The new scheme for bearing fault diagnosis based on wavelet variance spectrum entropy is proposed by combining wavelet analysis with entropy theory. The signal variance spectrum entropy in wavelet domain is used as fault diagnosis characteristics. The detection and diagnosis scheme of bearing vibration faults are proposed and the selection method for wavelet base based on discrimination ability factor is also presented. The results show that the wavelet variance spectrum entropy is particularly well adapted to describe bearings fault characteristics and fault diagnosis. The comparison with the method based on wavelet energy spectrum entropy is experimented which demonstrates that the proposed scheme outperforms efficiently other methods in terms of detection time and detection rate.