Linear precoding for finite-alphabet inputs over MIMO interference channels by exploiting SCSI

摘要

Based on the assumption of finite alphabet inputs, the precoding design for MIMO interference channels (MIMO-IC) is proposed by exploiting statistical channel state information (SCSI) at the transmitter. Most of the related precoding design relies on Gaussian input assumption instead of discrete constellation symbols. However, the input data consist of discrete constellation symbols in practice, which would inevitably lead to performance loss. Under the finite alphabet inputs, the upper bound of the sum average mutual information is firstly derived. Then, by maximizing this upper bound, the optimal precoding matrices are achieved. Simulation results demonstrate that the proposed precoding design could obtain higher mutual information than the existing schemes (such as the Max-SGINR scheme, D-SVD scheme and BM scheme).