Abstract: In order to rapidly identify and predict tea quality, the volatile odor components of Xinyang Maojian tea and the taste components of its infusion were analyzed with an electronic nose and an electronic tongue. Principal component analysis (PCA) indicated that the fusion of the electronic nose and tongue data allowed a better discrimination among tea samples. The fused data were used to develop mathematical models to predict the contents of tea polyphenol and caffeine. Results showed that the regression coefficients of the multiple linear regression (MLR), multivariate linear stepwise regression (MLSR) and quadratic polynomial stepwise regression (QPSR) models were all statistically significant (P < 0.01). The QPSR model presented the best prediction performance among these models. The determination coefficients of calibration and validation of this model were 0.999 and 0.975 for tea polyphenol, and 0.985 and 0.978 for caffeine, respectively; the root mean square error of calibration (RMSEC) were 0.083 and 0.174 for tea polyphenol, and 0.015 and 0.048 for caffeine, respectively. The combination of electronic nose and tongue is feasible to predict the quality and chemical composition of tea.

Key words: electronic nose, electronic tongue, principal component analysis (PCA), chemical composition