Lattice-Reduction-Aided Partial Marginalization for Soft Output MIMO Detector With Fixed and Reduced Complexity

In this letter, we propose lattice-reduction (LR)-aided partial marginalization (PM) for soft output multiple-input multiple-output (MIMO) detection. PM has the advantages of a fully predictable runtime and convenience in parallelization while offering a well-defined tradeoff between the performance and the computational complexity. However, the computational complexity of PM to achieve a high level of performance is considerably high. In order to reduce the complexity of PM, the proposed scheme performs low-complexity LR-aided marginalization instead of exact marginalization (EM) to avoid the exhaustive approach of the EM, which mainly increases the overall complexity. The experimental results demonstrate that the computational complexity of the proposed scheme is considerably lower with negligible performance degradation compared to conventional PM.

I. Introduction

In multiple-input multiple-output (MIMO) wireless communication systems, the high level of computational complexity for the MIMO detector on the receiver side is one of the most challenging issues in terms of implementation. In addition, because practical systems adopt error correction codes (ECC) to preserve the information bits from the noisy channel, MIMO detectors are required to provide not only the decision of the information bit but also the reliability information for each bit, i.e., soft information. In order to achieve (near-) maximum-likelihood (ML) performance with reduced complexity, list sphere detector (LSD)-based approaches have been discussed [1], [2]. However, their implementation is not favorable in practical systems where data should be processed at a constant rate because their complexity varies depending on the channel realizations and signal-to-noise ratio (SNR).