When machine learning meets compressive sampling for wideband spectrum sensing
Bassem Khalfi, Adem M. Zaid, Bechir Hamdaoui · 2017
This paper proposes a novel technique that exploits spectrum occupancy behaviors inherent to wideband spectrum access to enable efficient cooperative wideband spectrum sensing. Our technique requires lesser number of sensing measurements while still recovering spectrum occupancy information accurately. It does so by leveraging compressive sampling theory to exploit the block-like occupancy structure of wideband spectrum access. Our technique is also adaptive in that it accounts for the variability of spectrum occupancy over time. It exploits supervised learning to provide and use accurate realtime estimates of the spectrum occupancy. Using simulations, we show that our proposed technique outperforms existing approaches by making accurate spectrum occupancy decisions with lesser sensing communication and energy overheads.