Abstract: In order to detect the gas signals in complex environments of coal mine and solve the problem of the buried weal signal and the abnormal data because of surrounding noise interference on gas signal, a detection method for weak gas signal was introduced based on stochastic resonance. Sub-sampling method was used to transform large frequency signal scale and particle swarm optimization algorithm was used to optimize structural parameters. The resonance effect of large-parameter weak signal in a stochastic resonance system was analyzed. The results show that optimum matching between the nonlinear system, the input signal and the noise could be achieved adaptively with lower sampling frequency. The large-parameter multi-frequency weak signal can be distinguished from strong background noise effectively, and the detection sensitivity and dynamic range are enhanced. The research provides theory basic for early identification of gas outburst information.