基于模糊函数多域特征融合与集成学习的雷达辐射源信号识别

Radar emitter signal recognition based on ambiguity function multi-domain feature fusion and ensemble learning

针对复杂电磁环境下雷达辐射源信号识别方法中存在的抗噪性能差、识别准确率低等问题,提出一种融合模糊函数多域投影特征的集成深度学习识别方法.首先,对信号的模糊函数进行高斯平滑处理,从多域视角出发选取合适角度对模糊函数进行二维投影以构建特征数据集;然后,构建一种基于多域特征融合的两阶段识别分类方法,使用多个密集连接网络DenseNet 121作为初级分类器分别对3类特征数据集进行训练学习,得到初级分类结果;最后,通过Stacking策略对初级分类结果进行融合学习,得到最终类别信息.实验结果表明,所提出方法在信噪比为0 dB时对6类典型雷达信号的整体平均识别率均保持在97.24%以上,即使是在-4 dB环境中,识别率也稳定在87.16%以上,验证了所提出方法的有效性和可行性,具有一定的工程价值.

Aiming at the problems of pooranti-noise performance and low recognition accuracy of the radar emitter signal recognition method in the complex electromagnetic environment.An integrated deep learning recognition method based on multi-domain projection features of ambiguity function is proposed.Firstly,an ambiguity function is processed using a Gaussian operator,and the appropriate angle is selected to carry out two-dimensional projection to build a characteristic data set from the multi-domain perspective.Then,a two-stage recognition and classification method based on multi domain feature fusion is constructed.Multiple dense connected networks DenseNet121 are used as primary classifiers to train and learn the three kinds of feature data sets respectively,and the primary classification results are obtained.Finally,the results of the primary classification are integrated through the Stacking policy to obtain the final classification result.The experimental results show that the overall average recognition rate of the six types of typical radar signals is above 97.24%,when the signal-to-noise ratio is 0dB,even in the−4dB environment,the recognition rate is also stable in 87.16%,which verifies the effectiveness and feasibility of the proposed method,and its certain engineering value.