Regularized blind detection for MIMO communications

Multiple-Input Multiple-Output (MIMO) systems improve the throughput and reliability of wireless communications. Perfect Channel State Information (CSI) is needed at the receiver to perform coherent detection and achieve the optimal gain of the system. In fast fading and low SNR regimes, it is hard or impossible to obtain perfect CSI, which leads the receiver to operate without knowledge of the CSI and perform blind detection. In reality CSI may be available to the receiver but this CSI may be insufficient to support coherent detection. In this paper, we fill the gap between coherent and blind detection by considering a more realistic model where the receiver knows the statistics of the channel, that is Channel Distribution Information (CDI). We propose a new detection algorithm, called Regularized Blind Detection (RBD), where coherent and blind detection can be viewed as special cases in our model. The algorithm estimates CDI from any training symbols that are available and maximizes performance given the estimated CDI. Simulations demonstrate significant improvement in performance over blind detection. Our work can be viewed as a systematic exploration of space between coherent and blind detection with a strong Bayesian statistic flavor.