In order to solve the modal aliasing phenomenon of the empirical mode decomposition (EMD) in the bearing fault feature extraction, an ensemble empirical mode decomposition (EEMD), fast spectral kurtosis frequency selection and resonance demodulation technology combined rolling bearing fault diagnosis method was proposed. The original vibration signal was trained by EEMD and it was decomposed into multiple intrinsic mode functions (IMF). The IMF components meeting the kurtosis criterion were screened out, the signals were reconstructed for these components. The fast spectral kurtosis calculation was performed on the reconstructed signal to obtain a fast spectral kurtosis diagram, and the optimal frequency band center and bandwidth were obtained from the diagram, the design parameters of the FIR bandpass filter was determined. Finally, the envelope analysis of the filtered signal was carried out by resonance demodulation technology to obtain the envelope spectrum and determine the fault characteristic information of the rolling bearing. The feasibility of this method was verified through experimental analysis of rolling bearings.