摘要
针对战场实战电磁对抗中,多干扰机协同作战会导致多种不同雷达干扰信号同时存在,使用传统卷积神经网络对大量混叠干扰进行识别存在规模大、难以精细化识别干扰类型的问题,提出一种递进式卷积神经网络,通过分步算法分别提取存在的混叠类型特征以及干扰类型特征。通过对多种混叠干扰信号时频分析,构建训练集与测试集对网络进行训练。仿真实验表明,该网络对同时存在的3种混叠类型下的15种不同干扰信号,可以达到99.3889%以上的识别准确率,在不同干噪比条件下识别效能明显优于传统卷积神经网络。
In practical electromagnetic warfare scenarios,where multiple jammers operating in coordination create a complex environment with various radar jamming signals existing simultaneously,traditional convolutional neural networks.(CNNs)struggle with the scale and specificity required for effective interference classification.This paper introduces a stepwise CNN,which addresses these challenges by employing a staged approach to separately extract overlap type feature and interference type features.Through time-frequency analysis of various overlapping jamming signals,training and fest datasets are developed to train the network.Simulation experiments show that the proposed network achieves a recognition accuracy rate of over 99.3889%for fifteen different jamming signals under three types of mixed interference conditions.The recognition performance under different signal-to-noise ratio conditions is significantly better than traditional CNNs.
作者
付亦凡
阮航
穆贺强
周东平
潘黎
雷蕾
鲍嘉瑞
FU Yifan;RUAN Hang;MU Heqiang;ZHOU Dongping;PAN Li;LEI Lei;BAO Jiarui(Beijing Institute of Radio Measurement,Beijing 100854,China)
出处
《系统工程与电子技术》
北大核心
2025年第10期3251-3256,共6页
Systems Engineering and Electronics
关键词
递进神经网络
混叠干扰信号
特征提取
干扰识别
stepwise neural network
overlap interference
feature extraction
interference recognition
