ARMA信号的鲁棒自适应去卷滤波器

Robust Adaptive Deconvolution Filter for ARMA Signals

对于通过已知线性系统被观测的未知ARMA信号,本文用现代时间序列分析方法提出了一种新的鲁棒自适应去卷滤波器。它可处理非平稳ARMA信号及不稳定和非最小相位系统。其特点是,未知ARMA信号的自回归(AR)参数可单独通过ARMA新息模型的在线辨识得到,且由于AR参数估计误差所引起的未建模动态被用虚拟噪声补偿技术加以补偿,因而它对于模型误差是鲁棒的。仿真例子说明了其有效性。

Using the modern time series analysis method, this paper presents a new robust adaptive deconvolution filter for unknown ARMA signal observed through a linear system, which can handle nonstationary ARMA signal and unstable and/or nonminimum phase system. Its characteristic is that autoregressive(AR)parameters of ARMA signal can be identified on-line independently, and unmodeling dynamics caused by the errors of the AR parameter estimation is compensated using the fictitious noise compensating technique. It is robust for the model errors. Simulation example shows its usefulness.