The key of demodulation analysis is to accurately find a suitable frequency band to demodulation. A demodulation frequency band determination method based on the fast spectral correlation （Fast-SC） and the crest of envelope spectrum （EC） is proposed to extract the feature frequency of fault rolling bearings. Firstly， signal is performed by Fast-SC algorithm. Considering the difference between theoretical fault frequency and actual fault frequency caused by slip， the integrated spectrum correlation section of theoretical fault frequency interval of bearing is adopted as the target spectrum correlation curve. The center frequency search center of the demodulation frequency band according determined to its maximum value. Then， the EC is used as the bandwidth optimization indicator to adaptively obtain the optimal filter parameters. Finally， the signal is bandpass filtered according to the obtained filter parameter set， and its envelope spectrum is obtained to realize bearing fault feature extraction. Simulation and experiment results show that compared with the Autogram demodulation algorithm， the proposed method has better noise reduction ability and better choice of demodulation frequency band.