Please wait a minute...
浙江大学学报(工学版)  2022, Vol. 56 Issue (10): 1987-1994    DOI: 10.3785/j.issn.1008-973X.2022.10.010
自动化技术、信息工程     
基于相位融合特征和残差网络的调制智能识别
罗群平(),赵民建,赵明敏*(),苏智臻
浙江大学 信息与电子工程学院,浙江 杭州 310013
Modulation intelligent recognition based on phase fusion feature and residual network
Qun-ping LUO(),Min-jian ZHAO,Ming-min ZHAO*(),Zhi-zhen SU
College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310013, China
 全文: PDF(1693 KB)   HTML
摘要:

针对调制识别依赖人工提取的经验特征、识别准确率低的问题,提出基于相位星座图和矢量轨迹图融合聚类的智能识别算法.通过对接收信号进行滤波、定时同步预处理,恢复出较理想的基带信号,将信号的波形数据转化为星座图和矢量轨迹图特征. 利用深度学习(DL)将调制识别转换成图像的分类问题,将所提取的特征通过2路并行输入的轻量级残差结构网络,开展分层学习和特征融合训练,完成对目标调制方式的识别. 仿真实验表明,基于融合特征的识别结果优于目前的基于高阶累积量、星座图和波形数据的识别结果,当信噪比(SNR)高于2 dB时,对MPSK(调制的阶数为 2、4、8)、MQAM(调制的阶数为16、64)、MAPSK(调制的阶数为16、32)这7类调制的识别率可以达到95.14%.

关键词: 调制识别星座图矢量轨迹图融合训练残差网络    
Abstract:

An intelligent signal modulation recognition algorithm based on phase constellation and trajectory clustering was proposed aiming at the problem that the common recognition relied on manually extracted empirical features and suffered from low precision. An ideal baseband signal was recovered by preprocessing the received signal, such as filtering and timing synchronization. Then the waveform data of the signal was transformed into a constellation diagram and vector trajectory diagram. The signal modulation recognition problem was transformed into a featured image classification problem by exploiting the techniques in deep learning (DL). The extracted features were input into a lightweight residual structure network with two parallel inputs. Then hierarchical learning and feature fusion training were performed to achieve the target modulation recognition. The simulation experiments showed that the recognition rates achieved by the feature fusion method outperformed those achieved by the methods based on high-order statistics, constellation diagrams and signal waveform. The recognition rate for seven types of modulation consisting of MPSK (M = 2, 4, 8), MQAM (M = 16, 64), and MAPSK (M = 16, 32) can reach 95.14% when the signal-to-noise ratio (SNR) is larger than 2 dB.

Key words: modulation recognition    constellation diagram    vector trajectory diagram    fusion training    residual network
收稿日期: 2021-11-16 出版日期: 2022-10-25
CLC:  TN 911  
基金资助: 国防基础科研计划资助项目(JCKY2020210B021)
通讯作者: 赵明敏     E-mail: luojing123@zju.edu.cn;zmmblack@zju.edu.cn
作者简介: 罗群平(1989—),男,硕士生,从事通信信号处理及深度学习的研究. orcid.org/0000-0003-0770-4054. E-mail: luojing123@zju.edu.cn
服务  
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章  
罗群平
赵民建
赵明敏
苏智臻

引用本文:

罗群平,赵民建,赵明敏,苏智臻. 基于相位融合特征和残差网络的调制智能识别[J]. 浙江大学学报(工学版), 2022, 56(10): 1987-1994.

Qun-ping LUO,Min-jian ZHAO,Ming-min ZHAO,Zhi-zhen SU. Modulation intelligent recognition based on phase fusion feature and residual network. Journal of ZheJiang University (Engineering Science), 2022, 56(10): 1987-1994.

链接本文:

https://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2022.10.010        https://www.zjujournals.com/eng/CN/Y2022/V56/I10/1987

图 1  调制识别系统的流程图
图 2  信噪比为 15 dB时各调制的星座图、矢量图
图 3  残差基本单元
图 4  基于联合特征输入的卷积神经网络
序号 组件 尺寸维度 描述
1
2
3
4
5
6
7
8
9
10
输入
卷积层
卷积层
卷积层
残差单元
融合
卷积层
残差单元
全局平均
全连接
3×224×224
32×112×112
32×112×112
32×112×112
32×112×112
64×112×112
128×56×56
256×28×28
256
7
星座图和矢量图输入
32×Conv(3,3), Stride=2
32×Conv(3,3), Stride=1
32×Conv(3,3), Stride=1
Channel=32, Stride=1
Concat
128×Conv(3,3), Stride=2
Channel =256, Stride=2
Average pooling
输出
表 1  网络结构的细节
名称 含义 数值
Learning rate
Batch size
Epoch
Optimizer
Weight decay
Momentum
学习率
块包含的数据样本
最大迭代训练轮次
优化器
权重衰减
动量因子
0.001
64
300
SGD
0.001
0.9
表 2  神经网络的超参数
图 5  平均识别性能
调制方式 本文的方法 基于星座图的卷积
网络识别方法
基于高阶累积量的
KNN识别方法
基于IQ数据的卷积
网络识别方法
基于IQ数据的残差
网络识别方法
0 dB 5 dB 10 dB 0 dB 5 dB 10 dB 0 dB 5 dB 10 dB 0 dB 5 dB 10 dB 0 dB 5 dB 10 dB
BPSK 100 100 100 100 100 100 100 100 100 100 100 100 99.85 100 100
QPSK 98.33 100 100 98.33 100 100 100 100 100 44.25 100 100 81.80 100 100
8PSK 99.00 100 100 98.33 100 100 100 100 100 48.40 100 100 71.00 99.95 100
16APSK 77.67 100 100 77.67 100 100 68.50 75.83 82.33 35.60 83.40 99.95 47.05 97.65 100
32APSK 59.67 99.67 100 68.00 99.67 100 62.83 81.50 83.50 65.35 89.05 86.95 51.60 89.10 100
16QAM 100 100 100 76.00 100 100 84.17 90.50 95.33 17.65 60.85 100 38.90 86.05 100
64QAM 94.67 99.67 100 43.00 97.67 100 80.83 93.00 96.17 69.70 88.05 88.60 42.35 87.90 99.85
平均识别率 89.90 99.90 100 80.19 99.62 100 85.19 91.55 93.90 54.42 88.76 96.50 61.79 94.38 99.98
表 3  部分信噪比下的调制识别率
图 6  符号长度对识别结果的影响
图 7  基于单特征输入的卷积神经网络(星座图、矢量图)
图 8  基于不同特征的识别结果
图 9  残差模块对识别结果的影响
图 10  扩展性测试
识别方法 $ {t_{{\rm{gen}}}} $/s $ {t_{{\text{train}}}} $/s $ {N_{{\text{net}}}} $
本文的方法
基于星座图的卷积网络识别
基于高阶累积量的KNN识别
基于IQ数据的卷积网络识别
基于IQ数据的残差网络识别
0.758
0.168
0.0011
0.0023
0.0025
372.4
239.33
22.668
34.8
49.56
1070791
5967175

2748255
103337
表 4  不同算法的复杂度
1 DOBRE O A, ABDI A, BAR-NESS Y, et al Survey of automatic modulation classification techniques: classical approaches and new trends[J]. IET Communications, 2007, 1 (2): 137- 156
doi: 10.1049/iet-com:20050176
2 JDID B, HASSAN K, DAYOUB I, et al Machine learning based automatic modulation recognition for wireless communications: a comprehensive survey[J]. IEEE Access, 2021, 9 (1): 57851- 57873
3 李晨, 杨俊安, 刘辉 基于信息熵和GA-ELM的调制识别算法[J]. 系统工程与电子技术, 2020, 42 (1): 223- 229
LI Chen, YANG Jun-an, LIU Hui Modulation recognition algorithm based on information entropy and GA-ELM[J]. Systems Engineering and Electronics, 2020, 42 (1): 223- 229
doi: 10.3969/j.issn.1001-506X.2020.01.30
4 ZENG Y, ZHANG M, HAN F, et al Spectrum analysis and convolutional neural network for automatic modulation recognition[J]. IEEE Wireless Communications Letters, 2019, 8 (3): 929- 932
doi: 10.1109/LWC.2019.2900247
5 ALI A K, ERELEBI E Automatic modulation recognition of DVB-S2X standard-specific with an APSK-based neural network classifier[J]. Measurement, 2019, 151: 107257
6 XIE W, HU S, YU C, et al Deep learning in digital modulation recognition using high order cumulants[J]. IEEE Access, 2019, 7: 63760- 63766
doi: 10.1109/ACCESS.2019.2916833
7 WANG Y, LIU M, YANG J, et al Data-driven deep learning for automatic modulation recognition in cognitive radios[J]. IEEE Transactions on Vehicular Technology, 2019, 68 (4): 4074- 4077
doi: 10.1109/TVT.2019.2900460
8 李润东, 李立忠, 李少谦, 等 基于稀疏滤波神经网络的智能调制识别[J]. 电子科技大学学报, 2019, 48 (2): 161- 167
LI Run-dong, LI Li-zhong, LI Shao-qian, et al Intelligent modulation recognition based on sparse filtering neural network[J]. Journal of University of Electronic Science and Technology of China, 2019, 48 (2): 161- 167
doi: 10.3969/j.issn.1001-0548.2019.02.001
9 DENG L, YU D Deep learning: methods and applications[J]. Foundations and Trends in Signal Processing, 2014, 7 (3): 197- 387
10 MASI I, CHANG F J, CHOI J, et al Learning pose-aware models for pose-invariant face recognition in the wild[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 41 (2): 379- 393
11 CHOWDHURY A, ROSS A Fusing MFCC and LPC features using 1D triplet CNN for speaker recognition in severely degraded audio signals[J]. IEEE Transactions on Information Forensics and Security, 2020, 15: 1616- 1629
doi: 10.1109/TIFS.2019.2941773
12 HUANG C, LI Y, CHEN C L, et al Deep imbalanced learning for face recognition and attribute prediction[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42 (11): 2781- 2794
doi: 10.1109/TPAMI.2019.2914680
13 HUANG S, JIANG Y, GAO Y, et al Automatic modulation classification using contrastive fully convolutional network[J]. IEEE Wireless Communication Letters, 2019, 8 (4): 1044- 1047
doi: 10.1109/LWC.2019.2904956
14 TENG C F, CHOU C Y, CHEN C H, et al Accumulated polar feature-based deep learning for efficient and lightweight automatic modulation classification with channel compensation mechanism[J]. IEEE Transactions on Vehicular Technology, 2020, 69 (12): 15472- 15485
doi: 10.1109/TVT.2020.3041843
15 HUYNH-THE T, HUA C H, PHAM Q V, et al Mcnet: an efficient CNN architecture for robust automatic modulation classification[J]. IEEE Communications Letters, 2020, 244 (4): 811- 815
16 O'SHEA T J, ROY T, CLANCY T C Over-the-air deep learning based radio signal classification[J]. IEEE Journal of Selected Topics in Signal Processing, 2018, 12 (1): 168- 179
doi: 10.1109/JSTSP.2018.2797022
17 ANEDDA M, MELONI A, MURRONI M 64-APSK constellation and mapping optimization for satellite broadcasting using genetic algorithms[J]. IEEE Transactions on Broadcasting, 2015, 62 (1): 1- 9
18 HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition [C]// IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
19 MA J, LIN S C, GAO H, et al. Automatic modulation classification under non-Gaussian noise: a deep residual learning approach [C]// 2019 IEEE International Conference on Communications. Shanghai: IEEE, 2019: 1-6.
20 MORELLI M, ANDREA A N D, MENGALI U Feedforward Ml-based timing estimation with PSK signals[J]. IEEE Communications Letters, 1997, 1 (3): 80- 82
doi: 10.1109/4234.585803
21 ERUP L, GARDNER F M Interpolation in digital modems. II. implementation and performance[J]. IEEE Transactions on Communications, 1993, 41 (6): 998- 1008
doi: 10.1109/26.231921
22 PENG S L, JIANG H Y, WANG H X, et al Modulation classification based on signal constellation diagrams and deep learning[J]. IEEE Transactions on Neural Networks and Learning Systems, 2019, 30 (3): 718- 727
doi: 10.1109/TNNLS.2018.2850703
23 ASLAM M W, ZHU Z, NANDI A K Automatic modulation classification using combination of genetic programming and KNN[J]. IEEE Transactions on Wireless Communications, 2012, 11 (8): 2742- 2750
24 KRZYSTON J, BHATTACHARJEA R, STARK A. Complex-valued convolutions for modulation recognition using deep learning [C]// 2020 IEEE International Conference on Communications Workshops. Dublin: IEEE, 2020: 1-6.
[1] 刘近贞,陈飞,熊慧. 多尺度残差网络模型的开放式电阻抗成像算法[J]. 浙江大学学报(工学版), 2022, 56(9): 1789-1795.
[2] 明涛,王丹,郭继昌,李锵. 基于多尺度通道重校准的乳腺癌病理图像分类[J]. 浙江大学学报(工学版), 2020, 54(7): 1289-1297.
[3] 李红光,郭英,眭萍,齐子森. 基于时频特征的卷积神经网络跳频调制识别[J]. 浙江大学学报(工学版), 2020, 54(10): 1945-1954.