To solve the problem of bearing fault diagnosis, a fault feature extraction method combining wavelet denoising and HHT transformation was proposed. The cylindrical roller bearings with inner ring fault and roller fault were tested. Wavelet de-noising method was used to de-noising the time domain signal collected.The HHT transform was used for timefrequency analysis, and a series of eigenmode function components were obtained.Finally, the bearing failure condition was determined according to the test results.The experimental results show that the eigen value of the inner ring fault and roller fault bearing is basically the same as the theoretical calculation.