基于电子舌和多元数据分析的咖啡焙炒度检测方法研究

Method of Determination on Roasting Degree of Coffee Samples Combined with Electronic Tongue and Multivariate Data Analysis

利用电子舌技术结合多元数据分析对不同焙炒度(浅度、中度、深度)的咖啡豆进行区分。原始电子感官数据经归一化处理后,采用主成分分析(PCA)对其进行解析,结果表明:不同焙炒度的咖啡样品基本能够按各自特性聚为一类,扩展正则变量分析(ECVA)对样品的分类结果与PCA解析后的结果一致;比较不同的有监督模式识别方法:K-最近邻法(KNN)、偏最小二乘判别分析(PLS-DA)和最小二乘-支持向量机(LS-SVM)所建立模型对未知样品的预报能力,其中LS-SVM模型的预报结果较好,其识别率和预报率均为100%。

An electronic tongue technology combined with multivariate data analysis was utilized to classify coffee samples from different roasting degree（light, medium, dark）. Principal component analysis（PCA）was performed to the normalized data matrix to explore the space distribution of all samples. Results showed that samples could be clustered according to the respective properties, and similar results were obtained by extended canonical variates analysis（ECVA）. Different pattern recognition techniques, such as K-nearest neighbors（KNN）, partial least squares- discriminant analysis（PLS-DA）and least squares-support vector maehines（LS-SVM）were used to construct calibration models to compare the performance. The recognition rate and prediction rate were all 100% for LS-SVM model.