Aiming at how to ensure the integrity of the antifriction bearing fault vibration signal while its noise reduction, a signal de-noising method is proposed using complementary ensemble empirical mode decomposition (CEEMD) combined with wavelet semi-soft threshold, to make the noise reduction of vibration signal of antifriction bearing. First of all, CEEMD signal processing method is used for antifriction bearing fault vibration signal decomposition and different frequency bands of intrinsic mode components (IMFs) can be adaptively obtained according to the signal characteristics. Secondly, the correlation analysis of high frequency signal components with noise pollution is executed, so the high frequency IMFs with higher noise components are achieved and then be handled by wavelet semi-soft threshold noise reduction processing. Finally, the IMFs are recombined after noise reduction processing and the remaining components of signal reconstruction, completing the noise reduction process. Analysis results show that compared with the traditional wavelet threshold de-noising processing method and CEEMD signal de-noising processing method, the proposed method of noise reduction in this paper based on CEEMD and semi-soft threshold wavelet is better, and it can ensure the integrity of the signal.