Organic Compound Identification Based on Terahertz Spectrum

This paper brings an organic compound identification method based on terahertz time-domain spectrum. Firstly, the absorption coefficient spectrums of the substance samples are estimated depending on the time-domain signal and the features are extracted from these spectrums in the range of 0.2-2.5THz. Secondly, the classifier model of the extracted features is established using the support vector machine (SVM) for the training samples. Finally, the identified rate is calculated in terms of the trained model for the test samples. In the experiments, we compared the performance of the feature extraction using principal component analysis (PCA), linear discriminant analysis (LDA) and frequency-amplitude parameters respectively. The experimental results show that the support vector machine combined with principal component analysis performs more classification performance.