采用认知雷达架构有助于实现雷达抗干扰技术的智能化程度提升。针对认知抗干扰技术领域中雷达对电磁干扰环境的感知问题,文中提出了一种基于深度学习的多节点干扰调制类型识别方法。该方法针对雷达信号处理的不同节点,如数字波束形成、自适应副瓣对消、脉冲压缩前后以及动目标检测之后,采用多个节点的时频平面和距离多普勒平面作为干扰信号的联合特征提取对象,建立了基于深度学习的多节点干扰识别策略模型,以提高多种干扰场景下的干扰识别正确率。为了提升干扰特征的提取能力和网络的训练效率,用于干扰识别的深度学习算法在卷积神经网络(CNN)的基础上引入了注意力机制和残差网络,建立了针对多节点策略下的干扰类型识别网络结构,实现了对多种不同干扰场景下的干扰类型识别。仿真结果表明,在单一干扰场景下,当干噪比为14 dB 时,所提算法的干扰识别准确率可达92%。在多干扰场景下,所提算法在不同节点策略的加持下,识别准确率可达90%。

The cognitive radar architecture contributes to enhancing the intelligence of radar anti-jamming technology. Addressing the issue of radar perception of electromagnetic interference in the field of cognitive anti-jamming technology, a multi-node jamming modulation type recognition method based on deep learning is proposed. This method targets various nodes in radar signal processing, such as digital beamforming, adaptive sidelobe cancellation, before and after pulse compression, and post-moving target detection. The time-frequency plane and range-Doppler plane of multiple nodes are used as joint feature extraction objects for jamming signals. A deep learning-based multi-node jamming recognition strategy model is established to improve jamming recognition accuracy in various scenarios. To enhance the extraction capability of jamming features and the training efficiency of the network, the deep learning algorithm for jamming recognition incorporates attention mechanisms and residual networks into the convolutional neural network. This establishes an interference type recognition network structure for multi-node strategies, achieving recognition of various jamming types in different scenarios. Simulation results show that in a single jamming scenario, the proposed algorithm achieves a jamming recognition accuracy of up to 92?? at a jamming-to-noise ratio (JNR) of 14 dB. In multiple jamming scenarios, the proposed algorithm, supported by different node strategies, achieves an accuracy of up to 90%.

TN974