Spectrum Sensing based on an improved deep learning classification for cognitive radio

Cognitive Radio (CR) technology enables the efficient exploit of radio spectrum by utilizing existing unused frequencies. Spectrum Sensing is the most important process in CR by allowing users to detect unused communication channels. The paper proposes an enhancement of a Deep Learning (DL) classification approach for Spectrum Sensing. This approach introduces the Linear Support Vector Machine (SVM) as a classification layer in the DetectNet based DL technique. Simulation results of the proposed approach provide better performance in term of detection at low signal-to-noise ratio (SNR) compared to the DetectNet technique.