In order to solve the problem that the characteristic of the weak fault vibration signal of gearbox was not easy to be extracted in the strong noise background, a weak fault diagnosis method based on improved wavelet denoising pretreatment and Teager-kaiser energy operator is presented. The original signal is denoised by the method of wavelet improved threshold function; the signal-to-noise ratio is improved effectively compared to morphological filter method and traditional threshold function method. The denoised signal is composed into several intrinsic mode functions(IMFs) by ensemble empirical mode decomposition (EEMD). The correlation coefficients of each IMF component and the original signal are calculated, and the effective components are screened by combining the spectrum of each IMF component. A time-frequency analysis result of reconstruction signal that used the effective IMF components to get the reconstructed Teager energy spectrum, is compared with a marginal spectrum that used HHT to original signal. The comparison research results show that the proposed method is more effective to extract the week characteristics of gear fault. And the results also prove the proposed method worked.