Ieee Transactions on Vehicular Technology Compressive-sensing-based Multiuser Detector for the Large-scale Sm-mimo Uplink 2
Zhen Gao,L. Dai,Zhaocheng Wang,Sheng Chen,L. Hanzo
2015-01-01
Abstract:Conventional spatial modulation (SM) is typically consid5 ered for transmission in the downlink of small-scale multiple-input– 6 multiple-output (MIMO) systems, where a single antenna element (AE) of 7 a set of, e.g., 2 AEs is activated for implicitly conveying p bits. By contrast, 8 inspired by the compelling benefits of large-scale MIMO (LS-MIMO) 9 systems, here, we propose an LS-SM-MIMO scheme for the uplink (UL), 10 where each user having multiple AEs but only a single radio frequency 11 (RF) chain invokes SM for increasing the UL throughput. At the same time, 12 by relying on hundreds of AEs and a small number of RF chains, the base 13 station (BS) can simultaneously serve multiple users while reducing the 14 power consumption. Due to the large number of AEs of the UL users and 15 the comparably small number of RF chains at the BS, the UL multiuser sig16 nal detection becomes a challenging large-scale underdetermined problem. 17 To solve this problem, we propose a joint SM transmission scheme and 18 a carefully designed structured compressive sensing (SCS)-based multi19 user detector (MUD) to be used at the users and the BS, respectively. 20 Additionally, the cyclic-prefix single carrier (CPSC) is used to combat 21 the multipath channels, and a simple receive AE selection is used for the 22 improved performance over correlated Rayleigh-fading MIMO channels. 23 We demonstrate that the aggregate SM signal consisting of multiple UL 24 users’ SM signals of a CPSC block exhibits distributed sparsity. Moreover, 25 due to the joint SM transmission scheme, aggregate SM signals in the same 26 transmission group exhibit group sparsity. By exploiting these intrinsically 27 sparse features, the proposed SCS-based MUD can reliably detect the 28 resultant SM signals with low complexity. Simulation results demonstrate 29 that the proposed SCS-based MUD achieves a better signal detection 30 performance than its counterparts even with higher UL throughput. 31