Structured Random Codebook Design for GABP Iterative Detection in Massive SCMA

This paper proposes a random codebook design and its constraints for Gaussian belief propagation (GaBP) in massive Sparse Code Multiple Access (SCMA). In typical SCMA, a message passing algorithm (MPA) is applicable as the detection scheme for multi-user detection (MUD) on the assumption of a predefined optimal codebook. As the number of transmit streams increases, however, it becomes quite difficult to design optimal codebook due to a large number of its candidates. To address such an inconvenient problem, we employ the structured random codebook and GaBP iterative detection, which may achieve the near-optimal performance thanks to the diversity gain in the large-scale MUD. Moreover, an abnormal noise enhancement induced in GaBP can be suppressed by appropriate constraints in random codebook design. To improve the convergence property of GaBP iterative detection, we propose a novel random codebook design and its constraints for GaBP. Finally, we demonstrate the validity of our proposed method, in terms of improvement of bit error rate (BER) performances.