Abstract
We consider the problem of semi-blind channel estimation for Multi-Inputs Multi-Outputs (MIMO) channels. More precisely, we are interested here on the performance limit given by the channel estimate Cramer-Rao Bound (CRB). We start by reviewing the results related to the Bayesian CRB and show that this bound is in general inadequate to well characterize the channel estimation performance of a MIMO system. The purpose of this work is to propose alternative and more appropriate performance bounds by considering the stochastic CRBs. Different CRBs are computed and compared for different schemes of training sequences design and different assumptions on the source signal distributions.