Acta Optica Sinica, Volume. 43, Issue 19, 1906005(2023)

Optimization of Mach-Zehnder Interferometer Event Recognition Scheme Based on Empirical Mode Decomposition

Ming Wang1, Hao Feng1, Zhou Sha1、*, and Li Zhao2
Author Affiliations
  • 1State Key Laboratory of Precision Measuring Technology and Instruments, School of Precision Instrument and Opto-Electronics Engineering, Tianjin University, Tianjin 300072, China
  • 2Shandong Longquan Pipeline Engineering Co., Ltd., Zibo 255200, Shandong , China
  • show less
    Figures & Tables(14)
    Measurement optical path of the Mach-Zehnder interferometer
    Signal waveforms and Fourier frequency spectra. (a)-(f) Event 1-event 6
    Proportion of IMF energy per layer. (a)-(f) Event 1-event 6
    Waveforms of IMF2-IMF5 of time-domain signals in Fig. 2. (a)-(f) Event 1-event 6
    Frequency spectra of IMF2-IMF5 of time-domain signals in Fig. 2. (a)-(f) Event 1-event 6
    Deep learning classifier structure, two sets of one-dimensional convolutions with the same parameters
    Training process of Model 1. (a) train_accuracy; (b) test_accuracy; (c) train_loss; (d) test_loss
    Validation precision of each classifier. (a)-(f) Event 1-event 6
    Validation recall of each classifier. (a)-(f) Event 1-event 6
    • Table 1. Specific components of each classifier

      View table

      Table 1. Specific components of each classifier

      Model typeContour attributeFrequency attributeTime-frequency attribute
      Model 1IMF2 and IMF3FFT(IMF2-IMF5STFT(IMF2 and IMF3
      Model 2IMF2 and IMF3FFTSTFT
      Model 3Original signalFFTSTFT
      Model 4Original signalHilbert marginal spectrumHilbert spectrum
      Model 5Original signalDWTCWT
    • Table 2. Training and evaluation results of each classifier

      View table

      Table 2. Training and evaluation results of each classifier

      Model typeTrain accuracy /%Test accuracy /%Floating degree /%Average response time /ms
      Model 198.7497.02-1.7467.29
      Model 297.9494.02-4.0062.14
      Model 397.9492.09-5.9725.16
      Model 498.6190.89-7.83118.24
      Model 597.2091.13-6.24281.59
    • Table 3. Evaluation indexes of each model on validation set

      View table

      Table 3. Evaluation indexes of each model on validation set

      ParameterModel 1Model 2Model 3Model 4Model 5
      test_accuracy /%97.0294.0292.0990.8991.13
      val_accuracy /%94.8888.3287.4787.9285.41
      val_precision /%95.5090.1786.5887.5186.67
      val_recall /%92.8386.1777.6787.0084.16
      val_F1_score /%94.1588.1281.8487.2585.40
    • Table 4. Degree of precision improvement of Model 1 compared with other classifiers

      View table

      Table 4. Degree of precision improvement of Model 1 compared with other classifiers

      Model typeEvent 1Event 2Event 3Event 4Event 5Event 6
      2-1.02-1.0113.0917.9510.470
      34.30058.339.526.741.05
      43.195.3811.7615.0620.252.13
      52.102.0823.386.9818.7511.63
    • Table 5. Degree of recall improvement of Model 1 compared with other classifiers

      View table

      Table 5. Degree of recall improvement of Model 1 compared with other classifiers

      Model typeEvent 1Event 2Event 3Event 4Event 5Event 6
      28.977.613.266.9818.673.19
      3-9.5757.146.7424.3248.3312.79
      4-9.5712.529.2213.5814.103.19
      5-9.5713.796.7433.3330.88-1.02
    Tools

    Get Citation

    Copy Citation Text

    Ming Wang, Hao Feng, Zhou Sha, Li Zhao. Optimization of Mach-Zehnder Interferometer Event Recognition Scheme Based on Empirical Mode Decomposition[J]. Acta Optica Sinica, 2023, 43(19): 1906005

    Download Citation

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

    Category: Fiber Optics and Optical Communications

    Received: Mar. 24, 2023

    Accepted: Apr. 23, 2023

    Published Online: Sep. 28, 2023

    The Author Email: Zhou Sha (shazhou@tju.edu.cn)

    DOI:10.3788/AOS230698

    Topics