Rapid identification of Rhizoma Coptidis origin using terahertz spectroscopy and deep learning.

Rhizoma Coptidis (RC) from different geographical origins vary in terms of quality and efficacy. Terahertz time-domain spectroscopy facilitates rapid, non-destructive testing of samples without the need for complex sample pretreatment, which provides a new band for fingerprint and makes it possible to identify its origin quickly. In this study, four different-origin RC samples were selected as the research subjects. Their terahertz spectra showed subtle differences, making it quite challenging to directly determine their sources. The classification model ICOA-GRU-MSA was proposed for identification of RC origin. In this model, ICOA was utilized to optimize the hyperparameters of GRU, and MSA was introduced to enhance the model's sensitivity and ability in recognizing feature information. The results show that the proposed model has better learning ability and classification performance than other models in solving the origin identification of RC terahertz spectra. This study provides a novel insight and method for identification of traditional Chinese medicine origins.