摘要
随着通信技术的发展,信号体制、调制方式日趋复杂,例如CPM、OFDM等,这给调制识别技术带来了巨大挑战。近年来,深度学习技术由于其强大的特征提取能力和分类能力,被广泛应用到模式识别领域中。为了实现复杂调制方式的识别,文章将深度学习技术引入到调制识别领域,并提出一种基于改进的CLDNN模型的调制识别算法。CLDNN模型已被成功应用到语音识别领域,其表现出了强大的特征提取和分类能力。该方法在原有CLDNN模型的基础上,针对调制识别问题的特点,对CLDNN进行了改进。而且该方法不依赖于载波同步、码元同步等预处理。实验结果表明,该方法可同时识别12种信号调制方式和信号体制,信噪比在3dB以上时,整体识别准确率达到90%以上,并且可以较好地识别复杂调制方式和信号体制。
With the development of communication technologies, modulation formats are becoming more and more complex, such as CPM, OFDM, etc., which poses a huge challenge to automatic modulation recognition technology. In recent years, deep learning has been widely applied to the field of pattern recognition due to its powerful feature extraction and classification capabilities. In order to realize the recognition of complex modulation methods,deep learning technology is introduced into the field of modulation recognition, and a automatic modulation recognition algorithm based on improved CLDNN model is proposed. The CLDNN model has been successfully applied to the field of speech recognition, showing strong feature extraction and classification capabilities. Based on the original CLDNN model,this method improves the CLDNN based on the characteristics of the modulation identification problem. And the method does not rely on preprocessing such as carrier synchronization, symbol synchronization. The experimental results show that the proposed method can identify 12 kinds of signal modulation formats. When the signal-to-noise ratio is above 3 dB,the accuracy rate of modulation recognition is over90%,and the complex modulation method can be well recognized.
作者
李唱白
杨杰
黄知涛
王翔
LI Changbai;YANG Jie;HUANG Zhitao;WANG Xiang(Troop 31082 of the PLA,Beijing 100097,China;College of Electronic Science,National University of Defense Technology,Changsha 410073,China)
出处
《空间电子技术》
2019年第1期49-54,74,共7页
Space Electronic Technology
关键词
非合作通信
调制识别
深度学习
特征提取
Non-cooperative communication
Modulation recognition
Deep learning
Feature extraction