摘要
Automatic Modulation Recognition(自动调制识别) (AMR) is a critical component of(重要组成部分) smart communication(智能通信) and it contributed(促进了) to the development of many applications(许多应用的发展) such as Cognitive Radio(认知无线电) (CR). Therefore(因此), many researchers(许多研究者) have been interested in this field(对这一领域产生了兴趣). In this paper, a brief review(简要的综述) of AMR. More specifically(更具体地说), classification methods(分类方法) using Deep Learning (DL), especially those that give high accuracies(高精度), Such as Convolutional Neural Networks(卷积神经网络) (CNN), take into account(考虑) the database, the method of extracting features(提取特征的方法), and the number of modulation types for each(每种调试类型的数量) (SNR). Accordingly(此外), this paper demonstrates(论证) that the best classification accuracy results(最好的分类精度效果) are obtained using (DL) when using CNN. In recent research, the accuracy has been obtained reaching more than 90% when the classification is for 24 different types of modulation(24种不同的调制类型) and the size of the signal(信道大小) is 2.5 million.
关键词
Automatic Modulation Recognition,自动调制识别
Deep Learning,深度学习
Convolutional Neural Networks,卷积神经网络
引言
在大多数情况下,AMC技术主要分为两种方法:基于似然(likelihood)的方法(LB)和基于特征(features)的方法(FB),尽管LB技术获得了最好的分类结果,但它有相当大的局限性,包括高计算复杂度和对实时系统难以实现的挑战。另一方面,FB采取了不同的策略。
传统的FB技术分为特征提取(feature extraction)和特征分析(feature extraction )两部分。
信号数据集
Radio ML 2016.10A数据集
[5] T. O’Shea and N. West, “Radio Machine Learning Dataset Generation with GNU Radio”, Proceedings of the GNU Radio Conference, vol. 1, no. 1, PP.3-6. 2016.
Radio ML 2018.01A数据集
[6] T. J. O’Shea, T. Roy, and T. C. Clancy, “Over-the-Air Deep Learning-Based Radio Signal Classification”, vol. 12, no. 1, pp.168–179, 2018.
HisarMod2019.1数据集
[7] K. Tekbiyik, A. R. Ekti, A. Gorcin, G. K. Kurt, and C. Kececi, “Robust and Fast Automatic Modulation Classification with CNN under Multipath Fading Channels”, y Conference (VTC2020-Spring)PP.1-6,2020
模型架构
卷积网络CNN
卷积层、池化层、全连接层
模型输入数据
真实IQ数据
模拟IQ数据
图像数据
其他类型
论文链接
Automatic Modulation Classification Based Deep Learning: A Review