CN113238004A - Method for predicting sour taste and sweet taste by using MLP neural network model - Google Patents
Method for predicting sour taste and sweet taste by using MLP neural network model Download PDFInfo
- Publication number
- CN113238004A CN113238004A CN202110507386.5A CN202110507386A CN113238004A CN 113238004 A CN113238004 A CN 113238004A CN 202110507386 A CN202110507386 A CN 202110507386A CN 113238004 A CN113238004 A CN 113238004A
- Authority
- CN
- China
- Prior art keywords
- sample
- taste
- neural network
- sour
- sweet
- 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
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0001—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00 by organoleptic means
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Pathology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Food Science & Technology (AREA)
- Medicinal Chemistry (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Immunology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Seasonings (AREA)
Abstract
The invention discloses a method for predicting sourness and sweetness by using an MLP neural network model, which comprises the following steps: (1) scoring the sweet and sour taste strengths of the N samples according to an artificial sensory method for evaluating the strength of a 10-point scale in international standard ISO4121:2003 to obtain the sour and sweet taste strength values of different samples; (2) analyzing the taste profile of the sample by using an electronic tongue technology to obtain sweet and sour response values of different samples; (3) taking the sweet and sour response value obtained by the electronic tongue technology in the step (2) as input, taking the sweet and sour intensity value obtained by the artificial sensory method in the step (1) as output, and establishing an MLP neural network prediction model; (4) and predicting the sweet and sour intensity value based on an MLP neural network prediction model. The invention establishes a nonlinear MLP neural network prediction model by taking the response value of the electronic tongue sensor as input and the taste sense intensity as output, can realize quick and accurate prediction of the taste intensity of food, and perfects a quick taste evaluation system.
Description
Technical Field
The invention relates to the technical field of taste prediction methods, in particular to a method for predicting sour taste and sweet taste by using an MLP neural network model.
Background
The traditional evaluation of food taste and flavor is mainly finished by an evaluator through artificial sense, and the taste and flavor characteristics of food can be evaluated more three-dimensionally and abundantly. However, the time consumption of artificial sense is long, the evaluation cost is high, and the sense evaluation result is easily influenced by factors such as personal habits, health conditions and ages of the sensers. The electronic tongue is a sensing technology which can objectively identify and quantify food taste, and has the advantages of rapidness, stability and low fatigue. But compared with artificial sense organs, the electronic tongue has a single characterization effect on the taste sense of the food. Not asking a question here, whether there is a taste evaluation method that can take into account the stereo richness of artificial senses and the rapidity and stability of the electronic tongue? The neural network aims at simulating the structure and the function of the neural system to process data, and continuously adjusting the weight of chains among simulated neurons so that the whole network can better fit the relationship of training data. A nonlinear regression model is established by using a Multi-layer perceptron (MLP) neural network, so that the food types can be directionally distinguished and the sensory attributes of the food can be predicted. Therefore, based on electronic tongue analysis and artificial sensory evaluation, a prediction model of the sour taste and the sweet taste intensity of the food is established by utilizing the MLP neural network, and a fast and efficient method is provided for predicting the sour taste and the sweet taste intensity of various foods.
Disclosure of Invention
In order to solve the above problems, the present invention provides a method for predicting sourness and sweetness using an MLP neural network model.
In order to achieve the purpose, the technical scheme provided by the invention is as follows:
a method for predicting sourness and sweetness using an MLP neural network model, comprising the steps of:
(1) according to international standard ISO4121:2003, the sweetness and sourness strength of N samples are scored by a 10-point scale strength evaluation artificial sensory method to obtain sourness and sweetness strength values of different samples;
(2) analyzing the taste profile of the sample by using an electronic tongue technology to obtain the sour and sweet response values of different samples;
(3) taking the sour and sweet response values obtained by the electronic tongue technology in the step (2) as input, taking the sour and sweet strength values obtained by the artificial sensory method in the step (1) as output, and establishing an MLP neural network prediction model;
(4) and predicting the sensory sour and sweet taste intensity values of the food based on an MLP neural network prediction model.
Preferably, the sample is a food sample, a method for predicting the sourness and sweetness of a food using an MLP neural network model.
Preferably, N is equal to or greater than 50.
Preferably, in the step (1), a sensory evaluation group consisting of more than 14 sensory evaluators is established through the steps of recruitment, screening and training; sensory evaluation comprises the processes of early stage recruitment, screening and training of panelists. First, the panelists learn to discriminate and evaluate the basic tastes of both the sour and sweet criteria. Next, the sensitivity of the panelists to the above low concentrations of the two basic flavor substances was tested using the 3-AFC method, and the panelists were screened.
The screening steps are as follows: screening was performed according to international standard ISO 13301:2002 using triple point sorting (3-AFC) with one flavor as a group, one group comprising two pure water samples and one flavor sample, wherein the flavor sample comprises either a sour sample or a sweet sample; randomly presenting 6 groups of samples, each taste repeated three times, to a candidate evaluator in portions, asking them to select one sample with sensory differences from each group of three samples, and matching the difference sample with the taste; all samples were presented in a three-digit random number coded format; screening candidate evaluators with the accuracy reaching 100% and all correct matching rates; finally, selecting at least 7 females and at least 7 males, and requiring sensory evaluation experience;
these panelists were subjected to sensory training to become familiar with the sensory scale. And 6 training evaluation samples are taken in a course with 15 sections and 1 hour section each time until each evaluator can stably and consistently evaluate the samples by using descriptors and scales, and then formal evaluation is carried out on the test samples.
Wherein sensory samples are presented to the evaluator when the temperature of the samples is maintained at room temperature (25 ℃) on the day of preparation completion 2-3h before sensory testing.
The specific training steps are as follows: the sour taste or the sweet taste intensity of different samples is scored by taking 10 citric acid and sucrose with different concentrations as reference substances, wherein the corresponding sour taste or sweet taste intensity is 0-10 minutes.
In the step (1), according to the international standard ISO4121:2003, the step of scoring the sweetness and sourness intensity of N samples by using a 10-point scale intensity evaluation artificial sensory method comprises the following steps: the sensory evaluation personnel drink 10mL of sample into the mouth for 10s and spit the sample out, and then clean the oral cavity with ultrapure water to prevent aftertaste; the sensory evaluator needed to rest for at least 30s before tasting the next sample. All sensory evaluators need to use standardized methods and remember consistency of taste intensity; each panelist was asked to individually examine the samples in turn in a predetermined order and to score the flavor intensity of the samples on a 10-point scale.
Preferably, the sour sample is 0.2-0.3g/L citric acid; the sweet taste sample is sucrose at 13-15 g/L.
Preferably, the samples in step (2) are prepared differently according to their water-soluble properties. Classifying the samples in the step (2) according to different water solubility, and dissolving the completely water-soluble samples by distilled water at 25 ℃ to prepare test samples; and dissolving part of the water-soluble sample by distilled water at 60 ℃, centrifuging and taking supernatant to prepare a test sample.
Wherein the sample is divided into a completely water-soluble sample and a partially water-soluble sample. Examples of completely water-soluble samples include: fructus Broussonetiae extract, radix Ophiopogonis extract, radix scrophulariae extract, fructus Eriobotryae extract, Notoginseng radix substrate, Notoginseng radix extract, mume fructus spice, Fortunella margarita 1#, mume fructus liquor-4, fruit extract, fructus Siraitiae Grosvenorii extract, fructus Phyllanthi extract, mume fructus carbonized and fermented 3, flos Caryophylli extract, fructus Rosae Normalis extract, plateau folium gallus Domesticus Brisson extract, okra plant extract, fructus Chaenomelis extract, mume fructus fermented extract, dried fructus mume extract, cortex Pruni mume, fructus Punicae Granati extract, fructus Citri Sarcodactylis extract, herba Pileae Scriptae extract, fructus crataegi extract, fructus Jujubae spice, tree moss absolute oil, fructus Anisi Stellati extract, pericarpium Citri Tangerinae (5 years) extract, fructus anethi seed extract, fructus Zanthoxyli extract, Curcuma rhizome extract, coffee essence, sweet root extract, salted green lemon, wild plum, salted plum, compound essence with Notoginseng radix taste, compound essence with mume fructus mume taste, fructus Citri Tangerinae extract with fructus Pruni taste, fructus mume fructus extract, fructus mume extract with fructus mume taste, fructus mume extract with fructus mume taste, fructus mume extract with fructus mume extract and fructus mume taste, fructus mume extract with fructus mume taste and fructus extract with fructus mume taste, fructus mume extract with fructus mume taste and fructus extract with fructus mume taste, The total of 47 samples include compound essence with the taste of sugarcoated haws, L-malic acid, glycyrrhizin, maili sweet, high fructose syrup, sucralose, compound amino acid and the like. The method for processing the completely water-soluble sample comprises the following steps: the sample was prepared into a 5% solution by mass with distilled water at 25 ℃.
The partially water-soluble samples included: the main oil of fructus Broussonetiae, fructus Broussonetiae essential oil, lemon oil, cumquat essential oil, cedar essential oil (middle and first half), black tea, sweet osmanthus spice, pungent litse fruit oil, ice taste compound essence, neotame, IMP-GMP, cooling agent-23, cooling agent-3 and the like, and the total 13 types are provided. The method for processing the part of water-soluble sample comprises the following steps: mixing the sample with 60 ℃ distilled water according to the mass ratio of 50:50, centrifuging at the rotating speed of 6000rpm for 30 minutes, cooling to room temperature after the centrifugation is finished, and diluting the water layer by 20 times.
Preferably, in step (2), the electronic tongue takes 30s of time for each sample, and each sample comprises not less than 5 replicates and 3 repeats.
Preferably, in step (3), the MLP neural network model establishing step is:
selecting data of electronic tongue response values (acidity is greater than-13, sweetness is greater than 0), and dividing a sample data set into three sets, namely a training set, a testing set and a verification set, wherein the data amount in the three sets, namely the training set, the testing set and the verification set accounts for 70%, 30% and 0% of the total number of samples respectively; data were selected for electronic tongue response values of sour greater than-13 and sweet greater than 0, since these two values are electronic tongue threshold values for both tastes, above which a human is defined as perceptible, below which it is not perceptible. To prevent overfitting, the samples need to be split in the neural network, typically according to 7: 3 or 4: 3: 3, here SPSS default 7: 3.
the data in the training set is used for training the MLP neural network, the data in the testing set is used for monitoring errors in the training process to prevent over-training, and the data in the verifying set is used for evaluating the accuracy of the MLP neural network obtained by training.
Preferably, in the step (4), the food to be predicted is used as a prediction set sample, the prediction set sample electronic tongue data is input into the MLP neural network prediction model established in the step (3), the corresponding sour and sweet strength values are predicted, and the prediction condition of the MLP neural network prediction model is analyzed.
In general, after determining a good taste evaluator, a method for predicting sour taste and sweet taste by using an MLP neural network model mainly comprises the following steps:
in the step (1), the sensory method comprises the following steps: the sensory evaluation person took 10mL of the sample into the mouth for 10 seconds and spit it off, and then rinsed the mouth with ultrapure water to prevent aftertaste. The sensory evaluator needed to rest for at least 30s before tasting the next sample. All organoleptic evaluators need to use standardized methods and remember consistency of taste intensity. Each panelist was asked to individually examine the samples in turn in a predetermined order and to score the flavor intensity of the samples on a 10-point scale.
In the step (2), the electronic tongue collects each sample for 30s, and each sample comprises not less than 5 parallels and 3 repeats.
In the step (3), the method specifically comprises the following steps: selecting data of electronic tongue response values (acid is greater than-13, sweet is greater than 0), dividing a sample data set into a training set, a testing set and a verification set, wherein the data in the training set is used for training the MLP neural network, the data in the testing set is used for monitoring errors in the training process to prevent over-training, and the data in the verification set is used for evaluating the accuracy of the MLP neural network obtained by training.
The method comprises the steps of taking an artificial sensory evaluation score as output and an electronic tongue response value as input, importing related data into SPSS 26 software, taking the electronic tongue response value as independent variable input and the artificial sensory score as dependent variable output, selecting a multilayer radial basis function model, taking a hyperbolic tangent function as an activation function for a hidden layer, taking 1 neuron of the hidden layer, taking an identity equation as the activation function for an output layer, taking 1 neuron of the output layer, taking 70% of data in a test sample as a training set and taking 30% of data as a test set, and obtaining the MLP neural network model.
And inputting the prediction set sample electronic tongue data into a multilayer perceptron model to predict corresponding taste intensity evaluation, and analyzing the model prediction condition.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention relates to a method for predicting sour taste and sweet taste by using an MLP neural network model, and the taste of the traditional food is mainly finished by a sensory evaluator, so that the reaction mechanism is more complex. The perception of food taste is a complex physiological process, and there is not necessarily a linear relationship between the amount of the component and the sensory evaluation of the taste. The invention establishes a nonlinear MLP neural network prediction model by taking the response value of the electronic tongue sensor as input and the taste sense intensity as output, can realize quick and accurate prediction of the taste intensity of food, and perfects a quick taste evaluation system.
2. The prediction method can evaluate a plurality of characteristics of a plurality of samples at the same time, and is widely applied.
3. The sour taste and the sweet taste of each raw material are obtained by an intensity evaluation method of sensory evaluation. The strength evaluation sensory method is to evaluate the characteristics of the sample on a numerical scale by using a preset standard as an evaluation standard to obtain the score of the sample response sensory. The result is more standard, so that the accuracy of the whole prediction result is ensured.
4. As can be seen from the verification results of the examples, the prediction relative error of the invention is within 30% and accounts for about 59% of the total number of samples, and the prediction accuracy is high whether the sour samples or the sweet samples are used.
Drawings
FIG. 1 is a flow chart of a method for predicting sour and sweet tastes by using an MLP neural network model in the invention.
FIG. 2 is the prediction of the sourness intensity of a sample by an electronic tongue response value based on an MLP neural network model in the present invention.
FIG. 3 is the prediction of sweetness intensity of samples by electronic tongue response value based on MLP neural network model in the present invention.
Detailed Description
The present invention will be described in further detail with reference to examples.
All sensory evaluation tests were carried out in a constant temperature (25 ℃) sensory laboratory equipped with a ventilation system and a white light fixture. The temperature of the sample and the mouthwash (ultrapure water) was controlled around 25 ℃. During the experiment, sensory evaluators used a nose clip to protect against the effects of odors, samples were presented in 30mL tasteless taste cups and encoded with 3 random numbers.
The sensory evaluation person took 10mL of the sample into the mouth for 10 seconds and spit it off, and then rinsed the mouth with ultrapure water to prevent aftertaste. The sensory evaluator needed to rest for at least 30s before tasting the next sample. All organoleptic evaluators need to use standardized methods and remember consistency of taste intensity. Each panelist was asked to individually examine the samples in turn in a predetermined order and to score the flavor intensity of the samples on a 10-point scale.
Wherein, the 10-point scale (1-10 minutes) is as follows: sweet (sucrose, g/L): 10. 30, 60, 90, 120, 150, 180, 210, 240, 270; acid (citric acid, g/L): 0.01, 0.05, 0.1, 0.2, 0.4, 0.6, 0.8, 1,2, 3. And (4) respectively scoring the acidity and the sweetness of the sample by a grade evaluator, wherein two points can be scored after the score reaches a decimal point, and finally, the score is averaged.
TABLE 1 sample sourness, sweetness artificial sensory intensity
Note: "-" indicates that the sensory person does not experience this taste sensation.
Step (2): and analyzing the taste profile of the sample by means of an electronic tongue technology to obtain response values of different sensors. The sample treatment is the same as the step (1).
Wherein the electronic tongue equipment is TS-5000Z type electronic tongue of Japan Instrument company, and the sensor types thereof are AE1 sensor (for detecting astringency), C00 sensor (for detecting bitterness), GL1 sensor (for detecting sweetness) and CT0 sensor (for detecting saltiness); AAE sensor (detect umami taste); CAO sensor (detect sour taste). To ensure the accuracy and stability of the test data, activation, initialization, calibration and diagnosis of the electronic tongue are required.
The method for preparing the solution in the sensor for activating the sensor comprises the following steps: 248.2g of potassium chloride was dissolved in distilled water to a constant volume of 1L, and then 10mg of silver chloride was added and stirred for 8 hours to prepare an internal solution. The preparation method of the reference solution comprises the following steps: 2.23g of potassium chloride and 0.045g of tartaric acid are dissolved in distilled water and then the volume is adjusted to 1L. The electrode solution preparation method comprises the following steps: adding water into 300mL of ethanol and 8.3mL of concentrated hydrochloric acid to fix the volume to 1L to prepare an anion solution; adding water into 7.46g of KCl, 300ml of ethanol and 0.56g of KOH for metering to 1L to prepare a cation solution. The sensor activation method comprises the following steps: taking out the artificial bimolecular membrane sensor, unscrewing the electrode, adding 200 mu L of internal solution, reloading the electrode, and placing the electrode in a reference solution for activation for 24h for later use. The electrode activation method comprises the following steps: the electrode was removed, 400. mu.L of the internal solution was added, the electrode was replaced, and the electrode was placed in 3.33mol/L KCl solution for activation for 24 hours.
The electronic tongue adopts an artificial bimolecular membrane sensor, and is prepared by adopting a 10mmol/L potassium chloride solution in order to eliminate the influence caused by the inconsistency of potential differences caused by different cleaning solutions and detection solvents. The electronic tongue test procedure is as follows: and (3) cleaning the two groups of reference solutions for 90s, cleaning the two groups of reference solutions for 120s respectively, enabling the sensor to return to zero at the equilibrium position for 30s, and starting the test for 30s after the equilibrium condition is reached. And (5) cyclic testing, and taking the average value of the data of the last 3 times as a test result.
TABLE 2 sample sour, sweet electronic tongue response values
And (3): selecting data of electronic tongue response values (acid is greater than-13, sweet is greater than 0), dividing a sample data set into a training set, a testing set and a verification set, wherein the data in the training set is used for training the MLP neural network, the data in the testing set is used for monitoring errors in the training process to prevent over-training, and the data in the verification set is used for evaluating the accuracy of the MLP neural network obtained by training.
The method comprises the steps of taking an artificial sensory evaluation score as output and an electronic tongue response value as input, importing related data into SPSS 26 software, taking the electronic tongue response value as independent variable input and the artificial sensory score as dependent variable output, selecting a multilayer radial basis function model, taking a hyperbolic tangent function as an activation function for a hidden layer, taking 1 neuron of the hidden layer, taking an identity equation as the activation function for an output layer, taking 1 neuron of the output layer, taking 70% of data in a test sample as a training set and taking 30% of data as a test set, and obtaining the MLP neural network model.
And (4) inputting the prediction set sample electronic tongue data into a multilayer perceptron model to predict corresponding taste intensity evaluation, and analyzing the model prediction condition.
TABLE 3 sample sour sensory value prediction with MLP neural network model
TABLE 4 sample sweetness sensory value prediction with MLP neural network model
And inputting the sample electronic tongue data of the prediction set into a multi-layer sensor model to predict corresponding taste intensity evaluation, and comparing the actual value with the prediction result as shown in figures 1 and 2, so that the fitting effect of the prediction output curve and the actual curve is better and basically accords with the expected output change effect. Meanwhile, a relative error analysis model is used for predicting the situation (tables 3 and 4), wherein the relative error refers to a numerical value obtained by multiplying the ratio of an absolute error caused by measurement to an actual value by 100%, and is expressed by percentage, and the relative error can reflect the credibility degree of the model better. For the sour neural network prediction model, the prediction relative error of 19 samples in 45 samples is within 30% and accounts for about 42% of the total samples. For the sweet neural network prediction model, the prediction relative error of 22 samples in 37 selected samples is within 30 percent and accounts for about 59 percent of the total number of samples, and the prediction accuracy is high.
Claims (10)
1. A method for predicting sourness and sweetness using an MLP neural network model, comprising the steps of:
(1) according to international standard ISO4121:2003, the sweetness and sourness strength of N samples are scored by a 10-point scale strength evaluation artificial sensory method to obtain sourness and sweetness strength values of different samples;
(2) analyzing the taste profile of the sample by using an electronic tongue technology to obtain the sour and sweet response values of different samples;
(3) taking the sour and sweet response values obtained by the electronic tongue technology in the step (2) as input, taking the sour and sweet strength values obtained by the artificial sensory method in the step (1) as output, and establishing an MLP neural network prediction model;
(4) and predicting the sensory sour and sweet taste intensity values of the food based on an MLP neural network prediction model.
2. The method for predicting sourness and sweetness using the MLP neural network model of claim 1, wherein the sample is a food sample.
3. The method for predicting tartness and sweetness using an MLP neural network model of claim 1, wherein N is greater than or equal to 50.
4. The method for predicting sourness and sweetness using an MLP neural network model according to claim 1,
in the step (1), a sensory evaluation group consisting of more than 14 sensory evaluators is established through the steps of screening and training;
the screening steps are as follows: screening was performed according to international standard ISO 13301:2002 using triple point sorting (3-AFC) with one flavor as a group, one group comprising two pure water samples and one flavor sample, wherein the flavor sample comprises either a sour sample or a sweet sample; randomly presenting 6 groups of samples, each taste repeated three times, to a candidate evaluator in portions, asking them to select one sample with sensory differences from each group of three samples, and matching the difference sample with the taste; all samples were presented in a three-digit random number coded format; screening candidate evaluators with the accuracy reaching 100% and all correct matching rates; finally, selecting at least 7 females and at least 7 males, and requiring sensory evaluation experience;
the training steps are as follows: the sour taste or the sweet taste intensity of different samples is scored by taking 10 citric acid and sucrose with different concentrations as reference substances, wherein the corresponding sour taste or sweet taste intensity is 0-10 minutes.
5. The method for predicting sour and sweet tastes by using the MLP neural network model, as set forth in claim 1, wherein in the step (1), the step of scoring the sweet and sour tastes of the N samples by the artificial sensory evaluation method using the 10-point scale strength according to the international standard ISO4121:2003 comprises the steps of: the sensory evaluation personnel drink 10mL of sample into the mouth for 10s and spit the sample out, and then clean the oral cavity with ultrapure water to prevent aftertaste; the sensory evaluator needed to rest for at least 30s before tasting the next sample. All sensory evaluators need to use standardized methods and remember consistency of taste intensity; each panelist was asked to individually examine the samples in turn in a predetermined order and to score the flavor intensity of the samples on a 10-point scale.
6. The method for predicting sourness and sweetness using the MLP neural network model as set forth in claim 1, wherein the sourness sample is 0.2-0.3g/L of citric acid and the sweetness sample is 13-15g/L of sucrose.
7. The method for predicting sour taste and sweet taste using the MLP neural network model according to claim 1, wherein the sample of step (2) is classified according to its water solubility, and a test sample is prepared after the completely water-soluble sample is dissolved in distilled water at 25 ℃; and dissolving part of the water-soluble sample by distilled water at 60 ℃, centrifuging and taking supernatant to prepare a test sample.
8. The method for predicting sour taste and sweet taste using the MLP neural network model of claim 1, wherein in step (2), the electronic tongue takes 30s per sample, and each sample comprises not less than 5 replicates and 3 replicates.
9. The method for predicting sour taste and sweet taste using the MLP neural network model according to claim 1, wherein in the step (3), the MLP neural network model establishing step is:
selecting data with an electronic tongue response value of more than-13 and sweetness of more than 0, and dividing a sample data set into three sets, namely a training set, a testing set and a verification set, wherein the proportion of the data amount in the three sets, namely the training set, the testing set and the verification set, to the total sample amount is 70%, 30% and 0 respectively;
the data in the training set is used for training the MLP neural network, the data in the testing set is used for monitoring errors in the training process to prevent over-training, and the data in the verifying set is used for evaluating the accuracy of the MLP neural network obtained by training.
10. The method for predicting sour taste and sweet taste using the MLP neural network model according to claim 1, wherein in step (4), the food to be predicted is used as a prediction set sample, the prediction set sample electronic tongue data is inputted into the MLP neural network prediction model established in step (3), the corresponding sour taste and sweet taste intensity values are predicted, and the prediction condition of the MLP neural network prediction model is analyzed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110507386.5A CN113238004A (en) | 2021-05-10 | 2021-05-10 | Method for predicting sour taste and sweet taste by using MLP neural network model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110507386.5A CN113238004A (en) | 2021-05-10 | 2021-05-10 | Method for predicting sour taste and sweet taste by using MLP neural network model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113238004A true CN113238004A (en) | 2021-08-10 |
Family
ID=77133050
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110507386.5A Pending CN113238004A (en) | 2021-05-10 | 2021-05-10 | Method for predicting sour taste and sweet taste by using MLP neural network model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113238004A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114839297A (en) * | 2022-05-17 | 2022-08-02 | 云南中烟工业有限责任公司 | Evaluation method of cigarette taste sweet and sour balance characteristics |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120143804A1 (en) * | 2009-07-23 | 2012-06-07 | Yada Research And Development Co., Ltd. | Predicting odor pleasantness with an electronic nose |
| 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 |
| CN110823966A (en) * | 2019-12-10 | 2020-02-21 | 山东工商学院 | Grape wine SO based on electronic nose2Method for rapidly measuring concentration |
| CN111521641A (en) * | 2020-01-15 | 2020-08-11 | 新疆仪尔高新农业开发有限公司 | Method for identifying different distillation modes of brandy raw wine |
| US20200256837A1 (en) * | 2017-10-31 | 2020-08-13 | East China University Of Science And Technology | Electronic nose instrument for sensory quality evaluation of tobacco and tobacco product |
| CN112348232A (en) * | 2020-10-15 | 2021-02-09 | 上海交通大学 | Method for predicting sensory delicate flavor intensity of food based on BP neural network |
-
2021
- 2021-05-10 CN CN202110507386.5A patent/CN113238004A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120143804A1 (en) * | 2009-07-23 | 2012-06-07 | Yada Research And Development Co., Ltd. | Predicting odor pleasantness with an electronic nose |
| 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 |
| US20200256837A1 (en) * | 2017-10-31 | 2020-08-13 | East China University Of Science And Technology | Electronic nose instrument for sensory quality evaluation of tobacco and tobacco product |
| CN110823966A (en) * | 2019-12-10 | 2020-02-21 | 山东工商学院 | Grape wine SO based on electronic nose2Method for rapidly measuring concentration |
| CN111521641A (en) * | 2020-01-15 | 2020-08-11 | 新疆仪尔高新农业开发有限公司 | Method for identifying different distillation modes of brandy raw wine |
| CN112348232A (en) * | 2020-10-15 | 2021-02-09 | 上海交通大学 | Method for predicting sensory delicate flavor intensity of food based on BP neural network |
Non-Patent Citations (1)
| Title |
|---|
| 巩效伟等: "基于电子舌的电子烟甜度评价模型研究", 《中国烟草学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114839297A (en) * | 2022-05-17 | 2022-08-02 | 云南中烟工业有限责任公司 | Evaluation method of cigarette taste sweet and sour balance characteristics |
| CN114839297B (en) * | 2022-05-17 | 2024-01-26 | 云南中烟工业有限责任公司 | Evaluation method for sweet and sour balance characteristics of taste of cigarettes |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Quality analysis, classification, and authentication of liquid foods by near-infrared spectroscopy: A review of recent research developments | |
| Kiani et al. | Fusion of artificial senses as a robust approach to food quality assessment | |
| Gunness et al. | Sensory analysis of individual strawberry fruit and comparison with instrumental analysis | |
| CN103018293B (en) | A kind of electronic tongues detects the method for food liquid four kinds of basic taste sensation intensity fast | |
| Wei et al. | Tracing floral and geographical origins of honeys by potentiometric and voltammetric electronic tongue | |
| Marx et al. | Quantification of table olives' acid, bitter and salty tastes using potentiometric electronic tongue fingerprints | |
| CN103837587A (en) | Method for quickly evaluating taste of bayberry juice through electronic tongue system | |
| Liu et al. | A new comprehensive index for discriminating adulteration in bovine raw milk | |
| Gao et al. | Quality identification and evaluation of Pu‐erh teas of different grade levels and various ages through sensory evaluation and instrumental analysis | |
| Wang et al. | Evaluation of eight kinds of flavor enhancer of umami taste by an electronic tongue | |
| Davies et al. | A preliminary study on the characterisation of honey by near infrared spectroscopy | |
| CN102692488A (en) | Jinhua ham grading and identifying method based on electronic nose technology | |
| Pickering et al. | Thermal tasting and difference thresholds for prototypical tastes in wine | |
| Wang et al. | Electronic tongue for food sensory evaluation | |
| CN113238004A (en) | Method for predicting sour taste and sweet taste by using MLP neural network model | |
| Jurado-Campos et al. | Instrumental techniques to classify olive oils according to their quality | |
| CN107727711A (en) | A kind of evaluation method of beer wine body full weight | |
| CN110763806B (en) | Method for evaluating spicy grade of duck neck | |
| Han et al. | A novel quantitative prediction approach for astringency level of herbs based on an electronic tongue | |
| CN104062223A (en) | Determination method of citrus chewiness | |
| Li et al. | Application of multi‐element viscoelastic models to freshness evaluation of beef based on the viscoelasticity principle | |
| CN105738584A (en) | Method for judging excellent taste quality of late japonica rice | |
| Marjanska et al. | Combination of instrumental and qualitative descriptive analysis for evaluation of selected tonic waters quality features | |
| CN118033069A (en) | Method for measuring bitter and astringent composite taste intensity of liquid food | |
| CN110118809A (en) | Based on electronic nose ageing time Shanxi mature vinegar rapid detection method different from electronic tongues |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210810 |