Laser & Optoelectronics Progress, Volume. 57, Issue 2, 21009(2020)

Human Activity and IdentityMulti-Task Recognition Based on Convolutional Neural Network Using Doppler Radar

Hou Chunping, Jiang Tianli, Lang Yue*, and Yang Yang
Author Affiliations
  • School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China
  • show less
    Figures & Tables(13)
    Flow chart of building radar time-frequency image data set
    Radar time-frequency spectrograms of radar for six human activities. (a) Walking; (b) boxing; (c) crawling on the ground; (d) diving; (e) standing forward and jumping; (f) running
    Structural diagram of SE module
    Multi-task recognition model structure based on CNN
    Diagram of joint loss function generation
    Loss and accuracy curves. (a) Loss curves; (b) accuracy curves
    • Table 1. Results of multi-task and single-task recognitions

      View table

      Table 1. Results of multi-task and single-task recognitions

      Crossvalidationset No.Recognitionrate of MR /%Recognitionrate of SR /%
      HAHDHTHAHDHT
      199.9498.7898.7499.7095.9995.72
      299.9498.8098.7799.8896.7096.93
      399.9798.9098.8999.9097.9197.84
      4100.0099.9399.9399.9699.3799.34
      Mean99.9699.1099.0899.8697.4997.38
    • Table 2. Recognition results of different network structures

      View table

      Table 2. Recognition results of different network structures

      ModelRecognition rate /%
      HAHDHT
      D-MSSE99.9899.6299.61
      D-MS10099.8899.88
      D-SE99.9990.4790.47
    • Table 3. Recognition results of different loss optimization

      View table

      Table 3. Recognition results of different loss optimization

      LossRecognition rate /%
      HAHDHT
      LE99.9999.7499.73
      LE +LM99.9999.8699.86
      LE+LC10099.8799.86
    • Table 4. Recognition results under different STFT window lengths

      View table

      Table 4. Recognition results under different STFT window lengths

      L /sRecognition rate /%
      HAHDHT
      0.297.9978.5977.68
      0.499.7589.2589.13
      0.699.9192.4792.44
      0.899.9899.7399.71
    • Table 5. Recognition results under different NSR

      View table

      Table 5. Recognition results under different NSR

      NSR /dBRecognition rate /%
      HAHDHT
      -2067.5324.419.06
      -1096.9177.6376.49
      099.9190.890.75
      1099.9999.599.49
      +∞10099.9399.93
    • Table 6. Recognition results under different training data sizes

      View table

      Table 6. Recognition results under different training data sizes

      Data sizeRecognition rate /%
      HAHDHT
      40099.9497.5297.46
      60099.9799.7199.57
      80099.9799.6299.59
      100099.9999.7699.75
      130099.9999.8599.84
      150010099.9399.93
    • Table 7. Comparison of results of algorithms

      View table

      Table 7. Comparison of results of algorithms

      AlgorithmRecognition rate /%
      HAHDHT
      PRGC99.9498.998.84
      MRN99.9799.999.85.
      MCNN10099.9399.93
    Tools

    Get Citation

    Copy Citation Text

    Hou Chunping, Jiang Tianli, Lang Yue, Yang Yang. Human Activity and IdentityMulti-Task Recognition Based on Convolutional Neural Network Using Doppler Radar[J]. Laser & Optoelectronics Progress, 2020, 57(2): 21009

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Image Processing

    Received: Jun. 10, 2019

    Accepted: --

    Published Online: Jan. 3, 2020

    The Author Email: Lang Yue (langyue@tju.edu.cn)

    DOI:10.3788/LOP57.021009

    Topics