Dong Wenjiang, Zhao Jianping, Hu Rongsuo, Dong Yunping, Tan Lehe
Spice and Beverage Research Institute, Chinese Academy of Tropical Agricultural Sciences, Wanning, Hainan 571533, China; National Center of Important Tropical Crops Engineering and Technology Research, Wanning, Hainan 571533, China; Key Laboratory of Genetic Resources Utilization of Spice and Beverage Crops, Ministry of Agriculture, Wanning, Hainan 571533, China.
Spice and Beverage Research Institute, Chinese Academy of Tropical Agricultural Sciences, Wanning, Hainan 571533, China; National Center of Important Tropical Crops Engineering and Technology Research, Wanning, Hainan 571533, China.
Food Chem. 2017 Aug 15;229:743-751. doi: 10.1016/j.foodchem.2017.02.149. Epub 2017 Mar 2.
Electronic nose and tongue sensors and chemometric multivariate analysis were applied to characterize and classify 7 Chinese robusta coffee cultivars with different roasting degrees. Analytical data were obtained from 126 samples of roasted coffee beans distributed in the Hainan Province of China. Physicochemical qualities, such as the pH, titratable acidity (TA), total soluble solids (TSS), total solids (TS), and TSS/TA ratio, were determined by wet chemistry methods. Data fusion strategies were investigated to improve the performance of models relative to the performance of a single technique. Clear classification of all the studied coffee samples was achieved by principal component analysis, K-nearest neighbour analysis, partial least squares discriminant analysis, and a back-propagation artificial neural network. Quantitative models were established between the sensor responses and the reference physicochemical qualities, using partial least squares regression (PLSR). The PLSR model with a fusion data set was considered the best model for determining the quality parameters.
电子鼻和舌传感器以及化学计量多变量分析被用于对7个不同烘焙程度的中国罗布斯塔咖啡品种进行表征和分类。分析数据来自分布在中国海南省的126个烘焙咖啡豆样本。通过湿化学方法测定了诸如pH值、可滴定酸度（TA）、总可溶性固形物（TSS）、总固形物（TS）以及TSS/TA比值等理化性质。研究了数据融合策略以提高模型相对于单一技术性能的表现。通过主成分分析、K近邻分析、偏最小二乘判别分析以及反向传播人工神经网络实现了对所有研究咖啡样本的清晰分类。使用偏最小二乘回归（PLSR）建立了传感器响应与参考理化性质之间的定量模型。具有融合数据集的PLSR模型被认为是用于确定质量参数的最佳模型。
