Aiming at the problem that bearing fault signals are often mixed with noise and the fault features are difficult to accurately extract,a bearing fault feature extraction method was proposed based on wavelet threshold denoising (WTD) and complementary ensemble empirical mode decomposition (CEEMD).The WTD was used to denoise the original signal. The denoised signal was performed CEEMD decomposition to obtain a series of intrinsic mode functions (IMF);then the correlation between each IMF and the denoised signal was calculated.The useful IMFs were obtained by setting the threshold of the cross-correlation coefficient.Finally,the useful IMFs were reconstructed and the envelope spectrum was used to extract the fault characteristic frequency of the reconstructed signal.The measured signals show that the proposed method can reduce the noise interference and effectively extract fault characteristic information,which proves that it has high feasibility and strong practicability in a noisy environment.