Spectrum sensing using deep learning for proficient data transmission in wireless sensor networks for wireless communication

Cognitive radio has been recommended for enhancing wireless spectrum usage, and an important part of cognitive radio is spectrum sensing. Conventional spectrum sensing techniques rely on energy detection and feature extraction from a received signal at a specific position. The advancement of artificial intelligence and deep learning has provided a chance to increase spectrum sensing efficiency. The proposed investigation presented a hybrid model comprising CNN, BILSTM, and Transformer Networks (TN) for spectrum sensing. With the aid of Transformer networks, the proposed model improves sensing accuracy even for lower SNR signals, particularly at -20 dB. The results of the proposed technique show that CNN-BILSTM-TN enhances spectrum sensing in comparison to earlier models studied in this domain. The proposed approach outperforms prior methods in terms of F1 Score and has an improved probability of detection at -20 dB and a lower likelihood of missed detection. Analysis of ten modulation techniques' performance criteria, such as the Jaccard index, Cohen Kappa coefficient, and Matthew's correlation coefficient, demonstrates that the suggested method's spectrum sensing is better even at lower SNRs.