A new hybrid wavelet-Kalman filter method for the estimation of dynamic system is developed,Using this method, the real-time multiscale estimation of the dynamic system is implemented, and the observation equation established is for the object signal itself rather than its wavelet decomposition. The simulation results show that this method can be used to estimate the objects state of the stacked system, which is the foundation of multiscale data fusion; besides the performance of the new algorithm developed in the letter is almost optimal.