存在噪声和未建模动态系统传递函数的渐近无偏估计

Asymptotically Unbiased Estimation of Transfer Functions with Colored Noises and Unmodeled Dynamics

中文关键词  系统辨识  无偏估计  最小二乘法  未建模动态

英文关键词  identification  parameter estimation  least-squares estimation

中文摘要

      存在未建模动态时，如何去除系统传递函数估计中噪声引起的偏差是系统辨识关心的重要问题。本文将我们已提出的BELS方法进行推广得出：当系统未建模动态可用—FIR模型描述时，应用本文的方法可得到系统传递函数的无偏辨识，而不需知道有关噪声和未建模动态概率分布函数的先验知识。

英文摘要

      In the presence of the unmodeled dynamics, one common and important concern is how to remove the noise-induced bias in the estimate of the transfer function. An extension to the newly established BELS method is made in this paper. When the unmodeled dynamics is decribed by a finite impulse response (FIR) model, it can achieve a unbiased estimate of the transfer function without the priori knowledge about the probability density functions (PDFs) of the noise and the dynamics.