谱图数据融合结合模式识别算法鉴别苹果香精

采用拉曼光谱-离子迁移谱（ion mobility spectrometry，IMS）数据融合技术结合主成分分析（principal components analysis，PCA）-最近邻（nearest neighbor，NN）算法的模型鉴别9 种食用苹果香精。香精先经水溶液稀释处理，再经拉曼光谱和IMS分析，建立样品的拉曼光谱和IMS指纹图谱库，然后分别使用单谱数据结合PCA-NN模型以及拉曼光谱-IMS数据融合结合PCA-NN模型鉴别香精。结果表明：拉曼光谱-IMS结合PCA-NN模型对9 种食用苹果香精的识别率达98.35%，高于拉曼光谱的78.14%和IMS的94.18%。使用水溶液稀释技术，不存在副反应，无污染，操作简单快速，保留了样品的整体物质，保证了实验结果的可靠性和稳定性。拉曼光谱仪和离子迁移谱仪具有操作简单、分析速度快的优点。拉曼光谱-IMS结合PCA-NN模型为鉴别食用苹果香精提供了一种可靠、稳定、快速、全新的方法。

Discrimination of Apple Essences Based on Spectral Data Fusion Combined with Pattern Recognition Algorithm

In this paper, Raman spectroscopy and ion mobility spectrometry (IMS) were used in combination to characterize nine kinds of apple essences from different producers. A discrimination model was built using spectral data fusion combined with principal component analysis (PCA) and nearest neighbor (NN) algorithm. First, apple essences were diluted with ultra-pure water. Then, Raman and IMS fingerprint databases were established by Raman and IMS analyses, respectively. The single spectral data combined with PCA-NN models and the data fusion of Raman and IMS combined with PCA-NN model were used to distinguish apple essences, respectively. It was shown that the identification accuracy rate of the Raman-IMS combined with PCA-NN model for nine kinds of apple essences was 98.35%, which was higher than that of the Raman spectra combined with PCA-NN model (94.18%) and the IMS spectra combined with PCA-NN model (78.14%). Aqueous dilution technique was simple and fast, caused neither side effect nor pollution, and could retain all substances in the sample, ensuring the reliability and stability of the experimental results. Both Raman and IMS had the advantages of easy operation and quick analysis. The results from this study demonstrated that the Raman-IMS combined with PCA-NN model can be used as a reliable, stable and fast new method to discriminate among apple essences.