Abstract: In the terahertz time-domain spectroscopy technique, 16 common pathogens were experimentally studied and their characteristics absorption spectra in the frequency range of 0.1 to 2.2THz were obtained. The terahertz absorption spectra of 16 common pathogens were trained and identified by Siamese neural network method. First, the terahertz absorption spectra of the 16 pathogens were reduced by PCA to construct training data. Then, the constructed Siamese neural network model was trained by back propagation. Finally, the pathogen measured at different times was used as the target spectrum to evaluate the model, after comparing with the training data, the matching absorption spectrum was obtained, and the recognition rate reached 97.34%. The recognition results fully indicate that the identification of different kinds of pathogens can be recognized by Siamese neural network, which provides an effective method of the detection and identification of pathogens by terahertz spectroscopy.

Keywords: THz spectroscopy; machine learning; Siamese neural network; similarity learning