A de-noising method based on Adaptive Wavelet Correlation Analysis (AWCA) is attempted to be applied to the vibration signal for mine gearbox using characteristics of strongly correlative between adjacent layers of the wavelet coefficients of real signal. The wavelet coefficients can be divided into available coefficients and noise coefficients, and adaptive threshold method based on wavelet entropy is applied, the available coefficients are reconstructed to restore the original real signal after de-noising process. The method is applied to the simulated signal and vibration signal of mine gearbox respectively and is compared with methods of wavelet hard threshold and traditional correlation de-noising in the reconstructed waveforms, SNR and standard deviation, and the measured signal power spectral chart. The results indicate that the de-noising method based on AWCA can effectively remove noise and reserve available signal and can effectively improve the reliability of mine gearbox vibration signal.