Learning transformer network for effective radar chaff jamming suppression

As a typical and widely used passive jamming method, chaff has extremely strong interference effect on radar that remains a significant challenge effectively to counteract. Therefore, it is exceedingly necessary to suppress the chaff jamming and improve the anti-chaff jamming ability of radars. In this paper, to address this challenge, we propose an effective transformer network that can effectively suppress the radar chaff jamming. Specifically, the network can model the non-local information which is vital for high-quality radar echo signal reconstruction. we introduce the scaled dot-product multi-head self-attention mechanism that enables the network to use all similarities of the positions of each high-resolution range profile (HRRP) sequence from the query-key pairs for the feature aggregation, significantly improving the model's feature extraction capability and overall performance. In addition, we address the limitation of insufficient measured chaff radar echo signal by establishing the first remarkably rich and diverse dataset of measured HRRP sequences of chaff radar echo signal in the radar field through extensive anechoic chamber experiments. This dataset serves as a valuable resource and a critical foundation for advancing radar chaff jamming suppression research in this domain. Experimental results on measured HRRP sequences reveal that our proposed network can achieve the mean square error of 0.0258, the correlation coefficient of 93.26%, the signal to noise ratio of 11.1689 dB, and an improvement in the signal-to-noise ratio of 11.1475 dB after suppressing the chaff jamming. These results demonstrate that our proposed method achieves favorable performance against other approaches and establishes a new benchmark for radar chaff jamming suppression.