According to the characteristics of gear fault vibration signals, a method for gear fault diagnosis bused upon the empirical mode decomposition (EMD) is proposed in this paper. By using EMD,any complicated signal can be decomposed into a finite and often small number of intrinsic mode functions (IMFs), which are based upon the local characteristic time scale of the signal. Thus, EMD is perfectly suitable for non-stationary signal processing and fault characteristics extracting. It is well known that a gear vibration signal consists of a number of frequency family components, each of which is a modulated signal. Thus, we can use EMD to decompose a gear fault vibration signal into a number of IMF components, some of which correspond to the frequency families, and the others are noises.Therefore, the frequency families can be separated and the noise can be decreased at the same time.The proposed method has been applied to gear fault diagnosis. The results show that both the sensitivity and the reliability of this method are satisfactory.