用多尺度卡尔曼滤波法，对信号参数进行识别估计。高频信号和低频信号识别结合起来改进了算法识别的精确度和准确度。-It is an implementation of hierarchical (a.k.a. multi-scale) Kalman filter using belief propagation. The model parameters are estimated by expectation maximization (EM) algorithm. In this impleme

状态模型的极大似然估计，使用EM算法，以及卡尔曼滤波。-This supplementary note discusses the maximum likelihood esti-mation of state space models using Expectation-Maximization (EM) algorithm and bootstrap procedure for statistical inference. A Matlab program scr ipt impleme