A Spatiotemporal Dual-Attention Network for Automatic LPI Radar Modulation Recognition

Automatic modulation recognition (AMR) has become a basic task of reconnaissance receivers to process low probability of intercept (LPI) radar signals. Many traditional methods have been used in this field while the performance is seriously degraded in poor channel conditions. In this paper, an end-to-end spatiotemporal dual-attention network (STDAN) based on channel attention and multi-head self-attention (MHSA) is proposed to realize high recognition accuracy under low signal-to-noise ratio (SNR) conditions. STDAN can learn the potential correlation between spatial and temporal domains by the subtle features which are extracted through distinct attention mechanisms. The experimental results show that our method can achieve better performance compared with the state-of-the-art modulation recognition algorithms. When the SNR is as low as -4dB, the recognition probability of 13 typical modulated signals can still reach 96.6%.