The electrooculography (EOG) artifact removal for steady-state visual evoked potentials (SSVEP) is studied. An EOG artifact removal method for electroencephalogram (EEG) without ocular electrode based on adaptive neuro-fuzzy inference system (ANFIS) is proposed, and the experiments are carried out to prove the improvement of SSVEP-BCI′s stability under random blinking. It tackles the stability of brain control interface (BCI) under artifact interference and takes the stability of SSVEP-BCI under random blinking as the key point. The proposed method approximates the nonlinear transforming function from the EOG source to EOG artifact with ANFIS to cancel the EOG artifact from EEG. The EOG artifact is deleted in an adaptive noise cancellation (ANC) after the source obtained from the prefrontal lobe′s EEG passing through a tapped delay line (TDL) into the cancellation. All the parameters and functions are elaborated based on the experiments of steady-state visual stimulation under random blinking. The proposed method cancel EOG artifact in SSVEP while maintaining valid information on steady-state visual stimulation. The recognition accuracy is 6.25% and10% higher than that of the classical band-pass filter and traditional ICA respectively, thus ensure the stability of SSVEP-BCI under random blinking.