Dual-Branch Feature Fusion Attention Network with Multi-Scale Features for Radar Signal Intra-Pulse Modulation Recognition

Radar signal intra-pulse modulation recognition is a key issue in modern electronic countermeasures. With the rapid development of radar technology, the intra-pulse modulation recognition has become more and more complex and diverse, which has led to shortcomings in traditional recognition methods in feature extraction. This paper proposes a dual-branch feature fusion attention network with multi-scale features for radar signal intra-pulse modulation recognition. The multi-scale convolution kernel is designed to extract time-frequency features at different levels, and a multi-head attention mechanism is introduced after feature fusion aiming to enable the network to focus on more global time-frequency features. The test results show that the overall recognition accuracy of the proposed network for 10 types of radar signals with 0 dB Gaussian white noise reaches $\mathbf{9 8. 0 \%}$.