactive sensing vector could be an approach to accurately identify rice origin and distinguish rice type. The identification and discrimination ability of multi-sensor array could be enhanced by using the sensing interaction information based on information entropy. (C) 2022 The Electrochemical Society ("ECS"). Published on behalf of ECS by IOP Publishing Limited.