Deep Learning Detector and Two-Stage LLR Detector for MCIK-NOMA System

Multi Carrier Index Keying based non-orthogonal multiple access (MCIK-NOMA) system is a novel variant of IM-NOMA that significantly improves the error performance while maintaining a predetermined spectral efficiency. In this research, a deep learning (DL) detector and a two-stage LLR detector based MCIK-NOMA is introduced. To recover the information bits within the IM-NOMA system, we specifically propose a unique DL-based detector called Deep-MCIK-NOMA. This detector makes use of a deep neural network having completely connected layers. We preprocess the incoming signal and channel vectors depending on domain-specific knowledge before they enter the neural network to improve the performance of Deep-MCIK-NOMA. In comparison to the conventional detectors, simulation findings show that Deep-MCIK-NOMA achieves a BER close to the perfect level while demonstrating a quicker runtime. LLR based detector is also proposed for MCIK-NOMA system and shown that it performs same as ML detector, but with less complexity. We examine the bit error rate and efficiency of our proposed system, and the simulation results consistently demonstrate its superiority to the traditional methods like NOMA and OMA systems.