Evaluation of aroma quality using multidimensional olfactory information during black tea fermentation

Thus far, the intelligent evaluation of aroma quality during black tea fermentation remains an unsolved problem due to the hysteresis quality of traditional sensory evaluation methods. In our study, a combination of hyperspectral imaging technology and colorimetric sensing array (CSA) was used to collect the aroma information during black tea fermentation. Subsequently, different data fusion strategies coupled with the support vector regression (SVR) model were used to predict the aroma scores of finished tea at different fermentation times. The performance of the prediction model using data fusion strategies was better than that using each sensitive dye. The results demonstrated that the middle-level-competitive adaptive reweighted sampling (CARS) strategy showed the best performance, with the correlation coefficient of the prediction set (Rp) at 0.969, the relative percent deviation (RPD) at 4.091, and the variable compression rate at 96.83%. Based on the middle-level-CARS strategy, the discrimination rate of aroma quality for calibration and prediction set were 100% and 94.29%, respectively. The overall results sufficiently revealed that our proposed strategy provides a theoretical basis for the intelligent processing of black tea.