Two-stage algorithms of parameter estimation for the autoregressive moving average (ARMA) models are presented, which are called two-stage recursive least squares algorithm(2-RLS) and recursive least squares-pseudoinverse algorithm(RLS-PI). Based on online parameter estimation of the ARMA. innovation models, using the modern time series analysis method, the several self-tuning Kalman tracking filters are presented, where the three different algorithms of the Kalman tracking filter gains are used. The self-tuning white noise Wiener estimators and information fusion self-tuning white noise Wiener estimators are presented, and self-tuning Wiener filter and Wiener deconvolution filters for signals are also presented. Unified self-tuning Wiener state estimators are presented, which can handle the state filtering, smoothing and prediction problems in a unified framework, and their applications to the self-tuning tracking filters are given. The simulation examples show effectiveness of the proposed results. |