Acta Optica Sinica, Volume. 44, Issue 1, 0106009(2024)
Signal Processing in Smart Fiber-Optic Distributed Acoustic Sensor
Figures & Tables(25)
Fig. 1. Foundation of the next-generation fiber optic Internet of Things—fiber-optic distributed acoustic sensor based on optical communication cable sensing
Fig. 3. Smart fiber-optic DAS and its signal processing architecture in smart city monitoring applications
Fig. 4. One-dimensional convolution neural network based supervised recognition model[58]
Fig. 6. Recognition model combining multi-scale depth features with temporal relationships[67]
Fig. 7. Temporal sequential relationship among multi-scale deep features based on HMM[67]
Fig. 9. 2D spectrograms of different types of signals via STFT[59]. (a) Digging; (b) walking; (c) vehicle-passing; (d) damaging
Fig. 10. 2D images of different types of signals via GAF transform[71]. (a) Window partition signal; (b) schematic diagram of Gramian angular difference field (GADF) coding; (c) schematic diagram of Gramian angular summation field (GASF) coding
Fig. 11. Supervised recognition model based on attention mechanism and ResNet-CBAM[72]
Fig. 12. Time-space waterfall figures of vibration signals due to various third-party interference[73]. (a) Excavator; (b) electrical hammer; (c) shovel; (d) hammer; (e) pickaxe; (f) metro
Fig. 13. Recognition result fusion correction based on time-space label matrix[73]
Fig. 14. Supervised recognition model combining convolutional neural network and bi-directional long short term memory[75]
Fig. 15. Supervised recognition model based on dual path network[80]. (a) Dual path architecture; (b) equivalent block of RP; (c) equivalent block of DCP
Fig. 16. Supervised recognition model based on attention-based long short-term memory network[81]
Fig. 17. DAS recognition model based on fusion of manual features and deep features[83]
Fig. 23. Multi- radial-distance event classification method based on deep learning[100]
Table 1. Confusion matrix of binary classification
View tableTable 1. Confusion matrix of binary classification
Item Positive
（true label，abnormal）
Negative（true label，
normal）
Positive
（predicted label）
TP FP Negative
（predicted label）
FN TN
Table 2. Comparison of key DAS signal recognition algorithms and their performance
View tableTable 2. Comparison of key DAS signal recognition algorithms and their performance
Institution Information
extraction
Model/
method
A /% P /% R /% F1 FAR /
%
MAR /
%
IDT Application
scenario
Ref. No University of Electronic Science
and Technology of China
Time 1-D CNN 98.19 97.95 97.16 0.9753 — — 27.0 ms Pipeline ［ 58 ］Huazhong University of
Science and Technology
Time 1-D CNN +
DenseNet
98.40 — — — — — 2.00 ms Cable ［ 62 ］Beijing Jiaotong University Time DBN-GRU 96.72 — — — — 1.83 79.0 ms Cable ［ 63 ］Jinan University Time DRSN-NTF 92.82 — — 0.9167 — — — Perimeter
security
［ 64 ］Anhui University Time
（multi-scale）
MS 1-D CNN 96.59 — — — — — — Perimeter
security
［ 65 ］Tianjin University Time
（multi-sacle）
MSCNN+CPL 84.67 — — — — — 17.0 ms Perimeter
security
［ 66 ］University of Electronic Science
and Technology of China
Time（multi-scale，long-short-term） mCNN + HMM 98.10 98.08 98.08 0.9805 — — 67.0 ms Cable ［ 67 ］Tianjin University Time（multi-scale，long-short-time） LSTM+CNN 94.60 — — — — — — Perimeter
security
［ 68 ］University of Shanghai for Science and Technology Time SSGAN 88.94 — — — — — — Perimeter
security
［ 84 ］Beijing Jiaotong University Time Semi-
supervised
FixMatch
97.91 97.93 97.96 — — 2.04 - Railway ［ 86 ］UGES of Türkiye T-F 2D CNN 93.00 98.10 — — — — — Cable ［ 70 ］Beijing Institute of Technology T-F 2D CNN 97.18 98.02 97.99 0.9798 — — — Cable ［ 69 ］Hubei University of Technology T-F 2D CNN 97.22 93.66 91.90 0.9267 — 8.10 — Perimeter
security
［ 71 ］Zhejiang University T-F 2D CNN + SVM 93.30 — — — — — — Cable ［ 59 ］University of Electronic Science
and Technology of China
T-F ResNet+
CBAM
98.89 98.58 98.68 0.9863 — — 3.30 ms Cable ［ 72 ］University of Cologne T-F ALSTM 94.30 — — — — — 0.910 s Cable ［ 81 ］University of Electronic Science
and Technology of China
T-F Unsupervised
SNN
96.52 — — — — — 0.364 s Cable ［ 87 ］Russian Academy of Sciencesul T-S 2D CNN 91.20 92.06 — 0.9138 — — — Perimeter
security
［ 60 ］University of Applied Sciences，Austria T-S 2D CNN 99.91 — — — — — 34.3 μs Pipeline ［ 61 ］Tongji University T-S 2D CNN 98.00 — — — — — — Pipeline ［ 73 ］Tianjin University T-S 2D CNN+YOLO — 70.40 82.90 — — 17.1 — Pipeline ［ 74 ］University of Electronic Science
and Technology of China
T-S 1D CNNs-
BiLSTM
97.00 97.06 96.90 0.9706 — 3.10 49.0 ms Cable ［ 75 ］North China Electric Power University T-S MATCN 98.50 — — — — — 0.530 s Perimeter
security
［ 79 ］Sichuan University T-S 2D CNN + LSTM 85.60 — — 0.8870 8 — 1.24 s Railway ［ 82 ］Qilu University of Technology T-S 100G-Net 99.60 — — — — — 20.0 ms Perimeter
security
［ 76 ］Southern University of
Science and Technology
T-S Faster RCNN 96.32 — — — — — 0.160 s Cable ［ 77 ］T-S YOLO 96.14 — — — — — 43.8 ms Perimeter
security
［ 78 ］Shantou University T-S Transfer learning 96.16 — — — — — — Cable ［ 88 ］94.67 — — — — — 3.05 ms Cable ［ 89 ］Tsinghua University T-S SSAE 97.90 — 97.38 — — 2.62 1.73 ms Pipeline ［ 85 ］Shanghai Institute of Optics and Fine Mechanics，Chinese Academy of Sciences S-F DPN 99.28 99.28 99.28 0.9700 — 0.72 — Railway ［ 80 ］
Tools
Get Citation
Copy Citation Text
Huijuan Wu, Xinlei Wang, Haibei Liao, Xiben Jiao, Yiyu Liu, Xinjian Shu, Jinglun Wang, Yunjiang Rao. Signal Processing in Smart Fiber-Optic Distributed Acoustic Sensor[J]. Acta Optica Sinica, 2024, 44(1): 0106009
Paper Information
Category: Fiber Optics and Optical Communications
Received: Aug. 10, 2023
Accepted: Oct. 9, 2023
Published Online: Jan. 12, 2024
The Author Email: Wu Huijuan (hjwu@uestc.edu.cn), Rao Yunjiang (yjrao@uestc.edu.cn)
CSTR:32393.14.AOS231384
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20