Evaluation trend estimation device, evaluation trend estimation method, combined model generation method, and program

Abstract

To eliminate the need to measure brain waves when evaluating food and drink and to evaluate the food and drink with relatively high accuracy.SOLUTION: An evaluation tendency inference device includes a brain wave rhythm inference unit configured to receive an input of perception data indicating a perception of a person who has ingested food or drink with respect to the food or drink, and infer a brain wave rhythm that is a frequency spectrum of a brain wave of the person when the person is ingesting the food or drink, and an evaluation tendency inference unit configured to receive an input of the inferred brain wave rhythm, and infer a tendency in a plurality of persons regarding a subjective evaluation of the person who has ingested the food or drink with respect to the food or drink.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an evaluation tendency estimation device, an evaluation tendency estimation method, a combined model generation method, and a program. [Background technology]

[0002] One method for evaluating food and drink is to measure the electroencephalograms of people consuming the food and drink and use the results of the electroencephalogram to evaluate the palatability of the food and drink. Another method for evaluating food and drink is to use sensory evaluation of the food and drink to evaluate the palatability of the food and drink.

[0003] For example, Patent Document 1 describes a method for evaluating sensory (taste and smell) sensitivity using a hierarchical neural network. In this method, material attribute data, such as the amounts of components of a taste and smell ingredient, and biological information data, such as the power spectrum of the electroencephalogram of a subject who has tasted the taste and smell ingredient, are input in parallel, and a hierarchical neural network is trained to output food preference data, which represents the inertia of the subject who has tasted the taste and smell ingredient. Patent Document 1 also states that the material attribute data of a taste and smell ingredient may be used as an evaluation value for the taste ingredient, expressed using sensory evaluation terms, based on human sensations obtained by a predetermined sensory evaluation method.

[0004] In the method described in Patent Document 1, the trained hierarchical neural network is separated into a material attribute model based on variables of material attribute data and a bioinformation model based on variables of bioinformation data. In the method described in Patent Document 1, a biometric information model is used to determine the main factors of the biometric information data that are significantly related to the food preference data, and a map of the principal component loadings of the biometric information data relative to the food preference data is created, with the two elements of the biometric information data each on a coordinate axis. Furthermore, in the method described in Patent Document 1, a material attribute model is used to determine the main factors of material attribute data that are significantly related to food preference data. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2003-290179 Summary of the Invention [Problem to be solved by the invention]

[0006] In a method of measuring the brain waves of a person consuming food or drink and using the results of the brain wave measurement to evaluate the palatability of the food or drink, measuring a person's brain waves may be a burden for the person performing the brain wave measurement. Furthermore, in methods that use sensory evaluation of food and beverages to evaluate their palatability, the sensory evaluation is expressed subjectively, which may result in large individual differences in evaluation and lower accuracy in the evaluation of food and beverage palatability. It is preferable that there is no need to measure electroencephalograms when evaluating food and drink, and that food and drink can be evaluated with a relatively high degree of accuracy.

[0007] One example of the objective of the present disclosure is to provide an evaluation tendency estimation device, an evaluation tendency estimation method, a combined model generation method, and a program that do not require measuring electroencephalograms when evaluating food and beverages and can evaluate food and beverages with relatively high accuracy. [Means for solving the problem]

[0008] According to a first aspect of the present disclosure, an evaluation tendency estimation device includes an electroencephalogram (EEG) rhythm estimation unit that receives input of sensory data indicating the perception of a person who has consumed food or beverage toward the food or beverage, and estimates an EEG rhythm, which is the frequency spectrum of the person's EEG while consuming the food or beverage, and an evaluation tendency estimation unit that receives input of the estimated EEG rhythm, and estimates tendencies of multiple people in the subjective evaluations of the food or beverage by the people who have consumed the food or beverage.

[0009] According to a second aspect of the present disclosure, an evaluation tendency estimation method includes a computer receiving input of sensory data indicating a person's perception of a food or beverage after consuming the food or beverage, estimating an electroencephalogram rhythm, which is the frequency spectrum of the person's electroencephalograms while consuming the food or beverage, and receiving input of the estimated electroencephalogram rhythm, estimating a tendency for a plurality of people in the subjective evaluations of the food or beverage by the people who consumed the food or beverage.

[0010] According to a third aspect of the present disclosure, a method for generating a combined model includes: a computer using training data including perceptual data indicating a subject's perception of a food or drink that has been ingested, and electroencephalogram data indicating an electroencephalogram rhythm, which is a frequency spectrum of the subject's electroencephalograms while the subject is ingesting the food or drink, to train an electroencephalogram estimation model that receives input of perceptual data and outputs an estimated value of the electroencephalogram rhythm; for each of a plurality of subjects, using training data including the electroencephalogram data of the subject while ingesting the food or drink and subjective evaluation data, which is data indicating the subject's subjective evaluation of the food or drink, to train an evaluation tendency estimation model that receives input of the electroencephalogram rhythm and outputs an estimated value of the subjective evaluation of the food or drink; combining the trained electroencephalogram estimation model and the trained evaluation tendency estimation model so that the output of the electroencephalogram estimation model is input to the evaluation tendency estimation model, to generate a combined model that receives input of perceptual data and outputs an estimated value of tendencies in a plurality of people's subjective evaluation of the food or drink.

[0011] According to a fourth aspect of the present disclosure, the program causes a computer to receive input of sensory data indicating a person's perception of food or beverage after consuming the food or beverage, and estimate an electroencephalogram rhythm, which is the frequency spectrum of the person's electroencephalograms while consuming the food or beverage, and receive input of the estimated electroencephalogram rhythm, and estimate trends in the subjective evaluations of the food or beverage by the people who consumed the food or beverage across multiple people.

[0012] According to a fifth aspect of the present disclosure, a program causes a computer to execute the following steps: train an EEG rhythm estimation model that receives input of perceptual data and outputs an estimated EEG rhythm, using training data including perceptual data indicating a subject's perception of a food or drink after consuming the food or drink, and EEG rhythm data indicating an EEG rhythm, which is a frequency spectrum of the subject's electroencephalograms, while the subject is consuming the food or drink; train an evaluation tendency estimation model that receives input of an EEG rhythm and outputs an estimated value of a subjective evaluation of the food or drink, using training data for each of a plurality of subjects including the EEG rhythm data of the subject while consuming the food or drink and subjective evaluation data, which is data indicating the subject's subjective evaluation of the food or drink; and combine the trained EEG rhythm estimation model and the trained evaluation tendency estimation model so that the output of the EEG rhythm estimation model is input to the evaluation tendency estimation model, thereby generating a combined model that receives input of perceptual data and outputs an estimated value of tendencies in a plurality of people's subjective evaluation of the food or drink. [Effects of the Invention]

[0013] According to one aspect of the present disclosure, it is not necessary to measure electroencephalograms when evaluating food and drink, and it is expected that food and drink can be evaluated with relatively high accuracy. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a diagram illustrating an example of a configuration of an evaluation tendency estimation device according to an embodiment. [Figure 2] FIG. 10 is a diagram illustrating an example of sensory data according to an embodiment. [Figure 3] FIG. 10 is a diagram showing an example of a radar chart showing electroencephalogram rhythms and estimated values ​​of subjective assessment tendencies according to the embodiment. [Figure 4] FIG. 1 is a diagram illustrating an example of a binding model according to an embodiment. [Figure 5] 10 is a diagram showing an example of a processing procedure in which the evaluation tendency estimation device according to the embodiment estimates the evaluation tendency of a target food or drink. FIG. [Figure 6]FIG. 1 is a diagram illustrating an example of a configuration of an evaluation tendency estimation system according to an embodiment. [Figure 7] 1 is a diagram illustrating an example of a configuration of an evaluation tendency estimation device according to an embodiment. [Figure 8] 10 is a diagram showing an example of a processing procedure in which the evaluation tendency estimation device according to the embodiment estimates the evaluation tendency of a target food or drink. FIG. [Figure 9] FIG. 2 is a diagram showing a first example of the relationship between electroencephalogram data and sensory characteristics obtained in an experiment related to the embodiment. [Figure 10] FIG. 10 is a diagram showing a second example of the relationship between electroencephalogram data and sensory characteristics obtained in an experiment related to the embodiment. [Figure 11] FIG. 1 illustrates an example configuration of a computer according to at least one embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0015] The following describes embodiments of the present invention, but the following embodiments do not limit the scope of the invention as claimed. Furthermore, not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention.

[0016] Fig. 1 is a diagram showing an example of the configuration of an evaluation tendency estimation device according to an embodiment. In the configuration shown in Fig. 1, the evaluation tendency estimation device 100 includes a display unit 110, an operation input unit 120, a storage unit 130, and a processing unit 140. The processing unit 140 includes a class classification unit 141, an electroencephalogram rhythm estimation unit 142, and an evaluation tendency estimation unit 143.

[0017] The evaluation tendency estimation device 100 estimates the tendency of a plurality of individuals in terms of subjective evaluations of food and drink made by the individuals who have consumed the food and drink. In particular, the evaluation tendency estimation device 100 estimates the frequency spectrum of the electroencephalograms of the individuals when they are consuming the food and drink, based on the individuals' perceptions of the food and drink. Then, the evaluation tendency estimation device 100 estimates the tendency of a plurality of individuals in terms of subjective evaluations of the food and drink made by the individuals who have consumed the food and drink, based on the estimated frequency spectrum.

[0018] Estimation of evaluation trends across multiple people is also referred to as evaluation trend estimation. Foods and beverages that are the subject of subjective evaluation trend estimation are also referred to as target foods and beverages. The evaluation tendency estimation device 100 may be configured using a computer such as a personal computer (PC) or a workstation (WS).

[0019] The term "perception" here refers to the brain's awareness of stimuli sensed by the sensory organs. The term "sensation" here refers to stimuli sensed by the sensory organs. Data that indicates perception is also called "perceptual data." The sensory data may be obtained, for example, by a questionnaire given to people who have consumed the food or drink. The person in charge of generating the sensory data may refer to the questionnaire results and input the sensory data into the evaluation tendency estimation device 100. Alternatively, the person who has consumed the food or drink may input the sensory data into the evaluation tendency estimation device 100 themselves. Alternatively, the sensory data may be input into a device other than the evaluation tendency estimation device 100, such as by conducting a questionnaire for generating sensory data in the form of a mark sheet and using a mark sheet reading device to read the questionnaire results. The evaluation tendency estimation device 100 may then acquire sensory data from the other device. The sensory expression of the characteristics of food and drink can be considered a sensory evaluation. Sensory evaluation here is a method of analyzing the characteristics of an object using the five senses. The sensory expression of the characteristics of food and drink is also called the sensory expression of the characteristics of food and drink.

[0020] The frequency spectrum of an electroencephalogram is also called an electroencephalogram rhythm, and data indicating the electroencephalogram rhythm is also called electroencephalogram rhythm data. The word "subjective" here means that the way of looking at things (understanding), feeling (evaluation), or thinking (judgment) is based on or relies on a personal position, emotion, or experience. The subjective evaluation of food and drink here is an individual's evaluation of the food and drink.

[0021] Here, one possible method for estimating a subjective evaluation of a food or drink by a person who has consumed that food or drink is to directly estimate the person's subjective evaluation of the food or drink based on perceptual data that expresses the person's perception of the food or drink, without estimating the electroencephalogram rhythm.

[0022] However, with this method, since both the expression of perception and the subjective evaluation of food and drink are subjective expressions, it is thought that there is a large difference between individuals in the relationship between the expression of perception and the evaluation of food and drink.With this method, since there is a large difference between individuals in the relationship between the expression of perception and the evaluation of food and drink, it is thought that there is a low (poor) accuracy in estimating the subjective evaluation of food and drink.

[0023] For example, consider the case where the sweetness of a food or drink is used as the perception of the food or drink, and the preference for the food or drink (degree of like or dislike) is used as an example of the subjective evaluation of the food or drink. In this case, it is conceivable that the sweetness of the same food or drink will be expressed differently by different people due to individual differences in sensitivity to sweetness and individual differences in the sweetness standard when expressing sweetness quantitatively. Since there is variation in how people express the sweetness of food and drink due to individual differences, it is thought that the accuracy of estimation when estimating preference trends based on the expression of sweetness is low.

[0024] In contrast, the evaluation tendency estimation device 100 estimates an electroencephalogram rhythm and estimates the tendency of subjective evaluations of food and drink based on the estimated electroencephalogram rhythm. Since electroencephalogram rhythms can be acquired without the need for human expression, they are expected to have smaller individual differences than perceptual expressions. Because the electroencephalogram rhythms have smaller individual differences than perceptual expressions, it is expected that estimating the tendency of subjective evaluations of food and drink based on the electroencephalogram rhythm will have higher estimation accuracy than estimating the tendency of subjective evaluations of food and drink based on perceptual expressions.

[0025] The evaluation tendency estimation device 100 may estimate the electroencephalogram rhythm based on the perceptual expressions of a certain number of people, such as several hundred people. For example, the evaluation tendency estimation device 100 may calculate an average value of the perceptual expressions of a plurality of people toward food and drink, and estimate the electroencephalogram rhythm based on the calculated average value. Alternatively, the evaluation tendency estimation device 100 may estimate the electroencephalogram rhythm for each of a plurality of people based on the perceptual expressions of a food and drink, and calculate the average value of the estimated electroencephalogram rhythms. The evaluation tendency estimation device 100 may calculate the mode or median instead of the average value.

[0026] By having the evaluation tendency estimation device 100 estimate electroencephalogram rhythms based on the perceptual expressions of a certain number of people regarding food and drink, it is expected that the influence of individual differences in perceptual expressions on the estimated electroencephalogram rhythms will be relatively small, and it is expected that the evaluation tendency estimation device 100 will be able to estimate the tendency of subjective evaluations of food and drink with relatively high accuracy.

[0027] Alternatively, evaluation tendency estimation device 100 may estimate electroencephalogram rhythms and estimate the tendency of subjective evaluations of food and drink based on expressions of perception of food and drink made by the same people as those used in training a model that receives input of expressions of perception of food and drink and outputs estimated values ​​of electroencephalogram rhythms. In this case, the people who make expressions of perception of food and drink may be one or more people. Model learning can also be called model training.

[0028] As a result, the influence of individual differences in perceptual expression is small or absent when learning the model and when estimating the tendency of subjective evaluations of foods and drinks, and it is expected that evaluation tendency estimation device 100 will be able to estimate brain rhythms with relatively high accuracy. Because evaluation tendency estimation device 100 can estimate brain rhythms with relatively high accuracy, it is expected that it will be able to estimate the tendency of subjective evaluations of foods and drinks with relatively high accuracy.

[0029] Furthermore, subjective evaluations such as food preferences are direct and unique evaluations of the target food and beverage. In contrast, the evaluation tendency estimation device 100 estimates EEG rhythms and subjective evaluation trends for food and beverages using a model trained using training data acquired for not just one food but a variety of foods and beverages. The evaluation tendency estimation device 100 can indirectly and universally estimate EEG rhythms and subjective evaluation trends for food and beverages based on multiple foods and beverages. In this respect, the evaluation tendency estimation device 100 is expected to perform estimations with a relatively high degree of accuracy.

[0030] Estimating the subjective evaluation of food and drink can be considered as evaluating the food and drink. The estimated value of the subjective evaluation of food and drink can be used as the evaluation value of the food and drink. The estimation accuracy of the subjective evaluation of food and drink can be considered as the accuracy of the evaluation of the food and drink.

[0031] The display unit 110 has a display screen such as a liquid crystal panel or an organic EL panel, and displays various images. For example, the display unit 110 may display the estimated results of the tendency of subjective evaluations of food and drink. The operation input unit 120 includes input devices such as a keyboard and a mouse, and receives user operations. For example, the operation input unit 120 may receive input of an expression of perception of food and drink.

[0032] The storage unit 130 stores various data. For example, the storage unit 130 may store various models used when the evaluation tendency estimation device 100 estimates the tendency of subjective evaluations of foods and drinks. The storage unit 130 may be configured using a storage device included in the evaluation tendency estimation device 100.

[0033] The processing unit 140 performs various processes by controlling each unit of the evaluation tendency estimation device 100. The functions of the processing unit 140 may be performed by a CPU (Central Processing Unit) included in the evaluation tendency estimation device 100 reading and executing a program from the storage unit 130.

[0034] The classification unit 141 classifies foods and drinks into classes. The classification unit 141 may classify the sensory data. Classifying the sensory data into classes can be considered as classifying the foods and drinks whose perceptions are expressed by the sensory data. The classifying unit 141 may classify the food and drink or sensory data into classes of food and drink types, such as "anpan" (sweet bean bun) and "coffee."

[0035] The sensory data may be assigned with identification information indicating a class. The classification unit 141 may then classify the sensory data by reading the identification information. When generating the sensory data, an operator who generates the sensory data may assign identification information of the class of the food or drink used to generate the sensory data to the sensory data.

[0036] Alternatively, the classification unit 141 may classify the perceptual data based on the values ​​of the perceptual data (values ​​for each perceptual item indicated in the perceptual data). For example, the classification unit 141 may input the perceptual data to a classification model that receives the input of the perceptual data and outputs class identification information, thereby calculating the identification information of the class into which the perceptual data is to be classified.

[0037] This class classification model may be a model (machine learning model) that has been trained by machine learning using training data that is a combination of sensory data and identification information of the class (correct class) into which the sensory data is classified. In this case, machine learning can be performed by inputting the perceptual data contained in the training data into a classification model, and optimizing the learning parameter values ​​so that the likelihood of the input perceptual data being classified into the correct class is as high (large) as possible during classification by the classification model. The class classification model may be learned by the evaluation tendency estimation device 100. Alternatively, the class classification model may be learned by a device other than the evaluation tendency estimation device 100.

[0038] The machine learning model used as the classification model is not limited to a specific type. A known machine learning model, such as a perceptron, a regression model, or a support vector machine (SVM), may be used as the classification model. By using a known machine learning model as the classification model, a known algorithm can also be used as the machine learning algorithm, and in this respect, the learning of the classification model can be performed with a relatively small load.

[0039] The electroencephalogram (EEG) rhythm estimation unit 142 estimates the electroencephalogram (EEG) rhythm of a person consuming food or drink based on perceptual data indicating the person's perception of the food or drink. Specifically, the electroencephalogram (EEG) rhythm estimation unit 142 inputs the perceptual data into an electroencephalogram (EEG) rhythm estimation model to calculate the electroencephalogram. The electroencephalogram (EEG) rhythm estimation model is a model that receives input perceptual data and outputs an estimated value of the electroencephalogram rhythm.

[0040] Fig. 2 is a diagram showing an example of perceptual data. Fig. 2 shows an example of perceptual data for a specific bread. In this perceptual data, for each of N people (N is an integer greater than or equal to 1), the strength of the person's perception for each perceptual category, such as "chewy" or "fluffy," is shown as an integer value.

[0041] The number of subjects from which sensory data is acquired is not limited to a specific number, and may be one or more. As described above, by setting the number of subjects from which sensory data is acquired to a certain number or more, such as several hundred people, it is expected that the influence of individual differences in perceptual expression on the estimated electroencephalogram rhythm will be relatively small. As a result, it is expected that the evaluation tendency estimation device 100 will be able to estimate the tendency of subjective evaluations of food and drink with relatively high accuracy.

[0042] The perceptual items in the perceptual data are not limited to specific items, but can be various items depending on the input to the EEG rhythm estimation model. The perceptual items in the perceptual data can also be one or more in various numbers depending on the input to the EEG rhythm estimation model.

[0043] The electroencephalogram rhythm estimation section 142 may use an electroencephalogram rhythm estimation model provided for each type of food or drink. For example, an electroencephalogram rhythm estimation model may be provided for each class that is the target of classification by the classification section 141. The electroencephalogram rhythm estimation section 142 may then estimate the electroencephalogram rhythm using the electroencephalogram rhythm estimation model associated with the class selected by the classification by the classification section 141.

[0044] It is expected that the electroencephalogram rhythm estimation section 142 can estimate the electroencephalogram rhythm with a relatively high degree of accuracy by using sensory data in which sensory items corresponding to the type of food are set. In this case, the number of sensory items in the sensory data may differ for each type of food or drink.

[0045] The electroencephalogram rhythm estimation model is not limited to a specific type of model, and may be, for example, a model based on a mathematical formula such as Equation (1).

[0046]

number

[0047] θ i indicates a certain frequency component of the EEG in a certain part of the brain. i corresponds to the element of brain wave rhythm. i is an integer index, i=1, 2,..., that identifies the combination of brain region and frequency.

[0048] The brain regions and frequencies targeted by EEG rhythms are not limited to specific regions and frequencies. The number of combinations of brain regions and frequencies targeted by EEG rhythms is not limited to a specific number. For example, in the 10% method (extended 10-20 method), the combinations of brain regions and frequencies targeted by EEG rhythms may be, but are not limited to, three: a 4 Hz frequency component in FC1, a 4 Hz frequency component in FC2, and a 6 Hz frequency component in Pz.

[0049] The 4 Hz frequency component data in FC1 can be considered to be correlated with boredom. The 4 Hz frequency component data in FC2 can be considered to be correlated with attention. The 6 Hz frequency component data in Pz can be considered to be correlated with sensation.

[0050] x j indicates the value for each perceptual item in the perceptual data. j is an integer index, j=1, 2, . . . , M, that identifies the perceptual item. M is an integer greater than or equal to 1, indicating the number of perceptual items in the perceptual data.

[0051] β i,jis the perceived item value x j The weighting factor β i,j corresponds to the learning parameter. That is, the weighting coefficient β i,j The value of is determined by machine learning. The weighting coefficient β i,j The value of is determined for each type of food and drink, and the value according to the type of food and drink is used as the weighting coefficient β i,j This can be done by setting

[0052] The EEG rhythm estimation model can be trained using training data that combines perceptual data showing the perception of a person who has consumed the food or drink used in the training with measurement data of the EEG rhythm while the person is consuming the food or drink. In this case, the measurement data of the EEG rhythm corresponds to the correct answer data. In this case, machine learning can be performed by inputting the perceptual data contained in the training data into the EEG rhythm estimation model and optimizing the learning parameter values ​​so that the difference (estimation error) between the estimated EEG rhythm output by the EEG rhythm estimation model and the correct data is as small as possible. The electroencephalogram rhythm estimation model may be trained by the evaluation tendency estimation device 100. Alternatively, the electroencephalogram rhythm estimation model may be trained by a device other than the evaluation tendency estimation device 100.

[0053] The EEG rhythm estimation model may be a known machine learning model such as deep learning, perceptron, or regression model. By using a known machine learning model as the EEG rhythm estimation model, a known algorithm can also be used as the machine learning algorithm, and in this respect, the EEG rhythm estimation model can be trained with a relatively small load.

[0054] The evaluation tendency estimation unit 143 estimates the tendency of subjective evaluations of food and drink across multiple people based on the electroencephalogram rhythms. Specifically, the evaluation tendency estimation unit 143 inputs the electroencephalogram rhythms estimated by the electroencephalogram rhythm estimation unit 142 into an evaluation tendency estimation model to calculate a subjective evaluation value for food and drink. The evaluation tendency estimation model here is a model that has been trained using training data from multiple subjects and receives input of electroencephalogram rhythms to output a subjective evaluation value for food and drink. The subjective evaluation value output by the evaluation tendency estimation model can be considered as a tendency of subjective evaluation values ​​across multiple subjects who were the subject of the training data.

[0055] The evaluation tendency estimation unit 143 may use an evaluation tendency estimation model provided for each type of food and drink. For example, an evaluation tendency estimation model may be provided for each class that is the target of classification by the classification unit 141. The evaluation tendency estimation unit 143 may then estimate the tendency of subjective evaluation values ​​for food and drink using the evaluation tendency estimation model associated with the class selected by the classification by the classification unit 141. It is expected that the evaluation tendency estimation unit 143 can estimate the tendency of subjective evaluation values ​​for food and drink with a relatively high degree of accuracy by using an evaluation tendency estimation model according to the type of food.

[0056] The evaluation tendency estimation model is not limited to a specific type of model, and may be, for example, a model based on a mathematical formula such as Equation (2).

[0057]

number

[0058] y k indicates the evaluation value for each item of food and drink evaluation. k is an integer index where k=1, 2, . . . that identifies the food and drink evaluation item.

[0059] The evaluation items to be estimated by the evaluation tendency estimation unit 143 are not limited to specific items. For example, the evaluation items to be estimated by the evaluation tendency estimation unit 143 may include, but are not limited to, the degree of deliciousness, the degree of boredom, and the likelihood of becoming full, or some of these items.

[0060] The evaluation of a food or drink may be an evaluation made while the food or drink is being consumed. Furthermore, the evaluation items to be estimated by the evaluation tendency estimation unit 143 may include items that represent changes in evaluation over time. For example, the evaluation items to be estimated by the evaluation tendency estimation unit 143 may include, but are not limited to, duration of deliciousness, speed at which one becomes bored, duration of feeling full, or some of these.

[0061] The evaluation value of the item that shows the change over time in the evaluation is Δy k For example, if the evaluation value of deliciousness is y k When expressed as , the amount of change in palatability over time (for example, the speed of change in palatability over time) is expressed as Δy k For the items that represent the time change of the evaluation in equation (2), the left side can be expressed as y k Instead of Δy k It may be expressed as follows.

[0062] L is an integer greater than or equal to 1, and indicates the number of elements of the electroencephalogram rhythm (the number of combinations of brain regions and frequencies). w k,i is the element θ of the EEG rhythm i The weighting coefficient w k,i corresponds to the learning parameter. That is, the weight coefficient w k,i The value of is determined by machine learning. Setting up an evaluation tendency estimation model for each type of food and drink is done by using the weighting coefficient w k,i The value of is determined for each type of food and drink, and the value according to the type of food and drink is used as the weighting coefficient w k,i This can be done by setting

[0063] The evaluation tendency estimation model can be trained using training data that combines measurement data of the electroencephalogram rhythm of a person consuming the food or drink used in the training with data indicating the person's subjective evaluation of the food or drink. In this case, the data indicating the subjective evaluation of the food or drink corresponds to the correct answer data. In this case, machine learning can be performed by inputting the EEG rhythms contained in the training data into an evaluation tendency estimation model, and optimizing the learning parameter values ​​so that the difference (estimation error) between the estimated evaluation value output by the evaluation tendency estimation model and the correct data is as small as possible. The evaluation tendency estimation model may be learned by the evaluation tendency estimation device 100. Alternatively, the evaluation tendency estimation model may be learned by a device other than the evaluation tendency estimation device 100.

[0064] Data indicating subjective evaluations is also referred to as subjective evaluation data. The subjective evaluation data for a food or beverage is obtained, for example, by a questionnaire given to subjects who have consumed the food or beverage. The questionnaire for obtaining perceptual data and the questionnaire for obtaining subjective evaluation data may be conducted simultaneously or separately. The subjects in the test for obtaining training data for training the EEG rhythm estimation model and the subjects in the test for obtaining training data for training the subjective evaluation model may be the same person or different people. The training data for learning a model is also referred to as the training data of the model.

[0065] The person in charge of generating the subjective evaluation data may refer to the questionnaire results and input the subjective evaluation data into the evaluation tendency estimation device 100. Alternatively, the person who consumes the food or drink may input the subjective evaluation data into the evaluation tendency estimation device 100. Alternatively, the questionnaire for generating the subjective evaluation data may be conducted in the form of a mark sheet, and a mark sheet reading device may read the questionnaire results, so that the subjective evaluation data is input to a device other than the evaluation tendency estimation device 100. Then, the evaluation tendency estimation device 100 may acquire the subjective evaluation data from the other device.

[0066] The evaluation tendency estimation model may be a known machine learning model such as deep learning, perceptron, or regression model. By using a known machine learning model as the evaluation tendency estimation model, a known algorithm can also be used as the machine learning algorithm, and in this respect, the evaluation tendency estimation model can be trained with a relatively small load.

[0067] The subjects from which training data for the EEG rhythm estimation model is obtained and the subjects from which training data for the evaluation tendency estimation model is obtained may be the same person or different people. Furthermore, the number of subjects from which training data for the EEG rhythm estimation model is obtained and the number of subjects from which training data for the evaluation tendency estimation model is obtained may be the same or different.

[0068] Since the training data for both models requires measurement data of EEG rhythms, by measuring the EEG rhythms of the same person as the subject for both the EEG rhythm estimation model and the evaluation tendency estimation model, the measurement data of EEG rhythms can be shared between the two models, which allows for relatively efficient acquisition of training data.

[0069] When acquiring training data for an EEG rhythm estimation model, it is believed that individual differences will arise in perception and expression of food and drink. To avoid emphasizing the perception and expression of a particular person when learning the EEG rhythm estimation model, it is preferable to acquire training data for the EEG rhythm estimation model from a relatively large number of subjects.

[0070] Furthermore, when acquiring training data for a rating tendency estimation model, it is thought that individual differences will arise in the ratings and expressions of foods and drinks. In order to avoid emphasizing the ratings and expressions of a particular person when learning the rating tendency estimation model, it is preferable to acquire training data for the rating tendency estimation model from a relatively large number of subjects.

[0071] The subject who ingests food and drink when the evaluation tendency estimation device 100 estimates the tendency of subjective evaluations of food and drink and the subject from whom training data for the electroencephalogram estimation model is obtained may be the same person, or they may be different people. As described above, if the subject who ingests food and drink when the evaluation tendency estimation device 100 estimates the tendency of subjective evaluations of food and drink and the subject from whom training data for the electroencephalogram estimation model is obtained are the same person, the influence of individual differences in perceptual expression is expected to be zero or small, and the evaluation tendency estimation device 100 will be able to estimate brain rhythms with relatively high accuracy. The subject who consumes food and drink when the evaluation tendency estimation device 100 estimates the tendency of subjective evaluations of food and drink and the subject from whom training data for the subjective evaluation estimation model is obtained may be the same person or different people.

[0072] It is also preferable that the number of subjects who consume food and drink when the evaluation tendency estimation device 100 estimates the tendency of subjective evaluations of food and drink is relatively large so as not to emphasize the perceptions and expressions of a particular person.

[0073] When the evaluation items to be estimated by the evaluation tendency estimation unit 143 include an item that represents a change in evaluation over time, the evaluation tendency estimation unit 143 may also calculate the evaluation value of the item that represents a change in evaluation over time based on the electroencephalogram rhythm when the person is ingesting the target food or drink. For example, the evaluation tendency estimation unit 143 may also calculate the evaluation value of the item that represents a change in evaluation over time using formula (2).

[0074] Alternatively, the evaluation tendency estimation unit 143 may calculate an evaluation value for each evaluation item of the food or drink, or an evaluation value for an item that represents a change in the evaluation over time, based on the amount of change in the electroencephalogram rhythm in addition to the electroencephalogram rhythm when the person is ingesting the target food or drink. For example, the evaluation tendency estimation unit 143 may calculate an evaluation value for each evaluation item of the food or drink, or an evaluation value for an item that represents a change in the evaluation over time, using formula (3).

[0075]

number

[0076] Δθ i is the element θ of the EEG rhythm i For example, Δθ i However, the element θ of the EEG rhythm i Alternatively, the change in the value of the temperature may be expressed as the amount of change per unit time. w k,i ' is the change in the brain wave rhythm element Δθ i The weighting coefficient w k,i ' corresponds to the learning parameter. That is, the weight coefficient w k,i The value of ' is determined by machine learning.

[0077] When the evaluation tendency estimation unit 143 calculates the evaluation value of the item that represents the time change of the evaluation using the formula (3), the electroencephalogram rhythm estimation unit 142 calculates the element θ i In addition to the value of i The value of may also be estimated. To this end, we use the EEG rhythm element θ as training data for the EEG rhythm estimation model. i In addition to the measured values ​​of i Alternatively, training data may be used that also includes measurements of Regarding the equation (3), for the items that show the time change of evaluation, the left side is y k Instead of Δy k It may be expressed as follows.

[0078] The evaluation tendency estimation unit 143 may normalize or standardize the estimated values ​​of the subjective evaluation tendency for food and drink for each type of food and drink. For example, the storage unit 130 may store previously obtained estimates of subjective evaluation trends for food and drink before normalization or standardization. Then, each time the evaluation trend estimation unit 143 calculates an estimate of subjective evaluation trends for food and drink, it may normalize or standardize the estimate of subjective evaluation trends for food and drink for the class into which the food and drink is classified.

[0079] For example, the evaluation tendency estimation unit 143 may use a Min-Max Normalization method to scale the estimated value of the tendency of subjective evaluations of food and drink into a range from 0 to 1. Alternatively, the evaluation tendency estimation unit 143 may standardize the estimated value of the tendency of subjective evaluations of food and drink based on a standard normal distribution.

[0080] The evaluation tendency estimation unit 143 normalizes or standardizes the estimated values ​​of the subjective evaluation tendency for food and beverages for each type of food and beverage to make the magnitude of the estimated values ​​more or less uniform, thereby making it possible to compare evaluations of food and beverages between different types of food and beverages. The display unit 110 may display the evaluations of a plurality of foods and drinks in a comparable manner. For example, the evaluation tendency estimation unit 143 may generate a two-dimensional or three-dimensional radar chart showing the electroencephalogram rhythm and the estimated values ​​of the subjective evaluation tendency for each food and drink, and display the chart on the display unit 110.

[0081] Fig. 3 shows an example of a radar chart showing the estimated values ​​of the EEG rhythm and the subjective evaluation tendency. The radar chart in Fig. 3 shows three elements of the EEG rhythm, theta waves 1, theta waves 2, and theta waves 3, and the evaluation values ​​of the deliciousness of food and drink. Line L11 indicates the electroencephalogram rhythm and evaluation value for food and drink A. Line L12 indicates the electroencephalogram rhythm and evaluation value for food and drink B. Food and drink A and food and drink B may be the same type of food and drink, or may be different types of food and drink.

[0082] Display unit 110 displays the evaluation values ​​for each of a plurality of foods and drinks as a radar chart or the like, allowing the user to compare the evaluations for a plurality of foods and drinks by referring to the display on display unit 110. For example, by referring to the radar chart in Fig. 3, the user can visually understand that the evaluation of the palatability of food and drink A is higher than the evaluation of the palatability of food and drink B.

[0083] Here, if the scale of the evaluation value differs for each type of food and drink, the evaluation values ​​for different types of food and drink cannot be compared as they are. In contrast, the evaluation tendency estimation unit 143 normalizes or standardizes the evaluation value (e.g., estimated value of evaluation tendency) for each type of food and drink to make the size of the evaluation value uniform to a certain extent, making it possible to compare the evaluations of foods and drinks even if the types of food and drink are different, as in the example of Figure 3.

[0084] Furthermore, by displaying the element values ​​of the electroencephalogram rhythm and the evaluation value for food and drink on the display unit 110, the user can analyze the element values ​​of the electroencephalogram rhythm and the evaluation for food and drink, or confirm the validity of the evaluation. For example, in the example of Figure 3, we can hypothesize that the larger the data value of Theta Wave 1, the larger the evaluation value of deliciousness (the better the evaluation). Furthermore, if any relationship or hypothesis is found between the element values ​​of the electroencephalogram rhythm and the evaluation value for a food or drink, it can be determined that the evaluation value for the food or drink that conforms to that relationship or hypothesis is valid.

[0085] The electroencephalogram rhythm estimation model and the evaluation tendency estimation model may be combined into one model. A model formed by combining the electroencephalogram rhythm estimation model and the evaluation tendency estimation model is also called a combined model.

[0086] Fig. 4 is a diagram showing an example of a connection model. In the example of Fig. 4, the connection model is configured as a neural network 200. The neural network 200 includes an input layer 210, one hidden layer 220, and an output layer 230. For example, perceptual data for one person is input to the input layer 210. The intermediate layer 220 represents electroencephalogram (EEG) rhythm values. That is, the intermediate layer 220 represents electroencephalogram rhythm data. The output layer 230 outputs an estimated value of the evaluation of food and drink. Of the neural network 200, the partial model from the input layer 210 to the intermediate layer 220 corresponds to an example of an electroencephalogram rhythm estimation model, and the partial model from the intermediate layer 220 to the output layer 230 corresponds to an example of an evaluation tendency estimation model.

[0087] Now, consider the case where neural network 200 is trained using training data that combines sensory data and estimates of food and drink ratings, without guiding or constraining the values ​​of hidden layer 220. In this case, it is considered difficult to use the values ​​of the intermediate layer 220 to interpret the estimated results of the evaluation of food and drink.

[0088] In contrast, by having the intermediate layer 220 indicate the value of the electroencephalogram rhythm, the value of the intermediate layer 220 can be used to interpret the estimated result of the evaluation of a food or drink. For example, the weighting coefficient for the output from the node of the intermediate layer 220 can be considered to indicate the correlation between the node of the intermediate layer 220 and the node of the output layer 230. By referring to the value of this weighting coefficient, it is possible to grasp the correlation between the value of the element of the electroencephalogram rhythm (the magnitude of a certain frequency component of the electroencephalogram at a certain part of the brain) indicated as the value of the node of the intermediate layer 220 and the evaluation value of the food or drink indicated as the value of the node of the output layer 230.

[0089] FIG. 5 is a diagram showing an example of a processing procedure in which the evaluation tendency estimation device 100 estimates the evaluation tendency of a target food or drink. In the process of FIG. 5, the evaluation tendency estimation device 100 acquires sensory data of the target food or drink (step S101).

[0090] For example, the display unit 110 may display a sensory data input screen, and the operation input unit 120 may accept a user operation to input the sensory data. In this case, the sensory data may be input by the person who has consumed the target food or drink, or the sensory data may be input by a different person other than the person who has consumed the target food or drink. Alternatively, the evaluation tendency estimation device 100 may acquire already generated sensory data, for example, by receiving sensory data from another device.

[0091] Next, the classification unit 141 classifies the target food and drink (step S102). As described above, identification information indicating the class may be attached to the sensory data. The classification unit 141 may then classify the sensory data by reading the identification information. Alternatively, the classification unit 141 may classify the sensory data based on the value of the sensory data.

[0092] Next, the electroencephalogram rhythm estimation unit 142 selects an electroencephalogram rhythm estimation model that is associated with the class into which the target food or drink has been classified by the class classification unit 141, from among the electroencephalogram rhythm estimation models provided for each food or drink class (step S103). Furthermore, the evaluation tendency estimation unit 143 selects, from the evaluation tendency estimation models provided for each food and drink class, an evaluation tendency estimation model that is associated with the class into which the classifying unit 141 has classified the target food and drink (step S104).

[0093] Next, the electroencephalogram (EEG) rhythm estimation unit 142 estimates the electroencephalogram (step S105). Specifically, the electroencephalogram (EEG) rhythm estimation unit 142 inputs the perceptual data obtained in step S101 into the electroencephalogram (EEG) rhythm estimation model selected in step S103, and calculates an estimated value of the electroencephalogram.

[0094] Next, the evaluation tendency estimation unit 143 estimates the evaluation tendency for the target food or drink (step S106). Specifically, the evaluation tendency estimation unit 143 inputs the estimated value of the electroencephalogram rhythm obtained in step S105 into the evaluation tendency estimation model selected in step S104, and calculates an estimated value of the evaluation for the food or drink.

[0095] Then, the evaluation tendency estimation unit 143 normalizes or standardizes the estimated value of the evaluation tendency for food and drink (step S107). For example, the evaluation tendency estimation unit 143 may normalize or standardize the estimated value of the evaluation tendency for food and drink classified into a class to which the target impression object is classified. After step S107, the evaluation tendency estimation device 100 ends the processing of FIG.

[0096] If there is no model associated with the class into which the target food or drink is classified, the evaluation tendency estimation device may generate a model. 6 is a diagram showing an example of the configuration of an evaluation tendency estimating system 10. In the configuration shown in FIG.

[0097] The evaluation trend estimation device 300 estimates the evaluation trend for food and drink, similar to the evaluation trend estimation device 100. Furthermore, if there is no model associated with the class into which the target food and drink is classified, the evaluation trend estimation device 300 generates a model. The evaluation trend estimation device 300 generates a model by learning the model. The evaluation trend estimation device 300 is an example of the evaluation trend estimation device 100. The electroencephalogram sensor device 400 measures electroencephalograms and is used to acquire sensory data for generating training data that the evaluation tendency estimation device 300 uses to learn a model.

[0098] Fig. 7 is a diagram showing an example of the configuration of an evaluation tendency estimation device 300. In the configuration shown in Fig. 7, the evaluation tendency estimation device 300 includes a display unit 110, an operation input unit 120, a storage unit 130, a processing unit 340, and a communication unit 350. The processing unit 340 includes a class classification unit 141, an electroencephalogram rhythm estimation unit 142, an evaluation tendency estimation unit 143, and a model learning unit 344.

[0099] 7, parts that correspond to and have the same effect as the parts in FIG. 1 are given the same reference numerals (110, 120, 130, 141, 142, 143) as in FIG. 1, and detailed description thereof will be omitted here. The evaluation tendency estimation device 300 differs from the evaluation tendency estimation device 100 in that it includes a communication unit 350 and that the processing unit 340 includes a model learning unit 344. In other respects, the evaluation tendency estimation device 300 is similar to the evaluation tendency estimation device 100.

[0100] The communication section 350 communicates with other devices. For example, the communication section 350 receives brainwave rhythm data from the brainwave sensor device 400. Alternatively, the communication section 350 may receive brainwave data from the brainwave sensor device 400. The processing section 340 may then perform a Fourier transform on the brainwave data to generate brainwave rhythm data.

[0101] The model learning unit 344 learns the model. In particular, when there is no electroencephalogram estimation model associated with the class into which the target food or drink is classified, the model learning unit 344 generates an electroencephalogram rhythm estimation model by learning the electroencephalogram rhythm estimation model. Furthermore, when there is no evaluation tendency estimation model associated with the class into which the target food or drink is classified, the model learning unit 344 generates an evaluation tendency estimation model by learning the evaluation tendency estimation model.

[0102] FIG. 8 is a diagram showing an example of a processing procedure in which the evaluation tendency estimation device 300 estimates the evaluation tendency of a target food or drink. 8, the evaluation tendency estimation device 300 acquires sensory data of the target food and drink (step S201). The process performed by the evaluation tendency estimation device 300 in step S201 is similar to the process performed by the evaluation tendency estimation device 100 in step S101 in FIG. Next, the classification unit 141 classifies the target food and drink (step S202). Step S202 is the same as step S102 in FIG.

[0103] Next, the review tendency estimation unit 143 searches for a review tendency estimation model associated with the class into which the target food and drink has been classified by the classification unit 141 (step S203). Then, the evaluation tendency estimation unit 143 determines whether or not there is an evaluation tendency estimation model associated with the class into which the target food and drink has been classified by the class classification unit 141 (step S204).

[0104] When the evaluation tendency estimation unit 143 determines that there is no evaluation tendency estimation model (step S204: NO), the model learning unit 344 generates an evaluation tendency estimation model by learning the evaluation tendency estimation model (step S205). For example, the display unit 110 may display a message instructing a test to generate training data for the evaluation tendency estimation model. The test administrator may then conduct the test in accordance with the instructions and generate training data for the evaluation tendency estimation model. The model learning unit 344 may then use the obtained training data to learn the evaluation tendency estimation model.

[0105] Next, the electroencephalogram rhythm estimation section 142 searches for an electroencephalogram rhythm estimation model associated with the class into which the target food and drink has been classified by the classification section 141 (step S206). Then, the electroencephalogram rhythm estimation section 142 determines whether or not there is an electroencephalogram rhythm estimation model associated with the class into which the target food and drink has been classified by the classification section 141 (step S207).

[0106] If the electroencephalogram rhythm estimation section 142 determines that there is no electroencephalogram rhythm estimation model (step S207: NO), the model learning section 344 generates an electroencephalogram rhythm estimation model by learning the electroencephalogram rhythm estimation model (step S208). For example, the display unit 110 may display a message instructing a test to generate training data for the EEG rhythm estimation model. The test administrator may then conduct the test in accordance with the instructions and generate training data for the EEG rhythm estimation model. The model training unit 344 may then use the obtained training data to train the EEG rhythm estimation model.

[0107] Next, the electroencephalogram (EEG) rhythm estimation unit 142 estimates the electroencephalogram (step S209). Specifically, the electroencephalogram (EEG) rhythm estimation unit 142 inputs the perceptual data obtained in step S201 to the electroencephalogram (EEG) rhythm estimation model obtained in step S206 or step S208, and calculates an estimated value of the electroencephalogram.

[0108] Next, the evaluation tendency estimation unit 143 estimates the evaluation tendency for the target food or drink (step S210). Specifically, the evaluation tendency estimation unit 143 inputs the estimated value of the electroencephalogram rhythm obtained in step S209 into the evaluation tendency estimation model obtained in step S203 or step S205, and calculates an estimated value of the evaluation for the food or drink.

[0109] Then, the evaluation tendency estimation unit 143 normalizes or standardizes the estimated value of the evaluation tendency for food and drink (step S211). Step S211 is similar to step S107 in FIG. After step S211, the evaluation tendency estimation device 300 ends the processing of FIG.

[0110] On the other hand, in step S204, if the evaluation tendency estimation unit 143 determines that there is an evaluation tendency estimation model associated with the class into which the target food and drink has been classified by the class classification unit 141 (step S204: YES), the process proceeds to step S206. On the other hand, if the electroencephalogram rhythm estimation section 142 determines in step S207 that there is an electroencephalogram rhythm estimation model associated with the class into which the target food or drink has been classified by the class classification section 141 (step S207: YES), the process proceeds to step S209.

[0111] As described above, the electroencephalogram (EEG) rhythm estimation unit 142 receives input of sensory data indicating the perception of a person who has consumed a target food or beverage toward that food or beverage, and estimates the electroencephalogram rhythm, which is the frequency spectrum of the person's electroencephalogram while consuming that food or beverage. Evaluation tendency estimation section 143 receives the estimated electroencephalogram rhythm as input and estimates the tendency of a plurality of people in the subjective evaluation of the food or drink by the people who have consumed the food or drink. According to the evaluation tendency estimating device 100, there is no need to measure electroencephalograms when evaluating food and drink, and food and drink can be evaluated with a relatively high degree of accuracy.

[0112] Furthermore, the classification unit 141 classifies the target foods and drinks into classes. The evaluation tendency estimation unit 143 inputs the electroencephalogram rhythm into an evaluation tendency estimation model, among the evaluation tendency estimation models provided for each class, that corresponds to the class into which the target food or beverage is classified, and estimates the tendency of multiple people in the subjective evaluations of the target food or beverage by people who have consumed the food or beverage. The evaluation tendency estimation device 100 is expected to be able to estimate the tendency of subjective evaluation values ​​for food and drink with a relatively high degree of accuracy by using an evaluation tendency estimation model according to the type of food.

[0113] Furthermore, the evaluation tendency estimation unit 143 normalizes or standardizes the estimated value of the subjective evaluation tendency for the food or drink to be estimated for each class in the food classification. According to the evaluation tendency estimation device 100, through normalization or standardization, the magnitude of the estimated values ​​of the subjective evaluation tendency for food and beverages becomes more or less consistent between types of food and beverages, making it possible to compare evaluations of food and beverages between different types of food and beverages.

[0114] Furthermore, the classification unit 141 classifies the food and drink. The electroencephalogram estimating unit 142 inputs the electroencephalogram rhythm into an electroencephalogram estimating model, among the electroencephalogram estimating models provided for each class, that corresponds to the class into which the target food or drink is classified, and estimates the electroencephalogram rhythm of the person when ingesting the target food or drink. According to the evaluation tendency estimation device 100, it is possible to use sensory data in which sensory items corresponding to the type of food are set, and it is expected that this will enable estimation of electroencephalogram rhythms with a relatively high degree of accuracy.

[0115] The subjective evaluation of the food and drink also includes a subjective evaluation of the change over time in the subjective evaluation of the items set as subjective evaluation items. The evaluation tendency estimation device 100 can estimate not only the subjective evaluation of a person while ingesting a target food or drink, but also the change in the subjective evaluation thereafter.

[0116] (About the experiment) An experiment was conducted on the present invention. In the experiment, three monitors (subjects) ingested eight types of food and drink with different sensory characteristics (different perceptions obtained) until they became tired of them, and changes in preference and brain waves were measured. The sensory evaluation of each food was to be completed after the experiment (after they had finished consuming the food and drink).

[0117] The preference items were deliciousness, disgust, boredom, and satiety, and changes in preference were observed by repeatedly investigating these items. The eight types of food and drink used were as follows: Drinking fluids: water, juice, coffee Breads: Bean buns, cream buns Confectionery: dark chocolate, arare Niboshi fish: Niboshi In the experiment, the above-mentioned classification was used, and class estimation based on sensory expressions was not performed.

[0118] In the experiment, we decided to create a multiple regression equation to predict the "degree of boredom" among preferences. We decided to create a mathematical model using only the variables (EEG measurement location and frequency) that were statistically significant in the stepwise multiple regression analysis. As a result of the experiment, the F value (significance F), p value, and model for each type of food and drink were as follows.

[0119] <Cream bun> F=42.4, p=2.74e-16 Boredom = -0.012183 +0.16006*EEG(FC1, 6Hz) +0.32361*EEG(CP5, 8Hz) -0.10675*EEG(O2, 10Hz) +0.12771*EEG(FC5, 8Hz) +0.16006*EEG(FC2, 6Hz)

[0120] <Anpan> F=8.85, p=1.35e-05 Satiety = 4.7278 + 0.16359 * EEG(O2, 4Hz) - 0.16152 * EEG(T8, 6Hz) + 0.15119 * EEG(FC6, 8Hz) - 0.12817 * EEG(FC1, 12Hz) - 0.093596 * EEG(CP6, 12Hz)

[0121] <Chocolate> F = 6.99, p = 0.0135 Satiety = 5.5576 - 0.30222 * EEG(FC2, 10Hz)

[0122] <Dried small fish> F = 8.1, p = 2.24e - 06 Satiety = 6.6392 - 0.22535 * EEG(FC5, 6Hz) - 0.051487 * EEG(Fp2, 8Hz) - 0.15371 * EEG(FC1, 10Hz) + 0.31415 * EEG(O2, 6Hz) - 0.11392 * EEG(C4, 4Hz)

[0123] <Rice cracker> F = 6.41, p = 2.36e - 07 Satiety = 6.8781 - 0.081578 * EEG(CP2, 10Hz) - 0.077389 * EEG(F3, 10Hz) - 0.19536 * EEG(FC2, 10Hz) - 0.051104 * EEG(O2, 8Hz) - 0.056325 * EEG(C3, 12Hz)

[0124] <Juice> F = 7.76, p = 6.59e - 05 Satiety = 3.8122 + 0.0037298 * EEG(FC6, 6Hz) + 0.11952 * EEG(F3, 6Hz) + 0.04881 * EEG(Fp2, 4Hz) - 0.090773 * EEG(C3, 4Hz)

[0125] <Coffee> F=13.2, p=2.82e-07 Boredom = 4.6769 +0.63584*EEG(F8, 6Hz) -0.34477*EEG(F7, 4Hz) +0.17398*EEG(O1, 8Hz) -0.57515*EEG(FC6, 10Hz) -0.093573*EEG(P4, 6Hz)

[0126] <Water> F = 8.55, p = 0.000345 Boredom = 1.9105 + 0.20323*EEG(FC1, 4Hz) + 0.10676*EEG(F7, 6Hz) +0.12406*EEG(F8, 10Hz)

[0127] In the above formula, * denotes multiplication. The notation "EEG" represents the brain wave rhythm in the form of EEG (brain wave measurement position, frequency). The brain wave measurement position is indicated by the electrode position in the 10% method. As described above, it was verified that the "degree of boredom," one of the food and drink preferences, can be expressed by a mathematical formula that combines several electroencephalogram rhythms. This mathematical formula corresponds to the evaluation tendency estimation model. Generating this mathematical formula corresponds to generating the evaluation tendency estimation model in step S205 of FIG. 8. It was also confirmed that significant variables (electroencephalogram measurement location and frequency) differed depending on the type of food and drink.

[0128] In addition, in the experiment, a stepwise multiple regression analysis was performed to determine whether the EEG rhythms at the frontal electrodes FC1, FC2, FC5, FC6, F7, and F8, which were used in the model of "degree of boredom," could be estimated using the sensory evaluation of food and drink. The frequencies of theta and alpha waves, 4 Hz to 12 Hz, were used as the EEG rhythm frequencies. In addition, because the EEG rhythm of the occipital electrode O2 often remained in the regression equation, it was also analyzed as reference data.

[0129] for example, EEG (FC1, 6Hz) = constant term + coefficient * sensory evaluation (softness) Assuming that the above formula is obtained, this formula corresponds to an EEG rhythm estimation model that estimates the EEG rhythm (FC1, 6Hz) based on the sensory evaluation of "softness" (perceptual data indicating "softness"). In this case, "softness" can be evaluated as a sensory characteristic (type of perception) that expresses the EEG rhythm (FC1, 6Hz).

[0130] Fig. 9 is a diagram showing a first example of the relationship between electroencephalogram data and sensory characteristics obtained in an experiment. Fig. 9 shows, for standard sensory characteristics such as "soft" and "moist," sensory characteristics that were evaluated as having a significant correlation with the electroencephalogram rhythm (the combination of the electroencephalogram measurement position and frequency) for each electroencephalogram measurement position and each frequency. An "x" indicates that no sensory attribute had a significant correlation with that EEG rhythm.

[0131] Fig. 10 is a diagram showing a second example of the relationship between electroencephalogram data and sensory characteristics obtained in the experiment. Fig. 10 shows the sensory characteristics evaluated as having a significant correlation with the electroencephalogram rhythm (the combination of the electroencephalogram measurement position and frequency) for each electroencephalogram measurement position and each frequency, for the five tastes plus additional sensory characteristics such as "novelty" and "astringency." In the example of FIG. 10, an "x" indicates that there was no sensory characteristic that had a significant correlation with that electroencephalogram rhythm.

[0132] 9 and 10, it was found that there is a significant correlation between the EEG rhythm and the sensory characteristics in several combinations of the EEG rhythm and the sensory characteristics. This proves that it is possible to generate an EEG rhythm estimation model (a model that receives an input of sensory evaluation (perceptual data) and outputs an EEG rhythm). The process of generating the EEG rhythm estimation model is performed, for example, in step S208 of FIG. 8. It is possible to estimate not only the "degree of boredom" but also other preferences based on the electroencephalogram rhythm estimated from the sensory evaluation, and to estimate preferences based on the estimated electroencephalogram rhythm.

[0133] (Example of how to use the EEG rhythm estimation model and the evaluation tendency estimation model) The EEG rhythm estimation model can be considered as a model showing the relationship between the perception of food and drink and EEG rhythm, and the evaluation tendency estimation model can be considered as a model showing the relationship between EEG rhythm and the tendency to evaluate food and drink, and these models can be used as follows.

[0134] For example, consider the case where we want to develop a cream bun that people will not tire of (a cream bun that people will not tire of). We also assume that an evaluation tendency estimation model, such as the above formula, has been obtained for the degree of tiredness of cream buns. We also assume that a model showing a correlation, such as the examples in Figures 9 and 10, has been obtained as an electroencephalogram rhythm estimation model.

[0135] The relationship between EEG rhythms and evaluation tendencies for cream buns, as shown by the evaluation tendency estimation model, indicates that the greater the values ​​of the EEG (FC1, 6Hz), EEG (FC5, 8Hz), and EEG (FC2, 6Hz), the greater the degree of boredom (the greater the numerical value indicating the degree of boredom). Therefore, it can be estimated that the degree of boredom can be reduced by decreasing the values ​​of the EEG (FC1, 6Hz), EEG (FC5, 8Hz), and EEG (FC2, 6Hz).

[0136] Also, according to the example of FIG. 9, the combination of electroencephalogram rhythm and sensory characteristics EEG (FC1, 6Hz) ···Soft, EEG (FC5, 8Hz) Sticky, EEG (FC2, 6Hz) - Soft There is a positive correlation between each of these. Therefore, it can be seen that it may be possible to develop a cream bun that people will not tire of by reducing the chewiness, stickiness, and softness.

[0137] Alternatively, according to the example of FIG. 10, a combination of electroencephalographic rhythm and sensory characteristics EEG (FC1, 6Hz) Novelty, EEG (FC5, 8Hz) Taste, EEG (FC2, 6Hz)...General taste There is a positive correlation between each of these. Therefore, it can be seen that it may be possible to develop a cream bun that people will not tire of by reducing novelty, umami, and overall deliciousness.

[0138] FIG. 11 illustrates an example configuration of a computer according to at least one embodiment. In the configuration shown in FIG. 11, a computer 700 includes a CPU 710, a main memory device 720, an auxiliary memory device 730, an interface 740, and a non-volatile recording medium 750.

[0139] One or more of the evaluation tendency estimation device 100 and the evaluation tendency estimation device 300, or a part thereof, may be implemented in a computer 700. In this case, the operation of each of the above-mentioned processing units is stored in the auxiliary storage device 730 in the form of a program. The CPU 710 reads the program from the auxiliary storage device 730, loads it into the main storage device 720, and executes the above-mentioned processing in accordance with the program. The CPU 710 also allocates storage areas in the main storage device 720 corresponding to each of the above-mentioned storage units in accordance with the program. Communication between each device and other devices is executed by an interface 740 having a communication function and performing communication under the control of the CPU 710. The interface 740 also has a port for a non-volatile storage medium 750, and reads information from the non-volatile storage medium 750 and writes information to the non-volatile storage medium 750.

[0140] When the evaluation tendency estimation device 100 is implemented in a computer 700, the operations of the processing unit 140 and each unit thereof are stored in the form of a program in the auxiliary storage device 730. The CPU 710 reads the program from the auxiliary storage device 730, loads it into the main storage device 720, and executes the above-mentioned processing in accordance with the program.

[0141] Furthermore, the CPU 710 allocates a storage area for the storage unit 130 in the main storage device 720 in accordance with the program. Communication between the evaluation tendency estimation device 100 and other devices is carried out by the interface 740 having a communication function and operating under the control of the CPU 710. Display of images by the display unit 110 is carried out by the interface 740 having a display device and displaying various images under the control of the CPU 710. Reception of user operations by the operation input unit 120 is carried out by the interface 740 having an input device and receiving user operations under the control of the CPU 710.

[0142] When the evaluation tendency estimation device 300 is implemented in a computer 700, the operations of the processing unit 340 and each unit thereof are stored in the form of a program in the auxiliary storage device 730. The CPU 710 reads the program from the auxiliary storage device 730, loads it into the main storage device 720, and executes the above-mentioned processing in accordance with the program.

[0143] Furthermore, the CPU 710 allocates a storage area for the storage unit 130 in the main storage device 720 in accordance with the program. Communication with other devices by the communication unit 350 is performed by the interface 740 having a communication function and operating under the control of the CPU 710. Display of images by the display unit 110 is performed by the interface 740 having a display device and displaying various images under the control of the CPU 710. Reception of user operations by the operation input unit 120 is performed by the interface 740 having an input device and receiving the user operations under the control of the CPU 710.

[0144] One or more of the above-described programs may be recorded on nonvolatile recording medium 750. In this case, interface 740 may read the programs from nonvolatile recording medium 750. CPU 710 may then directly execute the programs read by interface 740, or may temporarily store the programs in main storage device 720 or auxiliary storage device 730 and then execute them.

[0145] Note that the processing of each part may be performed by recording a program for executing all or part of the processing performed by the evaluation tendency estimation device 100 and the evaluation tendency estimation device 300 on a computer-readable recording medium, and having a computer system read and execute the program recorded on this recording medium. Note that the term "computer system" here includes hardware such as an OS (Operating System) and peripheral devices. Furthermore, "computer-readable recording media" refers to portable media such as flexible disks, optical magnetic disks, ROMs (Read Only Memory), and CD-ROMs (Compact Disc Read Only Memory), as well as storage devices such as hard disks built into computer systems. The program may be one that realizes part of the aforementioned functions, or may be one that can realize the aforementioned functions in combination with a program already stored in the computer system.

[0146] Although the embodiments of the present invention have been described above in detail with reference to the drawings, the specific configuration is not limited to these embodiments and includes designs within the scope of the present invention. Furthermore, the above-described embodiments may be combined with other embodiments as appropriate.

[0147] Some or all of the above embodiments can be described as, but are not limited to, the following supplementary notes.

[0148] (Appendix 1) an electroencephalogram (EEG) rhythm estimation unit that receives input of perception data indicating perception of a person who has ingested food or drink, and estimates an electroencephalogram rhythm that is a frequency spectrum of the electroencephalogram of the person when ingesting the food or drink; an evaluation tendency estimation unit that receives an input of the estimated electroencephalogram rhythm and estimates a tendency of a plurality of people in subjective evaluations of the food and drink by people who have consumed the food and drink; An evaluation tendency estimation device comprising:

[0149] (Appendix 2) a classification unit that classifies the food and drink into classes, the evaluation tendency estimation unit inputs the electroencephalogram rhythm into an evaluation tendency estimation model associated with the class into which the food or drink is classified, among evaluation tendency estimation models provided for each class, and estimates tendencies of a plurality of persons in subjective evaluations of the food or drink made by persons who have consumed the food or drink. 2. The evaluation tendency estimation device according to claim 1.

[0150] (Appendix 3) the evaluation tendency estimation unit normalizes or standardizes the estimated value of the subjective evaluation tendency for the food or drink to be estimated for each class in the food classification; 3. An evaluation tendency estimation device according to claim 2.

[0151] (Appendix 4) a classification unit that classifies the food and drink into classes, the electroencephalogram estimating unit inputs the electroencephalogram into an electroencephalogram estimating model associated with the class into which the food or drink is classified, among electroencephalogram estimating models provided for each class, and estimates the electroencephalogram of the person when consuming the food or drink. 4. An evaluation tendency estimation device according to any one of appendices 1 to 3.

[0152] (Appendix 5) The subjective evaluation of the food or drink includes a subjective evaluation of changes over time in subjective evaluation of items set as subjective evaluation items. 5. An evaluation tendency estimation device according to any one of appendices 1 to 4.

[0153] (Appendix 6) The computer receiving input of perception data indicating a perception of the food or drink by a person who has consumed the food or drink, and estimating an electroencephalogram rhythm, which is a frequency spectrum of the electroencephalogram of the person when the person is consuming the food or drink; receiving an input of the estimated electroencephalogram rhythm, and estimating a tendency of a plurality of people in subjective evaluations of the food and drink by the people who ingested the food and drink; A method for estimating evaluation trends, including:

[0154] (Appendix 7) The computer training data including perceptual data indicating the perception of a subject who has ingested food and drink to the food and drink, and electroencephalogram (EEG) rhythm data indicating an electroencephalogram rhythm, which is a frequency spectrum of the electroencephalogram of the subject while ingesting the food and drink, to learn an electroencephalogram (EEG) estimation model that receives input of perceptual data and outputs an estimated value of the electroencephalogram; For each of a plurality of subjects, training data including electroencephalogram (EEG) rhythm data of the subject when ingesting food and beverages and subjective evaluation data that is data indicating the subject's subjective evaluation of the food and beverages is used to train an evaluation tendency estimation model that receives an input of the EEG rhythm and outputs an estimated value of the subjective evaluation of the food and beverages; the trained electroencephalogram rhythm estimation model and the trained evaluation tendency estimation model are combined so that an output of the electroencephalogram rhythm estimation model is input to the evaluation tendency estimation model, thereby generating a combined model that receives input of sensory data and outputs estimated values ​​of tendencies of subjective evaluations of food and drink for a plurality of people; A method for generating a combined model, comprising:

[0155] (Appendix 8) On the computer, receiving input of perception data indicating perception of the food or drink by a person who has consumed the food or drink, and estimating an electroencephalogram rhythm, which is a frequency spectrum of the electroencephalogram of the person when the person is consuming the food or drink; receiving an input of the estimated electroencephalogram rhythm, and estimating a tendency of a plurality of people in subjective evaluations of the food and drink by the people who ingested the food and drink; A program that executes the following.

[0156] (Appendix 9) On the computer, training data including perceptual data indicating the perception of a subject who has ingested food and drink to the food and drink, and electroencephalogram rhythm data indicating an electroencephalogram rhythm, which is a frequency spectrum of the electroencephalogram of the subject when ingesting the food and drink, to learn an electroencephalogram estimation model that receives input of perceptual data and outputs an estimated value of the electroencephalogram rhythm; learning an evaluation tendency estimation model that receives input of electroencephalogram rhythms and outputs estimated values ​​of subjective evaluations of food and drink for each of a plurality of subjects, using training data including electroencephalogram rhythm data of the subject when the subject is ingesting food and drink and subjective evaluation data that is data indicating the subject's subjective evaluation of the food and drink; combining the trained electroencephalogram rhythm estimation model and the trained evaluation tendency estimation model so that an output of the electroencephalogram rhythm estimation model is input to the evaluation tendency estimation model, and generating a combined model that receives input of sensory data and outputs estimated values ​​of tendencies of a plurality of people's subjective evaluations of food and drink; A program that executes the following. [Explanation of symbols]

[0157] 10 Evaluation trend estimation system 100, 300 Evaluation tendency estimation device 110 Display section 120 Operation input section 130 Storage section 140, 340 Processing section 141 Classification Unit 142 EEG rhythm estimation unit 143 Evaluation tendency estimation unit 344 Model Learning Department 350 Communications Department 400 Brain wave sensor device

Claims

1. an electroencephalogram (EEG) rhythm estimation unit that receives input of perception data indicating perception of a person who has ingested food or drink, and estimates an electroencephalogram rhythm that is a frequency spectrum of the electroencephalogram of the person when ingesting the food or drink; an evaluation tendency estimation unit that receives an input of the estimated electroencephalogram rhythm and estimates a tendency of a plurality of people in subjective evaluations of the food and drink by people who have consumed the food and drink; An evaluation tendency estimation device comprising:

2. a classification unit that classifies the food and drink into classes, the evaluation tendency estimation unit inputs the electroencephalogram rhythm into an evaluation tendency estimation model associated with the class into which the food or drink is classified, among evaluation tendency estimation models provided for each class, and estimates tendencies of a plurality of persons in subjective evaluations of the food or drink made by persons who have consumed the food or drink. The evaluation tendency estimation device according to claim 1 .

3. the evaluation tendency estimation unit normalizes or standardizes the estimated value of the subjective evaluation tendency for the food or drink to be estimated for each class in the food classification; The evaluation tendency estimation device according to claim 2 .

4. a classification unit that classifies the food and drink into classes, the electroencephalogram estimating unit inputs the electroencephalogram into an electroencephalogram estimating model associated with the class into which the food or drink is classified, among electroencephalogram estimating models provided for each class, and estimates the electroencephalogram of the person when consuming the food or drink. The evaluation tendency estimation device according to claim 1 .

5. The subjective evaluation of the food or drink includes a subjective evaluation of changes over time in subjective evaluation of items set as subjective evaluation items. The evaluation tendency estimation device according to claim 1 .

6. The computer receiving input of perception data indicating a perception of the food or drink by a person who has consumed the food or drink, and estimating an electroencephalogram rhythm, which is a frequency spectrum of the electroencephalogram of the person when the person is consuming the food or drink; receiving an input of the estimated electroencephalogram rhythm, and estimating a tendency of a plurality of people in subjective evaluations of the food and drink by the people who ingested the food and drink; A method for estimating evaluation trends, including:

7. The computer training data including perceptual data indicating the perception of a subject who has ingested food and drink to the food and drink, and electroencephalogram (EEG) rhythm data indicating an electroencephalogram rhythm, which is a frequency spectrum of the electroencephalogram of the subject while ingesting the food and drink, to learn an electroencephalogram (EEG) estimation model that receives input of perceptual data and outputs an estimated value of the electroencephalogram; For each of a plurality of subjects, training data including electroencephalogram (EEG) rhythm data of the subject when ingesting food and beverages and subjective evaluation data that is data indicating the subject's subjective evaluation of the food and beverages is used to train an evaluation tendency estimation model that receives an input of the EEG rhythm and outputs an estimated value of the subjective evaluation of the food and beverages; the trained electroencephalogram rhythm estimation model and the trained evaluation tendency estimation model are combined so that an output of the electroencephalogram rhythm estimation model is input to the evaluation tendency estimation model, thereby generating a combined model that receives input of sensory data and outputs estimated values ​​of tendencies of subjective evaluations of food and drink for a plurality of people; A method for generating a combined model, comprising:

8. On the computer, receiving input of perception data indicating perception of the food or drink by a person who has consumed the food or drink, and estimating an electroencephalogram rhythm, which is a frequency spectrum of the electroencephalogram of the person when the person is consuming the food or drink; receiving an input of the estimated electroencephalogram rhythm, and estimating a tendency of a plurality of people in subjective evaluations of the food and drink by the people who ingested the food and drink; A program that executes the following.

9. On the computer, training data including perceptual data indicating the perception of a subject who has ingested food and drink to the food and drink, and electroencephalogram rhythm data indicating an electroencephalogram rhythm, which is a frequency spectrum of the electroencephalogram of the subject when ingesting the food and drink, to learn an electroencephalogram estimation model that receives input of perceptual data and outputs an estimated value of the electroencephalogram rhythm; learning an evaluation tendency estimation model that receives input of electroencephalogram rhythms and outputs estimated values ​​of subjective evaluations of food and drink for each of a plurality of subjects, using training data including electroencephalogram rhythm data of the subject when the subject is ingesting food and drink and subjective evaluation data that is data indicating the subject's subjective evaluation of the food and drink; combining the trained electroencephalogram rhythm estimation model and the trained evaluation tendency estimation model so that an output of the electroencephalogram rhythm estimation model is input to the evaluation tendency estimation model, and generating a combined model that receives input of sensory data and outputs estimated values ​​of tendencies of a plurality of people's subjective evaluations of food and drink; A program that executes the following.

Images