The　estimation　scheme　exploring　the　channel　statistics　provides　advantage　estimation　performance　but　is　susceptible　to　the　accuracy　of　channel　parameters　obtained　and　thereby　the　estimation　model.　To　address　this　issue,　a　modified　Kalman　filter　to　fix　the　model　error　is　proposed.　According　to　the　analysis　of　the　model　errors　in　the　Kalman　recursion,　the　estimation　is　obtained　by　revising　the　prediction　residuals　and　observation　residuals.　The　simulation　results　show　that　in　comparison　with　the　standard　Kalman　filtering,　the　proposed　modified　Kalman　estimation　has　a　performance　improvement.