Abstract: Periodic impulses in vibration signals and its repeating frequency are the key factors for diagnosing rolling element bearing faults. A new method based on ensemble empirical mode decomposition(EEMD) and cross energy operator was proposed to extract the characteristic frequency of bearing fault. Firstly,the signal was decomposed into intrinsic mode function(IMF) by means of EEMD to satisfy the mono-component requirement by the cross energy operator. Next,the sensitive IMF was selected according to correlation and kurtosis,and instantaneous cross energy between the IMF and the original signal and its Fourier spectrum were calculated. Finally,the bearing faults were diagnosed by matching the repeating frequency of fault-induced periodic impulses with the fault characteristic frequency. By analyzing both a simulated faulty bearing vibration signal and the experimental data of bearing faults,the bearing faults were diagnosed and the effectiveness of the proposed method was validated.