Acoustic emission technique provide a new testing approach to the machinery fault diagnosis because of its characteristics of high sensitivity, wide frequency response range and information. The signal often is mixed with a variety of color noise when applied to the rotation machinery. Traditional noise reduction methods can not meet the requirements when the frequency of noise and acoustic emission are superposed. The morphology filter is applied to the frequency domain to eliminate the color noise. Then, according to the acoustic emission frequency response characteristic can fit the spectrum in order to smooth the white Gaussian noise. Finally, the time domain waveform will be reconstructed. The results of simulation and faulty bearing show that the proposed methodology can achieve good effect of noise reduction, and be beneficial to subsequent process and analysis.