CN107274065A - The subjective assessment of food taste and flavor and spectroscopic data modeling method and system - Google Patents
The subjective assessment of food taste and flavor and spectroscopic data modeling method and system Download PDFInfo
- Publication number
- CN107274065A CN107274065A CN201710340311.6A CN201710340311A CN107274065A CN 107274065 A CN107274065 A CN 107274065A CN 201710340311 A CN201710340311 A CN 201710340311A CN 107274065 A CN107274065 A CN 107274065A
- Authority
- CN
- China
- Prior art keywords
- food
- subjective assessment
- data
- mrow
- flavor
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0282—Rating or review of business operators or products
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Development Economics (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Educational Administration (AREA)
- Evolutionary Biology (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Tourism & Hospitality (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The present invention relates to the subjective assessment of food taste and flavor and spectroscopic data modeling method and system, wherein method includes obtaining the spectroscopic data of certain food, subjective assessment is obtained to participate in evaluation and electing people's information, set up the subjective assessment label of correspondence food, a large amount of collection data above, machine learning algorithm modeling is carried out, in the case of known food spectroscopic data and the people's information that participates in evaluation and electing, subjective assessment label result is predicted.Wherein system participates in evaluation and electing people's information collection module, spectra collection module, subjective assessment label input module, data memory module, data processing module and subjective assessment Tag Estimation output module including subjective assessment.Food taste and flavor subjective evaluation index of the present invention can quantify；The prediction of taste and flavor can be carried out；For the analysis of food, spectroscopic data and analysis process are obtained, it is quick, lossless；As data increase, category of model accuracy rate can be lifted；Food can be recommended and advised for specific crowd.
Description
Technical field
The present invention relates to pattern-recognition and spectrum analysis field, and in particular to the subjective assessment of food taste and flavor and light
Modal data modeling method and system.
Background technology
Taste and flavor is a kind of senior experience of the mankind to food.Mouthfeel refers to food in people oral cavity, by tactile
The direct feeling produced with chewing, is independently of another experience outside the sense of taste.And local flavor is the taste that people taste food
Feel and nasal receptor.Subjective assessment for food, it is necessary to which taste and flavor is united sets up appraisement system.In the past by
Do not set up and associate with objective data in the subjective assessment of food taste and flavor, the impression for food relies on descriptive language
Speech, there is no quantifiable subjective evaluation index.
The content of the invention
The technical problem to be solved in the present invention is：Subjective assessment in order to solve food taste and flavor lacks quantifiable
Subjective assessment problem, the present invention utilizes the characterization of molecules information of spectroscopic data, it is proposed that the subjective assessment of food taste and flavor
And spectroscopic data modeling method and system.
The technical solution adopted for the present invention to solve the technical problems is：A kind of subjective assessment of food taste and flavor and
Spectroscopic data modeling method, comprises the following steps：
Obtain the spectroscopic data D of certain foodi, obtain subjective assessment and participate in evaluation and electing people's information, set up correspondence food mouthfeel and wind
The subjective assessment label of taste；
Largely gather the spectroscopic data D of different foodsiParticipated in evaluation and electing people's information with subjective assessment, and set up correspondence food mouthfeel and
The subjective assessment label of local flavor, carry out machine learning algorithm modeling, known food spectroscopic data and subjective assessment participate in evaluation and electing people letter
In the case of breath, subjective assessment label result is predicted.
The spectroscopic data Di, it is the reflection infrared spectrum rate derivative data of every group of food, wherein i=1,2 ..., n are
Sampled point number；The difference between spectroscopic data vector is represented using COS distance, by rejecting abnormal data；Derivative calculations are used
Direct finite-difference method, is subtracted each other using spectroscopic data pointwise and obtained：
Wherein, r is the infrared spectrum primary reflection rate of every group of food, and i=1,2 ..., n are sampled point number, and λ is sampling
When spectral wavelength, Δ λ be wavelength sampling minimum interval.
The subjective assessment people's information that participates in evaluation and electing is converted into numerical value according to information content：
Sex Xa：Man, female, numerical value is 1,2, default value 0；
Age Xb：According to the numerical value filled in, 1,2 ..., do not set the upper limit, default value 0；
Native place Xc：China has 34 provinces, and numerical value is 1,2 ... 34, if subsequently newly-increased other countries province, suitable successively
Prolong, default value is 0；
National Xd：China has 56 nationalitys, and numerical value is 1,2 ... 56, if subsequently increased newly, other are national, postpone successively,
Default value is 0.
The subjective assessment label of the correspondence food taste and flavor, including, two basic sides of food mouthfeel ---
Cold and hot degree and soft or hard degree, and food flavor two category informations --- taste and smell.
Quantifiable index will be converted into by descriptive language to the subjective assessment of food mouthfeel, according to the two of food mouthfeel
Individual aspect, food hot/cold degree and soft or hard degree, carry out index decomposition, people is scored respectively by participating in evaluation and electing：
Food hot/cold degree α, 5 points of systems correspond to ice, cool, warm, hot, boiling hot five regions from low to high；
The soft or hard degree β of food, ten point system corresponds to dilute, thick, soft, glutinous, tender, sliding, bullet tooth, shortcake, crisp, hard ten from low to high
Individual region；
Set up food mouthfeel tag along sort YI=(α, β), corresponds cold and hot and soft or hard each region of degree；
Quantifiable index will be converted into by descriptive language to the subjective assessment of food flavor, according to the two of food flavor
Category information, to the taste and smell of food, carries out index decomposition, people is scored respectively by participating in evaluation and electing：
To the sense of taste δ of food, acid, sweet tea, hardship, peppery, salty 5 marking are carried out, each 5 points are made, altogether 25 kinds of situations；
To the smell ε of food, ten point system corresponds to smelly, tasteless, fragrant ten level conditions from low to high；
Set up food flavor tag along sort YII=(δ, ε), corresponds the various situations of taste and smell.
The machine learning algorithm modeling, for the subjective assessment label Y=(Y of every kind of food taste and flavorI,YII)=
(α, β, δ, ε), corresponding participate in evaluation and electing people's information and spectroscopic data X=(Xa, Xb, Xc, Xd, D1, D2..., Dn), sample set T=(x1, y1),
(x2, y2) ..., (xm, ym), wherein, xmRepresent m-th of sample, ymThe classification of m-th of sample is represented, m is number of samples；Set up
Sorting criterion between different subjective assessment labels and input data, algorithm selects nearest neighbour method or SVM (SVMs)；
The nearest neighbour method, the criterion k between generation is of all categories, for testing data, is counted one by one with the data of known class
Distance is calculated, is found according to unified distance discrimination criterion with known sample closest to data individual k, and with this datum
According to classification as testing data classification；
Specifically it is expressed as：Sample set T=(x1, y1), (x2, y2) ..., (xm, ym), xiRepresent i-th of sample, yiRepresent i-th
The classification of individual sample, m sample vector x1~xmCovariance matrix is designated as S, and judgment criterion function is differentiated using mahalanobis distance, its
Middle sample xiTo xjMahalanobis distance Δ (xi, xj) be,
In classification, k nearest samples, vote the classification belonging to this k sample, percentage of votes obtained is most before selection
Classification be testing sample classification, wherein k selection is continued to optimize according to training sample obtains, and k takes 2~5；KNN (neighbours
Method) relatively good place is, it can be evolved into a kind of similarity foundation what is calculated before us apart from criterion, to later I
The similarity weighed between different tree species there is one intuitively to recognize.In addition when our sample size is enough,
KNN result is that result is just more accurate.Algorithms T-cbmplexity is O (n), easy and effective without training, many in sample
In the case of, it is ensured that the higher degree of accuracy.
The SVM algorithm, is the method for discrimination of a two class problems in itself, but is a multiclass in data of the present invention classification
Problem.Therefore, from " one-to-one " method, for n class samples, one binary classifier of classification training, trains n* (n- altogether two-by-two
1)/2 a grader, is tested and is voted one by one to test sample, and who gets the most votes's classification is its class label；
The problem of classifying for two classes, the thought of svm classifier is to find this so that maximum super of two classification sample intervals
Plane, now, training data are (x1, y1), (x2, y2) ..., (xm, ym), y ∈ {+1, -1 }；The problem of asking for largest interval is just
The problem of being an optimizing, the equation of optimal classification surface：
Wx+b=0
Wherein w and b are unknown, the optimization problem to be solved after conversion：
So that meeting yi[(wxi)+b] -1 >=0 (i=1,2 ..., m)
M is number of samples, solves to obtain w and b, that is, obtains optimal classification surface；
For some in lower dimensional space linear inseparable situation, being mapped that to by kernel function in higher dimensional space makes
The problem of it is changed into linear separability；Conventional kernel function K (x, xi) there are multinomial inner product kernel function, Radial basis kernel function (RBF)
With Sigmoid inner product kernel functions：(K (x, xi) represent kernel function)
Multinomial inner product kernel function：
Radial basis kernel function：
Sigmoid inner product kernel functions：
The present invention carries out SVM kernel functional parameters from libsvm software kits and asked for, and the step of using is：
1) data set is prepared according to the form required by LIBSVM software kits；
2) simple zoom operations are carried out to data；
3) consider to select RBF kernel functions；
4) using cross validation selection optimal parameter C and g；
5) acquisition supporting vector machine model is trained to whole training set using optimal parameter C and g；
6) tested and predicted using the model of acquisition.
A kind of subjective assessment of food taste and flavor and spectroscopic data modeling, including：Subjective assessment participate in evaluation and electing people letter
Cease collection module, spectra collection module, subjective assessment label input module, data memory module, data processing module, Yi Jizhu
See and evaluate Tag Estimation output module；
The subjective assessment is participated in evaluation and electing people's information collection module, for collecting participate in evaluation and electing people's information, including sex Xa, age Xb, nationality
Pass through XcWith national Xd；Can be by various terminals, including the multiple terminal such as mobile phone A pp softwares, computer software, Intelligent bracelet is obtained.
The spectra collection module, the reflection infrared spectrum rate derivative data for obtaining food, by spectrum sensor,
And specific light source environment, the reflectance spectrum of food is collected, the primary reflection rate of corresponding wavelength is obtained, then by inserting calculation of group dividing
Obtain derivative data；
The subjective assessment label input module, quantifies for obtaining evaluation of the people to this kind of food taste and flavor of participating in evaluation and electing
Information, including food hot/cold, soft or hard degree, and the taste and smell of food score；
The data memory module, for storing training set data, and algorithm model parameter, can use and be locally stored
Space or cloud storage space；
The data processing module, for using algorithm model criterion, to new participate in evaluation and electing people and spectroscopic data, prediction is subjective
Evaluate label；
The subjective assessment Tag Estimation output module, for showing subjective assessment label, and will be by people's revision of participating in evaluation and electing
Subjective assessment label and correspondingly participate in evaluation and electing people's information and spectroscopic data is exported to data memory module, deposited as training set data
Storage, realizes the accumulation of data model.
The purpose of the present invention is exactly by the way that the spectroscopic data of intrinsic characteristic food item and food taste and flavor are set up into data
Model, realizes the quantifiable food mouthfeel of science and taste appraisement system.
Mouthfeel generally comprises two basic sides of cold and hot degree and soft or hard degree of food：The word of food hot/cold degree is described
For example warm cool blanching of language etc.；For example soft glutinous shortcake of the word of the description soft or hard degree of food is slided tender and crisp etc..The taste and smell that local flavor is included
Experience, generally comprises the five tastes --- and it is acid, sweet tea, hardship, peppery, salty and tasteless, fragrant, smelly.For two basic sides and wind of mouthfeel
Two category informations of taste, spectroscopic data contains key message therein --- the characterization of molecules of material.Food hot/cold degree, micro-
It is the severe degree of molecular thermalmotion in sight；And the soft or hard degree of food, it is the structural information of molecule.The taste and smell of food is then
It is all kinds of molecular compositions and the embodiment of content.Spectrum can characterize the characterization of molecules of material surface, when incident light and material surface
After molecule is had an effect, the characterization of molecules information of material is just contained in the spectrum of reflected light, data modeling can be carried out.By
The objective data of collection, after substantial amounts of spectroscopic data information is collected, with reference to the subjective assessment being associated, utilizes machine learning
Method realize the predictions of data.Spectrum analysis combination subjective assessment, the application potential of spectroscopic data will be played significantly.
The beneficial effects of the invention are as follows the subjective assessment of food taste and flavor of the invention and spectroscopic data modeling method
And system, compared with prior art, food taste and flavor subjective evaluation index of the present invention can quantify；Because of itself and intrinsic food point
The spectroscopic data of subcharacter is modeled, and can carry out the prediction of taste and flavor；For the analysis of food, spectroscopic data is obtained
And analysis process, it is quick, lossless；As data increase, category of model accuracy rate can be lifted；Can be right for specific crowd
Food is recommended and advised.
Brief description of the drawings
The present invention is further described with reference to the accompanying drawings and examples.
Fig. 1 is the flow chart of the subjective assessment and spectroscopic data modeling method of a kind of food taste and flavor of the present invention.
Embodiment
In conjunction with the accompanying drawings, the present invention is further explained in detail.
As shown in figure 1, subjective assessment and the spectroscopic data modeling method of a kind of food taste and flavor of the present invention, including
Following steps：
Obtain the spectroscopic data D of certain foodi, obtain subjective assessment and participate in evaluation and electing people's information, set up correspondence food mouthfeel and wind
The subjective assessment label of taste；
Largely gather the spectroscopic data D of different foodsiParticipated in evaluation and electing people's information with subjective assessment, and set up correspondence food mouthfeel and
The subjective assessment label of local flavor, carry out machine learning algorithm modeling, known food spectroscopic data and subjective assessment participate in evaluation and electing people letter
In the case of breath, subjective assessment label result is predicted.
The spectroscopic data Di, it is the reflection infrared spectrum rate derivative data of every group of food, wherein i=1,2 ..., n are
Sampled point number；The difference between spectroscopic data vector is represented using COS distance, by rejecting abnormal data；Derivative calculations are used
Direct finite-difference method, is subtracted each other using spectroscopic data pointwise and obtained：
Wherein, r is the infrared spectrum primary reflection rate of every group of food, and i=1,2 ..., n are sampled point number, and λ is sampling
When spectral wavelength, Δ λ be wavelength sampling minimum interval.
The subjective assessment people's information that participates in evaluation and electing is converted into numerical value according to information content：
Sex Xa：Man, female, numerical value is 1,2, default value 0；
Age Xb：According to the numerical value filled in, 1,2 ..., do not set the upper limit, default value 0；
Native place Xc：China has 34 provinces, and numerical value is 1,2 ... 34, if subsequently newly-increased other countries province, suitable successively
Prolong, default value is 0；
National Xd：China has 56 nationalitys, and numerical value is 1,2 ... 56, if subsequently increased newly, other are national, postpone successively,
Default value is 0.
The subjective assessment label of the correspondence food taste and flavor, including, two basic sides of food mouthfeel ---
Cold and hot degree and soft or hard degree, and food flavor two category informations --- taste and smell.
Quantifiable index will be converted into by descriptive language to the subjective assessment of food mouthfeel, according to the two of food mouthfeel
Individual aspect, food hot/cold degree and soft or hard degree, carry out index decomposition, people is scored respectively by participating in evaluation and electing：
Food hot/cold degree α, 5 points of systems correspond to ice, cool, warm, hot, boiling hot five regions from low to high；
The soft or hard degree β of food, ten point system corresponds to dilute, thick, soft, glutinous, tender, sliding, bullet tooth, shortcake, crisp, hard ten from low to high
Individual region；
Set up food mouthfeel tag along sort YI=(α, β), corresponds cold and hot and soft or hard each region of degree；
Quantifiable index will be converted into by descriptive language to the subjective assessment of food flavor, according to the two of food flavor
Category information, to the taste and smell of food, carries out index decomposition, people is scored respectively by participating in evaluation and electing：
To the sense of taste δ of food, acid, sweet tea, hardship, peppery, salty 5 marking are carried out, each 5 points are made, altogether 25 kinds of situations；
To the smell ε of food, ten point system corresponds to smelly, tasteless, fragrant ten level conditions from low to high；
Set up food flavor tag along sort YII=(δ, ε), corresponds the various situations of taste and smell.
The machine learning algorithm modeling, for the subjective assessment label Y=(Y of every kind of food taste and flavorI,YII)=
(α, β, δ, ε), corresponding participate in evaluation and electing people's information and spectroscopic data X=(Xa, Xb, Xc, Xd, D1, D2..., Dn), sample set T=(x1, y1),
(x2, y2) ..., (xm, ym), wherein, xmRepresent m-th of sample, ymThe classification of m-th of sample is represented, m is number of samples；Set up
Sorting criterion between different subjective assessment labels and input data, algorithm selects nearest neighbour method or SVM (SVMs)；
The nearest neighbour method, the criterion k between generation is of all categories, for testing data, is counted one by one with the data of known class
Distance is calculated, is found according to unified distance discrimination criterion with known sample closest to data individual k, and with this datum
According to classification as testing data classification；
Specifically it is expressed as：Sample set T=(x1, y1), (x2, y2) ..., (xm, ym), xiRepresent i-th of sample, yiRepresent i-th
The classification of individual sample, m sample vector x1~xmCovariance matrix is designated as S, and judgment criterion function is differentiated using mahalanobis distance, its
Middle sample xiTo xjMahalanobis distance Δ (xi, xj) be,
In classification, k nearest samples, vote the classification belonging to this k sample, percentage of votes obtained is most before selection
Classification be testing sample classification, wherein k selection is continued to optimize according to training sample obtains, and k takes 2~5；KNN (neighbours
Method) relatively good place is, it can be evolved into a kind of similarity foundation what is calculated before us apart from criterion, to later I
The similarity weighed between different tree species there is one intuitively to recognize.In addition when our sample size is enough,
KNN result is that result is just more accurate.Algorithms T-cbmplexity is O (n), easy and effective without training, many in sample
In the case of, it is ensured that the higher degree of accuracy.
The SVM algorithm, is the method for discrimination of a two class problems in itself, but is a multiclass in data of the present invention classification
Problem.Therefore, from " one-to-one " method, for n class samples, one binary classifier of classification training, trains n* (n- altogether two-by-two
1)/2 a grader, is tested and is voted one by one to test sample, and who gets the most votes's classification is its class label；
The problem of classifying for two classes, the thought of svm classifier is to find this so that maximum super of two classification sample intervals
Plane, now, training data are (x1, y1), (x2, y2) ..., (xm, ym), y ∈ {+1, -1 }；The problem of asking for largest interval is just
The problem of being an optimizing, the equation of optimal classification surface：
Wx+b=0
Wherein w and b are unknown, the optimization problem to be solved after conversion：
So that meeting yi[(wxi)+b] -1 >=0 (i=1,2 ..., m)
M is number of samples, solves to obtain w and b, that is, obtains optimal classification surface；
For some in lower dimensional space linear inseparable situation, being mapped that to by kernel function in higher dimensional space makes
The problem of it is changed into linear separability；Conventional kernel function K (x, xi) there are multinomial inner product kernel function, Radial basis kernel function (RBF)
With Sigmoid inner product kernel functions：(K (x, xi) represent kernel function)
Multinomial inner product kernel function：
Radial basis kernel function：
Sigmoid inner product kernel functions：
The present invention carries out SVM kernel functional parameters from libsvm software kits and asked for, and the step of using is：
1) data set is prepared according to the form required by LIBSVM software kits；
2) simple zoom operations are carried out to data；
3) consider to select RBF kernel functions；
4) using cross validation selection optimal parameter C and g；
5) acquisition supporting vector machine model is trained to whole training set using optimal parameter C and g；
6) tested and predicted using the model of acquisition.
A kind of subjective assessment of food taste and flavor and spectroscopic data modeling, including：Subjective assessment participate in evaluation and electing people letter
Cease collection module, spectra collection module, subjective assessment label input module, data memory module, data processing module, Yi Jizhu
See and evaluate Tag Estimation output module；
The subjective assessment is participated in evaluation and electing people's information collection module, for collecting participate in evaluation and electing people's information, including sex Xa, age Xb, nationality
Pass through XcWith national Xd；Can be by various terminals, including the multiple terminal such as mobile phone A pp softwares, computer software, Intelligent bracelet is obtained.
The spectra collection module, the reflection infrared spectrum rate derivative data for obtaining food, by spectrum sensor,
And specific light source environment, the reflectance spectrum of food is collected, the primary reflection rate of corresponding wavelength is obtained, then by inserting calculation of group dividing
Obtain derivative data；
The subjective assessment label input module, quantifies for obtaining evaluation of the people to this kind of food taste and flavor of participating in evaluation and electing
Information, including food hot/cold, soft or hard degree, and the taste and smell of food score；
The data memory module, for storing training set data, and algorithm model parameter, can use and be locally stored
Space or cloud storage space；
The data processing module, for using algorithm model criterion, to new participate in evaluation and electing people and spectroscopic data, prediction is subjective
Evaluate label；
The subjective assessment Tag Estimation output module, for showing subjective assessment label, and will be by people's revision of participating in evaluation and electing
Subjective assessment label and correspondingly participate in evaluation and electing people's information and spectroscopic data is exported to data memory module, deposited as training set data
Storage, realizes the accumulation of data model.
In the embodiment of the present invention, the subjective assessment people's information that participates in evaluation and electing can be obtained by mobile phone A pp user terminals；Light
Compose acquisition module to realize using shore pine FPI devices, coordinate bluetooth communication, spectroscopic data is transmitted to mobile phone A pp；Data storage mould
Block and data processing are in cloud server end, and data modeling tries to achieve modeling parameters using libsvm, and cloud service backstage carries out data and built
Mould, mobile phone A pp have subjective assessment label input, display and revision function, realize subjective assessment data continuous updating and
Accumulation.
Compared with prior art, food taste and flavor subjective evaluation index of the present invention can quantify；Because of itself and intrinsic food
The spectroscopic data of characterization of molecules is modeled, and can carry out the prediction of taste and flavor；For the analysis of food, spectrum number is obtained
According to and analysis process, it is quick, lossless；As data increase, category of model accuracy rate can be lifted；Can for specific crowd,
Food is recommended and advised.
Using the above-mentioned desirable embodiment according to the present invention as enlightenment, by above-mentioned description, relevant staff is complete
Various changes and amendments can be carried out without departing from the scope of the technological thought of the present invention' entirely.The technology of this invention
Property scope is not limited to the content on specification, it is necessary to its technical scope is determined according to right.
Claims (7)
1. subjective assessment and the spectroscopic data modeling method of a kind of food taste and flavor, it is characterised in that comprise the following steps：
Obtain the spectroscopic data D of certain foodi, obtain subjective assessment and participate in evaluation and electing people's information, set up the master of correspondence food taste and flavor
See and evaluate label；
Largely gather the spectroscopic data D of different foodsiParticipated in evaluation and electing people's information with subjective assessment, and set up correspondence food taste and flavor
Subjective assessment label, carry out machine learning algorithm modeling, participated in evaluation and electing people's information in known food spectroscopic data and subjective assessment
In the case of, predict subjective assessment label result.
2. subjective assessment and the spectroscopic data modeling method of food taste and flavor as claimed in claim 1, it is characterised in that：
The spectroscopic data Di, it is the reflection infrared spectrum rate derivative data of every group of food, wherein i=1,2 ..., n are sampled point
Number；The difference between spectroscopic data vector is represented using COS distance, by rejecting abnormal data；Derivative calculations use direct differential
Method, is subtracted each other using spectroscopic data pointwise and obtained：
<mrow>
<mfrac>
<mrow>
<mi>d</mi>
<mi>r</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>&lambda;</mi>
</mrow>
</mfrac>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>r</mi>
<mrow>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
</mrow>
<mrow>
<mn>2</mn>
<mi>&Delta;</mi>
<mi>&lambda;</mi>
</mrow>
</mfrac>
</mrow>
Wherein, r is the infrared spectrum primary reflection rate of every group of food, i=1,2 ..., n, and n is sampled point number, when λ is sampling
Spectral wavelength, Δ λ be wavelength sampling minimum interval.
3. subjective assessment and the spectroscopic data modeling method of food taste and flavor as claimed in claim 2, it is characterised in that：
The subjective assessment people's information that participates in evaluation and electing is converted into numerical value according to information content：
Sex Xa：Man, female, numerical value is 1,2, default value 0；
Age Xb：According to the numerical value filled in, 1,2 ..., do not set the upper limit, default value 0；
Native place Xc：China has 34 provinces, and numerical value is 1,2 ... 34, if subsequently newly-increased other countries province, postpones, write from memory successively
It is 0 to recognize value；
National Xd：China has 56 nationalitys, and numerical value is 1,2 ... 56, if subsequently increased newly, other are national, postpone, give tacit consent to successively
It is worth for 0.
4. subjective assessment and the spectroscopic data modeling method of food taste and flavor as claimed in claim 3, it is characterised in that：
The subjective assessment label of the correspondence food taste and flavor, including, two basic sides --- the cold and hot degree of food mouthfeel
With two category informations --- the taste and smell of soft or hard degree, and food flavor.
5. subjective assessment and the spectroscopic data modeling method of food taste and flavor as claimed in claim 4, it is characterised in that：
Quantifiable index will be converted into by descriptive language to the subjective assessment of food mouthfeel, according to two aspects of food mouthfeel,
Food hot/cold degree and soft or hard degree, carry out index decomposition, people is scored respectively by participating in evaluation and electing：
Food hot/cold degree α, 5 points of systems correspond to ice, cool, warm, hot, boiling hot five regions from low to high；
The soft or hard degree β of food, ten point system corresponds to dilute, thick, soft, glutinous, tender, sliding, bullet tooth, shortcake, crisp, Ying Shige areas from low to high
Domain；
Set up food mouthfeel tag along sort YI=(α, β), corresponds cold and hot and soft or hard each region of degree；
Quantifiable index will be converted into by descriptive language to the subjective assessment of food flavor, believed according to two classes of food flavor
Breath, to the taste and smell of food, carries out index decomposition, people is scored respectively by participating in evaluation and electing：
To the sense of taste δ of food, acid, sweet tea, hardship, peppery, salty 5 marking are carried out, each 5 points are made, altogether 25 kinds of situations；
To the smell ε of food, ten point system corresponds to smelly, tasteless, fragrant ten level conditions from low to high；
Set up food flavor tag along sort YII=(δ, ε), corresponds the various situations of taste and smell.
6. subjective assessment and the spectroscopic data modeling method of food taste and flavor as claimed in claim 5, it is characterised in that：
The machine learning algorithm modeling, for the subjective assessment label Y=(Y of every kind of food taste and flavorI,YII)=(α, β, δ,
ε), corresponding participate in evaluation and electing people's information and spectroscopic data X=(Xa, Xb, Xc, Xd, D1, D2..., Dn), sample set T=(x1, y1), (x2,
y2) ..., (xm, ym), wherein, xmRepresent m-th of sample, ymThe classification of m-th of sample is represented, m is number of samples；Set up different
Sorting criterion between subjective assessment label and input data, algorithm selects nearest neighbour method or SVM (SVMs)；
The nearest neighbour method, generate it is of all categories between criterion k, for testing data, with the data of known class calculate one by one away from
From, found according to unified distance discrimination criterion with known sample closest to data individual k, and with this given data
Classification as testing data classification；
Specifically it is expressed as：Sample set T=(x1, y1), (x2, y2) ..., (xm, ym), xiRepresent i-th of sample, yiRepresent i-th of sample
This classification, m sample vector x1~xmCovariance matrix is designated as S, and judgment criterion function is differentiated using mahalanobis distance, wherein sample
This xiTo xjMahalanobis distance Δ (xi, xj) be,
<mrow>
<mi>&Delta;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msqrt>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mi>T</mi>
</msup>
<msup>
<mi>S</mi>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</msqrt>
</mrow>
In classification, k nearest samples, vote the classification belonging to this k sample before selection, the most class of percentage of votes obtained
It is not the classification of testing sample, wherein k selection is continued to optimize according to training sample obtains, and k takes 2~5；
The SVM algorithm, from " one-to-one " method, for n class samples, one binary classifier of classification training, is instructed altogether two-by-two
Practice n* (n-1)/2 grader, test sample is tested and voted one by one, who gets the most votes's classification is its classification
Label；
The problem of classifying for two classes, the thought of svm classifier is to find this so that the maximum hyperplane of two classification sample intervals,
Now, training data is (x1, y1), (x2, y2) ..., (xm, ym), y ∈ {+1, -1 }；The problem of asking for largest interval is exactly one
The problem of optimizing, the equation of optimal classification surface：
Wx+b=0
Wherein w and b are unknown, the optimization problem to be solved after conversion：
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
</mtd>
<mtd>
<mrow>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<mo>|</mo>
<mo>|</mo>
<mi>w</mi>
<mo>|</mo>
<msup>
<mo>|</mo>
<mn>2</mn>
</msup>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
So that meeting yi[(wxi)+b] -1 >=0 (i=1,2 ..., m)
M is number of samples, solves to obtain w and b, that is, obtains optimal classification surface；
For some in lower dimensional space linear inseparable situation, being mapped that to by kernel function in higher dimensional space becomes it
The problem of for linear separability；SVM kernel functional parameters are carried out from libsvm software kits to ask for, the step of using is：
1) data set is prepared according to the form required by LIBSVM software kits；
2) simple zoom operations are carried out to data；
3) consider to select RBF kernel functions；
4) using cross validation selection optimal parameter C and g；
5) acquisition supporting vector machine model is trained to whole training set using optimal parameter C and g；
6) tested and predicted using the model of acquisition.
7. subjective assessment and the spectroscopic data modeling of a kind of food taste and flavor, it is characterised in that including：Subjective assessment
Participate in evaluation and electing people's information collection module, spectra collection module, subjective assessment label input module, data memory module, data processing mould
Block, and subjective assessment Tag Estimation output module；
The subjective assessment is participated in evaluation and electing people's information collection module, for collecting participate in evaluation and electing people's information, including sex Xa, age Xb, native place Xc
With national Xd；
The spectra collection module, the reflection infrared spectrum rate derivative data for obtaining food, by spectrum sensor, and
Specific light source environment, collects the reflectance spectrum of food, obtains the primary reflection rate of corresponding wavelength, then obtain by inserting calculation of group dividing
Derivative data；
The subjective assessment label input module, quantifies to believe for obtaining evaluation of the people to this kind of food taste and flavor of participating in evaluation and electing
Breath, including food hot/cold, soft or hard degree, and the taste and smell of food score；
The data memory module, for storing training set data, and algorithm model parameter；
The data processing module, for using algorithm model criterion, to new participate in evaluation and electing people and spectroscopic data, predicts subjective assessment
Label；
The subjective assessment Tag Estimation output module, for showing subjective assessment label, and the master that will be revised by the people that participates in evaluation and electing
See and evaluate label and correspondingly participate in evaluation and electing people's information and spectroscopic data is exported to data memory module, stored as training set data, it is real
The accumulation of existing data model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710340311.6A CN107274065A (en) | 2017-05-15 | 2017-05-15 | The subjective assessment of food taste and flavor and spectroscopic data modeling method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710340311.6A CN107274065A (en) | 2017-05-15 | 2017-05-15 | The subjective assessment of food taste and flavor and spectroscopic data modeling method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107274065A true CN107274065A (en) | 2017-10-20 |
Family
ID=60064098
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710340311.6A Pending CN107274065A (en) | 2017-05-15 | 2017-05-15 | The subjective assessment of food taste and flavor and spectroscopic data modeling method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107274065A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107944990A (en) * | 2017-12-29 | 2018-04-20 | 山大地纬软件股份有限公司 | A kind of integral counter-employee device and method of the precision push based on machine learning |
| CN108414471A (en) * | 2018-01-10 | 2018-08-17 | 浙江中烟工业有限责任公司 | A method of sense organ characterization information is differentiated based near infrared spectrum and sensory evaluating smoking's mutual information |
| CN110057975A (en) * | 2019-04-22 | 2019-07-26 | 广东工业大学 | A kind of olfactory descriptor acquiring method, system and relevant apparatus |
| CN111007176A (en) * | 2019-12-26 | 2020-04-14 | 冲之智能科技(天津)有限公司 | In-vehicle odor grade evaluation method based on gas chromatography and ion mobility spectrometry |
| CN113868835A (en) * | 2021-09-03 | 2021-12-31 | 厦门大学 | Modeling system based on rapid registration method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6978212B1 (en) * | 1999-11-01 | 2005-12-20 | Smiths Detection Inc. | System for portable sensing |
| CN1865943A (en) * | 2006-05-17 | 2006-11-22 | 江苏大学 | Visual detection method for food aroma and device therefor |
| CN101706488A (en) * | 2009-11-30 | 2010-05-12 | 内蒙古蒙牛乳业（集团）股份有限公司 | Method for establishing infant dairy product sensory evaluation expert system |
| CN104237420A (en) * | 2014-10-15 | 2014-12-24 | 云南中烟工业有限责任公司 | Method for building cigarette bitter taste index evaluation model based on taste activity value |
| CN104698044A (en) * | 2013-12-10 | 2015-06-10 | 上海闵临机电科技有限公司 | Electronic tongue for food detection |
| CN106501470A (en) * | 2016-11-23 | 2017-03-15 | 广东嘉豪食品有限公司 | Using gustatory system and the method for Electronic Nose association evaluation mustard chilli sauce local flavor grade |
-
2017
- 2017-05-15 CN CN201710340311.6A patent/CN107274065A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6978212B1 (en) * | 1999-11-01 | 2005-12-20 | Smiths Detection Inc. | System for portable sensing |
| CN1865943A (en) * | 2006-05-17 | 2006-11-22 | 江苏大学 | Visual detection method for food aroma and device therefor |
| CN101706488A (en) * | 2009-11-30 | 2010-05-12 | 内蒙古蒙牛乳业（集团）股份有限公司 | Method for establishing infant dairy product sensory evaluation expert system |
| CN104698044A (en) * | 2013-12-10 | 2015-06-10 | 上海闵临机电科技有限公司 | Electronic tongue for food detection |
| CN104237420A (en) * | 2014-10-15 | 2014-12-24 | 云南中烟工业有限责任公司 | Method for building cigarette bitter taste index evaluation model based on taste activity value |
| CN106501470A (en) * | 2016-11-23 | 2017-03-15 | 广东嘉豪食品有限公司 | Using gustatory system and the method for Electronic Nose association evaluation mustard chilli sauce local flavor grade |
Non-Patent Citations (3)
| Title |
|---|
| 侯媛彬、杜京义、汪梅: "《神经网络》", 31 August 2007, 西安电子科技大学出版社 * |
| 姜安、彭江涛、彭思龙、魏纪平、李长文: "酒香型光谱分析和模式识别计算分析", 《光谱学与光谱分析》 * |
| 戴琼海、张晶、李菲菲、范静涛: "光谱数据的特征挖掘降维方法", 《数据采集与处理》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107944990A (en) * | 2017-12-29 | 2018-04-20 | 山大地纬软件股份有限公司 | A kind of integral counter-employee device and method of the precision push based on machine learning |
| CN108414471A (en) * | 2018-01-10 | 2018-08-17 | 浙江中烟工业有限责任公司 | A method of sense organ characterization information is differentiated based near infrared spectrum and sensory evaluating smoking's mutual information |
| CN108414471B (en) * | 2018-01-10 | 2020-07-17 | 浙江中烟工业有限责任公司 | Method for distinguishing sensory characterization information based on near infrared spectrum and sensory evaluation mutual information |
| CN110057975A (en) * | 2019-04-22 | 2019-07-26 | 广东工业大学 | A kind of olfactory descriptor acquiring method, system and relevant apparatus |
| CN111007176A (en) * | 2019-12-26 | 2020-04-14 | 冲之智能科技(天津)有限公司 | In-vehicle odor grade evaluation method based on gas chromatography and ion mobility spectrometry |
| CN113868835A (en) * | 2021-09-03 | 2021-12-31 | 厦门大学 | Modeling system based on rapid registration method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107274065A (en) | The subjective assessment of food taste and flavor and spectroscopic data modeling method and system | |
| Yu et al. | Data mining-assisted short-term wind speed forecasting by wavelet packet decomposition and Elman neural network | |
| Angelov et al. | Autonomous learning multimodel systems from data streams | |
| Wu et al. | Evolving RBF neural networks for rainfall prediction using hybrid particle swarm optimization and genetic algorithm | |
| Qiu et al. | An integrated framework with feature selection for dropout prediction in massive open online courses | |
| CN107451101A (en) | It is a kind of to be layered integrated Gaussian process recurrence soft-measuring modeling method | |
| Xie et al. | Blockchain financial investment based on deep learning network algorithm | |
| CN108171209A (en) | A kind of face age estimation method that metric learning is carried out based on convolutional neural networks | |
| Nieto et al. | Support vector machines and neural networks used to evaluate paper manufactured using Eucalyptus globulus | |
| CN110990718B (en) | Social network model building module of company image lifting system | |
| CN102930163A (en) | Method for judging 2 type diabetes mellitus risk state | |
| CN110390565A (en) | The method and system of intelligent gateway adaptive management are realized by AI edge calculations | |
| CN103150611A (en) | Hierarchical prediction method of II type diabetes mellitus incidence probability | |
| CN103714261A (en) | Intelligent auxiliary medical treatment decision supporting method of two-stage mixed model | |
| CN112529638B (en) | Service demand dynamic prediction method and system based on user classification and deep learning | |
| Lin et al. | Intelligent decision support for new product development: a consumer-oriented approach | |
| CN106611016B (en) | A kind of image search method based on decomposable word packet model | |
| CN116959725A (en) | Disease risk prediction method based on multi-mode data fusion | |
| Lee et al. | Hidden markov models for forex trends prediction | |
| CN115796936A (en) | Cigarette sales prediction method and system based on combined model and storage medium | |
| Lauren et al. | A low-dimensional vector representation for words using an extreme learning machine | |
| CN117314006A (en) | Intelligent data analysis method and system | |
| Kai et al. | Notice of Retraction: A Novel Forecasting Model of Fuzzy Time Series Based on K-means Clustering | |
| CN113837266B (en) | Software defect prediction method based on feature extraction and Stacking ensemble learning | |
| Mao et al. | Naive Bayesian algorithm classification model with local attribute weighted based on KNN |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20180816 Address after: 100176 Chaoyang District, Beijing, Xiaohong gate Township, South Fourth Ring Road, No. 308, Cheng Shou Si Road, four floor, 2733 Applicant after: Beijing beye Technology Co., Ltd. Address before: 100020 Beijing Chaoyang District waterfront South Street 8, moist Pavilion 364-1-701 Applicant before: Cui Zhe |
|
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171020 |