Identification of amino acids in a mixture based on a CNN

THz-TDS systems are widely used in non-destructive inspection and chemical inspection and are receiving increasing recognition due to its wide spectrum and the ability to distinguish isomers that can not be easily identified by other types of spectrum. Convolutional neural network is the state-of-the-art technique for feature extraction and classification of single-dimensional and multidimensional matrices and is applicable to identify the THz absorption spectrum of chemicals. We present a method based on a CNN to identify the THz absorption spectrum of several types of amino acids measured by transmission spectroscopy of a THz-TDS system and receive a rate of accuracy close to 100% even for isomers such as L-Tyrosine and DL-Tyrosine. We also present the initial results of quantitatively identifying the amino acids in a mixture, which is among the first outcomes of such research. This technology will support the future application of THz-TDS in drug detection under complex states and environments.