Multi-Stage Message Passing Algorithm for SCMA Downlink Receiver

Abstract

Sparse code multiple access (SCMA) is a novel non- orthogonal access technique for future multiple access, which increases the spectral efficiency by directly mapping data streams of multiple users to sparse codewords in different SCMA layers, and these SCMA layers are imposed on the same SCMA block composed of limited resource elements (REs). As the sparsity of codewords, message passing algorithm (MPA) can be employed to detect multiple users' data with near optimal performance. But the complexity of MPA is still high, especially in the case of overload, where the downlink user receiver needs to implement the whole iterative process of MPA to get its data although the rest users' data is unnecessary for it. In this paper, we propose a novel multi-stage MPA algorithm, which sorts the level of signal to noise ratio (SNR) of imposed users to decide the detection order and iterations of each stage. Further by reconstructing the factor graph, it can realize downlink user detection without processing the whole MPA, which reduces the complexity by decreasing iterations. Combined with multi-stage MPA, power allocation should be reconsidered according to user's channel condition, which can further improve the performance of the proposed algorithm. The simulation results demonstrate that the performance of proposed detection algorithm is close to MPA at the low SNR region, and gains about 1dB over max-log-MPA, while the complexity is decreased around a half than MPA.