A New Processing Technique for the Identification of Chinese Herbal Medicine

Machine olfaction is widely used to classify and identify the Chinese Herbal Medicine (CHM). The traditional methods for identification were mostly used on the assumption of linear odor data that has variance with the reality. This work adopts a new processing technique of LLE+LDA: using the nonlinear algorithm called Locally Linear Embedding algorithm (LLE) to analyze the high-dimensional nonlinear data of Pungent CHM firstly, then combine with the Linear Discriminant Analysis (LDA) as classifier to complete the identification and classification. The result demonstrates that with this combinatorial theory, the machine olfaction can not only distinguish 6 types of Pungent Chinese Herbal Medicines, but also classify the 3 different production dates of the same kind and the same origin accurately. It provides a new technique for processing the odor data of Pungent CHM based in the machine olfaction.