Commonly Used In Herbal Medicine Near Infrared Spectroscopy Database And Identify The Research

Near-infrared spectroscopy (NIRS) is a kind of rapid, non-destructive and objective method for the quality monitoring and control of Chinese herbal medicines. As the indirect analysis techniques, NIRS technique need large amount samples, their near-infrared spectral data and predictive mathematical model. Therefore, it is necessary to establish the sample spectra databases and to explore suitable mathematical modeling methods for Chinese medicine quality control.In the present paper, NIRS database is established to identify and study the decoction pieces in common use. The Chinese herbal medicine samples were gathered for NIRS testing. Before experiments, the samples were identified by experienced apothecaries.The NIRS database of common used decoction pieces was establihed in the ISIS/Base Manage System (American MDL Information System Company). NIR Similarity-based identification approach are established using TQ spectra analysis software to make prediction on the unknown sample. Chemometrics analysis such as principal component analysis (PCA) , radial basis function (RBF) neural network, support vector machine (SVM) are explored to find the suitable modeling techniques. The main results obtained is as follows.1. Establishment of the NIRS database of traditional Chinese HerbsBased on the ISIS/Base Manage System (American MDL Information System Company), the NIRS database of decoction pieces in common use was established. 152 kinds of decoction pieces were included. Five samples for each kind decoction pieces and five spectrograms for each sample were included in the database. The number of samples is up to 760 and the number of spectrograms is up to 3800 spectrograms. The database can be used for storage, searching and management of the spectrogram and text. It is convenient for identification of decoction pieces.2.Qualitative analysis on gathered NIRSUsing multiple scattering correction (Multiplicative Scater Correction MSC), Norris derivative smoothing and a derivative, 152 similarity recognition model of decoction pieces were established. Using robust principal component analysis (RPCA), the near infrared spectroscopy of licorice and roast-licorice was analyzed and the chart of PC1 and PC2 was given. The result indicated that licorice and roast-licorice can be distinguished by this method. In addition, the recognition of licorice and roast-licorice are explored using RBF neural network. The initial spectral data are smoothed and zipped by wavelet algorithm to reduce the data dimensions. The key parameters in RBF neural network are discussed. The results are satisfying. The algorithm of Support Vector Machines (SVM) with RBF kernel was used for the recognition of licorice and roast-licorice based on NIRS data. The parameters in SVM are discussed. The accuracy rate of recognition is 100%。The results indicate that NIRS combined with chemometrics is a fast, non-destructive and effective analytical method for quality control of Chinese Herbal Medicine.