Classification And Identification Of Chinese Herbal Medicine By Data Fusion Based On Machine Olfactory And Gustatory

Chinese herbal medicines (CHMs) are the material base of disease prevention and treatment in the history of Traditional Chinese medicine (TCM), they have a profound cultural in thousands of years. TCM formed its unique theory system in the long-term of the medical practice, it has a small side-effect and excellent clinical effects. Chinese medicine resources are extremely rich in China, But the kind of medicinal materials is various and complicated, adulterations are always pretend to be precious and rare medicinal materials. Therefore, the quality control of CHMs has been concerned by more and more researchers. This paper proposed a new method of CHMs classification and Identification, using information fusion based on machine olfactory and gustatory identify.We select several typical CHMs as the research object, and use the electronic nose (E-nose) and Electronic Tongue (E-tongue) technology to collect the data of gas and taste of the CHMs with different varieties, growing areas, or production dates. First, only use the E-tongue technology to implement the classification of CHMs. Then, using the theory of data fusion get the multi-sensor data together, and make feature extraction and dimension reduction processing of the fusion data. Finally, choose the pattern recognition method of PCA+LDA to implement the training of samples and the classification of testing samples. This paper contrast the different effects between E-nose, E-tongue and together with them, the result is that using information fusion method with gas and taste based on machine olfactory and gustatory is more effective in CHMs classification and identification.According to the research target, we select six typical CHMs as the research object. They are White cardamomum, Atractylodes rhizome, Atractylodes lancea, Heracleum, Alpinia oxyphylla, and Turmeric root-tuber. Use the method of data fusion to make the classification and identification of samples. The results are as follows:1) Choosing PEN3E-nose and ASTREE E-tongue, which are made by AIRSENSE company in Germany and Alpha. MOS company in France, to get the sensors data of gas and taste. Then, using the pattern recognition methods (LDA, PCA+LDA), to get the recognition results from machine olfactory and machine gustatory.2) We used the information fusion theory of layers of data fusion to fuse the data of gas and taste collected. Then make the feature extraction and selection. At last, use the method of PCA to implement Dimension reduction processing, and use the LDA method to classify and identify the samples.3) Compared and analyzed the differences results from electronic nose, electronic tongue and the fusion technology with both of them. The two-dimension graphs of discriminant results clearly show that the classification result from the data fusion technology, which can make the characterization of CHMs both with gas and taste features, is more effective, the final recognition rate can reached100%by Euclidean distance discriminant.The results show that compared to using the electronic nose or the electronic tongue technology alone, choosing the data fusion method with both gas and taste information can achieve better classification and identification results of CHMs.