Early treatment is essential to improve the survival rate of patients. However, due to its hidden location and early symptoms similar to nasal inflammatory diseases, it is easy to ignore, and it is often found in the middle and late stages. In recent years, terahertz technology has attracted much attention in biomedical cancer detection due to its low energy, strong penetration, and fingerprint spectrum characteristics. In this study, nasopharyngeal carcinoma (NPC) and nasopharyngitis tissues were taken as the research objects to explore the application value of terahertz spectroscopy in the differential diagnosis of nasopharyngeal carcinoma and nasopharyngitis. The terahertz time-domain spectroscopy system was used to collect the spectrum of nasopharyngeal tissues in the range of 0.6~5.0 THz, and the absorption spectrum was obtained by parameter extraction. Based on the spectral data, the spectral characteristics of NPC and nasopharyngitis tissues were analyzed and compared. The spectral difference source between the two nasopharyngeal tissues was combined with the pathological H&E staining results. Through the application of the principal component analysis (PCA) method, the original power spectrum data collected in experiments were reduced. The features were extracted, and the scatter plot of samples in the three-dimensional coordinate space composed of the first, second, and third principal components was obtained. Based on the analysis of this scatter plot, a significant differentiation between NPC tissues and nasopharyngitis tissues in the feature space can be observed. The results show that the absorption of terahertz wave in NPC tissue is significantly higher than that in nasopharyngitis tissue in the range of 1.3 to 3.4 THz, and 2.7 THz is the best potential diagnostic frequency to distinguish NPC tissue from nasopharyngitis tissue. After further dimensionality reduction by principal component analysis, the cumulative variance contribution rate of the first four principal components reached 87.45%, which had a good clustering effect on the two groups of nasopharyngeal tissue samples. The principal component plot can clearly distinguishNPCtissue and nasopharyngitis tissue. K-nearest neighbor (KNN) algorithm and support vector machine (SVM) were combined to construct a classification model to realize the discrimination and classification of the two kinds of nasopharyngeal tissue THz spectra. Compared with the KNN algorithm, the accuracy of SVM classification model has an average classification reaches 92%. This study preliminarily verifies the effectiveness of THz spectra used to identify nasopharyngeal carcinoma from nasopharyngitis, which lays a foundation for further exploring their clinical value.