Spectral absorption method is an effective means to detect methane concentration. Through the prism chamber combined with the application of photonic crystal fiber, the on-line detection of methane concentration with a high precision by the spectral absorption is realized. Due to the environment temperature, pressure and the influence of the system itself, the received signal contains a lot of noises. According to the characteristics of strongly ability to generalize and seek the global optimal point, support vector machine (SVM) is used to process the signal of methane detection. Through Matlab simulation based on the SVM principle, it can effectively filter out noise, separate the effective signal and assess the effect of denoising with signal to noise ratio (SNR). Using this method to filter can make the SNR reach above 130 dB, which is greatly improved compared with the traditional wavelet denoising, to achieve the ideal denoising purpose.