摘要
水声通信信号识别为水声通信侦察和对抗的重要前提,具有重要作用。然而,传统的水声通信信号识别方法通常是基于信号处理和模式识别技术,依赖领域专家的专业知识和经验进行特征选择和提取,具有较强的主观性,且可能无法利用更复杂的信号特征。本文基于深度学习提出一种水声通信信号识别的智能方法。首先利用仿真数据对卷积神经网络进行训练,然后分别使用仿真和湖上试验数据对算法网络进行测试。仿真结果表明,在SNR=5dB时,该方法对2ASK、4ASK、BPSK、QPSK、2FSK、4FSK和OFDM等7种水下通信信号的识别率均能达到90%以上,7种湖上试验的通信信号类型平均识别率达到97.9%。这表明该方法具有良好的宽容性。此外,本文还通过对基于高阶累积量和深度学习方法的比较,验证了本文提出方法具有显著的优越性。
Underwater acoustic communication signal recognition is an important prerequisite for underwater acoustic communication reconnaissance and countermeasures and plays an important role.However,traditional underwater acoustic communication signal recognition methods are usually based on signal processing and pattern recognition technology.The selection and extraction of features mainly rely on the professional knowledge and experience of domain experts,which is highly subjective and may not be able to use more complex signal characteristics.In this paper,the convolutional neural network in deep learning is used to automatically extract the characteristics of communication signals.First,the network is trained using the simulated data,and then the algorithm network is tested using the simulation and lake test data.The results show that when the SNR is 5 dB,the recognition rates of seven underwater communication signals,including the 2ASK,4ASK,BPSK,QPSK,2FSK,4FSK and OFDM,can reach over 90%,and the average recognition rate of the seven types of communication signal types tested on the lake reaches 97.9%,which proves the good tolerance of the algorithm.At the same time,the comparison based on higher-order cumulant and deep learning method confirms the significant advantages of the proposed method.
作者
黄乐
夏志军
周胜增
原玉婷
王静怡
HUANG Le;XIA Zhi-jun;ZHOU Sheng-zeng;YUAN Yu-ting;WANG Jing-yi(Shanghai Marine Electronic Equipment Research Institute,Shanghai 201108,China;Dalian Naval Academy,Dalian 116018,China)
出处
《舰船科学技术》
北大核心
2024年第9期117-124,共8页
Ship Science and Technology
关键词
水声通信
信号识别
高阶累积量
深度学习
underwater acoustic communication
signal recognition
higher-order cumulant
deep learning
