高级搜索

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

基于深度卷积神经网络的协作频谱感知方法

盖建新 薛宪峰 吴静谊 南瑞祥

盖建新, 薛宪峰, 吴静谊, 南瑞祥. 基于深度卷积神经网络的协作频谱感知方法[J]. 电子与信息学报. doi: 10.11999/JEIT201005
引用本文: 盖建新, 薛宪峰, 吴静谊, 南瑞祥. 基于深度卷积神经网络的协作频谱感知方法[J]. 电子与信息学报. doi: 10.11999/JEIT201005
Jianxin GAI, Xianfeng XUE, Jingyi WU, Ruixiang NAN. Cooperative Spectrum Sensing Method Based on Deep Convolutional Neural Network[J]. Journal of Electronics and Information Technology. doi: 10.11999/JEIT201005
Citation: Jianxin GAI, Xianfeng XUE, Jingyi WU, Ruixiang NAN. Cooperative Spectrum Sensing Method Based on Deep Convolutional Neural Network[J]. Journal of Electronics and Information Technology. doi: 10.11999/JEIT201005

基于深度卷积神经网络的协作频谱感知方法

doi: 10.11999/JEIT201005
基金项目: 国家自然科学基金(61501150),黑龙江省自然科学基金(QC2014C074),黑龙江省省属本科高校基本科研业务费科研项目(2018-KYYWF-1656)
详细信息
    作者简介:

    盖建新:男,1980年生,博士,副教授,研究方向为频谱感知、机器学习、亚奈奎斯特采样理论、压缩感知等

    薛宪峰:男,1996年生,硕士生,研究方向为深度学习

    吴静谊:女,1996年生,硕士生,研究方向为压缩感知

    南瑞祥:男,1996年生,硕士生,研究方向为通信信号处理

    通讯作者:

    盖建新 jxgai@hrbust.edu.cn

  • 中图分类号: TN911

Cooperative Spectrum Sensing Method Based on Deep Convolutional Neural Network

Funds: The National Natural Science Foundation of China (61501150), The Natural Science Foundation of Heilongjiang Province (QC2014C074), The Fundamental Research Funds for the Universities in Heilongjiang Province (2018-KYYWF-1656)
  • 摘要: 针对传统卷积神经网络(CNN)频谱感知方法提取特征能力受限于网络结构简单,增加网络结构又容易出现梯度消失等问题,该文通过在传统卷积神经网络中添加捷径连接,实现输入层恒等映射更深的网络,提出一种基于深度卷积神经网络(DCNN)的协作频谱感知方法。该方法将频谱感知问题转化为图像二分类问题,对正交相移键控(QPSK)信号的协方差矩阵进行归一化灰度处理,并作为深度卷积神经网络的输入,通过残差学习训练深度卷积神经网络模型,提取2维灰度图像的深层特征,将测试数据输入到训练好的模型中,完成基于图像分类的频谱感知。实验结果表明:与传统的频谱感知方法相比,在低信噪比(SNR)下、多用户协作感知时,所提方法具有更高的检测概率和更低的虚警概率。
  • 图  1  频谱感知模型

    图  2  QPSK系统框图

    图  3  DCNN的块结构

    图  4  简化的DCNN结构

    图  5  DCNN的结构框图

    图  6  基本RLM

    图  7  DCNN, CNN准确率随网络层数的变化

    图  8  不同非授权用户数的DCNN分类准确率随网络层数的变化

    图  9  DCNN, CNN的准确率随迭代次数的变化

    图  10  DCNN, CNN的损失随迭代次数的变化

    图  11  DCNN, CNN和SVM在不同SNR下的检测概率

    图  12  DCNN, CNN和SVM频谱感知方法的ROC曲线

    表  1  DCNN的结构参数

     输入:采样协方差矩阵 (维度: 40 × 40)
     DCNN的各层卷积核大小
     输入层 Null
     卷积层 8@(3×3)
     RLM1 8@(1×1)8@(3×3)8@(1×1)
     RLM2 8@(1×1)8@(3×3)8@(1×1)
     捷径连接卷积层1 16@(1×1)
     RLM3 16@(1×1)16@(3×3)16@(1×1)
     RLM4 16@(1×1)16@(3×3)16@(1×1)
     捷径连接卷积层2 32@(1×1)
     RLM5 32@(1×1)32@(3×3)32@(1×1)
     RLM6 32@(1×1)32@(3×3)32@(1×1)
     全连接2 × 1
     输出:特征向量 (维度: 2 × 1)
    下载: 导出CSV

    表  2  DCNN协作频谱感知算法

     基于DCNN的协作频谱感知算法
     输入:训练样本$\{ ({ {{\mathit{\boldsymbol{x}}} }^{(1)} },{ {{\mathit{\boldsymbol{y}}} }^{(1)} }), ··· ({ {{\mathit{\boldsymbol{x}}} }^{(m)} },{ {{\mathit{\boldsymbol{y}}} }^{(m)} })\}$,权值${{\mathit{\boldsymbol{W}}}}$,测试样本$\{ ({ {{\mathit{\boldsymbol{x}}} }^{(m + 1)} },{ {{\mathit{\boldsymbol{y}}} }^{(m + 1)} }), ···, ({ {{\mathit{\boldsymbol{x}}} }^{(m + n)} },{ {{\mathit{\boldsymbol{y}}} }^{(m + n)} })\}$
     输出:检测概率${P_{\rm{d}}}$和虚警概率${P_{{\rm{af}}}}$
     步骤1:(训练阶段)输入训练样本$\{ ({ {{\mathit{\boldsymbol{x}}} }^{(1)} },{ {{\mathit{\boldsymbol{y}}} }^{(1)} }), ···,({ {{\mathit{\boldsymbol{x}}} }^{(m)} },{ {{\mathit{\boldsymbol{y}}} }^{(m)} })\}$。
     步骤2:Loop
        ${{{\mathit{\boldsymbol{x}}}}_{\rm{0}}}{\rm{ = }}f{(}{{{\mathit{\boldsymbol{W}}}}_{{\rm{ - 1}}}} \times {{{\mathit{\boldsymbol{x}}}}^{(m)}})$
        迭代计算l = 0, 3, 6, 9, 12, 15时的输出
        ${ {{\mathit{\boldsymbol{x}}} }_L} = { {{\mathit{\boldsymbol{x}}} }_l} + \displaystyle\sum\nolimits_{i = l}^{L - 1} {F({ {{\mathit{\boldsymbol{x}}} }_i},\{ { {{\mathit{\boldsymbol{W}}} }_i}{\rm{\} } })}$,L=l+3
        依次得到${{{\mathit{\boldsymbol{x}}}}_3}$, ${{{\mathit{\boldsymbol{x}}}}_6}$, ${{{\mathit{\boldsymbol{x}}}}_9}$, ${{{\mathit{\boldsymbol{x}}}}_{12}}$, ${{{\mathit{\boldsymbol{x}}}}_{15}}$, ${{{\mathit{\boldsymbol{x}}}}_{18}}$
        按照式(16)更新$ {\widehat{y}}^{(m)}$
        Until
        ${\rm{loss}} = \dfrac{1}{2}\displaystyle\sum\limits_{k = 1}^c { { {\left\| { { { {\hat{\mathit{\boldsymbol{y}}} } }^{(m)} }(k) - { {{\mathit{\boldsymbol{y}}} }^{(m)} }(k)} \right\|}^{ {\rm{ } }2} } }$收敛
     步骤3:(测试阶段)将测试集数据$\{ ({ {{\mathit{\boldsymbol{x}}} }^{(m + 1)} },{ {{\mathit{\boldsymbol{y}}} }^{(m + 1)} }),···, ({ {{\mathit{\boldsymbol{x}}} }^{(m + n)} },{ {{\mathit{\boldsymbol{y}}} }^{(m + n)} })\}$输入训练好的DCNN模型中,正确识别授权用户发射信号的样本数为${k_{{\rm{signal}}}}$,正确识别纯噪声的样本数为${k_{{\rm{noise}}}}$。
     步骤4:计算检测概率${P_{\rm{d} } } = { { {k_{ {\rm{signal} } } } } }/{n}$,虚警概率${P_{ {\rm{af} } } } = ({ {n - {k_{ {\rm{noise} } } } } })/{n}$。
    下载: 导出CSV

    表  3  3种算法的离线训练和在线检测时间(s)

    离线训练时间在线检测时间
    DCNN_5L19.261.66
    CNN_5L22.462.68
    DCNN_21L33.923.82
    CNN_21L229.934.87
    SVM14.004.70
    下载: 导出CSV
  • DIGHAM F F, ALOUINI M S, and SIMON M K. On the energy detection of unknown signals over fading channels[J]. IEEE Transactions on Communications, 2007, 55(1): 21–24. doi: 10.1109/TCOMM.2006.887483
    YANG Mingchuan, LI Yuan, LIU Xiaofeng, et al. Cyclostationary feature detection based spectrum sensing algorithm under complicated electromagnetic environment in cognitive radio networks[J]. China Communications, 2015, 12(9): 35–44. doi: 10.1109/CC.2015.7275257
    ZHANG Xinzhi, GAO Feifei, CHAI Rong, et al. Matched filter based spectrum sensing when primary user has multiple power levels[J]. China Communications, 2015, 12(2): 21–31. doi: 10.1109/CC.2015.7084399
    王磊, 郑宝玉, 李雷. 基于随机矩阵理论的协作频谱感知[J]. 电子与信息学报, 2009, 31(8): 1925–1929. doi: 10.3724/SP.J.1146.2008.01154

    WANG Lei, ZHENG Baoyu, and LI Lei. Cooperative spectrum sensing based on random matrix theory[J]. Journal of Electronics &Information Technology, 2009, 31(8): 1925–1929. doi: 10.3724/SP.J.1146.2008.01154
    许炜阳, 李有均, 徐宏乾, 等. 基于随机矩阵非渐近谱理论的协作频谱感知算法研究[J]. 电子与信息学报, 2018, 40(1): 123–129. doi: 10.11999/JEIT170309

    XU Weiyang, LI Youjun, XU Hongqian, et al. Study on cooperative spectrum sensing algorithm based on random matrix non-asymptotic spectral theory[J]. Journal of Electronics &Information Technology, 2018, 40(1): 123–129. doi: 10.11999/JEIT170309
    ZHOU Fuhui, BEAULIEU N C, LI Zan, et al. Feasibility of maximum eigenvalue cooperative spectrum sensing based on Cholesky factorisation[J]. IET Communications, 2016, 10(2): 199–206. doi: 10.1049/iet-com.2015.0252
    THILINA K M, CHOI K W, SAQUIB N, et al. Machine learning techniques for cooperative spectrum sensing in cognitive radio networks[J]. IEEE Journal on Selected Areas in Communications, 2013, 31(11): 2209–2221. doi: 10.1109/JSAC.2013.131120
    BAO Jianrong, NIE Jianyuan, LIU Chao, et al. Improved blind spectrum sensing by covariance matrix cholesky decomposition and RBF-SVM decision classification at low SNRs[J]. IEEE Access, 2019, 7: 97117–97129. doi: 10.1109/ACCESS.2019.2929316
    陈思吉, 王欣, 申滨. 一种基于支持向量机的认知无线电频谱感知方案[J]. 重庆邮电大学学报: 自然科学版, 2019, 31(3): 313–322. doi: 10.3979/j.issn.1673-825X.2019.03.005

    CHEN Siji, WANG Xin, and SHEN Bin. A support vector machine based spectrum sensing for cognitive radios[J]. Journal of Chongqing University of Posts and Telecommunications:Natural Science Edition, 2019, 31(3): 313–322. doi: 10.3979/j.issn.1673-825X.2019.03.005
    YU Chunyan, ZHAO Meng, SONG Meiping, et al. Hyperspectral image classification method based on CNN architecture embedding with hashing semantic feature[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2019, 12(6): 1866–1881. doi: 10.1109/JSTARS.2019.2911987
    YU Chunyan, HAN Rui, SONG Meiping, et al. A simplified 2D-3D CNN architecture for hyperspectral image classification based on spatial–spectral fusion[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 13: 2485–2501. doi: 10.1109/JSTARS.2020.2983224
    MAFFEI A, HAUT J M, PAOLETTI M E, et al. A single model CNN for hyperspectral image denoising[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(4): 2516–2529. doi: 10.1109/TGRS.2019.2952062
    CHEN Yushi, ZHU Kaiqiang, ZHU Lin, et al. Automatic design of convolutional neural network for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(9): 7048–7066. doi: 10.1109/TGRS.2019.2910603
    LIU Chang, WANG Jie, LIU Xuemeng, et al. Deep CM-CNN for spectrum sensing in cognitive radio[J]. IEEE Journal on Selected Areas in Communications, 2019, 37(10): 2306–2321. doi: 10.1109/JSAC.2019.2933892
    LEE W, KIM M, and CHO D H. Deep cooperative sensing: Cooperative spectrum sensing based on convolutional neural networks[J]. IEEE Transactions on Vehicular Technology, 2019, 68(3): 3005–3009. doi: 10.1109/TVT.2019.2891291
    张孟伯, 王伦文, 冯彦卿. 基于卷积神经网络的OFDM频谱感知方法[J]. 系统工程与电子技术, 2019, 41(1): 178–186. doi: 10.3969/j.issn.1001-506X.2019.01.25

    ZHANG Mengbo, WANG Lunwen, and FENG Yanqing. OFDM spectrum sensing method based on convolutional neural networks[J]. Systems Engineering and Electronics, 2019, 41(1): 178–186. doi: 10.3969/j.issn.1001-506X.2019.01.25
    MARSHALL J A. Neural networks for pattern recognition: By Albert Nigrin, MIT Press, Cambridge, MA: 1993, $45.00 413 pp., ISBN 0-262-14054-3[J]. Neural Networks, 1995, 8(3): 493–494. doi: 10.1016/0893-6080(95)90002-0
    JAMES D A. Reviewed work: Modern applied statistics with S-PLUS by W. N. Venables, B. D. Ripley[J]. Technometrics, 1996, 38(1): 77–78. doi: 10.2307/1268908
    WITTEK P. Pattern recognition and neural networks[M]. WITTEK P. Quantum Machine Learning. Amsterdam: Elsevier, 2014: 63–71. doi: 10.1016/b978-0-12-800953-6.00006-2.
    孙月驰, 李冠. 基于卷积神经网络嵌套模型的人群异常行为检测[J]. 计算机应用与软件, 2019, 36(3): 196–201, 276. doi: 10.3969/j.issn.1000-386x.2019.03.036

    SUN Yuechi and LI Guan. Abnormal behavior detection of crowds based on nested model of convolutional neural network[J]. Computer Applications and Software, 2019, 36(3): 196–201, 276. doi: 10.3969/j.issn.1000-386x.2019.03.036
  • 加载中
图(12) / 表(3)
计量
  • 文章访问数:  65
  • HTML全文浏览量:  22
  • PDF下载量:  16
  • 被引次数: 0
出版历程
  • 收稿日期:  2020-11-30
  • 修回日期:  2021-03-12
  • 网络出版日期:  2021-03-25

目录

    /

    返回文章
    返回

    官方微信,欢迎关注