• Infrared and Laser Engineering
  • Vol. 51, Issue 5, 20210419 (2022)
Hanlin Liu1、2, Jingtao Xin1、2, Wei Zhuang1、2、*, Jiabin Xia3, and Lianqing Zhu1、2
Author Affiliations
  • 1Key Laboratory of the Ministry of Education for Optoelectronic Measurement Technology and Instrument, Beijing University of Information Technology, Beijing 100192, China
  • 2Beijing Laboratory of Optical Fiber Sensing and Systems, Beijing Information Science & Technology University, Beijing 100016, China
  • 3School of Instrument Science and Opto-electronics Engineering, Hefei University of Technology, Hefei 230009, China
  • show less
    DOI: 10.3788/IRLA20210419 Cite this Article
    Hanlin Liu, Jingtao Xin, Wei Zhuang, Jiabin Xia, Lianqing Zhu. Demodulation method of overlapping spectrum based on convolutional neural network[J]. Infrared and Laser Engineering, 2022, 51(5): 20210419 Copy Citation Text show less
    Schematic diagram of the aliasing spectrum data acquisition system
    Fig. 1. Schematic diagram of the aliasing spectrum data acquisition system
    Program block diagram of the upper computer of the aliasing spectrum data acquisition
    Fig. 2. Program block diagram of the upper computer of the aliasing spectrum data acquisition
    Overlap process of the spectrum in the display interface of the host computer
    Fig. 3. Overlap process of the spectrum in the display interface of the host computer
    Schematic diagram of 1D convolutional neural network structure
    Fig. 4. Schematic diagram of 1D convolutional neural network structure
    CNN training time and test time under different training samples
    Fig. 5. CNN training time and test time under different training samples
    RMS and MAE error under different epoch times
    Fig. 6. RMS and MAE error under different epoch times
    Spectral demodulation results under different degrees of aliasing
    Fig. 7. Spectral demodulation results under different degrees of aliasing
    AlgorithmRMS/pmComputational time/sData source
    ELM0.9180.200Ref.[8]
    LS-SVR3.9550.0736Ref.[7]
    PSO-SA<5.000--Ref.[10]
    LSTM0.6740.574Ref.[8]
    CNN0.08260.338This work
    Table 1. Comparison between different algorithms
    CaseFBG1/nmFBG2/nmΔλ/nm RMS/pm
    a1536.84881536.15140.69740.0322
    b1537.15591536.15211.00380.0319
    c1535.75271536.1434−0.39070.0148
    d1536.13171536.1549−0.02320.0301
    Table 2. Demodulation results with different degrees of aliasing
    Hanlin Liu, Jingtao Xin, Wei Zhuang, Jiabin Xia, Lianqing Zhu. Demodulation method of overlapping spectrum based on convolutional neural network[J]. Infrared and Laser Engineering, 2022, 51(5): 20210419
    Download Citation