Deep Learning-Based Joint NOMA Signal Detection and Power Allocation in Cognitive Radio Networks

Abstract

Presently, Non-Orthogonal Multiple Access (NOMA) frequently uses Successive Interference Cancellation (SIC) with channel estimation to detect the receivers’ signal successfully. However, it’s a slow and sophisticated process. Consequently, the decoding precision of previous Primary Users (PUs) / Secondary Users (SUs) is degraded due to error propagation. These shortcomings are mitigated by the proposed Deep Learning (DL) schemes for signal detection and Power Allocation (PA) in NOMA-based futuristic Cognitive Radio Networks (CRNs). Hence, proposed technique is optimized jointly and determines the desired solution in one step without channel estimation through skilled Deep Neural Network (DNN) approach. It can understands the things because it is the core (so-called heart) of DL, which integrates user detection with channel estimation and PA. The optimization problem for fair PA is planned by combining proposed technique for optimizing systems threshold throughput. In addition, DNN training algorithms is also analyzed, and used to retrieve users’ information from the aired signal directly during the testing stage. The real-time experimental results and analytic outcomes are investigated and validate the supremacy of the proposed scheme over prevailing techniques, respectively. Furthermore, the robustness of the DL technique is additionally evaluated for its dominance over existing approaches.