Low-Complexity Large MIMO Detection Based on Beam-Domain Local LMMSE Filters

Linear minimum mean square error (LMMSE) filters are often utilized to achieve low-complexity multi-user detection (MUD) in uplink large multi-input multi-output (MIMO) systems. As the scale and density of MIMO systems grow towards truly massive setups, however, the LMMSE detection requiring high-dimensional matrix inversion operations becomes computationally expensive. As a promising approach to tackle this issue, the local LMMSE (LLMMSE) detector was proposed, where a contiguous block of beams can be selected for each user to construct the reduced beam-domain channels, assuming the use of digital beamforming at a base station (BS). A main issue is the performance degradation according to the angular spread of the received signal, due to the information loss induced by an excessive dimensionality reduction aiming at the computational reduction. To alleviate this issue, this paper proposes to selectively combine the information from the overlapped LLMMSE filters in the log-likelihood ratio (LLR) domain. In addition, this method is extended to probabilistic data association (PDA)-based iterative detection scheme for further enhancement of communication reliability. The efficacy of the proposed methods is demonstrated by simulation results in terms of the bit error rate (BER) performance and the computational cost.