Based on the discrimination method of the fruit juice essence and flavoring agent of Raman spectrum and SVM algorithmTechnical field
The invention belongs to food composition authentication technique field, particularly based on the discrimination method of the fruit juice essence and flavoring agent of Raman spectrum and SVM algorithm.
Background technology
Spices refers to the material with fragrance and/or fragrance of applicable human consumption.Relative molecular mass is generally less than 300, has sizable volatility, does not generally directly consume, but is mixed with essence for indirect consumption after perfumed products.Spices can be divided into daily and edible two large classes by purposes.The spices that can be used in blending food essence is called food spices, comprises natural flavor material, natural equivalent fragrance matter and artificial fragrance matter three class.
In food, use food spices, the object of essence is the local flavor making food generation, change or improve food.Food spices is generally mixed with for food perfuming after essence for food, and part also can be directly used in food perfuming.Essence for food, spices use in right amount by need of production in varieties of food items, in GB-2760, the food such as regulation bus sterilization breast, fresh fruit and vegetables, raw grain, rice, part dispensed food for baby and drinking pure water does not have necessity of perfuming, essence for food spices must not be added, except law, regulation or state food safety standard separately have clear stipulaties person.In this standard, also regulation allows the food natural perfume material list (totally 400 kinds of materials) used and the food synthetic perfume list (totally 1453 kinds of materials) allowing use.
In China, do not allow in food flavor production of flavor to use the food flavor outside GB-2760 food spices list.The quality of all necessary strict guarantee food flavor of food flavor manufacturer, dealer and user, user must ensure to use food flavor within the scope of the use amount allowed.
The quality control of flavorant essence is the problem that spices and essence manufacturing enterprise and the food enterprises are shown great attention to always.There is due to chemical fingerprint the features such as fingerprint characteristic analysis, macroscopical inference analysis, therefore be suitable for the stability of Analysis of Complex chemical substance composition, the effective means of spices and essence quality assessment can be become.Finger-print can reflect the overall picture of spices and essence substantially, makes its quality control index rise to the detection to whole interior quality by the mensuration of original single component, realizes the comprehensive evaluation to essence and flavoring agent inherent quality and control comprehensively.
The instrument analytical method of essence and flavoring agent finger-print mainly contains vapor-phase chromatography, liquid phase chromatography, ultraviolet spectroscopy, infra-red sepectrometry, and the research of various instrument analytical method is all comparatively ripe.The algorithm that essence and flavoring agent finger-print is generally commonly used has Nei Y-factor method Y, correlation coefficient process, Cosin method, the Furthest Neighbor of improvement, its algorithm is all generally the algorithm directly quoting traditional Chinese medicine fingerprint, to essence and flavoring agent finger-print algorithm carry out independent studies, comparative study, weight allocation research and each algorithm adaptability research less.At present, essence and flavoring agent fingerprint pattern technology is also in conceptual phase substantially, is not also directly applied to the report of juice production aspect.Therefore, necessary on the basis of prior art, research relates to a kind of simple operation, and energy is accurate, quick, the new method of comprehensive identification fruit juice essence and flavoring agent.
Summary of the invention
In order to solve the technical matters that above-mentioned background technology proposes, the present invention aims to provide the discrimination method of the fruit juice essence and flavoring agent based on Raman spectrum and SVM algorithm, and the method can accurately, fast and efficiently be differentiated the essence and flavoring agent in fruit juice.
In order to realize above-mentioned technical purpose, technical scheme of the present invention is:
Based on the discrimination method of the fruit juice essence and flavoring agent of Raman spectrum and SVM algorithm, comprise the following steps:
(1) pre-treatment is carried out to samples of juice, comprise extraction, enrichment and concentrated three processes successively;
(2) the Raman spectrum spectral data of samples of juice is gathered;
(3) adopt the Raman spectrum spectral data of SVM Algorithm Analysis samples of juice, obtain the classification results of samples of juice essence and flavoring agent, concrete steps are:
Smoothing and the normalized of (a) Raman spectrum spectral data to samples of juice;
B () adopts PCA algorithm to carry out Feature Dimension Reduction to Raman spectrum spectral data, projected to by the Raman spectrum spectral data after dimensionality reduction in the space that low dimensional feature vector opens, build training sample set X::
X＝[X(1),X(2),...,X(k)]∈Rn×k
Wherein, k is the kind of samples of juice, and every class samples of juice comprises n training sample data,
represent a jth training sample data of the i-th class samples of juice, j=1,2 ..., n, Rn × krepresent n × k solution space;
C () chooses kernel function K, build k (k-1)/2 SVM classifier, calculate the value of the kernel function K of proper vector in each SVM classifier, and calculates covariance matrix space according to the symmetric matrix that the value of kernel function K is formed;
D () carries out Householder conversion to covariance matrix, obtain corresponding lineoid matrix;
E (), according to covariance matrix and corresponding lineoid matrix, calculates the characteristic coefficient of each proper vector, and characteristic coefficient is carried out convergent-divergent to covariance matrix, obtain model parameter ω, b after inverting to the covariance matrix of convergent-divergent;
F the test sample book data x of samples of juice is substituted into model f (x, ω)=ω by ()tin x+b, obtain the classification results of samples of juice essence and flavoring agent according to f (x, ω).
Further, in step (1), samples of juice is carried out to the detailed process of pre-treatment: get samples of juice and add Extraction solvent, liquid-liquid extraction method is adopted to extract, combining extraction liquid, after solution crosses 0.45 μm of filter membrane, add anhydrous sodium sulfate drying to spend the night, after concentrated removing Extraction solvent, obtain the samples of juice after processing.
Further, said extracted solvent is one or more potpourri following: methyl alcohol, ethanol, ethyl acetate, acetone, methylene chloride, normal hexane, sherwood oil and ether.
Further, the kernel function in step (c) is gaussian kernel functionwherein, p is the bandwidth parameter of gaussian kernel function, x1for test sample book point, x2for kernel function center.
Further, the Raman spectrum spectral data of D3 portable color dispersion-type Frequency Stabilized Lasers Raman spectrometer collection samples of juice is adopted.
Further, the parameter of setting D3 portable color dispersion-type Frequency Stabilized Lasers Raman spectrometer is as follows: LASER Light Source is 785nm Frequency Stabilized Lasers, and laser power is 450mW, CCD temperature is refrigeration-85 DEG C, and spectral range is 100 ~ 3300cm-1, system analysis degree is 2.5 ~ 3.0cm-1, laser linewidth is < 0.15nm, and wave number is corrected as +/-1cm-1, intensity correction is YES, and signal quality is 12000:1, and weight is 11kg, and operating temperature is 0 DEG C ~ 50 DEG C.
Further, the Raman spectrum spectral data gathering samples of juice carries out in darkroom, and CCD chilling temperature controls between-75 DEG C ~-76 DEG C, and sweep time is 3s, and mean scan number of times is 3 times.
Adopt the beneficial effect that technique scheme is brought:
The present invention is analyzed by the Raman spectrum data gathering fruit juice, thus obtain the identification result of fruit juice essence and flavoring agent, Raman spectrum provides fast, simple, can repeat and the more important thing is undamaged qualitative and quantitative analysis, it is without the need to preparation of samples, and sample can directly by fibre-optical probe or by glass, quartz and optical fiber measurement.Raman Spectra peak is sharply clear, is more suitable for quantitative examination, database search and uses variance analysis to carry out qualitative examination.Essence and flavoring agent fingerprint pattern technology is directly applied to juice production aspect by the present invention, and has that detection speed is fast, simple to operate, favorable repeatability, sensitivity advantages of higher, has greater significance safely to juice production.
Accompanying drawing explanation
Fig. 1 is basic flow sheet of the present invention.
Embodiment
Below with reference to accompanying drawing, technical scheme of the present invention is described in detail.
The present embodiment adopts 3 kinds from the fruit juice of different brands as test sample: sample 1, sample 2, sample 3, and process by flow process shown in Fig. 1.
Step 1, respectively pre-treatment is carried out to sample 1, sample 2, sample 3:
Get samples of juice and add Extraction solvent, adopt liquid-liquid extraction method, combining extraction liquid, solution adds anhydrous sodium sulfate drying and spends the night after crossing 0.45 μm of filter membrane, obtains test sample after concentrated removing Extraction solvent;
Said extracted solvent is one or more potpourri following: methyl alcohol, ethanol, ethyl acetate, acetone, methylene chloride, normal hexane, sherwood oil, ether;
Above-mentioned liquid-liquid extraction method is, get 50mL samples of juice in 125mL separating funnel, add diethyl ether solution 10mL, and 1min is extracted in violent jolting, after stratification, reclaim juice solution from separating funnel bottom, top pours out diethyl ether solution, and the juice solution of recovery is according to above-mentioned extracting process re-extract twice again, merge the diethyl ether solution of three extractions, after 0.45 μm of membrane filtration, add anhydrous sodium sulfate drying and spend the night, also constant volume must need testing solution in 5mL measuring bottle to add methyl alcohol dissolving after filtering and concentrating.
The Raman spectrum spectral data of step 2, respectively collected specimens 1, sample 2, sample 3:
In the present embodiment, the D3 portable color dispersion-type Frequency Stabilized Lasers Raman spectrometer adopting company limited of En Wei SGEDT of the U.S. to produce carries out the collection of Raman spectrum spectral data, and its major parameter is: LASER Light Source: Frequency Stabilized Lasers (785nm); Laser power: 450mW; CCD temperature: freeze-85 DEG C; Spectral range: 100 ~ 3300cm-1; System analysis degree: 2.5 ~ 3.0cm-1; Laser linewidth: < 0.15nm; Wave number is corrected: +/-1cm-1; Intensity correction: YES; Signal quality: 12000:1; Weight: 11kg; Operating temperature 0 DEG C ~ 50 DEG C.
Raman spectrum is captured in darkroom and carries out, and CCD chilling temperature controls between-75 DEG C ~-76 DEG C, and sweep time is 3s, and mean scan number of times is 3 times.
The Raman spectrum spectral data of step 3, employing SVM Algorithm Analysis samples of juice, obtains the classification results of samples of juice essence and flavoring agent:
Smoothing and the normalized of (a) Raman spectrum spectral data to samples of juice;
B () adopts PCA algorithm to carry out Feature Dimension Reduction to Raman spectrum spectral data, projected to by the Raman spectrum spectral data after dimensionality reduction in the space that low dimensional feature vector opens, build training sample set X:
X＝[X(1),X(2),...,X(k)]∈Rn×k
Wherein, k is the kind of samples of juice, and every class samples of juice comprises n training sample data,represent a jth training sample data of the i-th class samples of juice, j=1,2 ..., n;
C () chooses kernel function K, build k (k-1)/2 SVM classifier, calculate the value of the kernel function K of proper vector in each SVM classifier, and calculates covariance matrix space according to the symmetric matrix that the value of kernel function K is formed;
D () carries out Householder conversion to covariance matrix, obtain corresponding lineoid matrix;
E (), according to covariance matrix and corresponding lineoid matrix, calculates the characteristic coefficient of each proper vector, and characteristic coefficient is carried out convergent-divergent to covariance matrix, obtain model parameter ω, b after inverting to the covariance matrix of convergent-divergent.
F the test sample book data x of samples of juice is substituted into model f (x, ω)=ω by ()tin x+b, obtain the classification results of samples of juice essence and flavoring agent according to f (x, ω).
In the present embodiment, the kernel function in step (c) is gaussian kernel functionwherein, p is the bandwidth parameter of gaussian kernel function, x1for test sample book point, x2for kernel function center.
Above embodiment is only and technological thought of the present invention is described, can not limit protection scope of the present invention with this, and every technological thought proposed according to the present invention, any change that technical scheme basis is done, all falls within scope.