Application of deep learning algorithm for spectrum sensing in cognitive radio

• Main Author: Zhang, Zihan
• Other Authors: Teh Kah Chan
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering
• Online Access: https://hdl.handle.net/10356/173534
• Institution: Nanyang Technological University

In recent years, with the development of related technologies, the requirements for frequency resources are also increasing day by day, but due to the scarcity of wireless resources, the utilization of resources is limited. With the increasing number of users and the emergence of many electromagnetic devices, radio frequency resources are increasingly scarce. Therefore, the effective utilization of spectrum-related techniques for the monitoring and management of frequencies has emerged as a prominent and actively pursued subject in contemporary scientific research. As a key resource in wireless communications, spectrum resources have become a hot spot for strategic competition and games among countries due to their scarcity. To improve the utilization of spectrum resources, cognitive radio technology has gradually emerged, and the secondary utilization of spectrum resources is realized through the method of opportunistic access to licensed frequency bands by unlicensed users. In the realm of cognitive radio, unlicensed users are required to employ spectrum sensing techniques to ascertain the presence of authorized users within the target frequency band before initiating their own transmissions. This proactive approach ensures avoidance of any interference with the communication of authorized users. Therefore, improving the accuracy of spectrum sensing has become the focus of research in cognitive radio. The deep learning technology that has achieved rapid development in recent years has demonstrated its superior performance in the feature extraction and classification of massive data. Fundamentally, spectrum sensing can be categorized as a binary classification task aimed at discerning the existence or absence of authorized user signals, employing numerous signal samples for accurate differentiation. Therefore, deep learning technology has great application potential in the field of spectrum sensing. This dissertation delves into research on deep learning-based spectrum sensing technology. Initially, it comprehensively analyzes the challenges and complexities encountered in existing spectrum sensing approaches, followed by a thorough exploration of the application of deep learning in addressing spectrum sensing problems. The dissertation also presents an overview of traditional spectrum sensing techniques, highlighting their respective strengths and limitations. Furthermore, a detailed exposition of deep learning theory is provided, elucidating the key components and their significance. Building upon the introduced deep learning framework, the dissertation proposes a novel spectrum sensing method that integrates feature extraction and convolutional neural networks, leveraging relevant algorithms from both traditional deep learning and spectrum sensing domains. Experimental validation is conducted to substantiate the efficacy of the proposed method. This dissertation also proposes a deep learning based orthogonal frequency-division multiplexing (OFDM) signal spectrum sensing algorithm for OFDM signals. A series of preprocessed data sets are input into the improved convolutional neural network to complete learning and build a spectrum sensing model.