上海大学学报(自然科学版) ›› 2019, Vol. 25 ›› Issue (5): 722-732.doi: 10.12066/j.issn.1007-2861.1991

• 研究论文 • 上一篇    下一篇

矩形积分双谱和半监督鉴别分析下的通信辐射源识别

韩国川1, 张金艺1,2(), 李科2, 何利康1, 姜玉稀3, 王涛2   

  1. 1.上海大学 微电子研究与开发中心, 上海 200444
    2.上海大学 特种光纤与光接入网重点实验室, 上海 200444
    3. 上海三思电子工程有限公司, 上海 201100
  • 收稿日期:2017-05-23 出版日期:2019-10-30 发布日期:2019-10-31
  • 通讯作者: 张金艺 E-mail:zhangjinyi@staff.shu.edu.cn
  • 基金资助:
    十三五国家重点研发计划项目(2017YFB0403500);上海市教委重点学科建设资助项目(J50104)

Communication emitter identification under square integral bispectra and semi-supervised discriminant analysis

Guochuan HAN1, Jinyi ZHANG1,2(), Ke LI2, Likang HE1, Yuxi JIANG3, Tao WANG2   

  1. 1. Microelectronic Research and Development Center,Shanghai University, Shanghai 200444, China
    2. Key Laboratory of Special Fiber Optics and Optical Access Networks, Shanghai University, Shanghai 200444, China
    3. Shanghai SANSI Electronic Engineering Co., Ltd., Shanghai 201100, China
  • Received:2017-05-23 Online:2019-10-30 Published:2019-10-31
  • Contact: Jinyi ZHANG E-mail:zhangjinyi@staff.shu.edu.cn

摘要:

针对传统同类通信辐射源识别方法中存在的指纹特征难以提取, 以及在先验标签信息较少时识别精度不高的问题, 以指纹特征差异微小的同厂、同批、同型号的通信辐射源为对象, 提出了一种基于矩形积分双谱和半监督鉴别分析的通信辐射源识别方法. 该方法采用矩形积分双谱算法提取通信辐射源双谱特征, 并将其作为指纹特征, 以表征所属通信辐射源; 同时, 采用半监督鉴别分析算法, 根据双谱特征数据的部分标签信息和非线性流形信息, 将高维双谱特征数据映射到低维子空间后进行分类识别, 来提升通信辐射源识别性能. 为验证该方法的有效性, 采用同厂、同批、同型号的FM电台作为同类通信辐射源的代表进行电台识别实验. 实验结果表明, 在FM电台训练样本中有先验标签信息的样本较少时, 该方法对电台测试样本的识别率最高达87.6%, 证明该方法在同类通信辐射源识别中指纹特征提取和识别精度方面具有优势.

关键词: 通信辐射源识别, 特征提取, 矩形积分双谱, 半监督鉴别分析

Abstract:

This paper is an attempt towards coping with the problem that the methods of traditional communication emitter identification with the same model performs poorly with extracting fingerprint feature and acquiring high accuracy recognition when the priori label information is insignificant. Targeting the same manufacturer, same batch and same type communication emitter identification of small fingerprint feature difference, an efficient algorithm based on square integral bispectra and semi-supervised discriminant analysis is proposed for communication emitter identification. This algorithm uses square integral bispectra for the extraction of the communication emitter signal bispectra feature as the fingerprint feature, which represents the communication emitter. Simultaneously, for the purpose of improving communication emitter identification recognition performance, the semi-supervised discriminant analysis algorithm is employed to map high dimensional bispectra feature data to a low dimensional subspace and identify in the low dimensional subspace by the nonlinear manifold information and partial label information of bispectra feature data. In order to verify the effect of the proposed algorithm, the same manufacturer, same batch and same type FM radios, as representative communication emitter with the same model, are used here to perform identification experiment. Experiment results show the highest recognition rate of proposed method for test sample is up to 87.6% when the labeled training FM radio samples are limited, which points to the effectiveness of this algorithm in extracting fingerprint feature and recognition accuracy on same type communication emitter identification.

Key words: communication emitter identification, feature extraction, square integral bispectra, semi-supervised discriminant analysis

中图分类号: