Abstract: A sufficient statistics based particle filter is proposed to deal with such problems in the resampling procedure as loss of diversity among particles and large computational complexity.If the posterior density function of the system state can be expressed with a set of sufficient statistics which are easy to update,the proposed method replaces the update of posterior density function with the propagation of sufficient statistics,so the resampling procedure can be avoided and the computation burden can be reduced.Simulation experiment is made by applying the proposed method to joint estimation of state and parameter of nonlinear system,and the results prove the validity of the proposed method.