This paper present a method toimprove the estimation accuracy of a tracking Karlman filter (TKF) by using a multilayed neural network (MNN). Estimation accuracy of the TKF is degraded due to the uncertainties which cannot be expressed by the linear state-space model given priori.. The MNN capable of leaning an arbitrary so that realized a mapping from measurement to the corrections of estimation of TKF. Simullation results show that the estimation accuracy is much improved by using MNN.