Cyclostationary based full-duplex spectrum sensing using adversarial training for convolutional neural networks

Spectrum sensing of orthogonal frequency division multiplex (OFDM) system has always been a challenge in cognitive radios (CR), especially for users which utilize the full-duplex(FD) mode. In this paper, we propose an advanced FD spectrum sensing scheme which can be successfully performed even when encountering severely self-interference from the user terminal. On the basis of ”classification converted sensing” framework, the cyclostationary periodogram generated by OFDM pilots is deduced in the form of images. These images are then plugged into the convolutional neural networks (CNNs) for classifications due to CNN’s strength in image recognition. More importantly, to achieve spectrum sensing against the residual self-interference, as well as the noise pollution and channel fading, we use the adversarial training where a CR-specific, modified training database is proposed. In addition, we propose a design plan of the signal structure for the CR terminal transmitting, which can fit in the proposed spectrum sensing scheme while benefiting its own transmission. Simulation results proved our method possesses an excellent sensing capability for the full-duplex system while achieving higher detection accuracy over the conventional method.