The　electroencephalography　(EEG)　is　easily　affected　by　the　ocular　artifact　(OA)　when　sampled,　and　this　will　produce　great　impact　on　the　performance　of　brain-computer　interface　system.　A　novel　method　was　proposed　based　on　the　combination　of　canonical　correlation　analysis　(CCA)　and　discrete　wavelet　transform　(DWT),　and　it　is　denoted　as　DCCA.　Firstly,　DWT　was　applied　to　the　collected　EEG　and　electrooculogram　(EOG)　signals　to　acquire　the　multiple　scale　wavelet　coefficients,　and　CCA　to　eliminate　the　correlation　among　the　coefficients.　Then,　the　correlation　coefficient　was　used　as　a　criterion　to　recognize　the　ocular　components,　and　the　corresponding　canonical　wavelet　coefficient　vectors　were　set　to　zero.　At　last,　the　inverse　algorithms　of　CCA　and　DWT　were　applied　in　sequence.　So,　the　OA　was　removed　from　EEG　in　this　way.　By　using　DCCA　and　other　methods,　experiment　research　was　finished　based　on　the　BCI　data　sets　which　contained　4　kinds　of　EOG　data　and　were　sampled　from　9　subjects　at　different　time.　The　significant　tests　show　that　the　proposed　method　has　obvious　superiority　in　the　aspects　of　root　mean　square　error　(RMSE)　and　signal　noise　rate　(SNR).　Furthermore,　it　has　good　real-time　performance　and　excellent　adaptive　capabilities.　©,　2014,　Science　Press.　All　right　reserved.