A recursive errors-in-variables method for tracking time varying autoregressive parameters from noisy observations

Abstract

Time Varying Autoregressive (TVAR) models play a key role in various applications such as radar processing, aeronautics and speech processing. Nevertheless, tracking TVAR parameters may be difficult, especially when the process is disturbed by an additive white noise. In this paper, we suggest the use of a recursive Errors-In-Variables method to estimate the variances of the driving process and the additive noise and to track TVAR parameters. This method is based on a Newton-Raphson algorithm. A comparative study with EKF, UKF and CDKF is also proposed.