A method of feature extraction for gear fault based on Ensemble empirical mode decomposition（EEMD）and incremental spectrum of singularity entropy is put forward for the non-stationary and non-linear characteristics of gear vibration signal.Firstly

the gear vibration signal is decomposed into several smooth intrinsic mode functions（IMFs）by EEMD.The method of EEMD could take advantage of dyadic scale decomposition characteristics of the normal distribution white noise to suppress the problem of mode confusion in EMD.Because of the interference of background noise and residual assisted white noise

the gear fault feature is not extracted exactly from IMF.The method of singular value decomposition is used to remove the noise and reconstruct the IMF.The reconstruction order is determined according to the incremental spectrum of singularity entropy.Therefore the gear fault feature frequency could be extracted exactly.Results of simulation analysis and the gear fault test indicated that this method is accurate and effective for gear fault feature extraction.