| 摘要: |
| 为了解决卫星通信系统在对抗电磁环境中的干扰实时检测识别问题，提出了一种基于自注意力(Self-attention，SA)机制的高效轻量化网络模型。通过采用DenseNet加速对原始IQ信号的特征提取，并引入自注意力模块代替参数量较大的多重卷积层，实现对卫星通信系统中常见的干扰样式进行分类识别。仿真结果表明，在识别准确率方面达到常规的神经网络模型和算法性能水平的条件下，所提模型在网络复杂度和运算时延方面得到有效压缩。 |
| 关键词: 卫星通信系统 干扰信号检测 自注意力机制 稠密卷积网络 轻量级模型 |
| DOI：10.20079/j.issn.1001-893x.220220002 |
|
| 基金项目: |
|
| Self-attention-based jamming signals detection and classification |
| WANG Ruidong,WANG Shilian,ZHANG Wei,ZHANG Yanlong |
| (1.College of Electronic Science and Technology,National University of Defense Technology,Changsha 410073,China;Unit 95869 of PLA,Huhhot 010020,China)) |
| Abstract: |
| To solve the problem of real-time detection and identification of interference in the adversarial electromagnetic environment for satellite communication systems,an efficient lightweight network model based on the self-attention(SA) mechanism is proposed.By using DenseNet model to accelerate the feature extraction of the raw IQ signal and introducing the SA module instead of the convolutional layer with a larger number of parameters,the classification and recognition of common interference patterns in satellite communication systems is achieved.The experimental results reveal that under the condition that the performance level of conventional neural network models and algorithms in terms of recognition accuracy is achieved,the proposed model is effectively compressed in terms of network complexity and computation delay. |
| Key words: satellite communication system jamming signal detection self-attention mechanism DenseNet lightweight model |
