基于深度学习的协作频谱感知算法

Spectrum sensing is a fundamental problem in cognitive radio. It can be thought of as a dichotomy problem, that is, whether authorized users use authorized spectrum or not. Collaborative spectrum sensing refers to the fact that secondary users infer the spectrum usage status of authorized users by using the signal information received by each other. Aiming at the problem that spectrum sensing is expensive when there are a large number of secondary users, a spectrum sensing algorithm is proposed that is suitable for a small number of secondary users. In this algorithm, the data information collected by a small number of secondary users is first stacked to construct features, and then the Inception-v1 network in deep learning is used for feature extraction. Finally, the trained network is used as a classifier for spectrum perception. Experimental results show that the proposed method has better spectrum sensing performance compared with other fusion methods in the case of fewer collaborative users.