PAPR Reduction for OFDM-Over-NOMA Using Hybrid Partial Selective Mapping and Artificial Neural Networks

Orthogonal Frequency Division Multiple Access (OFDMA) combined with Non-Orthogonal Multiple Access (NOMA) represents a promising hybrid approach for next-generation wireless communication systems. By leveraging the strengths of both OFDM and NOMA, this integration aims to improve system capacity and spectral efficiency while supporting Ultra-Reliable Low-Latency Communications (URLLC) for a wide range of devices. Despite its potential, one major limitation of this technology is the high peak-to-average power ratio (PAPR), which can negatively impact system performance. Traditional methods for PAPR reduction, such as Selective Mapping (SLM), rely on searching for optimal phase rotations across all sub carriers, but these methods are computationally intensive. In this work, we introduce a Partial SLM approach that reduces computational complexity by using a limited number of phase sequences in conjunction with an artificial neural network (ANN). The pre-trained ANN model is employed to identify signals with minimized PAPR efficiently. Results demonstrate that the proposed method effectively lowers PAPR to near-optimal levels while significantly reducing computational demands.