基于深度自适应小波网络的通信辐射源个体识别

Individual recognition of communication radiation source based on depth adaptive wavelet network

针对现有的通信辐射源个体识别方法中人工提取特征复杂以及深度学习网络的识别机制缺乏清晰解释的问题,提出了一种基于深度自适应小波网络(Deep Adaptive Wavelet Network,DAWN)的通信辐射源个体识别方法。首先分析了选择互调干扰作为辐射源间个体特征的原因;接着应用了可实现提升小波变换的卷积神经网络结构去提取特征,并在其基础上设计出可以同时完成特征提取和识别的DAWN;最后,选择Oracle数据集验证方法的可行性。实验结果表明:利用DAWN对5个通信辐射源个体识别的准确率为95.5%,并且方法具有良好的抗噪性。

Aiming at the problem of the complex artificial features extracted in the existing individual recognition methods of communication radiation sources and the lack of clear interpretation of the recognition mechanism of deep learning networks,an individual recognition method of communication radiation sources based on Deep Adaptive Wavelet Network(DAWN)is proposed.Firstly,the intermodulation interference is analyzed as the reason for individual characteristics between radiation sources.Then,the convolutional neural network structure that can realize lifting wavelet transform is applied to extract features,based on which DAWN can complete feature extraction and recognition at the same time.Finally,Oracle data sets are selected to verify the feasibility of the method.The experimental results show that the accuracy of identification of 5 communication radiation sources by DAWN is 95.5%,and the method has good anti-noise performance.