Abstract: We discuss the problems of coal mine polar gas existing in the quantitative analysis and detection methods at present. Based on the support vector machine algorithm, we initially establish the polar gas quantitative analysis of calibration model by using fourier transform infrared spectroscopy to test 10 kinds of coal mine polar gas such as CH₄, CO₂, SF₆, etc. We get 310 groups of test sample data through spectral data preprocessing and calibration model of SVM testing, and establish the nonlinear quantitative analysis model based on the SVM. By treating spectral data in advance, training and testing SVM calibration model, these make calibration model prediction accuracy meet the coal mine gas detection requirements. Model validation shows that all gas concentration of sample tests and the actual components concentration deviation, are no more than 10% of the full scale except C₃H₈，nC₄H₁₀ and CO₂, by analyzing the theoretical value compared with measured value, which meets the coal mine actual requirements, and initially implements the coal mine polar mixed gas quantitative analysis with the infrared spectrum technology, obtains a certain effect on the aspect of modeling speed and prediction accuracy, which achieves a purpose of rapid detection and analysis on polar coal mine gas concentration.