Considering the nonlinear and non-stationary features of cutting chatter signals, this paper proposed a chatter recognition method based on Hilbert-Huang transform. By this method, firstly, a chatter signal is decomposed into a series of intrinsic mode functions (IMFs) by empirical mode decomposition (EMD), and then the IMF with rich chatter information is selected and filtered by a band-pass filter, and then the Hilbert spectrum of the filtered IMF is obtained by Hilbert transform, lastly, cutting chatter is quantitatively recognized by the standard deviation of amplitude of Hilbert spectrum. The effectiveness of the proposed method was validated by simulated and actual vibration signals. For comparison, the method by wavelet packet decomposition and wavelet spectrum analysis was also provided.