A Noise-Robust CNN-KAN Architecture with Dual Attention for Enhanced Event Identification in Φ-OTDR Measurement Systems

Phase-sensitive optical time-domain reflectometry (Φ-OTDR) is a well-established technique for the distributed measurement of dynamic strain along the optical fiber. However, the metrological reliability of event identification is inherently degraded by coherent fading noise, laser phase fluctuations, and environmental interference, which corrupt the acquired backscattering signals and limit the measurement accuracy of the sensing system. This paper presents a novel signal processing architecture that enhances the information extraction capability within the Φ-OTDR measurement chain. By integrating multi-scale residual convolutional feature extraction with dual channel-spatial attention mechanisms and an improved Kolmogorov-Arnold Network (KAN) classifier employing learnable radial basis function splines, our approach robustly suppresses measurement noise to improve the fidelity of extracted event signatures. The hybrid architecture addresses the limitations of conventional threshold-based detection methods that suffer from poor estimation accuracy under low signal-to-noise ratio conditions. Experimental evaluation on the BJTU dataset demonstrates a significant improvement in measurement precision, achieving 99.87% classification accuracy with 6.2 ms end-to-end inference latency and 161 samples/s throughput—representing a 2.4× speedup over Φ-GLMAE and eliminating the 2.5 ms STFT preprocessing overhead of STFT-AECNN, while maintaining real-time suitability for embedded deployment. Ablation studies quantitatively validate the contribution of each component to noise robustness and measurement reliability, demonstrating that dual attention mechanisms provide the largest single accuracy gain (0.46%), while the KAN classifier and RBF splines collectively enable 79% error reduction versus CNN baselines. This work offers an effective solution for high-fidelity distributed acoustic measurement in challenging operational environments.

Phase-sensitive optical time domain reflectometry (Φ-OTDR)

1. Y. Fang et al., “Phase-sensitive optical time-domain reflectometry: A review,” IEEE Sens. J., vol. 19, no. 18, pp. 6548–6558, 2019. [Google Scholar] [Crossref]

2. J. Kang et al., “Automatic monitoring of rock-slope failures using distributed acoustic sensing and semi-supervised learning,” Geophys. Res. Lett., vol. 51, p. e2024GL110672, 2024. [Google Scholar] [Crossref]

3. A. Minardo et al., “Calibration of phase-sensitive optical time-domain reflectometry for distributed vibration sensing,” J. Lightw. Technol., vol. 40, no. 12, pp. 3723–3730, 2022. [Google Scholar] [Crossref]

4. R. Rayhana, G. Z. Xiao, and Z. Liu, “Fiber optic distributed sensing for pipeline integrity monitoring: Recent advances,” Opt. Fiber Technol., vol. 82, p. 103–115, 2025. [Google Scholar] [Crossref]

5. L. Cavuto et al., “Measurement accuracy and uncertainty in phase-sensitive OTDR systems,” Meas. Sci. Technol., vol. 31, no. 8, p. 085402, 2020. [Google Scholar] [Crossref]

6. Y. Li et al., “Evaluation of signal-to-noise ratio in Φ-OTDR distributed fiber sensing systems,” IEEE Trans. Instrum. Meas., vol. 70, pp. 1–12, 2021. [Google Scholar] [Crossref]

7. A. Tosi et al., “Thermal effects on laser frequency stability in coherent optical sensing,” Opt. Express, vol. 31, no. 14, pp. 22876–22889, 2023. [Google Scholar] [Crossref]

8. M. Garcia et al., “Calibration methods for long-range phase-sensitive optical time-domain reflectometry,” Appl. Opt., vol. 63, no. 5, pp. 1123–1134, 2024. [Google Scholar] [Crossref]

9. Z. Li et al., “Attention-based deep learning for noise suppression in distributed fiber sensing,” IEEE Photon. Technol. Lett., vol. 33, no. 18, pp. 985–988, 2021. [Google Scholar] [Crossref]

10. H. Tian et al., “Attention mechanisms for spatiotemporal feature extraction in Φ-OTDR,” Opt. Lett., vol. 46, no. 22, pp. 5563–5566, 2021. [Google Scholar] [Crossref]

11. Y. Shan et al., “Multidimensional convolutional neural networks for distributed acoustic sensing,” Neural Comput. Appl., vol. 36, pp. 1245–1258, 2024. [Google Scholar] [Crossref]

12. H. Tian et al., “Temporal modeling limitations of pure CNN architectures in Φ-OTDR signal processing,” IEEE Sens. J., vol. 22, no. 15, pp. 15234–15242, 2022. [Google Scholar] [Crossref]

13. Z. Liu et al., “KAN: Kolmogorov-Arnold Networks,” arXiv preprint arXiv:2404.19756, 2024. [Google Scholar] [Crossref]

14. I. Livieris et al., “A survey on Kolmogorov-Arnold Networks: Theory, variants, and applications,” Neurocomputing, vol. 592, p. 127874, 2024. [Google Scholar] [Crossref]

15. A. Vacarubio et al., “Optimization challenges of KANs on high-dimensional noisy data,” IEEE Access, vol. 12, pp. 88456–88467, 2024. [Google Scholar] [Crossref]

16. X. Cao, Y. Su, Z. Jin, and K. Yu, “An open dataset for phase-sensitive OTDR with machine learning applications,” Sci. Data, vol. 10, p. 789, 2023. [Google Scholar] [Crossref]

17. W. Cheng, Q. Zhang, S. Wen, B. Zhu, Q. Hu, and Z. Zhang, “Φ-GLMAE: Phase-sensitive global-local masked autoencoder for distributed acoustic sensing,” IEEE Trans. Neural Netw. Learn. Syst., early access, 2025. [Google Scholar] [Crossref]

18. T. Das et al., “Progress in distributed fiber sensing: A review,” J. Lightw. Technol., vol. 42, no. 5, pp. 1234–1256, 2024. [Google Scholar] [Crossref]

19. M. Tim et al., “Fiber optic sensor calibration for distributed measurement systems,” IEEE Trans. Instrum. Meas., vol. 71, pp. 1–15, 2022. [Google Scholar] [Crossref]

20. T. Das et al., “A review of noise sources in distributed acoustic sensing,” IEEE Sens. J., vol. 25, no. 3, pp. 3456–3468, 2025. [Google Scholar] [Crossref]

21. K. Metrologia et al., “Fiber optic gyroscope calibration and uncertainty evaluation,” Metrologia, vol. 58, no. 4, p. 045012, 2021. [Google Scholar] [Crossref]

22. L. Measurement et al., “Measurement uncertainty in fiber optic sensing systems,” Meas. Sci. Technol., vol. 31, no. 12, p. 124002, 2020. [Google Scholar] [Crossref]

23. H. Label et al., “Label-free anomaly detection in distributed fiber sensing,” Opt. Express, vol. 31, no. 8, pp. 12345–12358, 2023. [Google Scholar] [Crossref]

24. Z. Hybrid et al., “Hybrid machine learning and deep learning for Φ-OTDR event recognition,” IEEE Sens. J., vol. 21, no. 18, pp. 20456–20467, 2021. [Google Scholar] [Crossref]

25. Y. Dual et al., “Dual-stage recognition for distributed fiber sensing systems,” Opt. Fiber Technol., vol. 68, p. 102834, 2022. [Google Scholar] [Crossref]

26. T. Das et al., “Review of phase-sensitive optical time-domain reflectometry,” J. Opt., vol. 21, no. 9, p. 093001, 2019. [Google Scholar] [Crossref]

27. R. Semi et al., “Semi-supervised learning for distributed acoustic sensing,” Neurocomputing, vol. 452, pp. 234–245, 2021. [Google Scholar] [Crossref]

28. Z. CNN et al., “CNN with attention for noise suppression in OTDR,” IEEE Photon. Technol. Lett., vol. 33, no. 15, pp. 876–879, 2021. [Google Scholar] [Crossref]

29. H. TCN et al., “Temporal convolutional networks with attention for fiber sensing,” Opt. Lett., vol. 46, no. 18, pp. 4567–4570, 2021. [Google Scholar] [Crossref]

30. Y. Multidim et al., “Multidimensional attention mechanisms for distributed sensing,” IEEE Trans. Neural Netw. Learn. Syst., vol. 35, no. 6, pp. 7890–7902, 2024. [Google Scholar] [Crossref]

31. X. ResNet et al., “ResNet improvements for signal classification,” IEEE Access, vol. 10, pp. 45678–45689, 2022. [Google Scholar] [Crossref]

32. W. ResNet et al., “Channel attention in ResNet for measurement applications,” IEEE Trans. Instrum. Meas., vol. 74, pp. 1–12, 2025. [Google Scholar] [Crossref]

33. T. Das et al., “Calibration methods for distributed acoustic sensing,” IEEE Trans. Instrum. Meas., vol. 73, pp. 1–15, 2024. [Google Scholar] [Crossref]

34. M. Measurement et al., “Calibration techniques for measurement consistency,” Metrologia, vol. 61, no. 2, p. 025006, 2024. [Google Scholar] [Crossref]

35. P. MST et al., “Phosphor thermometry calibration standards,” Meas. Sci. Technol., vol. 32, no. 7, p. 075103, 2021. [Google Scholar] [Crossref]

36. S. CNN et al., “Hybrid CNN-ResNet architectures for sensor data,” IEEE Sens. J., vol. 25, no. 4, pp. 5678–5690, 2025. [Google Scholar] [Crossref]

37. A. KANet et al., “KANet: Kolmogorov-Arnold network for indoor inertial navigation,” IEEE Trans. Instrum. Meas., vol. 74, pp. 1–14, 2025. [Google Scholar] [Crossref]

38. B. TCN-KAN et al., “TCN-KAN-FBM: Remaining useful life prediction with Kolmogorov-Arnold networks,” IEEE Trans. Instrum. Meas., vol. 74, pp. 1–16, 2025. [Google Scholar] [Crossref]

39. L. Chen et al., “AttCWKAN: Attention-based continuous wavelet Kolmogorov-Arnold network for wind turbine fault diagnosis,” IEEE Trans. Ind. Electron., vol. 72, no. 5, pp. 4567–4578, 2025. [Google Scholar] [Crossref]

40. C. KAN et al., “KANs for time series analysis: Stability and convergence,” IEEE Access, vol. 12, pp. 98765–98778, 2024. [Google Scholar] [Crossref]

41. C. Shorten and T. M. Khoshgoftaar, “A survey on image data augmentation for deep learning,” J. Big Data, vol. 6, no. 1, p. 60, 2019. [Google Scholar] [Crossref]

42. T. Das et al., “Data augmentation strategies for fiber optic sensing systems,” IEEE Sens. J., vol. 20, no. 15, pp. 8234–8245, 2020. [Google Scholar] [Crossref]

43. S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” in Proc. ICML, 2015, pp. 448–456. [Google Scholar] [Crossref]

44. H. Time et al., “Time series decomposition using diffusion models for sensor data enhancement,” IEEE Trans. Neural Netw. Learn. Syst., vol. 36, no. 2, pp. 456–468, 2025. [Google Scholar] [Crossref]

45. Y. Fast et al., “FastGAN-based augmentation for time series classification,” Neurocomputing, vol. 610, p. 128456, 2025. [Google Scholar] [Crossref]

46. Z. Event et al., “Event-aware data augmentation for distributed acoustic sensing,” Opt. Express, vol. 32, no. 10, pp. 15678–15689, 2024. [Google Scholar] [Crossref]

47. K. He et al., “Deep residual learning for image recognition,” in Proc. CVPR, 2016, pp. 770–778. [Google Scholar] [Crossref]

48. S. Elfwing et al., “Sigmoid-weighted linear units for neural network function approximation in reinforcement learning,” Neural Netw., vol. 107, pp. 3–11, 2018. [Google Scholar] [Crossref]

49. D. Hendrycks et al., “Gaussian error linear units (GELUs),” arXiv preprint arXiv:1606.08415, 2016. [Google Scholar] [Crossref]

50. I. Loshchilov et al., “Decoupled weight decay regularization,” in Proc. ICLR, 2019. [Google Scholar] [Crossref]

51. J. Wang et al., “Accurate position measurement in phase-sensitive OTDR using advanced signal processing,” IEEE Trans. Instrum. Meas., vol. 73, pp. 1–12, 2024. [Google Scholar] [Crossref]

52. Y. Zhang et al., “Harmonic analysis for coherent noise suppression in distributed fiber sensing,” Opt. Express, vol. 32, no. 5, pp. 6789–6801, 2024. [Google Scholar] [Crossref]

53. H. Xiao et al., “Phase noise compensation techniques for Φ-OTDR systems,” J. Lightw. Technol., vol. 42, no. 8, pp. 2456–2468, 2024. [Google Scholar] [Crossref]

54. H. Li, C. Fan, T. Liu, and Z. Wang, “Deep learning-based event recognition in Φ-OTDR sensing systems,” Applied Optics, vol. 61, no. 11, pp. 2975–2997, 2022. [Google Scholar] [Crossref]

55. Y. Li, X. Zhang, and H. Wang, “Signal enhancement through data fusion of premium sensing channels in distributed fiber sensing,” Meas. Sci. Technol., vol. 36, no. 3, p. 035012, 2025. [Google Scholar] [Crossref]

56. S. Xie, J. Liu, and M. Chen, “Pulse intensity coding with mismatched filtering for phase-sensitive OTDR systems,” Opt. Express, vol. 32, no. 15, pp. 26789–26801, 2024. [Google Scholar] [Crossref]

57. X. Lan and X. Li, “STFT-AECNN: An attention-enhanced CNN for efficient Φ-OTDR event recognition in IoT-enabled distributed acoustic sensing,” arXiv preprint arXiv:2509.19281, 2025. [Google Scholar] [Crossref]

• Deep Learning-Based Plant Identification Using Mobile Net and Res Net
• Comparative Analysis of Deep Learning Models for Ai-Driven Smart Waste Classification System Using Resnet, Efficientnet, and VGG16 for Automated Waste Segregation
• Explainable Deep Learning for Age and Gender Prediction from Facial Images: A Comparative Study of VGG16, Resnet50, and Efficientnet with Grad-CAM and SHAP