Maximum likelihood detection and constellation multiplexing in correlated MIMO channel

We propose a precoder based on constellation multiplexing for spatial multiplexing in the presence of spatial correlation. Given the long term statistical channel information at the transmitter, a precoder is selected from the candidates that correspond to different level of constellation. The precoder selection is based on minimizing the union bound on the symbol error probability and applied to the transmitted symbol vector. The optimal precoder varies with different correlation level and is evaluated with maximum likelihood detection and ZF-SIC detection schemes under different correlation scenarios. It is shown by simulation that both the detection schemes with precoding outperform their counterpart without precoding.