A fast Ensemble Empirical Mode Decomposition (FEEMD) was proposed. The amplitude of white noise and ensemble trails were fixed as 0.2 and 2 respectively. The signal was decomposed by changing the density of added white noise. To enhance decomposition effective, the residue was decomposed by EMD when abnormal component were extracted. This method overcome drawback of selecting parameters by experience, and reduce computational cost effectively. Finally, characteristic frequency of bearing inner fault motors were identified by FEEMD with Hilbert envelope, and compared with traditional EEMD method. The result indicates that FEEMD can identified the character frequency effectively.