Objective Based on the nonstationary and fuzzy characteristics of electroencephalogram ( EEG) , we proposed an electrooculogram ( EOG) removal method, in which discrete wavelet transform ( DWT) is combined with second-order blind identification ( SOBI) , and fuzzy entropy is used for discriminant of ocular artifacts ( OA) . Methods DWT is used to analyze each channel of EEG with noise to obtain more stable multi-scale wavelet coefficients. Then, the wavelet coefficients in the same layer are selected to construct the wavelet coefficient matrix, and it is further separated by using SOBI to get the estimation of source signals, whose fuzzy entropies are calculated and employed to realize the automatic identification and elimination of OA.The experimental data comes from the publication of Data Sets 2 b database in BCI Competition IV. The commonly used performance indicators such as signal-to-noise ratio, correlation coefficient and root mean square error are used to measure the effect of artifact removal. Results This method achieves the maximum values over multiple performance indicators relative to the commonly used methods. Conclusions The automatic and accurate identification and elimination of OA is realized by this method, yielding more stable experimental results.