By combining the model-based analysis for dynamic systems with multiscale signal transformation

we propose a new multiscale dynamic model based on dynamic system with single model and single sensor at one given scale.By developing Kalman filtering we put froward a new multiscale algorithm for stochastic signal on the basis of the multiscale dynamic model

and we can simultaneously implement decomposition and estimation for stochastic signal.The estimates obtained by use of the new algorithm is more accurate than the results based on Kalman filtering at the finest scale.A set of Monte Carlo simulations is performed

and the resluts show the efficiency of our new algorithm.