As a generalization of the spectral correlation density, modulation intensity distribution (MID) can extract the amplitude modulations of either discrete carrier signals or random ones satisfactorily. However, the computational effort of modulation intensity factor in MID is so heavy that it can not satisfy the real time request of industrial applications. Then, a new method, that is combination slice analysis of MID (C-SMID), is proposed in the paper to detect the characteristic frequency of bearing faults. Firstly, the speed variation range of the rolling bearing is used to decide the possible frequency fluctuation range of the bearing fault signal in the new method. Then, the value range of the selectivity factor △f is set and the character slices of MID is calculated out. Finally, the bearing fault type is judged based on the signal energy comparison between different slices. Comparing with the original MID, the new method has better anti-noise property and less amount of calculation. In order to verify the feasibility and superiority of the new method, several comparative experiments between it with the common envelope demodulation method had been executed based on the simulate signals and the measured ones that acquired from the rolling bearing fault simulation experiment platform of QPZZ-II. The experiment results show that the new method has higher detection precision for such bearing faults as inner race, outer race and rolling parts.