Spectrum Sensing Under Unknown Channel Division
Yirui Du, Bin Li, Xin Jiang, Ruonan Zhang · 2024
In recent spectrum sensing technologies, deep learning methods are receiving increasing attention, mainly due to their ability to effectively improve sensing accuracy in low signal-to-noise ratio (SNR) environments. However, most of the existing deep learning technologies rely on known channel divisions for spectrum sensing, which is often difficult to achieve in real-world applications. To address this issue, this study proposes two innovative spectrum sensing methods: multi-scale serial networks and multi-scale parallel networks. Compared to traditional methods, these two networks can perform spectrum sensing without prior knowledge of channel division, significantly improving the recognition ability of wireless signals and spectrum utilization. Simulation results show that the proposed serial and parallel networks achieve higher sensing accuracy compared to traditional CNN networks under unknown channel division conditions.