• Laser & Optoelectronics Progress
  • Vol. 61, Issue 5, 0506002 (2024)
Huikang Liang, Haoshen Xie, Hongbin Huang, and Weiping Liu*
Author Affiliations
  • Department of Electronic Engineering, College of Information Science and Technology, Jinan University, Guangzhou 510623, Guangdong , China
  • show less
    DOI: 10.3788/LOP230884 Cite this Article Set citation alerts
    Huikang Liang, Haoshen Xie, Hongbin Huang, Weiping Liu. Distributed Optical Fiber Acoustic Sensing Signal Recognition Based on Improved Depth Residual Shrinkage Network[J]. Laser & Optoelectronics Progress, 2024, 61(5): 0506002 Copy Citation Text show less
    Comparison of three threshold functions
    Fig. 1. Comparison of three threshold functions
    Structure of RSBU-NTF based on new threshold function
    Fig. 2. Structure of RSBU-NTF based on new threshold function
    Block diagram of DAS perimeter security system
    Fig. 3. Block diagram of DAS perimeter security system
    Time domain signal waveforms of different vibration events. (a) Clap; (b) walk; (c) rolling; (d) noise; (e) climbing; (f) watering
    Fig. 4. Time domain signal waveforms of different vibration events. (a) Clap; (b) walk; (c) rolling; (d) noise; (e) climbing; (f) watering
    When RSNR=1 dB, accuracy and loss iteration curves of the two models in the training process. (a) Accuracy; (b) loss iteration curves
    Fig. 5. When RSNR=1 dB, accuracy and loss iteration curves of the two models in the training process. (a) Accuracy; (b) loss iteration curves
    CNNResNetDRSNDRSN-NTFOutput size
    InputInputInputInput1×8000×1
    Conv(4,3,/2)Conv(4,3,/2)Conv(4,3,/2)Conv(4,3,/2)4×4000×1
    CBU(4,3,/2)RBU(4,3,/2)RSBU(4,3,/2)RSBU-NTF(4,3,/2)4×2000×1
    CBU(4,3)RBU(4,3)RSBU(4,3)RSBU-NTF(4,3)4×2000×1
    CBU(8,3,/2)RBU(8,3,/2)RSBU(8,3,/2)RSBU-NTF(8,3,/2)8×1000×1
    CBU(8,3)RBU(8,3)RSBU(8,3)RSBU-NTF(8,3)8×1000×1
    CBU(16,3,/2)RBU(16,3,/2)RSBU(16,3,/2)RSBU-NTF(16,3,/2)16×500×1
    CBU(16,3)RBU(16,3)RSBU(16,3)RSBU-NTF(16,3)16×500×1
    BN,ReLU,GAPBN,ReLU,GAPBN,ReLU,GAPBN,ReLU,GAP16×1×1
    FCFCFCFC6×1×1
    Table 1. Structure-related superparameters of CNN, ResNet, DRSN and DRSN-NTF
    NetworkAverage training accuracy /%Average test accuracy /%Average training time /s
    DRSN-NTF97.9992.82238
    DRSN97.3891.77220
    ResNet95.9189.30154
    CNN91.4986.16150
    Table 2. Average recognition accuracy of CNN, ResNet, DRSN and DRSN-NTF
    Signal-to-noise ratio /dBTest accuracy of two networks /%Difference between DRSN and DRSN-NTF /percentage point
    DRSNDRSN-NTF
    594.7395.350.62≤0.90
    493.5594.450.90
    393.3394.080.75
    291.4792.801.33≥1.27
    190.1891.631.45
    087.3588.621.27
    Average value91.7792.821.05
    Table 3. Comparison results of test accuracy of DRSN and DRSN-NTF
    LabelVibration eventAverage F1-score value
    DRSNDRSN-NTF
    1Clap0.980.98
    2Walk0.940.95
    3Rolling0.870.89
    4Noise0.990.99
    5Climbing0.790.83
    6Watering0.840.86
    Table 4. When RSNR=1 dB, the comparison results of the average F1-score values of DRSN and DRSN-NTF
    Huikang Liang, Haoshen Xie, Hongbin Huang, Weiping Liu. Distributed Optical Fiber Acoustic Sensing Signal Recognition Based on Improved Depth Residual Shrinkage Network[J]. Laser & Optoelectronics Progress, 2024, 61(5): 0506002
    Download Citation