Improved Design and Implementation of Variational Bayesian Iterative Receivers

Abstract

It was recently shown that detection performance can be significantly improved if the statistics of channel estimation errors are available and properly used at the receiver. Although deriving the statistics of channel estimation errors is rather straightforward for pilot-only channel estimation methods, it is not the case for semi-blind receivers such as variational Bayesian (VB) receivers. We have shown in a recent contribution that by a modified formulation of the VB formalism, one can reduce the impact of channel uncertainty on the decoder performance. In this paper, we propose different practical VB receiver implementation techniques that lead to further performance improvement. The adequacy of the proposed receiver design compared to classically-used VB receivers is demonstrated by simulations for orthogonal frequency-division multiplexing (OFDM) systems.