一种基于轻量级神经网络的调制识别方法

李晓威; 朴春慧; 杨兴雨

doi:10.3969/j.issn.2096-6091.2022.24.002

您当前的位置：

首页 >

文章列表页 >

一种基于轻量级神经网络的调制识别方法

研究论文 | 更新时间：2023-08-03

- 一种基于轻量级神经网络的调制识别方法
- A Modulation Recognition Method Based on Lightweight Neural Network
- 新一代信息技术 2022年5卷第24期页码：5-11
- 作者机构：
  
  1.石家庄铁道大学信息科学与技术学院，河北石家庄 050043
  2.河北省电磁环境效应与信息处理重点实验室，河北石家庄 050043
- 作者简介：
  
  [ "李晓威（1998‒），男，硕士研究生，研究方向：信号调制识别、联邦学习。" ]
  [ "朴春慧（1964‒），女，博士，教授，研究方向：大数据技术及应用、数据安全与隐私保护、区块链技术及应用。" ]
  [ "杨兴雨（1988‒），男，博士，讲师，研究方向：激光成像雷达、机器视觉、人工智能、智能信号处理、图像处理。" ]
- 基金信息：
  
  河北省重点研发计划项目(21355902D)
- DOI：10.3969/j.issn.2096-6091.2022.24.002
  中图分类号：
扫描看全文
李晓威, 朴春慧, 杨兴雨. 一种基于轻量级神经网络的调制识别方法[J]. 新一代信息技术, 2022, 5(24): 05-11

LI Xiao-wei, PIAO Chun-hui, YANG Xing-yu. A Modulation Recognition Method Based on Lightweight Neural Network[J]. New Generation of Information Technology, 2022, 5(24): 05-11
李晓威, 朴春慧, 杨兴雨. 一种基于轻量级神经网络的调制识别方法[J]. 新一代信息技术, 2022, 5(24): 05-11 DOI： 10.3969/j.issn.2096-6091.2022.24.002.

LI Xiao-wei, PIAO Chun-hui, YANG Xing-yu. A Modulation Recognition Method Based on Lightweight Neural Network[J]. New Generation of Information Technology, 2022, 5(24): 05-11 DOI： 10.3969/j.issn.2096-6091.2022.24.002.

摘要

针对目前用于信号调制方式识别的深度学习模型参数量和计算量较大的问题，提出了一种称为TBCLDNN-ECA的轻量级神经网络模型。该模型首先使用两个Bottleneck结构CNN分支并行提取信号的不同特征，两分支合并后使用注意力机制和BiLSTM，进一步提取信号的序列特征。在数据集RadioML2016.10a上进行了仿真实验，结果表明提出的模型在信噪比0~18 dB时平均识别准确率达到了92.1%，在信噪比18 dB时达到93.53%，识别准确率优于目前流行的调制识别模型，同时模型参数量和计算量均有了一定幅度的降低。因此该模型在不损失识别准确率的情况下，具有模型参数量和计算量少的优势。

Abstract

In order to solve the problem that the deep learning model used for signal modulation recognition has a large number of parameters and computation, a lightweight neural network model called TBCLDNN-ECA (Two-Stream Bottleneck Convolution Long-Short Term Memory Fully Connected Deep Neural Network with Efficient Channel Attention Module) is proposed. The model first uses two Bottleneck CNN (Convolutional Neural Networks) branches to extract different features of the signal in parallel. After the two branches are merged, attention mechanism and BiLSTM (Bi-directional Long Short-Term Memory) are used to further extract the sequence features of the signal. Simulation experiments are conducted on the dataset RadioML2016.10a. The results show that the average recognition accuracy of the proposed model reaches 92.1% at the signal to noise ratio of 0~18 dB, and 93.53% at the signal to noise ratio of 18 dB. The recognition accuracy is better than the current popular modulation recognition model. At the same time, the model parameters and calculation amount have been reduced to a certain extent. Therefore, this model has the advantage of less model parameters and computation without loss of identification accuracy.

关键词

调制识别轻量级神经网络瓶颈型结构

Keywords

modulation recognitionlightweight neural networkbottleneck structure

references

石彦坤. 基于深度学习的信号调制识别研究[D]. 哈尔滨: 哈尔滨工业大学, 2020.

SHI J B, ZHAO H J, WANG M Y, et al. Signal recognition based on federated learning[C]//IEEE INFOCOM 2020 - IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS). Piscataway: IEEE, 2020: 1105-1110.

HUANG S, YAO Y Y, XIAO Y, et al. Cumulant based maximum likelihood classification for overlapped signals[J]. Electronics Letters, 2016, 52(21): 1761-1763.

SHI Q H, KARASAWA Y. Noncoherent maximum likelihood classification of quadrature amplitude modulation constellations: Simplification, analysis, and extension[J]. IEEE Transactions on Wireless Communications, 2011, 10(4): 1312-1322.

DONG S L, LI Z P, ZHAO L F. A modulation recognition algorithm based on cyclic spectrum and SVM classification[C]//2020 IEEE 4th Information Technology, Networking, Electronic and Automation Control Conference (ITNEC). Piscataway: IEEE, 2020: 2123-2127.

LIANG Y, XIANG X, SUN Y, et al. Novel modulation recognition for WFRFT-based system using 4th-order cumulants[J]. IEEE Access, 2019, 7: 86018-86025.

O’SHEA T J, CORGAN J, CLANCY T C. Convolutional radio modulation recognition networks[C]//JAYNE C, ILIADIS L. International Conference on Engineering Applications of Neural Networks. Cham: Springer, 2016: 213-226.

WU Y L, LI X J, FANG J. A deep learning approach for modulation recognition via exploiting temporal correlations[C]//2018 IEEE 19th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC). Piscataway: IEEE, 2018: 1-5.

张军, 符杰林, 林基明. 基于CLDNN的调制信号识别方法[J]. 计算机应用与软件, 2021, 38(10): 216-220, 277.

XU J L, LUO C B, PARR G, et al. A spatiotemporal multi-channel learning framework for automatic modulation recognition[J]. IEEE Wireless Communications Letters, 2020, 9(10): 1629-1632.

ZHANG Z F, LUO H, WANG C, et al. Automatic modulation classification using CNN-LSTM based dual-stream structure[J]. IEEE Transactions on Vehicular Technology, 2020, 69(11): 13521-13531.

HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE, 2016: 770-778.

WANG Q L, WU B G, ZHU P F, et al. ECA-net: Efficient channel attention for deep convolutional neural networks[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE, 2020: 11531-11539.

HUANG Z H, XU W, YU K. Bidirectional LSTM-CRF models for sequence tagging[EB/OL]. [2022-10-07]. https://arxiv.org/pdf/1508.01991.pdfhttps://arxiv.org/pdf/1508.01991.pdf.

HU J, SHEN L, ALBANIE S, et al. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(8): 2011-2023.

HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780.

ELMAN J L. Finding structure in time[J]. Cognitive Science, 1990, 14(2): 179-211.

O’SHEA T, WEST N E. Radio machine learning dataset generation with GNU radio[C]//Proceedings of the Gnu Radio Conference, 2016.

KINGMA DP,BA J. Adam: a method for stochastic optimization[EB/OL].[2022-10-20]. DOI:10.48550/arXiv.1412.6980http://dx.doi.org/10.48550/arXiv.1412.6980.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

暂无数据