To solve the filtering accuracy reduction and divergence problems in the nonlinear system when the noise characteristics are unknown,a hybrid filter composed of Sage-Husa filter and particle filter is proposed.Firstly,a preliminary estimate of the state variables is provided by particle filter,which is then taken as the input measurement value of the secondary filter,and forms a new system with the state equation.After that,the modified Sage-Husa filter is used for estimating the statistic property of system noise in real time,and the final system state estimated value is obtained.The calculation complexity of the algorithms is calculated out quantitatively to compare the algorithms performance further,the result shows that the calculation complexity keeps unchanged in new algorithm.Finally,target tracking simulation results demonstrated the availability of the new algorithm.