A Modified Kalman filter

based on the model of filter bank for tracking maneuvering targets（see referenced）

is presented in this paper. It detects the target maneuver by judging whether observation residue exhibits bias. The maneuver acceleration command is estimated by damp least squares method only when the maneuver is detected

and then the filter will correct the state-prediction and its error covariance with the estimate. Otherwise the filter works as a single Kalman filter with zero maneuver acceleration command. Thus

a better compromise between the steady filtering accuracy and rapid response to maneuver is obtained. The computer simulation results show that the filtering accuracy is somewhat better than that of Moose’s complicated filter bank and the computation burden is only 1/3.6 of Moose’s filter.