【摘要】 目前基于频域构造的残差网络干扰识别算法在提取压制干扰特征时，泛化能力较弱，在低干噪比下受噪声影响大。为此，针对5类频谱图相似的压制干扰信号，提出一种基于灰度时域图和Fused Feature Aware Transformer(FFA-Transformer)网络的压制干扰识别算法。将一维的时域信号构造成体现信号特征的灰度时域图，使得网络能够提取更多信号特征，并搭建FFA-Transformer网络，利用该网络实现5类压制干扰信号的识别。该网络通过加权平均的方式融合编码器输出的全局特征和局部特征，网络同时关注纹理等全局特征和边缘等局部特征，加强了全局特征和局部特征的关联性，较传统ViT网络和传统CNN网络训练频谱图和时域图更能高精度地实现5类压制干扰信号的识别。更多还原

【Abstract】 At present, when the residual network interference identification algorithm of the frequency domain structure is extracting the suppressed interference characteristics, the generalization ability is weak, and the noise ratio is greatly affected by the low noise ratio. To this end, for the 5-type spectrum diagrams similar to the suppressed interference signals, a suppressed interference identification algorithm is proposed based on gray time domain diagram and Fused Feature Aware Transformer(FFA-Transformer) network. Constructing the one-dimensional time domain signal to the gray time domain diagram that reflects the signal characteristics so that the FFA-Transformer can extract more signal features. And construct FFA-Transformer network to recognition of five types of suppressed interference signals, which combines global and local features of encoder output by weighted average. Focusing on the local characteristics and edges such as texture and edges, etc., it has strengthened the correlation between global characteristics and local characteristics. Compared with the traditional ViT network and traditional CNN network training spectrum diagram and time domain map, it can achieve high-precision recognition of five types of suppressed interference signals.更多还原