Research on Intelligent Modulation Recognition Technology of Radar Emitters

To address the problems of insufficient accuracy and robustness in traditional radar emitter modulation recognition methods under complex electromagnetic environments, this paper proposes an MFE-CNN-MaSA recognition framework. This method converts seven typical radar modulation signals into time-frequency images via STFT, achieves multi-scale feature enhancement through the MFEblock, extracts features combined with CNN, and uses the MaSA mechanism to improve discriminability. Experiments show that the framework achieves a recognition accuracy of 98.93% at 20 dB SNR, with an average accuracy of 75.19% in the range of 0-20 dB. Its anti-noise performance and recognition accuracy are superior to those of the comparison networks. Moreover, it requires no manual feature design and has practical engineering application value.