A Gauss-Markov model formulation for the estimation of ARMA model of time-varying signals and systems

Abstract

A Gauss-Markov model is formulated to estimate the model of a non-stationary signal. The time-varying parameters of the model are modelled as stochastic processes. A time-varying ARMA model is considered to represent the non-stationary process. Furthermore, in this work, a unified method for the optimal estimation of both the time-varying parameters and their corresponding stochastic model parameters is presented. This method utilises the proposed Gauss-Markov model for the estimation process through the extended Kalman filter (EKF).