FAN-F: A Fourier Analysis Network-based Framework for Efficient and Reliable Spectrum Sensing for 5G and Beyond

Spectrum sensing plays a major role for dynamic spectrum access in Cognitive Radio Networks as we move towards 6G networks characterized by high-frequency operation, ultralow latency requirements, and dense spectrum usage. Most state-of-the-art spectrum sensing techniques suffer from high computational complexity and degraded performance in low Signal-to-Noise Ratio (SNR) environments. To overcome these shortcomings, we propose a Fourier Analysis Network (FAN)-based spectrum sensing approach that specifically addresses the challenge of low detection probability in low-SNR conditions while significantly reducing computational cost. The proposed FAN-based framework integrates Fourier transform properties within the network layers, allowing it to better model periodic and frequency-localized characteristics of wireless signals, such as those found in 5G and 6G-based waveforms. We have evaluated the proposed approach on three datasets that include 5G-NR, DeepSig, and Radar. The experimental results show that the proposed method achieves an improvement in the detection probability of 32.88% and 91.43% compared to the baseline models KAN-SS and PU-DetNet, respectively, at SNR = -10dB. The proposed scheme also outperforms the baseline models in terms of F1-score, with a reduction of 50.60% in FLOPs.