Electroencephalogram　(EEG)　contains　a　lot　of　physiological　characteristics　and　disease　information,　and　the　existence　of　various　artifacts　is　a　major　obstacle　in　EEG　analysis.　Artifacts　are　mostly　generated　by　the　unconscious　eye　activity　and　muscle　activity　of　the　human　body　and　the　most　serious　effect　is　the　strong　electrooculogram　(EOG)　generated　by　eye　movement.　The　existing　removal　of　EOG　requires　various　pretreatment　and　calibration　procedures,　which　are　inconvenient　and　time-consuming.　Therefore,　for　complex　EOG　artifacts,　we　proposed　a　method　combining　the　discrete　wavelet　transform　(DWT)　and　the　least　mean　square　(LMS)　adaptive　filter.　The　EOG　component　decomposed　by　DWT　is　used　as　the　reference　signal　of　the　adaptive　filter　to　estimate　EOG　artifacts.　The　experimental　results　show　that　the　average　mean　square　error　(MSE)　of　the　subjects　is　reduced　by　79.58%　and　the　average　signal-to-noise　ratio　(SNR)　is　increased　by　143.6S%.　We　also　implement　the　proposed　filtering　algorithm　in　hardware　and　finally　form　a　mature　EEG　filtering　system　for　the　brain-computer　interface.　©　2023　IEEE.