一种基于特征融合Transformer的频谱感知方法研究  

A Study on a Spectrum Sensing Method Based on Feature-Fusion Transformer

作　　者：刘思佚 徐东辉 刘丁胤 胡国杰 康凯 LIU Si-yi;XU Dong-hui;LIU Ding-yin;HU Guo-jie;KANG Kai(Rocket Force Engineering University Combat Support College,Xi’an 710025,China)

机构地区：[1]火箭军工程大学作战保障学院,陕西西安710025

出　　处：《中国电子科学研究院学报》2024年第7期658-666,共9页Journal of China Academy of Electronics and Information Technology

摘　　要：随着无线通信技术的迅速发展,电磁频谱资源日益紧张,高效的频谱感知和信号分类技术变得至关重要,如何实现智能频谱感知与分类识别是当前研究的热点方向。文中针对复杂电磁环境下如何提高信号分类性能,提出了一种基于特征融合Transformer的频谱智能感知方法。该方法设计了特征融合层和改进的位置编码方案,通过优化Transformer架构,增强了模型对不同类型信号的识别能力。实验结果表明,改进模型的准确率表现出显著优势,分类准确率达到99.3%,较现有模型高出4.1个百分点。此外,在不同信噪比条件下,模型展现了卓越的抗噪性能,进一步证明了其在复杂电磁环境中的应用潜力和研究价值。With the rapid development of wireless communication technology,electromagnetic spectrum resources are becoming increasingly scarce,making efficient spectrum sensing and signal classification technologies crucial.Intelligent spectrum sensing and classification have become a hot research topic.This paper proposes a spectrum intelligent sensing method based on feature fusion Transformer to improve signal classification performance in complex electromagnetic environments.The proposed method designs a feature fusion layer and an improved positional encoding scheme,optimizing the Transformer architecture to enhance the model’s ability to recognize different types of signals.Experimental results show that the improved model demonstrates significant advantages in metrics such as accuracy,achieving a classification accuracy of 99.3%,which is 4.1 percentage points higher than existing models.Furthermore,the model exhibits excellent noise resistance under different signal-to-noise ratio conditions,further proving its application potential and research value in complex electromagnetic environments.

关 键 词：特征融合 频谱感知 TRANSFORMER 分类准确率 抗噪声性能 

分 类 号：TN92[电子电信—通信与信息系统]