Fingerprint Built And Classification Of Chinese Herbal Medicine Based On Bionic Olfactory Technique

Chinese herbal medicine (CHM) is rich in resources and it played a huge role in human medicine development and promotion of human health. However, due to the large variety, plenty of fake and shoddy products appear on the market, even ordinary Chinese herbal medicines are a lot of adulterants, and it seriously affecting the development of Chinese medicine. Therefore, quality judgment of CHM has been a focus for researchers, among the origin of the factors is an important criterion to judge CHM quality. However, presently CHM quality is mainly evaluated by a human taste panel in domestic and foreign. The organic evaluating may vary with many factors, which request to establish more scientific CHM quality evaluating rules.Smell plays an important role in the quality analysis of CHM, the bionic olfactory technology simulates human’s sense of smell mechanism, which automatic complete the identification by detecting volatile odor of CHM. Present the reports of application of the bionic olfactory technology on the CHM quality evaluating are few. So we intend to detect CHM volatile information by bionic olfactory technology and built a new technology.Three groups of typical pungent CHM samples, concluding Zingiberaceae, Umbelliferae and Asteraceae, are selected as the research object. Many feature values were extracted by electronic nose (PEN3) and the original feature vector was consisted. Principle component analysis (PCA) was used to extract corresponding principal components to constitute the input of pattern recognition. Cluster analysis (CA) and BP neural network were used in the pattern recognition to classify the different origin and confusing CHM, and finally establish an appropriate amount of odor fingerprint database.The results of CA showed that it can correctly classify all test samples. By BP neural network method, it worked out that only a few samples of Anhui Atractylodes were classified incorrectly, however, the classification rate of testing sets reached86.67%, and the training sets reached100%; the correct rate of confusing CHM is 100%, and only Amomums were classified incorrectly and classification rate reached93.33%.The feasibility and accuracy of two pattern recognition methods were analyzed and compared. The conclusion came out:CA can fast classify the samples due to the simplicity of the algorithms, the calculation might be augmented because of the complexity of the similarity functions of CA, the increased calculation might lead to the fact that CA were no longer the fast and easy algorithms. BP has highly non-linearity. Although the classification boundaries in any shape could be approached by BP in theory, the calculation was of great complexity.CA was concluded as the most suitable approach for the experiment in the thesis.Finally, the fingerprint databases were built by three methods based on statistical characteristics (mean, variance, and peak). The results show the fingerprint curve has a high degree of distinction, and the fingerprint curve of the test samples is consistent with the fingerprint library.It works out that it’s possible to identify CHM accurately by portable electronic nose and built fingerprint library.