抄録
This paper proposes a new radar tracking filter called Noise-estimate Particle Filter (NPF) which combines kalman and particle filters. The tracking performance of the kalman filter depends on the setting of several parameters such as a system noise and an observation noise. However, it is quite difficult to choose proper parameters for various scenarios beforehand, and thus they are regulated in trial-and-error manner in some cases. The proposed filter estimates proper noise parameters of a kalman filter based on a scheme of a particle filter on-line. Simulation results show that the proposed technique has a higher tracking performance in various scenarios than conventional kalman and particle filters.
