一种融合流形学习和深度学习的时变信道自动调制识别技术  

Automatic Modulation Recognition in Time-Varying Channels Combining Manifold Learning and Deep Learning

作　　者：邢怀志 李汀[1] 李飞[1] XING Huaizhi;LI Ting;LI Fei(College of Telecommunications and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing,Jiangsu 210003,China)

机构地区：[1]南京邮电大学通信与信息工程学院,江苏南京210003

出　　处：《信号处理》2022年第7期1517-1524,共8页Journal of Signal Processing

基　　金：国家自然科学基金(61771254,61871238)。

摘　　要：自动调制识别在军事领域和民用领域都发挥了巨大作用。现有的大多数研究都是基于高斯白噪声信道,但是时变信道下的自动调制识别才更符合实际并且具有挑战性。该文针对时变信道提出了一种融合流形学习和深度学习的自动调制识别方法,第一次将格拉斯曼流形引入到信号的特征提取,通过将信号星座图建模到格拉斯曼流形上完成特征提取。分类网络由基于流形学习和深度学习的两部分组成,流形数据先经过流形学习网络进行降维,然后映射到平滑子空间,最后通过简单的卷积神经网络完成分类。实验结果表明,与传统的卷积神经网络相比该文所提出的方案具有良好的性能,同时为自动调制识别提供了新的解决思路。Automatic modulation recognition has played an important role in both military and civilian fields.Most of the existing researches are based on Gaussian white noise channels,but automatic modulation recognition in time-varying channels is more realistic and challenging.This paper proposes an automatic modulation recognition method that integrates manifold learning and deep learning for time-varying channels.For the first time,the Grassmann manifold is introduced into the feature extraction of the signal,and the signal constellation is modeled to the Grassmann manifold to extract feature.The classification network is composed of two parts based on manifold learning and deep learning,The manifold data is first reduced by the manifold learning network,then mapped to the smooth subspace,finally classified by a simple convolutional neural network.The experimental results show that compared with the traditional convolutional neural network,the proposed scheme has good performance,and at the same time provides a new solution for automatic modulation recognition.

关 键 词：自动调制识别 时变信道 星座图 深度学习 格拉斯曼流形 流形学习 

分 类 号：TN929.5[电子电信—通信与信息系统]