A Novel Orthogonal Superimposed Training-Based Channel Estimation Method for MIMO Channels
The authors propose a novel superimposed training-based channel estimator for frequency-flat block-fading Multiple-Input Multiple-Output (MIMO) systems. The superimposed training sequence is designed to be orthogonal to the information-bearing data and the orthogonal property is obtained by exploiting the random space-time symbol interleaver at transmitter. As a result, the interference from the unknown data is totally eliminated during the process of channel estimation. The proposed design is then proved to be optimal to minimize the squared error criterion. Furthermore, the optimal power allocation for training is also analytically derived. Simulation results demonstrate the excellent performance of the proposed scheme.