Bayesian Channel Estimation for Doubly Correlated Mimo Systems

In this paper we derive a Bayesian estimator for doubly correlated MIMO channels. The Bayesian estimator has clearly superior normalized mean squared error performance compared to parametric approaches especially when the channel is strongly correlated. However, since the computational costs may exceed practical limits we present a class of fix point algorithms significantly reducing the numerical effort. The convergence behavior of the fixed point algorithm is examined and a regularization is proposed in order to guarantee general convergence.