航空学报 > 2016, Vol. 37 Issue (7): 2251-2258   doi: 10.7527/S1000-6893.2015.0313

基于压缩协作表示的辐射源识别算法

周志文, 黄高明, 高俊   

  1. 海军工程大学 电子工程学院, 武汉 430033
  • 收稿日期:2015-07-13 修回日期:2015-11-19 出版日期:2016-07-15 发布日期:2015-12-07
  • 通讯作者: 周志文 男,博士研究生。主要研究方向:辐射源识别和信息融合。Tel:027-65461265,E-mail:mini_paper@sina.com E-mail:mini_paper@sina.com
  • 作者简介:黄高明 男,博士,教授,博士生导师。主要研究方向:盲信号处理和无源探测。Tel:027-65461251,E-mail:hgaom@126.com;高俊 男,博士,教授,博士生导师。主要研究方向:数字信号处理和数字通信技术。Tel:027-65461126,E-mail:gaojunnj@163.com
  • 基金资助:

    国家“863”计划(2014AAXXXX061)

Emitter recognition algorithm based on compressively collaborative representation

ZHOU Zhiwen, HUANG Gaoming, GAO Jun   

  1. College of Electronic Engineering, Naval University of Engineering, Wuhan 430033, China
  • Received:2015-07-13 Revised:2015-11-19 Online:2016-07-15 Published:2015-12-07
  • Supported by:

    National High-tech Research and Development Program of China (2014AAXXXX061)

摘要:

针对低信噪比(SNR)条件下传统辐射源识别算法性能下降的问题,提出了基于压缩协作表示的识别算法,分别从特征提取和分类器设计两方面进行描述。首先将时域辐射源信号变换到二维时频域,通过图像处理方法提取高维特征列向量。经随机矩阵压缩到一定维度后,输入到提出的压缩协作表示分类器中得到识别结果。进而,对协作表示系数进行非负约束,提出了更符合实际应用场景的算法。仿真结果验证了所提算法的可行性与有效性,且在低信噪比条件下稳健性强、抗噪声干扰性能好、计算量较小、易于工程实现。

关键词: 压缩分类器, 协作表示, 非负约束, 辐射源识别, 鲁棒性

Abstract:

Aimed at performance decline of traditional emitter recognition algorithms under the low signal-to-noise ratio (SNR) conditions, an algorithm based on compressively collaborative representation is proposed, which is described from feature extraction and classifier design. The time-domain emitter signal is transformed into a 2-dimensional image through time-frequency analysis, and then a high-dimensional feature vector is extracted with image processing methods. After random compression, the features are input into the proposed classifiers to obtain recognition results. Furthermore, the non-negative constraint is posed to the collaborative representation coefficients, which makes the proposed algorithm more suitable for the practical applications. The simulation results verify the validity of the proposed algorithm and show that its robustness to noise is better when SNR is low and it can be easily realized in engineering due to the low computation burden.

Key words: compressive classifier, collaborative representation, non-negative constraint, emitter recognition, robustness

中图分类号: