Learning-Based Near-Orthogonal Superposition Code for MIMO Short Message Transmission

Massive machine type communication (mMTC) has attracted new coding schemes optimized for reliable short message transmission. In this paper, a novel deep learning-based near-orthogonal superposition (NOS) scheme is proposed to transmit messages in multiple-input multiple-output (MIMO) channels mMTC applications. the MIMO-NOS scheme, neural network-based encoder via end-to-end learning with corresponding detector/decoder superposition-based auto-encoder framework including MIMO channel. The spreads information bits multiple high dimensional vectors be combined (superimposed) into single vector and reshaped space-time For receiver, we propose looped K -best tree-search algorithm cyclic redundancy check (CRC) assistance enhance error correcting ability block-fading comprehensive understanding of further quantify gain from individual components/modules framework, analyze decoding complexity measured by floating point operations (FLOPs). Simulation results show outperforms maximum likelihood (ML) detection polar code CRC-assisted list 1 – 2 dB various systems (32 64 bit)