[Determination of geographical origins of Chinese medical herbs by NIR and pattern recognition]

Geographical origin of medical herbs is an important factor of the quality of many traditional Chinese herbal medicines. The objective of the present study is to investigate whether NIR 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 and 380 Salviae Miltiorrhizae Radix (SMR) samples from 6 provinces in China. The multiclass classifier achieves leave-one-out cross-validation accuracy of 99% for (ADR) and 95% (SMR). This classification scheme can be a highly accurate approach to the rapid and nondestructive discrimination of medical herbs of different origins.