首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于高阶累积量特征学习的调制识别方法
引用本文:袁莉芬,宁暑光,何怡刚,吕密,路健.基于高阶累积量特征学习的调制识别方法[J].系统工程与电子技术,2019,41(9):2122-2131.
作者姓名:袁莉芬  宁暑光  何怡刚  吕密  路健
作者单位:1. 合肥工业大学电气与自动化工程学院, 安徽 合肥 230009; 2. 德州农工大学工程学院, 美国 德克萨斯州 卡城 TX77843; 3. 国家电网合肥供电公司, 安徽 合肥 230009
基金项目:国家重点研发计划“重大科学仪器设备开发”(2016YFF0102200);国家自然科学基金(61102035,51577046,51607004);国家自然科学基金重点项目(51637004);中国博士后特别项目(2015T80651);中国博士后面上项目(2014M5517)资助课题
摘    要:自动调制分类是确保通信安全、可靠的关键技术之一。在低信噪比(low signal-to-noise ratio,Low-SNR)环境中,自动调制分类识别率低且识别类型受限。利用零均值高斯白噪声(white Gaussian noise,WGN)的高阶累积量理论值等于0的性质,在信号分析过程中,引入高阶累积量,可使系统免受WGN的影响。同时,引入深度学习网络结构完成微弱特征的表征,可有效解决调制方式受限及Low-SNR情况下的识别率下降问题。实验结果表明,所提方法在高斯信道环境下的分类精度比现有方法要高,在Low-SNR的不同信道环境下有较高的识别率,且使模型在时间、相位和频率偏移量方面具有更好的鲁棒性。

关 键 词:调制分类  高阶累积量  深度学习  低信噪比  调制特征

Modulation recognition method based on high-order cumulant feature learning
YUAN Lifen,NING Shuguang,HE Yigang,LYU Mi,LU Jian.Modulation recognition method based on high-order cumulant feature learning[J].System Engineering and Electronics,2019,41(9):2122-2131.
Authors:YUAN Lifen  NING Shuguang  HE Yigang  LYU Mi  LU Jian
Institution:(College of Electrical Engineering and Automation,Hefei University of Technology,Hefei 230009,China;College of Engineering,Texas Agricultural&Mechanical University,College Station,Texas TX77843,United States of America;State Grid Hefei Power Supply Company,Hefei 230009,China)
Abstract:Automatic modulation classification is one of the key technologies to ensure communication security and reliability. In a low signal-to-noise ratio (Low-SNR) environment, the automatic modulation classification recognition rate is low and the recognition type is limited. By using the property that the high-order cumulant is equal to zero of zero-mean white Gaussian noise (WGN), the high-order cumulant is introduced to protect the system from WGN in the signal analysis process. Moreover, the deep learning network structure is introduced to complete the characterization of weak features, which can effectively solve the problem of the limited modulation method. And it can also solve the problem of low recognition rate under Low-SNR. The experimental results show that the classification accuracy of the proposed method is better than the existing methods in the Gaussian channel environment, and it has a higher recognition rate in different channel environments with Low-SNR. And, it makes the model’s time, phase and frequency offset more robust.
Keywords:modulation classification  high-order cumulant  deep learning  low signal-to-noise ratio (Low-SNR)  modulation characteristics  
本文献已被 维普 等数据库收录!
点击此处可从《系统工程与电子技术》浏览原始摘要信息
点击此处可从《系统工程与电子技术》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号