Advanced Photonics, Volume. 7, Issue 2, 026008(2025)

Time-wavelength multiplexed photonic neural network accelerator for distributed acoustic sensing systems

Fuhao Yu1,2、†, Kangjian Di3, Wenjun Chen3, Sen Yan1,2, Yuanyuan Yao1,2, Silin Chen3, Xuping Zhang1,2, Yixin Zhang1,2、*, Ningmu Zou2,3,4、*, and Wei Jiang1,2、*
Author Affiliations
  • 1Nanjing University, College of Engineering and Applied Sciences, Nanjing, China
  • 2Nanjing University, Ministry of Education Key Laboratory of Intelligent Optical Sensing and Manipulation, Nanjing, China
  • 3Nanjing University, School of Integrated Circuits, Suzhou, China
  • 4Nanjing University, Interdisciplinary Research Center for Future Intelligent Chips (Chip-X), Suzhou, China
  • show less
    Figures & Tables(11)
    (a) Setup of DAS system; (b) raw signals collected by the system; (c) spatiotemporal maps of various events after denoising and quantization. NLL, narrow linewidth laser; OC, optical coupler; AOM, acousto-optic modulator; DAQ, digital acquisition card; Cir, circulator; EDFA, erbium-doped fiber amplifier; BPD, balanced photodetector; PG, pulse generator; BPF, bandpass filter; LNA, low-noise amplifier; DM, demodulation module.
    Photonic neural network accelerator.
    Experimental setup of the TWM-PNNA. AWG, arbitrary waveform generator; MZM, Mach–Zehnder modulator; WSS, wavelength-selective switch; OSW, optical switch; WDM, wavelength-division multiplexer; PD, photodetector.
    Results of convolution operations in the TWM-PNNA. (a1)–(a10) 1D feature maps after optical convolution; (b1)–(b10) 10 different convolution kernels; (c1)–(c10) restored 2D feature maps.
    Classification accuracy under different image sizes. (a) Electrical neural networks and optical neural networks in binary images; (b) electrical neural networks in binary images and nonbinary images.
    Comparison of chirp effects. (a) Without chirp. (b) With chirp.
    Confusion matrix for the DAS classification task with image size of 36×36. (a) Electrical CNN confusion matrix; (b)–(e) photonic convolution kernel confusion matrices; (b) push–pull modulation with wavelength spacing of 2 nm; (c) single-drive modulation with wavelength spacing of 2 nm; (d) push–pull modulation with wavelength spacing of 0.4 nm; (e) single-drive modulation with wavelength spacing of 0.4 nm.
    Relationship between wavelength spacing and classification accuracy under different modulation schemes.
    Pruning scheme in the TWM-PNNA.
    Pruning results in TWM-PNNA.
    • Table 1. Database construction for vibration events.

      View table
      View in Article

      Table 1. Database construction for vibration events.

      Event typeTrainTestTotalLabel
      Event 1 (background)82620610320
      Event 2 (climbing)6761708461
      Event 3 (vehicle passing)7121788902
      Event 4 (knocking)7571899463
      Event NABA + BN1
    Tools

    Get Citation

    Copy Citation Text

    Fuhao Yu, Kangjian Di, Wenjun Chen, Sen Yan, Yuanyuan Yao, Silin Chen, Xuping Zhang, Yixin Zhang, Ningmu Zou, Wei Jiang, "Time-wavelength multiplexed photonic neural network accelerator for distributed acoustic sensing systems," Adv. Photon. 7, 026008 (2025)

    Download Citation

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

    Category: Research Articles

    Received: Nov. 4, 2024

    Accepted: Feb. 13, 2025

    Posted: Feb. 13, 2025

    Published Online: Apr. 1, 2025

    The Author Email: Yixin Zhang (zyixin@nju.edu.cn), Ningmu Zou (nzou@nju.edu.cn), Wei Jiang (weijiang@nju.edu.cn)

    DOI:10.1117/1.AP.7.2.026008

    CSTR:32187.14.1.AP.7.2.026008

    Topics