单传感器单模型动态系统多尺度分解与估计新算法

本文将基于模型的动态系统分析方法与具有统计特性的多尺度信号变换方法相结合,基于某一尺度上给定的单传感器单模型动态系统,建立起一个新的多尺度动态模型. 基于建立的多尺度模型和标准Kalman 滤波,提出一个能同时对随机信号进行多尺度分解与多尺度估计的新算法. 获得比仅在原始尺度上进行Kalman 滤波好的处理效果,应用Monte Carlo 仿真验证了算法的有效性.
关键词: 　多尺度估计; 动态系统; 随机信号; Kalman 滤波; 小波变换
Abstract : 　By combining the model2based 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 fil2
tering we put froward a new multiscale algorithmfor stochastic signal on the basis of the multiscale dynamic model ,and we can simul2 taneously implement decomposition and estimation for stochastic signal. The estimates obtained by use of the new algorithm is more ac2 curate 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.
Key words : 　multiscale estimation ;dynamic system;stochastic signal ;Kalman filtering ;wavelet transform