At present, the palpation examination of oral cancer relies on doctor′s experience; the detection method of intraoperative frozen section is invasive detection and the accuracy is low. In order to meet the actual needs of clinical surgical oral cancer detection, the oral cancer detection method based on highwavenumber (2 400～4 000 cm-1) fiber laser Raman spectroscopy is studied. The principle using fiber Raman spectroscopy to realize oral cancer detection is studied. The feasibility of the Principal Component AnalysisLinear Discriminant Analysis (PCALDA) statistical algorithm is analyzed. A portable oral cancer tissue detection fiber Raman system was built, and the histopathological sections of the oral cancer tissues and normal tissues were prepared. The highwavenumber Raman spectra of the oral cancer tissues and normal tissues were collected, and the spectra were processed and analyzed with PCALDA algorithm. The spectral analysis results show that the average spectra of the cancer tissues and normal tissues are significantly different at 2 780, 2 890, 2 936, 3 180, 3 285, 3 300 and 3 650 cm-1. The Raman spectroscopy of the normal tissues and oral cancer tissues were dimensionreduced and classified with the PCALDA analysis, and the detection accuracy is 966%. Compared with the result of confocal microscopy Raman spectroscopy, the detection efficiency is improved. Compared with the result of fingerprint Raman spectrum detection, the detection accuracy is improved by about 06%. The study result indicates that combining statistical method, the highwavenumber Raman spectroscopy method can nicely distinguish between normal tissues and cancerous tissues, which is expected to be an oral cancer detection method with high sensitivity, high timeefficiency and high accuracy, and has important significance for intraoperative oral cancer detection.