On Developing an Infrared-Spectrum Fingerprint Classifier for Fast Non-destructive Identification of Chinese Herbal Qualities

Traditional Chinese medicine (TCM) has been recently receiving wide attention in the medical field because of its unique effects in chronic and consumptive diseases with little side effects. In general, the effectiveness of TCM greatly depends on the herbal qualities (HQ), whose identification therefore plays an important role in the TCM development. In the literature, one typical approach is to identify the HQ based on its spectrum fingerprint, e.g., Ultraviolet Spectrum, Infrared Spectrum (IRS), Nuclear Magnetic Resonance, and so forth, in which IRS is a favorite one because of its simple sampling, quick experimental operations, and distinct fingerprint characteristics in comparison to other spectra. However, the existing IRS based approaches need a destructive sampling pre-processing, which not only is time-consuming, but also completely damage the sample during the identification process. Further, the existing IRS fingerprint recognition is to visually compare the difference of some top peaks in the fingerprints without utilizing advanced feature extraction techniques. Consequently, their recognition rates are still far away from expectation. This project therefore aims to develop a classifier for recognizing IRS fingerprints acquired via a non-destructive sampling way. We will conduct the project with two main steps. The first step is to extract the underlying features from an IRS fingerprint by formulating it as either one of two problems: (1) Time Series Indexing, and (2) Attributed Graph Indexing. The next step is then to classify the fingerprints by utilizing a robust clustering algorithm in terms of cluster number. With labeling the HQ of each class, this project will eventually give out a fast non-destructive way to recognize the IRS fingerprints for identifying Chinese HQ.