基于U-Net的雷达脉冲信号一体化检测、分选和识别

Integrated Detection,Sorting and Identification of Radar Pulse Signal Based on U-Net

检测、分选和识别通常表现为电子战信号处理流程中相对独立的环节。全卷积网络在图像语义分割领域取得了巨大成功。受其启发,提出了一种基于U-Net网络的全新信号处理方法,可以同步实现端到端的信号检测、分选和识别。详细介绍了算法原理和U-Net网络的训练方法,并完成了仿真试验。结果表明,U-Net网络可以有效分割噪声和目标信号,来自于不同已知目标的信号,以及分别来自于已知或未知目标的信号。

Detection,sorting and identification are usually represented as relatively independent links in the signal processing process of electronic warfare(EW).Full convolutional network(FCN)has achieved great success in image semantic segmentation.Inspired by it,a new signal processing method based on U-Net is proposed.End-to-end signal detection,sorting and identification can be realized synchronously.The algorithm principle and the training method of U-Net network are introduced in detail,and the simulation experiments are completed.The results show that the U-Net network can effectively segment noise and target signal,signals from different known targets,and signal from known or unknown targets.