Deep Denoising and Clustering-Based Cooperative Spectrum Sensing for Non-Orthogonal Multiple Access

Abstract

Non-orthogonal multiple access (NOMA) technology offers higher communication throughput than its orthogonal multiple access counterpart. However, it also poses new challenges for spectrum sensing technology. Accurate spectrum sensing of a channel occupied by multiple users is challenging, especially in low signal-to-noise ratio environments. To improve the spectrum sensing performance, a spectrum sensing algorithm based on deep denoising and clustering is developed for power domain NOMA. First, a novel auto-encoder for deep denoising that can filter out the noise of signals is proposed. Then the auto-encoder is transplanted to a variational auto-encoder for extracting features with high separability. Finally, a ring K-means++ algorithm is proposed to classify features. In the experiments, simulations of algorithms are carried out in various scenarios with different numbers of primary users. The results show that the proposed algorithm outperforms other algorithms.