留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于改进AdaBoost.M2算法的自动调制识别方法

王沛 刘春辉 张多纳

王沛,刘春辉,张多纳. 基于改进AdaBoost.M2算法的自动调制识别方法[J]. 北京航空航天大学学报,2023,49(8):2089-2098 doi: 10.13700/j.bh.1001-5965.2021.0577
引用本文: 王沛,刘春辉,张多纳. 基于改进AdaBoost.M2算法的自动调制识别方法[J]. 北京航空航天大学学报,2023,49(8):2089-2098 doi: 10.13700/j.bh.1001-5965.2021.0577
WANG P,LIU C H,ZHANG D N. Automatic modulation recognition method based on improved weight AdaBoost.M2 algorithm[J]. Journal of Beijing University of Aeronautics and Astronautics,2023,49(8):2089-2098 (in Chinese) doi: 10.13700/j.bh.1001-5965.2021.0577
Citation: WANG P,LIU C H,ZHANG D N. Automatic modulation recognition method based on improved weight AdaBoost.M2 algorithm[J]. Journal of Beijing University of Aeronautics and Astronautics,2023,49(8):2089-2098 (in Chinese) doi: 10.13700/j.bh.1001-5965.2021.0577

基于改进AdaBoost.M2算法的自动调制识别方法

doi: 10.13700/j.bh.1001-5965.2021.0577
基金项目: 北京市自然科学基金(4204102)
详细信息
    通讯作者:

    E-mail:liuchunhui2134@buaa.edu.cn

  • 中图分类号: TN911.3

Automatic modulation recognition method based on improved weight AdaBoost.M2 algorithm

Funds: Beijing Natural Science Foundation (4204102)
More Information
  • 摘要:

    针对同族调制类型通信信号识别难度大、深度学习模型普遍存在泛化能力弱的问题,基于经典AdaBoost.M2算法,提出改进样本权重的AdaBoost.M2算法,用于解决大样本情况下学习率与加权后样本数据难以相适应的问题。改进后的新样本权重确保训练样本数据的数量级在加权后不变,并使算法更迅速地关注到难分类样本,提高了弱分类器综合性能,降低了加权投票模型中弱分类器重要性之间的差异。针对部分样本的统计特性易淹没于噪声中造成难分类问题,提出随机特征裁剪方法,使算法避免过度关注异常特征,降低了极难分类样本对AdaBoost.M2算法性能的负面影响,提升了算法的泛化能力,并以低信噪比数据进行实验验证。针对调制类型同族信号难分类的问题,选取同族调制类型的通信信号开展模型训练和测试。实验结果表明:相比于单一卷积长短时记忆全连接深度网络(CLDNN)算法,改进AdaBoost.M2算法对低信噪比PSK族类和QAM族类通信信号的测试集准确率分别提高了8.5%和11.25%,相比于直接集成CLDNN的经典AdaBoost.M2算法,测试集准确率分别提高了8.25%和6.5%。

     

  • 图 1  AdaBoost.M2算法框架

    Figure 1.  Framework of AdaBoost.M2 algorithm

    图 2  样本权重调整流程

    Figure 2.  Process of sample weight adjustment

    图 3  弱分类器CLDNN模型

    Figure 3.  Weak classifier CLDNN model

    图 4  分类器权重统计特性对比

    Figure 4.  Comparison of statistical characteristics of classifier weights

    图 5  分类器训练集准确率对比

    Figure 5.  Comparison of accuracy of training set

    图 6  分类器测试集准确率对比

    Figure 6.  Comparison of accuracy of testing set

    图 7  特征裁剪方法示意图

    Figure 7.  Schematic diagram of feature clipping method

    图 8  CLDNN算法对PSK族类分类结果混淆矩阵

    Figure 8.  Confusion matrix of PSK family classification by CLDNN algorithm

    图 9  经典AdaBoost.M2算法对PSK族类分类结果混淆矩阵

    Figure 9.  Confusion matrix of PSK family classification by classical AdaBoost.M2 algorithm

    图 10  本文算法对PSK族类分类结果混淆矩阵

    Figure 10.  Confusion matrix of PSK family classification by proposed algorithm

    图 11  CLDNN算法对QAM族类分类结果混淆矩阵

    Figure 11.  Confusion matrix of QAM family classification by CLDNN algorithm

    图 12  经典AdaBoost.M2算法对QAM族类分类结果混淆矩阵

    Figure 12.  Confusion matrix of QAM family classification by classical Adaboost.M2 algorithm

    图 13  本文算法对QAM族类分类结果混淆矩阵

    Figure 13.  Confusion matrix of QAM family classification by proposed algorithm

    图 14  PSK族类不同信噪比下分类准确率对比

    Figure 14.  Comparison of classification accuracy of PSK family with different SNRs

    图 15  不同信噪比下8PSK和16PSK信号分类准确率对比

    Figure 15.  Comparison of 8PSK and 16PSK signal classification accuracy with different signal-to-noise ratios

    图 17  不同信噪比下16QAM和32QAM信号分类准确率对比

    Figure 17.  Comparison of 16QAM and 32QAM signal classification accuracy with different signal-to-noise ratios

    图 16  QAM族类不同信噪比下分类准确率对比

    Figure 16.  Comparsion of classification accuracy of QAM family with different signal-to-noise ratios

    表  1  信号族类及调制类型

    Table  1.   Signal family and modulation type

    族类调制类型
    PSK8PSK、16PSK、32PSK、16APSK、32APSK
    QAM16QAM、32QAM、64QAM、128QAM、256QAM
    下载: 导出CSV

    表  2  完整的参数组合及强分类器性能对比

    Table  2.   Complete parameter combination and comparison of strong classifier performance %

    算法训练集准确率测试集准确率
    无参数组合A$\psi $无参数组合A$\psi $
    2351023510
    CLDNN 88.85 79.75
    经典AdaBoost.M2 85.45 80.00
    改进样本权重的
    AdaBoost.M2
    0.5 92.80 93.60 92.85 94.85 0.5 84.75 84.00 85.50 84.50
    1 94.45 93.85 94.70 94.25 1 84.50 85.25 85.50 84.25
    2 94.65 95.40 95.60 96.10 2 82.75 81.75 81.50 83.50
    下载: 导出CSV

    表  3  不同特征裁剪方法参数下性能对比

    Table  3.   Performance comparison of different feature clipping methods with different parameters %

    cs训练集准确率测试集准确率
    cf=64cf=128cf=256cf=512cf=64cf=128cf=256cf=512
    12592.4593.1093.7093.6080.5078.2583.2580.50
    6295.3095.5095.5095.8084.5084.5084.2586.75
    3091.5091.6091.6089.1585.5085.5085.2584.25
    2090.0092.2091.2591.1086.0085.2583.2586.25
    1092.6092.8092.7593.8584.7587.2588.2584.25
    下载: 导出CSV

    表  4  算法复杂度对比

    Table  4.   Complexity comparison of algorithm

    算法训练2 000个
    样本用时/min
    单样本测试
    平均用时/ms
    CLDNN1.3517.45
    经典AdaBoost.M226.31349.50
    改进样本权重的AdaBoost.M227.85390.13
    本文算法30.66398.63
    下载: 导出CSV

    表  5  各算法对不同族类信号分类结果对比

    Table  5.   Comparison of classification results of different signals by each algorithm %

    算法训练集准确率 测试集准确率
    PSKQAMPSKQAM
    CLDNN88.8593.40 79.7575.75
    经典AdaBoost.M285.4587.9580.0080.50
    本文算法92.7595.7088.2587.00
    下载: 导出CSV
  • [1] 李晓峰, 周宁, 周亮. 通信原理[M]. 2版. 北京: 清华大学出版社, 2014: 1-10.

    LI X F, ZHOU N, ZHOU L. Communication principle[M]. 2nd ed. Beijing: Tsinghua University Press, 2014: 1-10(in Chinese).
    [2] 高文欢. 浅析无线电信号通用解调和调制识别技术[J]. 中国无线电, 2015(3): 61-62. doi: 10.3969/j.issn.1672-7797.2015.03.037

    GAO W H. Analysis on general demodulation and modulation identification technology of radio signals[J]. China Radio, 2015(3): 61-62(in Chinese). doi: 10.3969/j.issn.1672-7797.2015.03.037
    [3] 程汉文, 朱雷, 吴乐南. 基于累计量的干扰信号调制识别算法[J]. 电子与信息学报, 2009, 31(7): 1741-1745.

    CHENG H W, ZHU L, WU L N. Modulation classification algorithm for jamming signal based on cumulant[J]. Journal of Electronics & Information Technology, 2009, 31(7): 1741-1745(in Chinese).
    [4] 屈明, 张效义. 基于频域处理的信号检测和自动调制识别算法[J]. 数据采集与处理, 2006, 21(S1): 19-21. doi: 10.3969/j.issn.1004-9037.2006.z1.005

    QU M, ZHANG X Y. Signal detection and classification algorithm based on frequency region processing[J]. Journal of Data Acquisition & Processing, 2006, 21(S1): 19-21(in Chinese). doi: 10.3969/j.issn.1004-9037.2006.z1.005
    [5] 潘宝凤. 通信侦察系统总体设计技术[J]. 电讯技术, 2011, 51(6): 1-5. doi: 10.3969/j.issn.1001-893x.2011.06.001

    PAN B F. Overall design of communication reconnaissance system[J]. Telecommunication Engineering, 2011, 51(6): 1-5(in Chinese). doi: 10.3969/j.issn.1001-893x.2011.06.001
    [6] WU H C, SAQUIB M, YUN Z F. Novel automatic modulation classification using cumulant features for communications via multipath channels[J]. IEEE Transactions on Wireless Communications, 2008, 7(8): 3098-3105. doi: 10.1109/TWC.2008.070015
    [7] 张天骐, 范聪聪, 葛宛营, 等. 基于ICA和特征提取的MIMO信号调制识别算法[J]. 电子与信息学报, 2020, 42(9): 2208-2215. doi: 10.11999/JEIT190320

    ZHANG T Q, FAN C C, GE W Y, et al. MIMO signal modulation recognition algorithm based on ICA and feature extraction[J]. Journal of Electronics & Information Technology, 2020, 42(9): 2208-2215(in Chinese). doi: 10.11999/JEIT190320
    [8] NANDI A K, AZZOUZ E E. Modulation recognition using artificial neural networks[J]. Signal Processing, 1997, 56(2): 165-175. doi: 10.1016/S0165-1684(96)00165-X
    [9] MUSTAFA H, DOROSLOVACKI M. Digital modulation recognition using support vector machine classifier[C]//Conference Record of the Thirty-Eighth Asilomar Conference on Signals, Systems and Computers. Piscataway: IEEE Press, 2005: 2238-2242.
    [10] FEHSKE A, GAEDDERT J, REED J H. A new approach to signal classification using spectral correlation and neural networks[C]//First IEEE International Symposium on New Frontiers in Dynamic Spectrum Access Networks. Piscataway: IEEE Press, 2005: 144-150.
    [11] HEADLEY W C, REED J D, DA SILVA C R C M. Distributed cyclic spectrum feature-based modulation classification[C]//2008 IEEE Wireless Communications and Networking Conference. Piscataway: IEEE Press, 2008: 1200-1204.
    [12] O’SHEA T J, HOYDIS J. An introduction to deep learning for the physical layer[J]. IEEE Transactions on Cognitive Communications and Networking, 2017, 3(4): 563-575. doi: 10.1109/TCCN.2017.2758370
    [13] KARRA K, KUZDEBA S, PETERSEN J. Modulation recognition using hierarchical deep neural networks[C]//2017 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN). Piscataway: IEEE Press, 2017: 1-3.
    [14] 陈曦. 基于深度学习的数字信号调制方式识别方法研究[D]. 西安: 西安电子科技大学, 2018: 9-67.

    CHEN X. Research on digital signal modulation recognition method based on deep learning[D]. Xi'an: Xidian University, 2018: 9-67 (in Chinese).
    [15] 张劭. 基于深度学习的未知调制类型的信号识别[D]. 西安: 西安电子科技大学, 2018: 7-54.

    ZHANG S. Signal recognition of unknown modulation type based on deep learning[D]. Xi'an: Xidian University, 2018: 7-54(in Chinese).
    [16] O’SHEA T J, CORGAN J, CLANCY T C. Convolutional radio modulation recognition networks[C]//International Conference on Engineering Applications of Neural Networks. Berlin: Springer, 2016: 213-226.
    [17] ZHANG D N, DING W R, ZHANG B C, et al. Automatic modulation classification based on deep learning for unmanned aerial vehicles[J]. Sensors, 2018, 18(3): 924. doi: 10.3390/s18030924
    [18] RAJENDRAN S, MEERT W, GIUSTINIANO D, et al. Deep learning models for wireless signal classification with distributed low-cost spectrum sensors[J]. IEEE Transactions on Cognitive Communications and Networking, 2018, 4(3): 433-445. doi: 10.1109/TCCN.2018.2835460
    [19] 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
    [20] LIU X Y, YANG D Y, EL GAMAL A. Deep neural network architectures for modulation classification[C]// 2017 51st Asilomar Conference on Signals, Systems, and Computers. Piscataway: IEEE Press, 2018: 915-919.
    [21] 张宇. 基于深度学习的通信信号调制识别算法研究[D]. 哈尔滨: 哈尔滨工程大学, 2019: 7-59.

    ZHANG Y. Research on modulation recognition algorithm of communication signal based on deep learning[D]. Harbin: Harbin Engineering University, 2019: 7-59(in Chinese).
    [22] 陶桂平, 韩立岩. 最大熵原理在不完全信息博弈中的应用[J]. 首都经济贸易大学学报, 2011, 13(3): 67-71. doi: 10.3969/j.issn.1008-2700.2011.03.010

    TAO G P, HAN L Y. The application of the maximum entropy principle in the incomplete information game[J]. Journal of Capital University of Economics and Business, 2011, 13(3): 67-71(in Chinese). doi: 10.3969/j.issn.1008-2700.2011.03.010
    [23] 廖红文, 周德龙. AdaBoost及其改进算法综述[J]. 计算机系统应用, 2012, 21(5): 240-244. doi: 10.3969/j.issn.1003-3254.2012.05.056

    LIAO H W, ZHOU D L. Review of AdaBoost and its improvement[J]. Computer Systems & Applications, 2012, 21(5): 240-244(in Chinese). doi: 10.3969/j.issn.1003-3254.2012.05.056
  • 加载中
图(17) / 表(5)
计量
  • 文章访问数:  165
  • HTML全文浏览量:  35
  • PDF下载量:  21
  • 被引次数: 0
出版历程
  • 收稿日期:  2021-09-28
  • 录用日期:  2021-10-29
  • 网络出版日期:  2021-11-11
  • 整期出版日期:  2023-08-31

目录

    /

    返回文章
    返回
    常见问答