This paper presents a new diagnostic approach based on the signal pre-whitening method and empirical mode decomposition (EMD) for rolling element bearings. Cepstrum pre-whitening method is used as a pretreatment for removing all the discrete components from the signal. The whitened signal is then decomposed into several Intrinsic Mode Functions (IMFs) by using the EMD which is an effective tool for isolating the high frequency modulation components generated by the roller bearing defects. Envelope analysis is applied to the selected IMFs in high-frequency section for extracting bearing fault frequency component. Practical vibration signals measured from rolling element bearings with inner race fault, outer race fault and ball fault are analyzed by the proposed method. The result shows that the proposed method can successfully detect the three types of bearing faults at early stage of damage development.