This paper deals with a synchronous sensing analysis for five tastes namely sweetness, bitterness,sourness, saltiness, astringency by qualitatively and quantitatively analyzing taste stimuli withnonmodified multi-metal sensor, which has the advantage of being rapid, easy and simultaneous. Eightmetal electrodes that were easily available constituted the electrode array. Cyclic voltammetry (CV)and differential pulse voltammetry (DPV) were employed. The original voltammetric signals weretransformed by Continuous Wavelet Transform (CWT) in order to reveal more feature information forsensing taste stimuli. The data of feature points from the transformed signal as the input were used forneural network model. Polar Phasor Approach was used to classify and locate single taste. It wasresulted that Layer-Recurrent neural network could effectively identify the types of mixed tastes. Theresponse areas of eight electrodes for five tastes in oxidation part and reduction part of CV and DPVwere visually characterized. Combining CV with DPV, the accuracies of the training set and test setwere both more than that of single voltammetry, confirming that Back Propagation neural networkcould quantitatively analyze each taste of the mixture. Moreover, the advantages of nonmodifiedsensor compared with modified sensor were discussed. It was concluded that synchronous sensinganalysis has strong ability to sensing the five tastes accurately and simultaneously as human gustation.
展开▼
