To enhance the rapid detection and accurate identification of radio signals, we propose using a deep learning method called YOLOv5s-CWMDSQ for signal identification. This method involves sending the waterfall map of a radio signal spectrum to an improved target detection network for classification. Our algorithm builds upon YOLOv5 by introducing improvements such as the CA attention mechanism to enhance the model’s accuracy in locating and identifying targets, and the boundary loss function WIOU to improve overall detector performance. We also replaced the MESwish activation function, utilized Decoupled Head to speed up network convergence, and incorporated the SPPFCSPC module to elevate the model’s receptive field and feature expression ability through multi-scale spatial pyramid pooling. Lastly, we integrated data enhancement to improve the diversity, robustness, and generalization ability of the model, achieving higher accuracy and performance. Experimental results exhibited an increase in the mAP value from 82.2% to 90.1% and detection speed of 44.488FPS in the dataset with 3,000 samples of 6 signal types, proving the model’s superior accuracy and real-time capabilities.

Qinjun Li, Yan Zhao, Tianwei Cui, Yuying Wu. The Research of Spectrum Signal Recognition Based on Improved YOLOv5 Algorithm. Academic Journal of Computing & Information Science (2023), Vol. 6, Issue 7: 101-113. https://doi.org/10.25236/AJCIS.2023.060714.

[1] Zhao Q, Mao X, Pang K. Development status and trend of civilian aircraft air-to-ground broadband communication system (in Chinese) [J]. Communications Technology, 2019, 52(10): 2428-2432. 

[2] Wang W. Design and implementation of transmit and receive channels for broadband signal processing module (in Chinese) [J]. Communications Technology, 2017, 50(07): 1560-1563. 

[3] Dobre, Octavia A. Signal Identification for Emerging Intelligent Radios: Classical Problems and New Challenges [J]. IEEE instrumentation & measurement magazine, 2015, 18(2):11-18. 

[4] Tang L. Overview of FPGA-based broadband signal detection design (in Chinese) [J]. Science and Informationization, 2021(12): 13-15. 

[5] Wang H, Tao M, Zhao H, Jiang C. Simulation and implementation of multi-antenna signal detection based on FPGA and DSP (in Chinese) [J]. TV Technology, 2014, 38(15): 76-79. DOI: 10. 16280/j. videoe. 2014. 15. 023. 

[6] Lei Z , Jiang M , Yang G ,et al. Towards recurrent neural network with multi-path feature fusion for signal modulation recognition[J].Wireless Networks, 2022.DOI:10.1007/s11276-021-02877-8.

[7] Lin Lin, Xuezhi He, Jinbao Xie. Radio Signals Modulation Mode Recognition Based on Semisupervised Deep Learning[C]. //2018 3rd International Conference on Computer Science and Information Engineering (ICCSIE 2018). Proceedings of the 2018 3rd International Conference on Computer Science and Information Engineering (ICCSIE 2018). 2018: 299-305. 

[8] Zeng Y , Zhang M , Han F , et al. Spectrum Analysis and Convolutional Neural Network for Automatic Modulation Recognition [J]. IEEE wireless communications letters, 2019, 8(3): 929-932. DOI:10. 1109/LWC. 2019. 2900247.

[9] Qian Mao, Fei Hu, Qi Hao. Deep Learning for Intelligent Wireless Networks: A Comprehensive Survey[J]. Communications surveys & tutorials, 2018, 20(4):2595-2621. DOI: 10. 1109/COMST. 2018. 2846401. 

[10] Ya Tu, Yun Lin, Haoran Zha, Ju Zhang, Yu Wang, Guan Gui, Shiwen Mao. Large-scale real-world radio signal recognition with deep learning[J]. Chinese Journal of Aeronautics, 2022, 35(09):35-48. 

[11] J. L. Ziegler, R. T. Arn and W. Chambers, "Modulation recognition with GNU radio, keras, and HackRF," 2017 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN), Baltimore, MD, USA, 2017. 

[12] Parnas, David Lorge. The Real Risks of Artificial Intelligence Incidents from the early days of AI research are instructive in the current AI environment [J]. Communications of the ACM, 2017, 60(10):27-31. 

[13] O'Shea T J, Corgan J, Clancy T C. Convolutional Radio Modulation Recognition Networks[J]. Springer, Cham, 2016. 

[14] O'Shea T J, Corgan J, Clancy T C. Unsupervised representation learning of structured radio communication signals [J]. IEEE, 2016. 

[15] O'Shea T J, West N, Vondal M, et al. Semi-supervised radio signal identification[J]. IEEE, 2017. 

[16] O'Shea T J, West N. Radio Machine Learning Dataset Generation with GNU Radio[C]// Proceedings of the GNU Radio Conference. 2016. 

[17] Kulin M, Kazaz T, Moerman I, et al. End-to-end Learning from Spectrum Data: A Deep Learning approach for Wireless Signal Identification in Spectrum Monitoring applications [J]. IEEE Access, 2018, 6:18484-18501. 

[18] Selim A, Paisana F, Arokkiam J A, et al. Spectrum Monitoring for Radar Bands using Deep Convolutional Neural Networks[J]. arXiv e-prints, 2017. 

[19] Maglogiannis V, Shahid A, Naudts D, et al. Enhancing the Coexistence of LTE and Wi-Fi in Unlicensed Spectrum Through Convolutional Neural Networks [J]. IEEE Access, 2019, 7:28464-28477. 

[20] Girshick R, Donahue J, Darrell T, et al. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation[J]. IEEE Computer Society, 2014. 

[21] Redmon J, Farhadi A. YOLO9000: Better, Faster, Stronger [C]// IEEE Conference on Computer Vision & Pattern Recognition. IEEE, 2017:6517-6525. 

[22] Wei L, Dragomir A, Dumitru E, et al. SSD: Single Shot MultiBox Detector[J]. Springer, Cham, 2016. 

[23] Li X, Chen T. Radio frequency spectrum artificial intelligence recognition technology based on TensorFlow (in Chinese) [J]. China Radio, 2018(9): 55-56, 60. DOI:10. 3969/j. issn. 1672-7797. 2018. 09. 047. 

[24] Zhou X, He X, Zheng C. Radio signal identification based on image deep learning (in Chinese) [J]. Journal of Communications, 2019, 40(7): 114-125. DOI:10. 11959/j. issn. 1000-436x. 2019167. 

[25] Zhou Y, Hou J, Li J, et al. Spectrum signal identification based on frequency domain stacking and deep learning (in Chinese) [J]. Journal of Computer Applications Research, 2023, 40(3): 874-879. DOI:10. 19734/j. issn. 1001-3695. 2022. 07. 0586.