Acta Optica Sinica, Volume. 38, Issue 4, 0430005(2018)
De-Noising of Near Infrared Spectra Based on Generalized
Figures & Tables(13)
Fig. 1. (a) Original “Bump” signal and (b) its time-frequency spectrum from generalized S transform
Fig. 2. (a) Noised “Bump” signal (signal noise ratio rSNR=20 dB) and (b) its time-frequency spectrum from generalized S transform
Fig. 3. (a) Singular value sequence of time-frequency matrix of noised “Bump” signal and (b) k-means clustering results of eigenvector
Fig. 4. (a) De-noised “Bump” signal and (b) its time-frequency spectrum from generalized S transform
Fig. 5. Near infrared spectra of wheat gluten samples. (a) Raw spectra; (b) first derivative spectra
Fig. 6. (a) First derivative spectrum of No. 9 wheat gluten sample and (b) its time-frequency spectrum from generalized S transform
Fig. 7. (a) Singular value sequence of time-frequency matrix of derivative spectra and (b) k-means clustering results of eigenvector
Fig. 8. (a) De-noised first derivative spectrum of No. 9 sample and (b) its time-frequency spectrum from generalized S transform
Fig. 10. Comparison of near infrared moisture content prediction results obtained from three de-noising methods and measured values of wheat gluten samples. (a) 9-point smoothing method; (b) wavelet soft thresholding method; (c) proposed method in this paper
Table 1. Evaluation parameters comparison of de-noising effect of 9-point smoothing method, wavelet soft threshold method, and proposed method in this paper
View tableTable 1. Evaluation parameters comparison of de-noising effect of 9-point smoothing method, wavelet soft threshold method, and proposed method in this paper
Parameters before de-noising De-noising method Parameters after de-noising r SNRx RMSEr SNRx RMSEr SI10 0.0463 9 point smoothing method 23.784 0.01826 0.967 Wavelet soft thresholding method 30.932 0.01423 0.972 Proposed method in this paper 37.098 0.01185 0.978 20 0.0313 9 point smoothing method 45.671 0.01022 0.983 Wavelet soft thresholding method 57.983 0.00938 0.987 Proposed method in this paper 61.327 0.00800 0.990 30 0.0165 9 point smoothing method 75.024 0.00654 0.991 Wavelet soft thresholding method 84.921 0.00534 0.993 Proposed method in this paper 88.673 0.00409 0.997
Table 2. Data of wheat gluten samples
View tableTable 2. Data of wheat gluten samples
Sample set Total sample number Moisture content range /% Average of moisture content /% Calibration set 35 4.50-5.90 5.39 Validation set 30 4.50-5.85 5.41
Table 3. Evaluation parameters comparison of de-noising effect of No. 9 spectrum sample using 9-point smoothing method, wavelet soft thresholding method, and proposed method in this paper
View tableTable 3. Evaluation parameters comparison of de-noising effect of No. 9 spectrum sample using 9-point smoothing method, wavelet soft thresholding method, and proposed method in this paper
De-noising method Noise reduction ratio Magnitude attenuation ratio /% Time consuming /s 9-point smoothing method 13.18 28.66 0.031 Wavelet soft thresholding method 15.69 24.66 0.910 Proposed method in this paper 21.54 19.76 5.750
Tools
Get Citation
Copy Citation Text
Jianhua Cai, Yongliang Xiao, Xiaoqin Li. De-Noising of Near Infrared Spectra Based on Generalized
Paper Information
Category: Spectroscopy
Received: Jul. 10, 2017
Accepted: --
Published Online: Jul. 10, 2018
The Author Email: Cai Jianhua (cjh1021cjh@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20