北京航空航天大学学报 ›› 2018, Vol. 44 ›› Issue (3): 583-592.doi: 10.13700/j.bh.1001-5965.2017.0146

• 论文 • 上一篇    下一篇

基于BDS-GD的低截获概率雷达信号识别

王星, 呙鹏程, 田元荣, 王玉冰   

  1. 空军工程大学航空航天工程学院, 西安 710038
  • 收稿日期:2017-03-13 出版日期:2018-03-20 发布日期:2017-09-18
  • 通讯作者: 王星 E-mail:wangx1965@163.com
  • 作者简介:王星,男,博士,教授,博士生导师。主要研究方向:电子对抗理论与技术;呙鹏程,男,硕士研究生。主要研究方向:智能告警系统及其算法;田元荣,男,博士研究生。主要研究方向:电子侦察与压缩感知;王玉冰,女,硕士研究生。主要研究方向:雷达工作模式判定。
  • 基金资助:
    航空科学基金(20152096019)

LPI radar signal recognition based on BDS-GD

WANG Xing, GUO Pengcheng, TIAN Yuanrong, WANG Yubing   

  1. Aeronautics and Astronautics Engineering College, Air Force Engineering University, Xi'an 710038, China
  • Received:2017-03-13 Online:2018-03-20 Published:2017-09-18
  • Supported by:
    Aeronautical Science Foundation of China (20152096019)

摘要: 针对在信号特征提取与识别中使用双谱估计数据量大、维度高的问题,提出了双谱对角切片(BDS)与广义维数(GD)相结合的识别方法。通过提取信号双谱对角切片减少数据量,并利用多重分形理论中的广义维数降低数据维度,对切片内部特性进行细微描述,基于距离测度提出特征评价指标,从而选出最具有区分度的3个阶数对应的广义维数作为特征向量,输入到最小二乘支持向量机中进行分类识别。使用4种低截获概率(LPI)雷达信号作为待识别信号,仿真结果表明,本文方法提取的信号特征在特征空间中有良好的聚集性和离散性,在0 dB信噪比下,识别准确率能达到92.2%,与选取的其他方法对比说明其具有很好的识别性能。

关键词: 低截获概率(LPI), 双谱对角切片(BDS), 多重分形, 广义维数(GD), 特征提取

Abstract: Regarding the deficiencies of bispectrum with big data and high dimension in signal feature extraction and recognition, a method combining bispectra diagonal slice (BDS) with generalized dimension (GD) was proposed. First, BDS was used to reduce data volume and GD in the multi-fractal theory was taken to reduce dimension in order to make subtle description for slice. Second, generalized dimension corresponding to three ranks is treated as feature vectors by feature evaluation index based on distance measure. Finally, the feature vectors will be input into least squares support vector machine for recognition. Four sorts of low probability of intercept (LPI) signal are used to be recognized, and the simulation results show that the signal features in feature space have good aggregation and discreteness, and the accuracy rate of recognition can reach 92.2% when the SNR is 0 dB, which shows that it has good performance in recognition compared with other algorithms.

Key words: low probability of intercept (LPI), bispectra diagonal slice (BDS), multi-fractal, generalized dimension (GD), feature extraction

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发