• Spectroscopy and Spectral Analysis
  • Vol. 42, Issue 5, 1553 (2022)
Zhong WANG, Dong-dong WAN, Chuang SHAN, Yue-e LI, and Qing-guo ZHOU*;
Author Affiliations
  • School of Information Science and Engineering, Lanzhou University, Lanzhou 730000, China
  • show less
    DOI: 10.3964/j.issn.1000-0593(2022)05-1553-08 Cite this Article
    Zhong WANG, Dong-dong WAN, Chuang SHAN, Yue-e LI, Qing-guo ZHOU. A Denoising Method Based on Back Propagation Neural Network for Raman Spectrum[J]. Spectroscopy and Spectral Analysis, 2022, 42(5): 1553 Copy Citation Text show less
    Comparison of different denoising methods for 20 db noise
    Fig. 1. Comparison of different denoising methods for 20 db noise
    Comparison of different denoising methods for 30 db noise
    Fig. 2. Comparison of different denoising methods for 30 db noise
    RMSE and SNR changes of different wavelets and different thresholds for 20 db noise
    Fig. 3. RMSE and SNR changes of different wavelets and different thresholds for 20 db noise
    RMSE and SNR changes of different wavelets and different thresholds for 30 db noise
    Fig. 4. RMSE and SNR changes of different wavelets and different thresholds for 30 db noise
    RMSE and SNR changes of different wavelet decomposition levels for different noise
    Fig. 5. RMSE and SNR changes of different wavelet decomposition levels for different noise
    RMSE and SNR changes of different denoising methods for 20 db noise
    Fig. 6. RMSE and SNR changes of different denoising methods for 20 db noise
    RMSE and SNR changes of different denoising methods for 30 db noise
    Fig. 7. RMSE and SNR changes of different denoising methods for 30 db noise
    Raman Spectrum of experimental sample(a): Noisy Raman spectrum; (b): Denoised spectrum by analyst; (c): Denoised spectrum by BP network
    Fig. 8. Raman Spectrum of experimental sample
    (a): Noisy Raman spectrum; (b): Denoised spectrum by analyst; (c): Denoised spectrum by BP network
    滑动窗口平均值S-G滤波法小波阈值
    窗口大小窗口大小阶数tptrsorhscalnwname
    20 db15213heursuressln8sym7
    30 db30212sqtwologhmln6sym5
    Table 1. Parameter settings of various methods
    原始含噪神经网络滑动窗口均值S-G滤波傅里叶变换小波阈值
    20 dbRMSE9.719 02.616 42.990 83.372 44.379 92.317 7
    SNR12.570 923.969 122.807 721.764 719.494 025.022 2
    30 dbRMSE31.605 36.772 78.127 811.054 513.855 86.183 7
    SNR2.328 115.708 214.123 911.452 59.490 716.498 4
    Table 2. Comparison of various methods for RMSE and SNR
    神经网络傅里叶变换小波阈值滑动窗口均值S-G滤波器
    RMSE¯2.787 94.362 62.434 47.745 63.912 1
    20 dbSNR¯22.950 818.819 324.031 616.958 019.989 8
    S99219699
    RMSE¯7.197 413.460 86.692 811.268 410.688 7
    30 dbSNR¯14.565 09.020 315.285 011.543 011.043 6
    S1002793100
    Table 3. Comparison of various methods for RMSE and SNR on different Noise
    Zhong WANG, Dong-dong WAN, Chuang SHAN, Yue-e LI, Qing-guo ZHOU. A Denoising Method Based on Back Propagation Neural Network for Raman Spectrum[J]. Spectroscopy and Spectral Analysis, 2022, 42(5): 1553
    Download Citation