Deep learning-based spectrum sensing in cognitive radio

Abstract

With the increase in demand for spectrum resources, cognitive radio is dependent heavily to efficiently managing the overwhelming radio spectrum scarcity. To ensure that the spectrum resources in cognitive radio are fully utilized, spectrum sensing is involved to identify the presence and absence of authorized primary users in the network and allow unauthorized secondary users to access when the spectrum is left idle.

Conventional energy detection is a popular method used as it does not require prior information about the signal however it has limitations on its detection performance due to the uncertainty of noise. Hence, deep learning methods such as convolutional neural networks and long short-term memory has been introduced as it is able to identify patterns of the signal.

In this project, we will be comparing the performance of the conventional and deep learning methods in identifying weak signals under low signal-to-noise ratio levels to prove that the deep learning method is more effective in tackling the problem of spectrum shortage.