Convolution Neural Network-Based Spectrum Sensing for Cognitive Radio Systems Using USRP with GNU Radio
Gyu-Hyung Lee, Young-Doo Lee, Insoo Koo · 2018
Spectrum sensing is the core technology in cognitive radio systems to find the available channel. In spectrum sensing, the energy detection has a disadvantage that it is difficult to detect the signal of the primary user in the low SNR. In this paper, we use a convolution neural network to enhance the performance in low SNR. For the practical test, the proposed scheme is implemented with Universal Software Radio Peripheral National Instruments 2900 devices. The experimental results of the proposed scheme are compared with the energy detection using accuracy metric according to SNR. With simulation results, we demonstrate that the proposed scheme shows much better performance in low SNR.