Aiming at the problem that it is difficult to effectively extract the acoustic signal in the strong noise environment of underground mine, a signal filtering algorithm of piecewise three steady-state stochastic resonance was proposed based on three steady-state stochastic resonance and combined with piecewise bi-stable potential function. Aiming at the problems of low signal-to-noise ratio, severe parameter coupling and output saturation of traditional stochastic resonance output, a piecewise three steady-state nonlinear system model was constructed. The system model can independently adjust the well depth, well position and barrier steepness to induce the optimal three steady-state stochastic resonance. Based on the output signal-to-noise ratio, the parameters were optimized by artificial fish swarm algorithm (AFSA) to improve the signal filtering effect of piecewise three steady-state stochastic resonance. The results show that the signal filtering algorithm of segmented three steady-state stochastic resonance is more effective than other classical algorithms, improving the output signal-to-noise ratio of processing downhole acoustic signal and providing a better method for the extraction of downhole acoustic communication signal.