Steady state visual evoked potential (SSVEP) has been widely applied in brain computer interface (BCI) systems and brain’s cognitive research, phase feature is one of the important characteristics of SSVEP. Aiming at the uncertainty principle constraint of the fast Fourier transform (FFT) in SSVEP phase extraction, a phase extraction method based on HilbertHuang transform is proposed. In this method, the electroencephalogram signal is decomposed into a series of intrinsic mode functions (IMF) by empirical mode decomposition, and the mean value of the instantaneous frequency of each modal function is analyzed to determine whether the IMF component belongs to noise. Filtering the noise IMF components to get the electroencephalogram IMF components, on which is performed Hilbert transform to obtain the phase of SSVEP by operating with reference signal. Compared with FFT phase extraction, the experimental results show that the proposed mothed can extract the phase information of SSVEP while removing the noise components. It also has better accuracy, precision and adaptability.