Kernel　adaptive　algorithm　is　an　extension　of　adaptive　algorithm　in　nonlinear,　and　widely　used　in　the　field　of　non-stationary　signal　processing.　But　the　distribution　of　classic　data　sets　seems　relatively　regular　and　simple　in　time　series.　The　distribution　of　the　electroencephalograph　(EEG)　signal　is　more　randomness　and　non-stationarity,　so　online　prediction　of　EEG　signal　can　further　verify　the　robustness　and　applicability　of　kernel　adaptive　algorithms.　What＇s　more,　the　purpose　of　modeling　and　analyzing　the　time　series　of　EEG　signals　is　to　discover　and　extract　valuable　information,　and　to　reveal　the　internal　relations　of　EEG　signals.　The　time　series　prediction　of　EEG　plays　an　important　role　in　EEG　time　series　analysis.　In　this　paper,　kernel　RLS　tracker　(KRLST)　is　presented　to　online　predict　the　EEG　signals　of　motor　imagery　and　compared　with　other　13　kernel　adaptive　algorithms.　The　experimental　results　show　that　KRLST　algorithm　has　the　best　effect　on　the　brain　computer　interface　(BCI)　dataset.　Copyright　©　by　HIGH　TECHNOLOGY　LETTERS　PRESS.