Identification is very important for the development of Chinese traditional medicines. In recent years, rapid progress in ultrafast laser technology provides a steady and available source for terahertz pulses generation, which greatly promotes the development of THz spectroscopy and imaging technique. SVM is a method for recognition of two kinds of samples. Appling SVM to the identification of Chinese traditional medicines through THz spectrum is a new way. The experiment on three groups of Chinese traditional medicines (zhigancao and shengancao, nanchaihu and beichaihu, shandougen and beidougen) was studied. The THz frequency spectrum and absorptivity were obtained and used to construct the feature space of Chinese traditional medicines. Three kinds of SVM were build, which used three kinds of kernel functions. By comparison, a model of BP artificial neural network was constructed. The result of using three kinds of SVM and BP artificial neural network to identify the Chinese traditional medicines showed that both methods have good prediction ability, but obviously the effect of SVM is better than BP artificial neural network for small samples. Using SVM in terahertz spectrum is a efficacious way for classification of Chinese traditional medicines.