The study on the temperature evolutionary trend under both the ground stress and gas pressure is significant in the process of coal gas absorption and desorption. However, the valid signal is difficult to gain because of much noise in the experimental data, so according to the characteristics of the temperature signal gained under different experimental conditions of coal gas absorption and desorption, this essay filtrates noise and restores the valid signal by using the wavelet denoising of MATLAB to obtain the valid temperature evolutionary curve. Through contrasting and analyzing wavelet basis function and different decomposition levels, the essay chooses the sym8 wavelet basis function which has a higher order of vanishing moments and better smoothness. The reconstructed signal filtrates most of high frequency signal, and keep low frequency, as well as smooth signal. The result shows that the de-noising effect is good and the temperature evolution rule can be well reflected in the process of the coal gas absorption and desorption.