小样本情景下基于特征融合的辐射源个体识别

Specific Emitter Identification Based on Feature Fusion in Few-Shot Scene

针对小样本情景下辐射源个体识别困难等难题,提出一种在小样本情景下基于特征融合和深度学习的辐射源个体识别方法。将基于希尔伯特-黄变换、魏格纳-威利分布、连续小波变换所提取的信号个体特征进行融合并构建ResNet完成训练和识别。采用该方法,对3部实采辐射源数据进行测试,测试结果表明,相比于使用单一特征进行网络分类识别的主流算法,该方法能提高小样本情境下辐射源个体识别准确率、改善神经网络性能。

Aiming at the difficulty of specific emitter identification in few-shot scene,a specific emitter identification method based on feature fusion is proposed.The signal features extracted from different signal processing methods,such as Hilbert Huang Transform,Wigner-Ville Distribution and Continuous Wavelet Transform,are fused,and the ResNet is constructed to train and recognize the fused emitter individual features.Using this method,three emitter data are tested.The test results show that,compared with the mainstream network classification algorithm using single feature,the accuracy of emitter individual recognition could be improved in few-shot scene and also the performance of neural network.