Skip to Main Content
In this article, we deal with the training-based channel estimation (TBCE) scheme in the spatially correlated Rician flat fading multiple-input multiple-output (MIMO) channels. The performance of the least squares (LS) and linear minimum mean square error (LMMSE) estimators is investigated. Moreover, optimal training signals in the minimum mean square error (MMSE) sense are achieved. It is shown that the traditional LS estimator cannot exploit the knowledge of the first and second-order statistics about the Rician fading MIMO channel. However, the LMMSE estimator uses the knowledge of spatial correlation. Furthermore, when the Rice factor increases, the mean square error (MSE) of this estimator significantly decreases. Theoretical analysis and simulation results show that the performance of the LMMSE estimator in the Rician model compared with Rayleigh one is much better.