Linear Precoding with Minimum BER Criterion for MIMO-OFDM Systems Employing ML Detection

This paper proposes a new MIMO-OFDM preceding method that minimizes BER of the maximum likelihood detection (MLD). Conventional MIMO precoding methods can approximately minimize BER of the minimum mean square error (MMSE) based detector by using singular value decomposition (SVD) of the channel matrices. However, they cannot minimize BER of MLD because the orthogonalization does not increase the Euclidean distance between a pair of channel-distorted replicas of transmitted signals. To optimize the Euclidean distance effectively, the proposed method controls its precoding parameters by minimizing the upper bound of BER based on the pairwise error. Computer simulations demonstrate that the proposed precoding method outperforms the MMSE precoder and the conventional minimum BER (MBER) precoder that assumes the precoding matrix structure similar to the solution of the weighted MMSE precoder.