The qualitative and quantitative assessment of tea quality based on E-nose, E-tongue and E-eye signals combining with chemometrics methods

Abstract. In this work, electronic nose (E-nose), electronic tongue (E-tongue) and electronic eye (E-eye) were jointly applied as intelligent instruments to acquire aroma, taste and color signals of tea samples. Features were severally extracted from E-nose, E-tongue and E-eye signals and were fused for analysis. The polyphenols, catechins, caffeine and amino acid as quality indices were detected by traditional methods as reference. For qualitative identification, support vector machine (SVM) and random forest (RF) were comparatively employed in modeling severally based on individual and fusion signals. The SVM and RF models based on the fusion signals achieved perfect classification results with the accuracy of 100%. For quantitative prediction of tea quality indices, partial least squares regression (PLSR), SVM and RF were applied based on individual and fusion signals to establish regression models between electronic signals and the amount of polyphenols, catechins, caffeine and amino acid. The RF prediction models reached higher correlation coefficients (R2) and lower root mean square errors (RMSE) than the PLSR and SVM models did. Meanwhile, the fusion signals had a better performance than the individual signals in PLSR, SVM and RF regression models. This work indicated that the simultaneous utilization of E-nose, E-tongue and E-eye based on appropriate chemometrics method could be successfully applied for qualitative and quantitative analysis of tea quality.