E-mail Alert
RSS
Article navigation >
Opto-Electronic Advances
>
Early View
| Citation: |
Zhang SX, Li H, Fan CZ et al. Adaptive decentralized AI scheme for signal recognition of distributed sensor systems. Opto-Electron Adv 7, 240119 (2024). doi: 10.29026/oea.2024.240119
|
Article Open Access
Adaptive decentralized AI scheme for signal recognition of distributed sensor systems
-
Abstract
Artificial intelligence (AI) plays a critical role in signal recognition of distributed sensor systems (DSS), boosting its applications in multiple monitoring fields. Due to the domain differences between massive sensors in signal acquisition conditions, such as manufacturing process, deployment, and environments, current AI schemes for signal recognition of DSS frequently encounter poor generalization performance. In this paper, an adaptive decentralized artificial intelligence (ADAI) method for signal recognition of DSS is proposed, to improve the entire generalization performance. By fine-tuning pre-trained model with the unlabeled data in each domain, the ADAI scheme can train a series of adaptive AI models for all target domains, significantly reducing the false alarm rate (FAR) and missing alarm rate (MAR) induced by domain differences. The field tests about intrusion signal recognition with distributed optical fiber sensors system demonstrate the efficacy of the ADAI scheme, showcasing a FAR of merely 4.3% and 0%, along with a MAR of only 1.4% and 2.7% within two specific target domains. The ADAI scheme is expected to offer a practical paradigm for signal recognition of DSS in multiple application fields. -
-
References
[1] Abdollahzadeh S, Navimipour NJ. Deployment strategies in the wireless sensor network: a comprehensive review. Comput Commun 91 – 92 , 1–16 (2016). [2] Johnson KJ, Rose-Pehrsson SL. Sensor array design for complex sensing tasks. Annu Rev Anal Chem 8, 287–310 (2015). doi: 10.1146/annurev-anchem-062011-143205 [3] Li J, Bao RR, Tao J et al. Recent progress in flexible pressure sensor arrays: from design to applications. J Mater Chem C 6, 11878–11892 (2018). doi: 10.1039/C8TC02946F [4] Wang HT, Hao CL, Lin H et al. Generation of super-resolved optical needle and multifocal array using graphene oxide metalenses. Opto-Electron Adv 4, 200031 (2021). doi: 10.29026/oea.2021.200031 [5] Hürlimann M, Coviello V, Bel C et al. Debris-flow monitoring and warning: review and examples. Earth-Sci Rev 199, 102981 (2019). doi: 10.1016/j.earscirev.2019.102981 [6] Kandris D, Nakas C, Vomvas D et al. Applications of wireless sensor networks: an up-to-date survey. Appl Syst Innov 3, 14 (2020). doi: 10.3390/asi3010014 [7] Zhu WQ, Biondi E, Li JX et al. Seismic arrival-time picking on distributed acoustic sensing data using semi-supervised learning. Nat Commun 14, 8192 (2023). doi: 10.1038/s41467-023-43355-3 [8] Yang HY, Wang YC, Peng HY et al. Breath biopsy of breast cancer using sensor array signals and machine learning analysis. Sci Rep 11, 103 (2021). doi: 10.1038/s41598-020-80570-0 [9] Leal-Junior A, Avellar L, Biazi V et al. Multifunctional flexible optical waveguide sensor: on the bioinspiration for ultrasensitive sensors development. Opto-Electron Adv 5, 210098 (2022). doi: 10.29026/oea.2022.210098 [10] Hou LQ, Bergmann NW. Novel industrial wireless sensor networks for machine condition monitoring and fault diagnosis. IEEE Trans Instrum Meas 61, 2787–2798 (2012). doi: 10.1109/TIM.2012.2200817 [11] Krantz-Rülcker C, Stenberg M, Winquist F et al. Electronic tongues for environmental monitoring based on sensor arrays and pattern recognition: a review. Anal Chim Acta 426, 217–226 (2001). doi: 10.1016/S0003-2670(00)00873-4 [12] Zheng XT, Yang ZJ, Sutarlie L et al. Battery-free and AI-enabled multiplexed sensor patches for wound monitoring. Sci Adv 9, eadg6670 (2023). doi: 10.1126/sciadv.adg6670 [13] Alsheikh MA, Lin SW, Niyato D et al. Machine learning in wireless sensor networks: algorithms, strategies, and applications. IEEE Commun Surv Tut 16, 1996–2018 (2014). doi: 10.1109/COMST.2014.2320099 [14] Lee H, Lee S, Kim J et al. Stretchable array electromyography sensor with graph neural network for static and dynamic gestures recognition system. npj Flex Electron 7, 20 (2023). doi: 10.1038/s41528-023-00246-3 [15] Ding Y, Elsayed EA, Kumara S et al. Distributed sensing for quality and productivity improvements. IEEE Trans Autom Sci Eng 3, 344–359 (2006). doi: 10.1109/TASE.2006.876610 [16] Duan LX, Xu D, Tsang IWH. Domain adaptation from multiple sources: a domain-dependent regularization approach. IEEE Trans Neural Netw Learn Syst 23, 504–518 (2012). doi: 10.1109/TNNLS.2011.2178556 [17] Quqa S, Li SJ, Shu YN et al. Crack identification using smart paint and machine learning. Struct Health Monit 23, 248–264 (2024). doi: 10.1177/14759217231167823 [18] Kim KK, Kim M, Pyun K et al. A substrate-less nanomesh receptor with meta-learning for rapid hand task recognition. Nat Electron 6, 64–75 (2023). [19] Liu SQ, Yu FH, Hong R et al. Advances in phase-sensitive optical time-domain reflectometry. Opto-Electron Adv 5, 200078 (2022). doi: 10.29026/oea.2022.200078 [20] Lyu CG, Hu XY, Niu ZH et al. A light-weight neural network for marine acoustic signal recognition suitable for fiber-optic hydrophones. Exp Syst Appl 235, 121235 (2024). doi: 10.1016/j.eswa.2023.121235 [21] Xin LP, Li ZY, Gui X et al. Surface intrusion event identification for subway tunnels using ultra-weak FBG array based fiber sensing. Opt Express 28, 6794–6805 (2020). doi: 10.1364/OE.387317 [22] Tejedor J, Macias-Guarasa J, Martins HF et al. A contextual GMM-HMM smart fiber optic surveillance system for pipeline integrity threat detection. J Lightw Technol 37, 4514–4522 (2019). doi: 10.1109/JLT.2019.2908816 [23] Wu HJ, Liu XR, Xiao Y et al. A dynamic time sequence recognition and knowledge mining method based on the hidden markov models (HMMs) for pipeline safety monitoring with Φ-OTDR. J Lightw Technol 37, 4991–5000 (2019). doi: 10.1109/JLT.2019.2926745 [24] Zhang SX, He T, Li H et al. Modified data augmentation integration method for robust intrusion events recognition with fiber optic DAS system. J Lightw Technol 42, 453–462 (2024). doi: 10.1109/JLT.2023.3301557 [25] Shiloh L, Eyal A, Giryes R. Efficient processing of distributed acoustic sensing data using a deep learning approach. J Lightw Technol 37, 4755–4762 (2019). doi: 10.1109/JLT.2019.2919713 [26] Shi Y, Dai SW, Liu XY et al. Event recognition method based on dual-augmentation for a Φ-OTDR system with a few training samples. Opt Express 30, 31232–31243 (2022). doi: 10.1364/OE.468779 [27] Huang XD, Wang BY, Liu K et al. An event recognition scheme aiming to improve both accuracy and efficiency in optical fiber perimeter security system. J Lightw Technol 38, 5783–5790 (2020). doi: 10.1109/JLT.2020.3003396 [28] Lyu CG, Huo ZQ, Liu YG et al. Robust intrusion events recognition methodology for distributed optical fiber sensing perimeter security system. IEEE Trans Instrum Meas 70, 9505109 (2021). [29] Yang F, Ling ZN, Zhang YH et al. Event detection, localization, and classification based on semi-supervised learning in power grids. IEEE Trans Power Syst 38, 4080–4094 (2023). doi: 10.1109/TPWRS.2022.3209343 [30] Li YJ, Cao XM, Ni WH et al. A deep learning model enabled multi-event recognition for distributed optical fiber sensing. Sci China Inf Sci 67, 132404 (2024). doi: 10.1007/s11432-023-3896-4 [31] Wu HJ, Gan DK, Xu CR et al. Improved generalization in signal identification with unsupervised spiking neuron networks for fiber-optic distributed acoustic sensor. J Lightw Technol 40, 3072–3083 (2022). doi: 10.1109/JLT.2022.3144147 [32] Wang SL, Liu F, Liu B. Semi-supervised deep learning in high-speed railway track detection based on distributed fiber acoustic sensing. Sensors 22, 413 (2022). doi: 10.3390/s22020413 [33] Yang YY, Zhang HF, Li Y. Long-distance pipeline safety early warning: a distributed optical fiber sensing semi-supervised learning method. IEEE Sens J 21, 19453–19461 (2021). doi: 10.1109/JSEN.2021.3087537 [34] Shi Y, Li YH, Zhang YC et al. An easy access method for event recognition of Φ-OTDR sensing system based on transfer learning. J Lightw Technol 39, 4548–4555 (2021). doi: 10.1109/JLT.2021.3070583 [35] Lyu CG, Huo ZQ, Cheng X et al. Distributed optical fiber sensing intrusion pattern recognition based on GAF and CNN. J Lightw Technol 38, 4174–4182 (2020). doi: 10.1109/JLT.2020.2985746 [36] Ben-David S, Blitzer J, Crammer K et al. A theory of learning from different domains. Mach Learn 79, 151–175 (2010). doi: 10.1007/s10994-009-5152-4 [37] Borgwardt KM, Gretton A, Rasch MJ et al. Integrating structured biological data by kernel maximum mean discrepancy. Bioinformatics 22, e49–e57 (2006). doi: 10.1093/bioinformatics/btl242 [38] Tzeng E, Hoffman J, Zhang N et al. Deep domain confusion: maximizing for domain invariance. arXiv: 1412.3474, 2014. https://arxiv.org/abs/1412.3474 [39] Scudder H. Probability of error of some adaptive pattern-recognition machines. IEEE Trans Inf Theory 11, 363–371 (1965). doi: 10.1109/TIT.1965.1053799 [40] McLachlan GJ. Iterative reclassification procedure for constructing an asymptotically optimal rule of allocation in discriminant analysis. J Am Stat Assoc 70, 365–369 (1975). doi: 10.1080/01621459.1975.10479874 [41] Tarvainen A, Valpola H. Mean teachers are better role models: weight-averaged consistency targets improve semi-supervised deep learning results. In Proceedings of the 31st International Conference on Neural Information Processing Systems 1195–1204 (Curran Associates Inc. , 2017). https://dl.acm.org/doi/10.5555/3294771.3294885 [42] Lu YL, Zhu T, Chen L et al. Distributed vibration sensor based on coherent detection of phase-OTDR. J Lightw Technol 28, 3243–3249 (2010). [43] Sohn K, Berthelot D, Li CL et al. FixMatch: simplifying semi-supervised learning with consistency and confidence. In Proceedings of the 34th International Conference on Neural Information Processing Systems 51 (Curran Associates Inc. , 2020). https://dl.acm.org/doi/10.5555/3495724.3495775 [44] Masci J, Meier U, Cireşan D et al. Stacked convolutional auto-encoders for hierarchical feature extraction. In Proceedings of the 21st International Conference on Artificial Neural Networks and Machine Learning–ICANN 2011 52–59 (Springer, 2011); https://doi.org/10.1007/978-3-642-21735-7_7. -
Supplementary Information
Supplementary information for Adaptive decentralized AI scheme for signal recognition of distributed sensor systems
-
Access History
Figures(12)
Tables(2)
Article Metrics
Export File
Citation
Zhang SX, Li H, Fan CZ et al. Adaptive decentralized AI scheme for signal recognition of distributed sensor systems. Opto-Electron Adv 7, 240119 (2024). doi: 10.29026/oea.2024.240119
Format
Content
DownLoad:
-
Figure 1.
(a) The characteristics of unlabeled data. (b) The model transferring tasks by ADAI scheme. (c) The process of ADAI scheme.
-
Figure 2.
The detailed process of ADAI scheme. (a) The model training by DA method. (b) The model selecting by EPL method.
-
Figure 3.
A typical application of ADAI scheme about intrusion signal recognition with DOFS system.
-
Figure 4.
(a) The experimental scenes of three domains. (b) The demodulated waveforms of four typical events in three domains.
-
Figure 5.
The architecture of AI model.
-
Figure 6.
(a) The recognition accuracies of Model-(DA)-T1' and Model-(Fixmatch)-T1' on Dataset-S and Unlabeled dataset-T1' at different training epochs. The recognition accuracies of (b) Model-(DA)-T1', and (c) Model-(Fixmatch)-T1' on Dataset-S and Unlabeled dataset-T1' at the 1~10 training epochs.
-
Figure 7.
(a) Accuracies of pseudo labels created by EPL method in three training times. (b) Performance of EPL method for model selecting.
-
Figure 8.
(a) The accuracies of Model-(Baseline), Model-(ADAI)-T1', and Model-(ADAI)-T2' on Dataset- T1 and Dataset-T2. The MAR and FAR of (b) Model-(Baseline) and Model-(ADAI)-T1' on Dataset-T1, and (c) Model-(Baseline) and Model-(ADAI)-T2' on Dataset-T2, respectively.
-
Figure 9.
The confusion matrices of (a) the Model-(Baseline) on Dataset-T1, (b) the Model-(Baseline) on Dataset-T2, (c) the Model-(ADAI)-T1' on Dataset-T1, and (d) the Model-(ADAI)-T2' on Dataset-T2.
-
Figure 10.
The accuracies of (a) Model-(ADAI)-T1' on Dataset-T1, and (b) Model-(ADAI)-T2' on Dataset-T2 under the conditions of different time-shift points and different ratio noise.
-
Figure 11.
The comparison of ROC curves between ADAI scheme and Baseline on (a) Dataset-T1, and (b) Dataset-T2.
-
Figure 12.
(a) The performance comparison of Model-(ADAI)-T1'-T1, Model-(Baseline)-T1, and Model-(AE)-T1'-T1 when providing different number of labeled samples from Domain-T1. The T-SNE visualizations of the (b) Model-(ADAI)-T1'-T1, (c) Model-(Baseline)-T1, and (d) Model-(AE)-T1'-T1 when providing four labeled samples for each class.
