Two-stage Active User Detection with False Alarm Correction for GF-NOMA System

Grant-free (GF) low-density signature orthogonal frequency division multiplexing (LDS-OFDM) is one of the most generic grant-free non-orthogonal multiple access (GF-NOMA) schemes. This paper proposes a two-stage active user detection (AUD) consisting of the initial AUD stage and the false alarm correction stage. In the initial AUD stage, an initial active user set is efficiently estimated by the conventional cover decoder, which may still contain a few false alarms. Hence, a false alarm correction stage is further invoked, which consists of two cooperative and iterative detection components, namely, message passing algorithm (MPA) based data decoder and belief propagation (BP) based false alarm corrector. Based on users' signature spreading sequences in the initial active user set, a tanner graph with potential redundant-edges is constructed, on which MPA is executed for data decoding. In turn, with the aid of decoded data symbols, the remaining false alarms in the initial active user set are further removed by the false alarm corrector. Hence, the tanner graph could evolve iteratively. Furthermore, the false alarm performance of the initial AUD stage is theoretically analyzed. Based on this analysis, the LDS matrix used in GF LDS-OFDM is optimized. Finally, the complexity of our proposal is also provided.