Abstract: There are a lot of active noise reduction equipment under the mine to reduce the noise.But it has a better effect mainly for the high-frequency noise.Active noise reduction algorithm is adopted to decrease the noise effect on the miners’ health of body and mind based on noise signals of downhole environment.LMS algorithm is analyzed and there are two problems.One is a secondary source signal feedback and another is voice signal transmission delay.In view of these two major problems, the FX-LMS algorithm is adopted.In the secondary channel a transfer function is introduced to preprocess the original reference signal, and the solving method which is to find the scope of the FX-LMS algorithm convergence step length is also gave.It provides references for looking for the optimal experimental parameters.In the MATLAB development environment, noise reduction is carried out for the sound signal mixed with white noise, and simulation results under different convergence steps are gave.The results show that FX-LMS algorithm is effective, and verify that convergence steps will affect convergence speed and steady-state error.