基于改进残差网络的雷达辐射源多标签识别

Multi-label Recognition of Radar Emitter Based on Improved Residual Network

针对现有的雷达辐射源识别方法具有低信噪比情形下识别精度低、无法识辩多个辐射源等缺点,文中提出融入注意力机制的残差网络用于雷达辐射源多标签识别。首先,利用残差网络学习过程数据的时序特征并提取相应的深层特征;然后,引入注意力机制对提取的特征进一步地分类和识别;最后,在雷达辐射源数据集上进行仿真实验。结果表明所提的Atten-Resnet模型不但可以在多标签条件下进行雷达辐射源的准备识别,而且在信噪比为6 dB时仍然可以保持95%以上的准确率。Atten-Resnet模型具有较高的实用性和较强的鲁棒性。

With the development of radar technology,the problem of radar radiation comes along.However,the existing radar radiation source identification methods have some disadvantages such as low recognition accuracy and inability to change multiple radiation sources in time in the case of low SNR.To solve the above problem,this paper proposes a residual network integrated with attention mechanism for multi-tag recognition of radar emitters.Firstly,residual network is used to learn temporal features of process data and extract corresponding deep features.Then,attention mechanism is introduced to further classify and recognize the extracted features.Finally,simulation experiments are carried out on the data set of radar emitter.The results show that the proposed Atten-Resnet model can not only perform the preparation identification of radar emitter under multi-label conditions,but also maintain an accuracy of more than 95%when the signal-to-noise ratio is 6 dB.Therefore,the proposed Atten-Resnet model has high practicability and strong robustness.