基于特征融合网络的特定短波信号体制识别算法  

Specific Shortwave Signal System Recognition Algorithm Based on Feature Fusion Network

作　　者：孙鹏 李保国[1] 鹿旭 SUN Peng;LI Baoguo;LU Xu(National University of Defense Technology,Changsha 410005,China)

机构地区：[1]国防科技大学,湖南长沙410005

出　　处：《信息工程大学学报》2023年第6期641-648,共8页Journal of Information Engineering University

摘　　要：短波信号体制识别在非合作通信领域起着非常重要的作用。但新体制信号的广泛应用及短波信道日益复杂的电磁环境的影响,给短波信号体制识别工作带来许多困难。提出了一种基于特征融合网络的特定短波信号体制识别算法,采用小波变换的相关去噪系数和倒谱系数作为特征,分别利用残差网络和长短时记忆网络进行特征提取,最后进行特征融合,实现了特定短波信号体制识别的功能。在短波信道条件下9类特定体制信号,该网络0 dB信噪比的总体识别率达到了95.9%。Shortwave signal system identification plays a very important role in the field of non-coop-erative communication.However,due to the extensive use of new system signals and the increasingly complex electromagnetic environment of short-wave channels,the identification of short-wave signal systems is difficult.In this paper,a signal system recognition algorithm based on feature fusion net-work is proposed to realize the function of specific shortwave signal system recognition.It uses the correlation denoising coefficient and cepstrum coefficient of wavelet transform as features,and uses residual network and long short-term memory network for feature extraction and feature fusion respec-tively.By using this network under the condition of shortwave channel,the average recognition rate of 9 kinds of specific system signals with 0 dB signal-to-noise ratio reaches 95.9%.

关 键 词：短波信号 体制识别 特征融合 深度学习 

分 类 号：TN911.7[电子电信—通信与信息系统]