Aiming at the problem of fault diagnosis of rolling bearings of gearbox using acoustic emission, the method based on SVD and Fast Kurtogram algorithm is proposed. First, background noise was restained to increase signal-to-noise ratio by SVD, then best parameters of bank-pass filter were extracted by the Fast Kurtogram algorithm, and rolling bearing fault features were extracted by energy operator demodulation envelop spectrum, the method can effectively reduce uncertainty in artificial selection of filter parameter of resonance demodulation. The results of simulation data and testing data show that the proposed method was effective.