Skip to Main Content
We consider a terrestrial wireless channel, whose statistical model under flat-fading conditions is due to Clarke. A lot of papers in the literature deal with receivers for this scenario, aiming at estimating and tracking the time-varying channel, possibly with the aid of known (pilot) symbols. A common approach to derive receivers of reasonable complexity is to resort to a Kalman filter which is based on an approximation of the actual fading process as autoregressive moving-average (ARMA) of a given order. The aim of this paper is to show that the approximation of the actual fading process, usually exploited in the literature, is far from effective. Thus, we present a novel technique, based on an off-line minimization of the mean square error of the channel estimate, which ensures a considerable gain in terms of bit-error rate for Kalman-based receivers without increasing the receiver complexity. Moreover, we also propose a novel approximation, to be employed in Kalman smoothers proposed for iterative detection schemes, which allows further performance improvements without a significant increase of the computational complexity.