为实现白酒质量的快速鉴别，本研究采用两个不同的电子鼻和电子舌融合系统采集不同品牌白酒样品的气-味信息，采用主成分分析法、K均值法对检测结果进行聚类分析，采用支持向量机法对白酒的品牌进行预测分类分析。经主成分分析聚类分析后，应用基于TGS型气敏传感器的电子鼻和电子舌融合系统可以将3种酒进行很好的区分，其余5种酒有交叉，而应用基于MQ、MP型气敏传感器的电子鼻和电子舌融合系统可以将8种白酒基本区分开来；K均值法分别用于两融合系统，前融合系统的错分类概率为33.3%，后融合系统的错分类概率为23.75%。采用支持向量机法对白酒的品牌进行预测分类，应用前融合系统的识别率为93.75%，而应用后融合系统的识别率为98.75%。结果表明：气-味信息融合技术可以实现对白酒品牌的鉴别，且应用基于MQ、MP型气敏传感器的电子鼻和电子舌融合系统对白酒类别的识别结果较好于应用基于TGS型气敏传感器的电子鼻和电子舌融合系统。

Two different fusion systems of electronic nose and tongue were used to collect data on aroma and taste of Chinese liquors of different brands: (A) a fusion system of electronic tongue and nose, based on Taguchi gas sensors (TGS); (B) a fusion system of electronic tongue and nose, based on MQ-MP-type gas sensors. Clustering analysis of the aroma and taste was performed using principal component analysis (PCA) and the k-means algorithm. Next, prediction and classification of the different brands of Chinese liquors were performed using a support vector machine (SVM). After clustering analysis by PCA, fusion system A could distinguish three brands of Chinese liquors, while the rest of the brands showed overlapping results. Fusion system B could distinguish eight brands of Chinese liquors. The k-means algorithm was applied to the two fusion systems, and the wrong classification rates of fusion systems A and B were found to be 33.3% and 23.75%, respectively. The accuracy of prediction and classification of the Chinese liquors by the fusion systems A and B for was 93.75% and 98.75%, respectively. The study showed that the combination of information on aroma and taste could be used for identification of different brands of Chinese liquors. Additionally, identification by the fusion system B was better than that by the fusion system A.

国家自然科学基金资助项目（31041569）；吉林省科技发展计划资助项目（20130101053JC；20140204020GX）