模式识别和红外光谱法相结合鉴别中药材产地  

Use of FTIR and Pattern Recognition to Determine Geographical Origins of Chinese Medical Herbs

作　　者：刘沭华[1] 张学工[1] 周群[2] 孙素琴[2] 

机构地区：[1]清华大学自动化系,北京100084 [2]清华大学化学系,北京100084

出　　处：《光谱学与光谱分析》2005年第6期878-881,共4页Spectroscopy and Spectral Analysis

基　　金：国家中医药管理局科技重大项目(国中医药科2001ZDZX01);国家自然科学基金项目(60275007)资助

摘　　要：用红外光谱法取得来自不同产地的中药材的中红外光谱,结合模式识别的方法实现中药产地的快速鉴别。采用了近邻法和多类支持向量机方法对采集自4个不同产地的2 6 9个白芷样本和6个不同产地的380个丹参样本进行了产地鉴别,得到的交叉验证准确率达99% ,为中药材产地的自动鉴别探索了一条有效的途径。Geographical origin of medical herbs is an important factor of the quality of many traditional Chinese herbal medicines. The objective of this study is to investigate whether F-FIR spectroscopy coupled with pattern recognition techniques could effectively discriminate geographical origins of medical herbs. Nearest neighbor method (NNM) and a SVM-based multiclass classifier were employed to discriminate 269 angelicae dahuricae radix (ADR) samples from 4 provinces in China and 380 salviae miltiorrhizae radix (SMR) samples from 6 provinces. A leave-one-out cross-validation accuracy of 99% was achieved by the multiclass classifier. The study shows this classification scheme can be a highly accurate approach for the discrimination of medical herbs of different origins.

关 键 词：傅里叶红外光谱 模式识别 中药材鉴别 支持向量机 

分 类 号：O61[理学—无机化学]