Nowadays, coffee has a broad highlight in the international commerce and its qualities are certified according to the Global Quality Program developed by specialized coffee agencies. Expert tasters provide a score classifying the coffees as Gourmet, Superior, Traditional and Inferior. However, due to relatively high-cost, possibility of misclassification, difficulty of large-scale analysis, and mainly to the subjectivity generated by the tasters, the international commerce requires a fast and reliable analysis. Fourier Transform Infrared Spectroscopy (FTIR) is a widely disseminated technique in laboratories routine for direct analysis besides successfully applied to measure the coffee quality. In the present work, coffee FTIR spectra were assisted by chemometric tools based on pattern recognition such as principal component analysis (PCA) and linear discriminant analysis (LDA) to find a suitable classification model for authentication purposes. Spectral predictors did not favor the discrimination of coffee samples. LDA performance was greatly improved using the PCs obtained from PCA as input variables. Several models were fitted from these PCs, so that the optimal classifier achieved 95 % accuracy in the overall predictions. This model also showed sensitivities in the range 83–100 % and specificities ranged from 93 to 100 %, indicating that the tuned model is trustworthy to successfully evaluate the commercial coffee authenticity. The developed method is simple, rapid, and environmentally friendly, since not demanding any sample preparation and does not producing waste after analysis. Its application can help to protect coffee market and companies by product quality certification.

如今，咖啡在国际贸易中具有广泛的亮点，其质量根据专业咖啡机构制定的全球质量计划进行认证。专业品酒师提供一个评分，将咖啡分为美食、上等、传统和下等。然而，由于成本相对较高，容易误分类，难以大规模分析，主要是品尝者产生的主观性，国际贸易需要快速可靠的分析。傅里叶变换红外光谱 (FTIR) 是实验室常规中广泛传播的技术，除了成功应用于测量咖啡质量外，还用于直接分析。在目前的工作中，咖啡 FTIR 光谱由基于模式识别的化学计量工具辅助，例如主成分分析 (PCA) 和线性判别分析 (LDA)，以找到适合身份验证的分类模型。光谱预测器不利于咖啡样品的区分。使用从 PCA 获得的 PC 作为输入变量大大提高了 LDA 性能。从这些 PC 中拟合了几个模型，因此最佳分类器在整体预测中达到了 95% 的准确率。该模型还显示出 83-100% 范围内的灵敏度和 93-100% 范围内的特异性，表明调整后的模型是值得信赖的，可以成功评估商业咖啡的真实性。所开发的方法简单、快速、环保，因为不需要任何样品制备，分析后不产生废物。它的应用可以通过产品质量认证帮助保护咖啡市场和公司。