Deep learning-based efficient spectrum sensing framework for enhanced data transmission in cognitive radio sensor networks

Abstract Cognitive Radio (CR) is a wireless communication technology that is intelligent and designed to maximize the use of wireless spectrum, improve communication system performance, and optimize wireless spectrum usage. The rapid growth of wireless communication has led to an increase in the demand for spectrum. Spectrum Sensing (SS) is a feature that allows CR to preserve Primary Users spectrum access from SUs. SS improves the efficiency and flexibility of CR devices through maximizing spectrum use. As IoT, wireless technology and other technologies are becoming more prevalent, adaptability becomes more and more critical. The traditional methods of frequency distribution are inefficient and can lead to under-use of bandwidth, as bandwidth demands increase. AI and deep learning (DL) have improved the spectrum sensing capabilities. This allows CR to adapt dynamically to changes in radio frequency environments. In the proposed model recurrent neural networks and transformer networks DL techniques integration is used. RNNs are useful for detecting temporal dependencies. Transformer networks, on the other hand, have better abilities to detect long-range dependences in spectrum data. This integrated approach aims to improve performance by combining these two architectures. The proposed approach uses digital modulations to improve SS while assessing model performance with metrics such as the Jaccard Index (JI), CKC Score, and F1 Score. The model shows a better detection probability and reduced errors in sensing, especially at low signal-to-noise ratios.