Iterative Gaussian-Approximated Message Passing Receiver for MIMO-SCMA System

Abstract

Sparse code multiple access (SCMA) and multiple-input multiple-output (MIMO) are viewed as two key techniques to address the high spectral efficiency requirements for the 5G wireless system. However, the complexity of standard message passing algorithm (MPA) for the MIMO-SCMA detection yet increases exponentially with the degree of resource nodes $d_f$ and the number of transmit antennas $N_t$ . In this paper, to tackle this problem, we design two types of low-complexity MPA over the MIMO-SCMA extended factor graph to perform joint multiuser and MIMO detection. The main idea of the proposed algorithms is approximating the discrete interferences at function nodes (FNs) as the continuous Gaussian messages so that the traversal enumeration at the FN update can be avoided. The symbol level Gaussian approximated MPA (GAMPA-S) updates the symbol likelihood during the detector's inner-loop iterations which is similar to the standard MPA, and this algorithm is compatible to any SCMA codebooks. The bit level Gaussian approximated MPA (GAMPA-B) updates the bit likelihood ratio which shows lower complexity compared to the GAMPA-S. However, the GAMPA-B requires that the SCMA codebooks can satisfy the bit significance expansion property. In the receiver, the proposed detectors are combined with users’ decoders to form the outer-loop structure, in which a critical damping factor (DF) is designed to compress the amplitude of extrinsic information exchanged between the detector and decoders. The extrinsic information transform chart is introduced to optimize the DF. Consequently, the complexity of the proposed receiver just increases linearly with $d_f$ and $N_t$ while its performance is comparable or even better than the conventional receiver. This would effectively boost the practical use of MIMO-SCMA techniques.