제2권1호
-
David W.Hopkins 1
Derivatives often present the data in spectra in a manner that brings out the information that we are interested in, such as the number and position of bands, and their relative intensity, and removes unwanted baseline variation. Two major methods of calculating derivative spectra are the Savitzkly-Golay method of polynomial curve-fitting, and the Norris Segment-Gap method. Both of these methods can be presented as convolution processes. The definition of the Norris Derivatives is clarified by presenting the Norris Derivatives as convolution functions. A new, normalized form of Norris derivatives is derived that preserves the basic shapes of the calculated derivatives but brings the results into agreement with the Savitzky-Golay results over a range of parameters in both methods. The effectiveness of the convolution functions in removing the random high frequency noise in the spectra can be evaluated by calculating an index called RSSK/Norm, the square root of the sum of the squares of the convolution coefficients, divided by the normalization constant. he agreement and utility of the 2 methods of calculating derivatives is demonstrated by evaluating the results of using derivatives up to the fourth order on the second overtone aromatic CH band of polystyrene as an examples, in the 1680 nm region of the NIR. -
M.A.Hana;W.F.McClure;T.B.Whitaker 15
Designing (specifying the number of nodes in each layer) and training (calibration and validation) back-propagation (BP) for analyzing NIR data can be an arduous and time-consuming task. Actually, training is somewhat trivial. A BP network may be trained by randomly dividing the data set (DS) into two parts, training the network with one part and checking its performance with the other part. However, this procedure is plagued with the lack of objective information about network characteristics - the required number of nodes in the hidden layer(s) and the number of epochs needed to train for optimal performance. Work reported in this paper compares a BP network tuning procedure with a conventional reference (training and testing) procedure. The tuning procedure, believed to have several novel attributes, involved randomly dividing a data set into five groups. Each of the five groups was randomly subdivided into two groups with 80% in a training set and 20% in a tuning set. Training was interrupted periodically after every 100 epochs. During each interruption, network performance was checked against the tuning set - each time recording the mean-squared error (MSE) and the number of epochs (K) needed to reach this point. This procedure continued until a plot of MSE vs total epochs identified a minimum MSE. The number of epochs required achieve minimum MSE was noted. Now optimized (or tuned), network performance was determined by testing the network with all available data. One nice feature of using the tuning method is that the entire process can be executed without user input - i.e., the whole process of developing and training a BP network becomes objective. Four different near infrared data sets (A, B, C and D) were used in this work. Tow of the data sets (A and B) were used to determine the concentration of nicotine in tobacco samples. The other two sets (C and D) were used as a basis for classifyign tobaccos. The optimum BP architecture for each of the four data sets were those consisting of 1, 5, 2 and 1 hidden units in the hidden layer, respectively. The suggested tuning method improved, though marginally in some cases, the true performances of all calibration models as well as their standard deviations. since this work was dependent upon the artificial neural network (ANN) literature, a glossary of terms is given at the end of this paper. Results indicate improved performance using the tuning procedure. In addition, BP network calibrations were better than multiple linear regression (MLR) calibrations on the same data. -
R.James Berry 25
The purpose of this paper is to review the advances in combining near infrared spectroscopy (NIR) with artificial neural networks (NNs). The paper briefly introduces NNs, explains why they are of particular use in NIR research, then reviews their presence in published research. -
Yukiteru Katsumoto;Jian Hui Jiang R.;James Berry;Yukihiro Ozaki 29
This review paper outlines modern pretreatment methods used in NIR spectroscopy. The pretreatment methods can be divided into four categories. One method in is noise reduction. Smoothing is a representative method for the noise reduction. Another is baseline correction. The second derivative and multiplicative scatter correction (MSC) are most frequently employed for baseline correction. The third is centering and normalization and the last is resolution enhancement. Difference spectra, mean centering and second derivative are used in NIR spectroscopy as resolution enhancement methods. In this paper advantages and drawbacks of pretreatment methods currently used in NIR spectroscopy are discussed with many examples of NIR spectra. -
Daisuke Masui;Suehara, Ken-ichiro;Yasuhisa Nakano;Takuo Yano 37
Near infrared spectroscopy (NIRS) was applied to determination of the lipid content of the compost during the compost fermentation of tofu (soybean0curd) refuse. The absorption of lipid observed at 5 wavelengths, 1208, 1712, 1772, 2312 and 2352 nm on the second derivative spectra. To formulated a calibration equation, a multiple linear regression analysis was carried out between the near-infrared spectral data and on the lipid content in the calibration sample set (sample number, n=60) obtained using Soxhlet extraction method. The value of the multiple correlation coefficient (R) was 0.975 when using the wavelengths of 1208 and 1712 nm were used in the calibration equation. To validate the calibration equation obtained, the lipid content in the validation sample set (n=35) not used for formulating the calibration equation was calculated using the calibration equation, and compared with the value obtained using the Soxhlet extraction method. Good agreement was observed between the results of the Soxhlet extraction method and those values of the NIRS method. The simple correlation coefficient (r) and standard error of prediction (SEP) were 0.964 and 0.815 %, respectively. suitability of the lipid content as an indicator of the compost fermentation of tofu refuse was also studied. The decrease of the lipid content in the compost corresponded to the decrease of the total dry weight of the compost in the composter. The lipid content was a significant indicator of the compost fermentation. The NIRS method was applied to measure the time course of the lipid content in the compost fermentation and good results were obtained. The study indicates that NIRS is a useful method for process management of the compost fermentation of tofu refuse. -
Larry Leonardi;David H.Burns;Luis Openheimer;Rene P.Michel 43
A non-invasive spectroscopic method is presented for the measurement of pulmonary edema. Both early diagnosis and quantitative edema estimates were investigated. The spectroscopic determination of pulmonary edema involved the acquisition of diffuse reflectance spectra in the near-infrared (NIR) region with change in water concentration - water is the main constituent of edema fluid. Pulmonary edema was induced into the excised perfused lungs of seven animals by elevating the hydrostatic pressure. Estimates of edema were ascertained from a partial least squares regression of the measured spectral response. Actual edema was determined from the change (increase) in total lung weight. Estimates in relative lung weight increases due to in vitro edema were made with the near infrared spectra. The results revealed that fluid accumulation produced spectral changes in the O-H and C-H absorptions as well as scattering changes in the spectra. Histology of the lung was used to verify the presence or absence of interstitial and alveolar edema. Results demonstrated that near infrared spectroscopy might provide a new tool for clinical assessment of pulmonary edema. -
Sohn, Mi-Ryeong;Kwon, Young-Kill;Cho, Rae-Kwang 55
The object of this work was to investigate the influence of growing district and harvest year on calibration for sweetness (Brix) determination of Fuji apple fruit using near infrared (NIR) reflectance spectroscopy, and to develop the robust calibration across these variation. The calibration models was based on wavelength range of 1100∼2500 nm using a stepwise multiple linear regression. A calibration model by sample set of one growing district was not transferable to other growing districts. The combined calibration (data of three growing districts) predicted reasonable well against a population set drawn from all growing districts (SEP=0.69, Bias=0.075). A calibration model by sample set of one harvest year was not also transferable to other harvest years. The combined calibration (data of three harvest years) predicted well against a population set drawn from all harvest years (SEP=0.53, Bias=0.004). -
Roumiana Tsenkova;Stefka Atanassova;Kiyohiko Toyoda 59
Nowadays, medical diagnostics is efficiently supported by clinical chemistry and near infrared spectroscopy is becoming a new dimension, which has shown high potential to provide valuable information for diagnosis. The investigation was carried out to study the influence of mammary gland inflammation, called mastitis, on cow´s milk spectra and milk composition measured by near infrared spectroscopy (NIRS). Milk somatic cell counts (SCC) in milk were used as a measure of mammary gland inflammation. Naturally occurred variations with milk composition within lactation and in the process of milking were included in the experimental design of this study. Time series of unhomogenized, raw milk spectral data were collected from 3 cow along morning and evening milking, for 5 consecutive months, within their second lactation. In the time of the trial, the investigated cows had periods with mammary gland inflammation. Transmittance spectra of 258 milk samples were obtained by NIRSystem 6500 spectrophotometer in 1100-2400 nm region. Calibration equations for the examined milk components were developed by PLS regression using 3 different sets of samples: samples with low somatic cell count (SCC), samples with high SCC and combined data set. The NIR calibration and prediction of individual cow´s milk fat, protein, and lactose were highly influenced by the presence of mil samples from animals with mammary gland inflammation in the data set. The best accuracy of prediction (i.e. the lower SEP and the higher correlation coefficient) for fat, protein and lactose was obtained for equations, developed when using only “healthy” samples, with low SCC. The standard error of prediction increased and correlation coefficient decreased significantly when equations for low SCC milk were used to predict examined components in “mastitis” samples with high SCC, and vice versa. Combined data set that included samples from healthy and mastitis animals could be used to build up regression models for screening. Further use of separate model for healthy samples improved milk composition measurement. Regression vectors for NIR mild protein measurement obtained for “healthy” and “mastitic” group were compared and revealed differences in 1390-1450 nm, 1500-1740 nm and 1900-2200 nm regions and thus illustrated post-secretory breakdown of milk proteins by hydrolytic enzymes that occurred with mastitis. For the first time it has been found that monitoring the spectral differences in water bands at 1440 nm and 1912 nm could provide valuable information for inflammation diagnosis.