基于贝叶斯深度学习的一维雷达有源干扰信号识别方法

雷达干扰信号准确识别是雷达抗干扰的前提，对于雷达生存至关重要。针对传统雷达干扰信号识别方法需要繁琐的分析计算提取特征，通用性差，泛化能力弱，难以适应复杂的雷达工作环境问题。本文考虑无需人工提取特征信息且具有较好的分类识别效果的深度学习网络。考虑到传统的深度学习网络由于使用点估计方式，不能够很好的衡量预测结果中的不确定性，本文提出了一种基于贝叶斯深度学习的干扰识别方法。首先，通过概率建模代替网络参数模型的点估计，解决了不确定性随机数据引起的网络过拟合问题。其次，考虑有效利用雷达回波信号的时序特性设计了LSTM层，同时解决训练过程中的梯度消失问题。基于线性调频雷达有源干扰实测数据完成了网络训练与测试，实验结果表明，引入贝叶斯方法可以在加快网络收敛速度的同时有效提高识别准确率。

One-dimensional Radar Active Jamming Signal Recognition Method Based on Bayesian Deep Learning

? ?Accurate identification of radar jamming signal is the premise of radar jamming， which is very important for radar survival. In view of the traditional radar jamming signal recognition method needs complicated analysis and calculation to extract features， and it is difficult to adapt to the complex radar working environment because of its poor universality and weak generalization ability. In this paper， deep learning networks with better classification and recognition effect without manual extraction of feature information are considered. Considering that the traditional deep learning networks cannot measure the uncertainty in the prediction results well due to the use of point estimation method， this paper proposes a jamming identification method based on Bayesian deep learning. Firstly， the problem of network overfitting caused by uncertain random data is solved by using probabilistic modeling instead of point estimation of network parameter model. Secondly， the LSTM layer is designed to effectively utilize the timing characteristics of radar echo signals， and solve the problem of gradient disappearance during training. The network training and testing are completed based on the measured data of LFM radar active jamming. The experimental results show that the Bayesian method can accelerate the convergence speed of the network and effectively improve the recognition accuracy.