Abstract: Aiming at the difficulty of weak fault feature extraction for rolling bearing early fault,a fault diagnosis method based on Hermitian time-wavelet energy spectrum was proposed for rolling bearing fault diagnosis. First,the Hermitian wavelet to original signal was applied to acquire the continues wavelet transformation. Then the signal energy distribution along the time axis was calculated according to the wavelet transform results. The spectrum kurtosis was used as the selection criterion of optimal cumulation scale to obtain the time-wavelet energy distribution. Finally,the spectrum of time-energy distribution was calculated to obtain the time-energy spectrum and extract the fault feature. The results of the simulation signal and vibration signals of outer and inner rings pitting fault show that this method can effectively extract weak fault feature in strong noise background,and is superior compared with ordinary timewavelet energy spectrum; therefore,it is suitable for early fault feature extraction of bearings.