Deep learning techniques for robust non-cooperative spectrum sensing in cognitive radio networks

Abstract

The objective of this thesis is to exploit recent advances in deep learning-based techniques for spectrum sensing in cognitive radio networks. Firstly, we propose a deep learning-based approach for denoising and non-cooperative spectrum sensing under Rayleigh fading channels, where the model learns to sequentially denoise the signal and perform spectrum sensing, thereby improving detection accuracy, particularly under low signal-to-noise ratio (SNR) conditions. However, as the detection performance is highly affected by the denoising stage, we introduce a joint learning strategy that simultaneously performs denoising and spectrum sensing for orthogonal frequency-division multiplexing (OFDM) systems, optimizing the trade-off between the two tasks and further enhancing performance. Following that, to enhance the robustness and adaptability of spectrum sensing in dynamic wireless environments, we develop a contrast learning-based supervised spectrum sensing technique that extracts channel-invariant features, enabling the model to adapt to new environments without extensive retraining. To tackle the issue of limited labeled data, we extend our work by proposing a dual contrast self-supervised learning-based spectrum sensing (DC4S) framework. This approach leverages unlabeled data to learn meaningful representations, reducing the dependency on labeled samples and enabling effective spectrum sensing even in scenarios where labeled data are scarce. Additionally, it is designed to be robust against different sub-types of tapped-delay line (TDL) channels and certain OFDM signal hyper-parameters. Overall, the proposed methods progressively address the challenges of high-level noise, dynamic environments, and limited labeled data. This thesis illustrates the potential of deep learning-based techniques in enhancing the performance and robustness of spectrum sensing in cognitive radio networks, contributing to the development of more efficient and reliable wireless communication systems.