Maneuver detection is a key issue in the multiple-model target tracking applications.In Kalman filter

the conventional maneuver detection algorithm fails to work well when the maneuver is almost buried in noise.The multiresolutional approach performs well for maneuver detection

but the time delay limits its application.In this paper

a new algorithm for maneuver detection is proposed

which is based on the third-order cumulant of the measurements.As the higher-order cumulants are blind to Gaussian noise (white or colored)

the behavior of the maneuver becomes obvious and easy to be detected in the higher-order cumulant domains.Taking the sliding window processing

the maneuver detection algorithm is real-time without any time delay.Simulation results demonstrate the effectiveness of the algorithm over the conventional and multiresolutional one's

especially at low SNR.