| 引用本文: | 晏行伟， 孔令轩， 刘 坤， 刘安娜. 基于MobileNet⁃DOA的无人机射频信号识别方法[J]. 雷达科学与技术, 2025, 23(1): 57-66.[点击复制] |
| YAN Xingwei, KONG Lingxuan, LIU Kun, LIU Anna. Drone Radio Frequency Signal Identification Method Based on MobileNet⁃DOA[J]. Radar Science and Technology, 2025, 23(1): 57-66.[点击复制] |
|
| 摘要: |
| 针对当前低信噪比环境下，基于射频信号的无人机型号和飞行模式识别率低的问题，本文提出了一种基于MobileNet?DOA模型的无人机射频信号识别方法。该方法首先对原始无人机射频信号进行基于变分模态分解的信号预处理，降低背景噪声和同频干扰，然后利用短时傅里叶变换将预处理信号转换为时频图，最后利用MobileNet?DOA模型完成无人机射频信号识别。在模型方面，本文首先将DOConv卷积融合到MobileNetv4模型中，在增强模型特征提取能力的同时，提高了训练和运算速度。其次，使用FA注意力机制进一步提升了模型在低信噪比环境下的识别准确率。实验结果表明，该方法在-15~15 dB信噪比范围内的平均检测准确率达到了94.83%，可应用于无人机实时检测识别系统中。 |
| 关键词: 无人机射频信号识别 变分模态分解 MobileNet模型 DOConv卷积 FA注意力机制 |
| DOI：DOI:10.3969/j.issn.1672-2337.2025.01.006 |
| 分类号:TN911.7 |
| 基金项目:国家自然科学基金（No.62173124， 62101563）； 河北省自然科学基金（No.F2022202064） |
|
| Drone Radio Frequency Signal Identification Method Based on MobileNet⁃DOA |
|
YAN Xingwei, KONG Lingxuan, LIU Kun, LIU Anna
|
|
1. College of Electronic Science and Technology, National University of Defense Technology, Changsha 410073, China;2. Tianjin Advanced Technology Research Institute, Tianjin 300459, China;3. School of Artificial Intelligence, Hebei University of Technology, Tianjin 300400, China;4. Beijing Institute of Tracking and Telecommunications Technology, Beijing 100094, China
|
| Abstract: |
| Aiming at the problem of low identification rates of drone models and flight modes based on radio frequency signals in low SNR environments, a drone radio frequency signal identification method based on MobileNet?DOA model is proposed in this paper. Firstly, the raw drone radio frequency signal is preprocessed by using variational mode decomposition to reduce background noise and co?frequency interference. Then, the preprocessed signal is transformed into time?frequency images by short?time Fourier transform. Finally, the MobileNet?DOA model is employed to complete the identification of drone radio frequency signals. In terms of the model, this paper integrates DOConv into MobileNetv4 model, enhancing the feature extraction capability while improving the training and computation speed. Additionally, the FA attention mechanism is utilized to further improve the identification accuracy of the model in low SNR environment. Experimental results show that the average detection accuracy of the proposed method achieves 94.83% in the SNR range from ?15 dB to 15 dB, making it applicable to real?time detection and identification system of drone. |
| Key words: drone radio frequency signal identification variational mode decomposition MobileNet model DOConv FA attention mechanism |
