Smart Sensing Framework Using Kolmogorov-Arnold Networks for Enhanced Detection in 5G Environments
Prapti Patel, Dhaval K. Patel · 2025
Spectrum sensing is crucial for dynamic spectrum access in Cognitive Radio Networks (CRNs), particularly as we advance to 5G technology, which operates across higher frequency bands and supports ultra-reliable low latency communications. In this study, we propose a Kolmogorov-Arnold Network (KAN)-based spectrum sensing scheme specifically designed to address the challenges of low probability of detection in low Signal-to-Noise Ratio (SNR) environments commonly found in 5G networks. The proposed scheme improves detection performance by effectively handling the increased noise and interference associated with higher frequency bands and massive MIMO setups. Our experimental results demonstrate a significant improvement in the probability of detection at SNR = -10 dB. Compared to Long Short-Term Memory (LSTM) networks, and PU-DetNet, the KAN-based scheme shows an improvement in detection probability by 67.43%, and 47.29%, respectively. This enhancement is validated across diverse datasets, including empirical, 5G New Radio (NR), DeepSig, and radar datasets. The performance of the proposed KAN-based scheme is analyzed based on Accuracy, Precision, and F1-score which is crucial for optimizing spectrum utilization and ensuring reliable communications in the complex and high-demand 5G framework.