Metric Selection for Reduced-Complexity MAP Detectors in MIMO Systems

Abstract

Reduced-complexity maximum a posteriori probability (MAP) signal detection is a promising technique for multiple-input multiple-output space division multiplexing (MIMO-SDM) transmission. These detectors avoid exhaustive searches for all possible transmitted symbol vectors by generating a set of candidate symbol vectors. One problem with the reduced-complexity MAP detectors when used in conjunction with soft input decoders is the inaccuracy of log likelihood ratio (LLR) values since they are computed from a handful of candidate symbol vectors, which degrades the subsequent decoding process. To rectify this weakness, this paper proposes an LLR computation scheme for reduced complexity MAP detectors. The unique feature of the proposed scheme is that it utilizes the statistical property of the MIMO channel metric to narrow down further the number of candidate symbol vectors. Toward this goal, metric selection is performed to select only a statistically good portion of the candidate symbol vectors. Computer simulation results show that the proposed LLR computation scheme is more effective than the existing schemes especially when the number of candidate symbol vectors becomes smaller in reduced-complexity MAP detection.