The collected vibration signal of hob in engineering site is contaminated with noise. It is difficult to extract features contained in vibration signal. In this study, ensemble empirical mode decomposition (EEMD) is applied to denoise vibration signals. To solve the problem of selecting and processing of intrinsic mode function (IMF) after EEMD decomposition, a denoising method of hob vibration signal based on grey criterion and EEMD is proposed. Firstly, the original signal is decomposed into several IMF components by EEMD. Then, according to the proposed grey criterion, each IMF component is processed by polarity consistency and mean processing. The grey correlation between IMF1 and other IMF components is calculated. All IMF components are arranged in descending order according to the grey correlation degree. The first half of IMF components in the descending order are selected for soft threshold processing. Finally, processed IMF components, unprocessed IMF components and residual components are reconstructed to obtain the denoised signal. The feasibility and validity of the method are verified by the simulation signal with different initial signaltonoise ratios and the vibration signals of the hob in actual machining. Meanwhile, the proposed method is compared with EEMD combined with correlation coefficient and wavelet soft threshold denoising. Experimental results show that this method has better denoising effectiveness.