An algorithm for estimating the time varying AR parameters of a nonstationary process

A nonstationary scalar random process represented by an autoregressive (AR) model with rapidly time-varying coefficients is considered. Assuming the observation of this AR process is contaminated by additive white Gaussian noise, the authors estimate the coefficients using the Kalman filtering technique and the solution of the Ricatti equation. Computer simulations have shown favorable results that converge to the true solution.<
>