STRUCTURED SPARSITY-AWARE MESSAGE PASSING DECODER FOR GENERALIZED SPATIAL MODULATION-BASED COMMUNICATION SYSTEM

Abstract

Generalised Spatial Modulation (GSM) is a modification of a massive multipleinput multiple-output (MIMO) system wherein only a few antennas are active at a time, and the active antennas are chosen based on some sequence of bits. In this report, we analyse the performance of a GSM communication system with different decoding algorithms. Since only a few antennas are active, we have sparsity in the transmitted vectors. To exploit this sparsity, we have used Orthogonal Matching Pursuit (OMP) algorithm derived from the compressive sensing (CS) field as one of the options for the decoder. We compare the performance of conventional decoders with OMP. Further, we consider the Temporally correlated GSM system (TC-GSM) which includes a convolution encoder block before the antenna selection block which induces temporal correlation across the active antenna sequence. For the TC-GSM communication system, we target to develop a decoding algorithm that can exploit the temporal correlation and sparsity present in the transmitted vectors to perform better than the OMP-based decoder. We derive a factor graph-based message passing decoder for TC-GSM, which is able to exploit the underlying temporal correlation across the active antennas in multiple transmission epochs. Through extensive simulations, we demonstrate the efficacy of our proposed structured sparsity-aware TC-GSM decoder and compare its performance to OMP and conventional MMSE-based decoding schemes.