The Pulse Signal Reconstruction Method Against Broadband Continuous Wave Interference

In the acoustic confrontation scenario of noncooperative localization, a ship needs to receive continuous wave (CW) pulse signals from other nodes for localization. At the same time, the ship emits broadband interference used to jam and deceive an enemy ship. The interference creates an extremely strong interference background at the hydrophone close to the ship, thus damaging subsequent localization. Therefore, to localize other nodes, the problem of CW pulse signal reconstruction under strong interference background needs to be solved on a priority basis. Focusing on the signal reconstruction problem under strong interference conditions, this article proposes a parallel convolutional neural network with skip connections. The network mainly consists of a target subnet and an interference subnet. The input to the network contains a CW pulse signal, interference, and background noise. The target subnet is designed to estimate the target component, that is, the CW pulse signal. Additionally, the interference subnet is tasked with estimating the interference component. Ultimately, the acquired target and interference components are used to reconstruct the CW pulse signal of interest. The performance of the proposed network is evaluated using the signal-to-interference ratio (SIR) gain and signal-to-distortion ratio (SDR). According to simulation results, when the input SIR and signal-to-noise ratio are in the range of −8 to 10 dB, the SIR gain and the SDR of our method surpass comparative algorithms. Experimental results show that our network outperforms other benchmark algorithms in reconstructing underwater CW pulse signals with low input SIR. The calculated SIR gain and SDR are 37.80 dB and 6.82 dB, respectively.