Deep Learning Approaches for Spectrum Sensing in Cognitive Radio Networks
Sadaf Nazneen Syed, Pavlos I. Lazaridis, Faheem A. Khan, Qasim Zeeshan Ahmed, Maryam Hafeez, Violeta Holmes, Ioannis P. Chochliouros, Zaharias D. Zaharis
Abstract
The number of network devices continues to rise as we advance towards 6G communication systems. A new range of frequencies is allocated while the earlier resources remain underutilized. Cognitive Radio (CR) enables the dynamic spectrum management of frequencies while detecting the unoccupied bands with the aid of spectrum sensing. By adopting Deep Learning (DL) for spectrum sensing, the performance of the 6G networks can be made more robust. This paper presents a survey of several DL algorithms such as, Multilayer Perceptrons (MLPs), Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks and combination of CNNs and LSTMs that have been applied for performing spectrum sensing. The application of DL to spectrum sensing is then demonstrated by considering it as a binary classification problem. An MLP with its hyperparameters optimized by using Grid Search algorithm is proposed to classify a dataset consisting of RadioML 2018.01A and noise samples with high accuracy.