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

基于深度学习声谱图分类的"听声识风"

杨昊岩 栾涛 韩仲志 倪建功 高霁月

杨昊岩, 栾涛, 韩仲志, 倪建功, 高霁月. 基于深度学习声谱图分类的'听声识风'[J]. 华南师范大学学报(自然科学版), 2021, 53(5): 10-16. doi: 10.6054/j.jscnun.2021069
引用本文: 杨昊岩, 栾涛, 韩仲志, 倪建功, 高霁月. 基于深度学习声谱图分类的"听声识风"[J]. 华南师范大学学报(自然科学版), 2021, 53(5): 10-16. doi: 10.6054/j.jscnun.2021069
YANG Haoyan, LUAN Tao, HAN Zhongzhi, NI Jiangong, GAO Jiyue. 'Identify Wind Force by Listening' Based on Deep Learning Spectrogram Classification[J]. Journal of South China normal University (Natural Science Edition), 2021, 53(5): 10-16. doi: 10.6054/j.jscnun.2021069
Citation: YANG Haoyan, LUAN Tao, HAN Zhongzhi, NI Jiangong, GAO Jiyue. "Identify Wind Force by Listening" Based on Deep Learning Spectrogram Classification[J]. Journal of South China normal University (Natural Science Edition), 2021, 53(5): 10-16. doi: 10.6054/j.jscnun.2021069

基于深度学习声谱图分类的"听声识风"

doi: 10.6054/j.jscnun.2021069
基金项目: 

国家自然科学基金项目 31872849

山东省重点研发计划项目 2019GNC106037

青岛市科技惠民计划项目 19-6-1-66-nsh

青岛市科技惠民计划项目 19-6-1-72-nsh

详细信息
    通讯作者:

    栾涛,Email: tluan@qau.edu.cn

    韩仲志,Email: hanzhongzhi@qau.edu.cn

  • 中图分类号: TP399

"Identify Wind Force by Listening" Based on Deep Learning Spectrogram Classification

  • 摘要: 将深度学习与声谱图相结合,提出了一种新型的风级识别方法——"听声识风". 在实验室条件下模拟1~4级风并记录对应风声音频. 通过傅里叶变换等方法将风声音频转换成声谱图,共得到2 608幅二维声谱图像用作数据集. 将声谱图数据集导入深度卷积神经网络GoogLeNet中进行风力等级识别,测试准确率达到了99.6%. 为了进一步证明实验结果的可靠性,将声谱图数据集分别导入ResNet18、ShuffleNet中进行训练,均获得了99.2%的测试准确率,结果表明该方法可以有效地进行风级识别. "听声识风"研究首次通过深度学习声谱图分类实现了对风级的识别,这是一种智能的、快速的风级识别新方法.
  • 图  1  风声声波图

    Figure  1.  The wind sound sonogram

    图  2  风声声谱图

    Figure  2.  The wind sound spectrogram

    图  3  GoogLeNet网络结构图

    Figure  3.  The GoogLeNet network structure diagram

    图  4  3种模型训练结果对比

    Figure  4.  The comparison of training results in three models

    图  5  ROC曲线

    Figure  5.  The ROC curve

    表  1  1~4级风级

    Table  1.   The level 1~4 wind

    风级 名称 标准风速/(m·s-1) 模拟风速/(m·s-1)
    1 软风 0.3~1.5 0.90
    2 轻风 1.6~3.3 2.45
    3 微风 3.4~5.4 4.40
    4 和风 5.5~7.9 6.70
    下载: 导出CSV

    表  2  声谱图数据集

    Table  2.   The spectrogram data set

    类别 训练集 验证集 测试集 总计
    1级风 534 67 67 668
    2级风 506 63 63 632
    3级风 519 65 65 649
    4级风 527 66 66 659
    总计 2 086 261 261 2 608
    下载: 导出CSV

    表  3  混淆矩阵

    Table  3.   The confusion matrix

    设置风级 GoogLeNet ResNet18 ShuffleNet
    4级 1级 3级 2级 4级 1级 3级 2级 4级 1级 3级 2级
    4级 65 0 1 0 64 0 2 0 64 0 2 0
    1级 0 67 0 0 0 67 0 0 0 67 0 0
    3级 0 0 65 0 0 0 65 0 0 0 65 0
    2级 0 0 0 63 0 0 0 63 0 0 0 63
    准确率/% 99.6 99.2 99.2
    下载: 导出CSV

    表  4  模型性能评价指标

    Table  4.   The model performance evaluation

    指标 GoogLeNet ResNet18 ShuffleNet
    1级 2级 3级 4级 1级 2级 3级 4级 1级 2级 3级 4级
    灵敏度/% 100 100 100 98 100 100 100 97 100 100 100 97
    特异性/% 100 100 99.5 100 100 100 99 100 100 100 99 100
    准确率/% 100 100 98.5 100 100 100 97 100 100 100 97 100
    下载: 导出CSV

    表  5  3种网络分类结果的对比

    Table  5.   The comparison of three network classification results

    模型 层数 存储空间/MB 参数量/M 训练时间/s 准确率/%
    GoogLeNet 22 27.0 7.0 490 99.6
    ResNet18 18 44.0 11.7 354 99.2
    ShuffleNet 50 6.3 1.4 466 99.2
    下载: 导出CSV
  • [1] 孙力, 毕硕本, 潘威, 等. 明清时期东南沿海地区的台风重建及影响因素分析[J]. 华南师范大学学报(自然科学版), 2019, 51(6): 88-97. doi: 10.6054/j.jscnun.2019106

    SUN L, BI S B, PAN W, et al. Reconstructing typhoons in southeast coastal areas during Ming and Qing dynasties and analyzing the factors[J]. Journal of South China Normal University(Natural Science Edition), 2019, 51(6): 88-97. doi: 10.6054/j.jscnun.2019106
    [2] FRAUKE T, FLORIS V D M, LUEDER V B, et al. Minute-scale power forecast of offshore wind turbines using long-range single-Doppler Lidar measurements[J]. Wind Energy Science, 2020, 5(4): 1449-1468. doi: 10.5194/wes-5-1449-2020
    [3] LIU B M, GUO J P, GONG W, et al. Characteristics and performance of wind profiles as observed by the radar wind profiler network of China[J]. Atmospheric Measurement Techniques, 2020, 13(8): 4589-4600. doi: 10.5194/amt-13-4589-2020
    [4] 仇丽凤, 李雷茹. 基于GPS的GTS1型探空仪测风技术分析[J]. 科技创新导报, 2011(10): 101-101. doi: 10.3969/j.issn.1674-098X.2011.10.081

    QIU L F, LI L R. Analysis of GPS-based GTS1 radiosonde wind measurement technology[J]. Science and Technology Innovation Herald, 2011(10): 101-101. doi: 10.3969/j.issn.1674-098X.2011.10.081
    [5] 董德保, 张统明, 芮斌. 风廓线雷达大气风场观测误差分析[J]. 气象科技, 2014, 42(1): 48-53. doi: 10.3969/j.issn.1671-6345.2014.01.007

    DONG D B, ZHANG T M, RUI B. Wind profiler radar observation error analysis of atmospheric wind fields[J]. Meteorological Science and Technology, 2014, 42(1): 48-53. doi: 10.3969/j.issn.1671-6345.2014.01.007
    [6] SUBRAHMANYAM K, BABY S R. C-band Doppler weather radar observations during the passage of tropical cyclone 'Ockhi'[J]. Natural Hazards, 2020, 104: 2197-2211. doi: 10.1007/s11069-020-04268-2
    [7] PIRHALLA M, HEIST D, PERRY S, et al. Urban wind field analysis from the Jack rabbit Ⅱ special sonic anemometer study[J]. Atmospheric Environment, 2020, 243: 117871/1-14. doi: 10.1016/j.atmosenv.2020.117871
    [8] CADENAS E, RIVERA W, CAMPOS-AMEZCUA R, et al. Wind speed prediction using a univariate Arima model and a multivariate Narx model[J]. Energy Weekly News, 2016, 9(2): 109-123. http://www.chemie.de/fachpublikationen/921785/
    [9] 颜晓娟, 龚仁喜, 张千锋. 优化遗传算法寻优的SVM在短期风速预测中的应用[J]. 电力系统保护与控制, 2016, 44(9): 38-42. https://www.cnki.com.cn/Article/CJFDTOTAL-JDQW201609006.htm

    YAN X J, GONG R X, ZHANG Q F. Application of optimization svm based on improved genetic algorithm in short-term wind speed prediction[J]. Power System Protection and Control, 2016, 44(9): 38-42. https://www.cnki.com.cn/Article/CJFDTOTAL-JDQW201609006.htm
    [10] GUO Z, CHI D, WU J, et al. A new wind speed forecasting strategy based on the chaotic time series modelling technique and the Apriori algorithm[J]. Energy Conversion and Management, 2014, 84: 140-151. doi: 10.1016/j.enconman.2014.04.028
    [11] TIAN Z. Chaotic characteristic analysis of short-term wind speed time series with different time scales[J]. Energy Sources Part A: Recovery Utilization and Environmental Effects, 2019(3): 1649757/1-15. http://www.researchgate.net/publication/334973472_Chaotic_characteristic_analysis_of_short-term_wind_speed_time_series_with_different_time_scales
    [12] 廖恩红, 李会芳, 王华, 等. 基于卷积神经网络的食品图像识别[J]. 华南师范大学学报(自然科学版), 2019, 51(4): 113-119. doi: 10.6054/j.jscnun.2019075

    LIAO E H, LI H F, WANG H, et al. Food image recognition based on convolutional neural network[J]. Journal of South China Normal University(Natural Science Edition), 2019, 51(4): 113-119. doi: 10.6054/j.jscnun.2019075
    [13] 杜云梅, 黄帅, 梁会营. 基于深度卷积神经网络的脑电图异常检测[J]. 华南师范大学学报(自然科学版), 2020, 52(2): 122-128. doi: 10.6054/j.jscnun.2020035

    DU Y M, HUANG S, LIANG H Y. The detection of anomaly in electroencephalogram with deep convolutional neural networks[J]. Journal of South China Normal University(Natural Science Edition), 2020, 52(2): 122-128. doi: 10.6054/j.jscnun.2020035
    [14] SARHAN S, NASR A A, SHAMS M. Multipose face recognition-based combined adaptive deep learning vector quantization[J]. Computational Intelligence and Neuroscience, 2020, 2020: 8821868/1-11. http://www.researchgate.net/publication/344388125_Multipose_Face_Recognition-Based_Combined_Adaptive_Deep_Learning_Vector_Quantization
    [15] 蔡冬丽, 钟清华, 朱永升, 等. 基于混合神经网络的脑电情感识别[J]. 华南师范大学学报(自然科学版), 2021, 53(1): 109-118. doi: 10.6054/j.jscnun.2021017

    CAI D L, ZHONG Q H, ZHU Y S, et al. EEG emotion recognition based on hybrid neural networks[J]. Journal of South China Normal University(Natural Science Edition), 2021, 53(1): 109-118. doi: 10.6054/j.jscnun.2021017
    [16] YANG H Y, NI J G, GAO J Y, et al. A novel method for peanut variety identification and classification by improved VGG16. [J]. Scientific Reports, 2021, 11(1): 15756/1-17. http://www.nature.com/articles/s41598-021-95240-y?utm_source=xmol&utm_medium=affiliate&utm_content=meta
    [17] 张泽浩, 周卫星. 基于深度学习的图像去雾算法[J]. 华南师范大学学报(自然科学版), 2019, 51(3): 123-128. doi: 10.6054/j.jscnun.2019055

    ZHANG Z H, ZHOU W X. Image dehazing algorithm based on deep learning[J]. Journal of South China Normal University(Natural Science Edition), 2019, 51(3): 123-128. doi: 10.6054/j.jscnun.2019055
    [18] 马琼雄, 余润笙, 石振宇, 等. 基于深度强化学习的水下机器人最优轨迹控制[J]. 华南师范大学学报(自然科学版), 2018, 50(1): 118-123. http://journal-n.scnu.edu.cn/article/id/4199

    MA Q X, YU R S, SHI Z Y, et al. Autonomous underwater vehicles optimal trajectory control based on deep reinforcement learning[J]. Journal of South China Normal University(Natural Science Edition), 2018, 50(1): 118-123. http://journal-n.scnu.edu.cn/article/id/4199
    [19] AK R, VITELLI V, ZIO E. An interval-valued neural network approach for uncertainty quantification in short-term wind speed prediction[J]. IEEE Transactions on Neural Networks and Learning Systems, 2015, 26(11): 2787-2800. doi: 10.1109/TNNLS.2015.2396933
    [20] KHODAYAR M, KAYNAK O, KHODAYAR M E. Rough deep neural architecture for short-term wind speed forecasting[J]. IEEE Transactions on Industrial Informatics, 2017, 13(6): 2770-2779. doi: 10.1109/TII.2017.2730846
    [21] CHEN J, ZENG G Q, ZHOU W, et al. Wind speed forecasting using nonlinear-learning ensemble of deep lear-ning time series prediction and extremal optimization[J]. Energy Conversion and Management, 2018, 165: 681-695. doi: 10.1016/j.enconman.2018.03.098
    [22] LIU H, MI X, LI Y. Wind speed forecasting method based on deep learning strategy using empirical wavelet transform, long short term memory neural network and elman neural network[J]. Energy Conversion and Management, 2018, 156: 498-514. doi: 10.1016/j.enconman.2017.11.053
    [23] SZEGEDY C, LIU W, JIA Y, et al. Going deeper with convo-lutions[C]//Computer Vision and Pattern Recognition. Boston: IEEE, 2015: 1-9.
    [24] LANDIS H R, KOCH G G. The measurement of observer agreement for categorical data[J]. Biometrics, 1977, 33(1): 159-174. doi: 10.2307/2529310
    [25] BLANCHE P, JEAN-FRANCOIS D, RIOU J. A closed max-t test for multiple comparisons of areas under the ROC curve[J]. Biometrics, 2020, 76(4): 1-12. doi: 10.1111/biom.13401
  • 加载中
图(5) / 表(5)
计量
  • 文章访问数:  221
  • HTML全文浏览量:  144
  • PDF下载量:  33
  • 被引次数: 0
出版历程
  • 收稿日期:  2021-04-12
  • 网络出版日期:  2021-11-11
  • 刊出日期:  2021-10-25

目录

    /

    返回文章
    返回