In light of the difficulties of feature extraction of early faults in rolling element bearings, a method based on baby wavelet deployment and coefficients integration is proposed for fault diagnosis. First, the baby wavelets are deployed, and wavelet transform is conducted. The wavelet coefficients at different scales are then integrated according to the proposed peakedness indicator. Finally, the autocorrelation spectrum is adopted to restrain noise and highlight fault information. The effectiveness and advantages of the proposed method are proven through simulated and experimental signals. The results show that the weak fault features can be extracted, and the early fault diagnosis of rolling bearings is realized through the proposed method.