Learning device, stress estimation device, learning method, stress estimation method, and program

Abstract

In the present invention, an information processing device 1X primarily includes: a first dividing means 14X; a second dividing means 15X; a feature amount selection means 16X; and a learning means 17X. The first dividing means 14X performs a first division by which an observation feature amount of a subject is divided on the basis of an attribute of the subject and / or the environment. The second dividing means 15X performs a second division by which the observation feature amount is divided on the basis of the observation subject of the observation feature amount and / or an activity state of the subject. From the observation feature amounts divided on the basis of the first division and the second division, the feature amount selection means 16X selects a stress estimation feature amount which is a feature amount used in stress estimation. The learning means 17X trains a stress estimation model, for at least each set obtained by the first division, on the basis of the stress estimation feature amount and a correct stress value corresponding to the stress estimation feature amount.

Description

[Technical field]

[0001] The present disclosure relates to the technical fields of a learning device, a stress estimation device, a learning method, a stress estimation method, and a storage medium that perform processing related to estimation of a stress state. [Background technology]

[0002] There are known devices or systems that determine the stress state of a subject based on data measured from the subject. For example, Patent Document 1 discloses a portable stress measurement device that determines the temporary stress level of a subject on each day based on the biological data of the subject. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2007-275287 A Summary of the Invention [Problem to be solved by the invention]

[0004] When estimating a subject's stress level from their biometric data, there was a problem that the estimation accuracy was not stable for unknown data.

[0005] In view of the above-mentioned problems, one objective of the present disclosure is to provide a learning device, a stress estimation device, a learning method, a stress estimation method, and a storage medium that perform processing to obtain stress estimation results with stable estimation accuracy. [Means for solving the problem]

[0006] One aspect of the learning device is The observed feature quantity of the subject is calculated based on at least one of the attributes or environment of the subject. For each cluster A first division means for performing a first division; The above Each of the clustersBased on at least one of the observed object of the observation feature or the activity state of the subject, For each sub-cluster A second division means for performing a second division; The observed feature Each sub-cluster of from, Observation feature quantities of a type that are ranked in a top predetermined number of correlation scores based on a correlation coefficient between each of the sub-clusters and the stress data corresponding to the observation feature quantities of each of the sub-clusters, Stress estimation feature, which is a feature used for stress estimation As A feature selection means for selecting a feature; The stress estimation feature amount is the input data to the stress estimation model, Stress corresponding to the stress estimation feature The set of data with the correct answer is used as training data. At least the first division The cluster Every The above A learning means for learning a stress estimation model; It is a learning device having the following.

[0007] One aspect of the stress estimation device includes: A classification means for classifying an observed feature amount of a subject to be estimated based on at least one of an attribute or an environment of the subject to be estimated; Referring to feature selection information in which the types of observed features used as input data of training data used in machine learning of a stress estimation model used in the stress estimation are associated with the classification, a feature selection method is provided for selecting the type of observed feature corresponding to the classification. The observed feature of , the stress estimation feature, which is the feature used for stress estimation As A feature selection means for selecting a feature; a stress estimation means for selecting a stress estimation model based on the classification and inputting the stress estimation feature amount into the selected stress estimation model, thereby estimating a stress value of the estimation subject; The stress estimation device has the following features.

[0008] One aspect of the learning method includes: The computer The observed feature quantity of the subject is calculated based on at least one of the attributes or environment of the subject. For each cluster Do the first division, The above Each of the clusters Based on at least one of the observed object of the observation feature or the activity state of the subject, For each sub-cluster Do a second division to divide, The observed feature Each sub-cluster of from, Observation feature quantities of a type that are ranked in a top predetermined number of correlation scores based on a correlation coefficient between each of the sub-clusters and the stress data corresponding to the observation feature quantities of each of the sub-clusters,Stress estimation feature, which is a feature used for stress estimation As choose, The stress estimation feature amount is the input data to the stress estimation model, Stress corresponding to the stress estimation feature The set of data with the correct answer is used as training data. At least the first division The cluster Every The above Train the stress estimation model, It is a learning method. Note that the "computer" includes any electronic device (it may be a processor included in an electronic device), and may be configured from multiple electronic devices.

[0009] One aspect of the stress estimation method includes: The computer classifying the observed feature amount of the subject to be estimated based on at least one of the attributes or the environment of the subject to be estimated; Referring to feature selection information in which the types of observed features used as input data of training data used in machine learning of a stress estimation model used in the stress estimation are associated with the classification, a feature selection method is provided for selecting the type of observed feature corresponding to the classification. The observed feature of , the stress estimation feature, which is the feature used for stress estimation As choose, selecting a stress estimation model based on the classification, and inputting the stress estimation feature amount into the selected stress estimation model, thereby estimating a stress value of the estimation subject; A stress estimation method.

[0010] One aspect of the program is The observed feature quantity of the subject is calculated based on at least one of the attributes or environment of the subject. For each cluster Do the first division, The above Each of the clusters Based on at least one of the observed object of the observation feature or the activity state of the subject, For each sub-cluster Do a second division to divide, The observed feature Each sub-cluster of from, Observation feature quantities of a type that are ranked in a top predetermined number of correlation scores based on a correlation coefficient between each of the sub-clusters and the stress data corresponding to the observation feature quantities of each of the sub-clusters, Stress estimation feature, which is a feature used for stress estimation As choose, The stress estimation feature amount is the input data to the stress estimation model,Stress corresponding to the stress estimation feature The set of data with the correct answer is used as training data. At least the first division The cluster Every The above This is a program that causes a computer to execute a process for learning a stress estimation model.

[0011] One aspect of the program is classifying the observed feature amount of the subject to be estimated based on at least one of the attributes or the environment of the subject to be estimated; Referring to feature selection information in which the types of observed features used as input data of training data used in machine learning of a stress estimation model used in the stress estimation are associated with the classification, a feature selection method is provided for selecting the type of observed feature corresponding to the classification. The observed feature of , the stress estimation feature, which is the feature used for stress estimation As choose, The program causes a computer to execute a process of estimating a stress value of the person to be estimated by selecting a stress estimation model based on the classification and inputting the stress estimation feature into the selected stress estimation model. Effect of the Invention

[0012] It is possible to perform stress estimation with stable estimation accuracy, or to learn a stress estimation model for performing such stress estimation. [Brief description of the drawings]

[0013] [Figure 1] 1 shows a schematic configuration of a stress estimation system according to a first embodiment. [Diagram 2] 2 illustrates an example of a hardware configuration of a stress estimation device common to each embodiment. [Diagram 3] 4 is an example of a functional block in a learning phase of the information processing device according to the first embodiment. [Figure 4] 4 is an example of a functional block of a feature amount selection unit. [Diagram 5] 1 shows a histogram of correlations for a certain type of observed feature. [Figure 6]This figure clearly shows the timing of the stress questionnaire to measure PSS values ​​during the measurement period of observational data for a sample of subjects. [Figure 7] 1 is an example of a flowchart illustrating a procedure of a learning process executed by the information processing device in a learning phase in the first embodiment. [Figure 8] 4 is an example of a functional block in an estimation phase of the information processing device according to the first embodiment. [Figure 9] 1 is an example of a flowchart illustrating a procedure of a stress estimation process executed by the information processing device in an estimation phase in the first embodiment. [Figure 10] 1 shows a schematic configuration of a stress estimation system according to a second embodiment. [Figure 11] FIG. 13 is a block diagram of a learning device according to a third embodiment. [Figure 12] 13 is an example of a flowchart executed by the learning device in the third embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Hereinafter, embodiments of a learning device, a stress estimation device, a learning method, a stress estimation method, and a storage medium will be described with reference to the drawings.

[0015] First Embodiment (1) System Configuration FIG. 1 shows a schematic configuration of a stress estimation system 100 according to the first embodiment. The stress estimation system 100 learns a model (also called a "stress estimation model") that estimates a person's stress, and performs stress estimation based on the learned estimation model. Hereinafter, a person who is the subject of stress estimation will be called an "estimation subject", and a person who is measured in generating training data (learning samples) necessary for learning the stress estimation model will be called a "sample subject". When there is no particular distinction between the estimation subject and the sample subject, they will also be simply called a "subject". Note that the "estimation subject" may be an athlete or employee whose stress state is managed by an organization, or an individual user.

[0016] The stress estimation system 100 mainly includes an information processing device 1, an input device 2, a display device 3, a storage device 4, and a sensor 5.

[0017] The information processing device 1 performs data communication with the input device 2, the display device 3, and the sensor 5 via a communication network or by direct wireless or wired communication. The information processing device 1 collects information required for learning a stress estimation model or for estimating the stress of the estimation subject using the stress estimation model based on an input signal "S1" supplied from the input device 2 and a sensor signal "S3" supplied from the sensor 5, and stores the collected information in the storage device 4. The information processing device 1 also generates a display signal "S2" based on the estimation result of the stress state of the estimation subject (specifically, a stress value indicating the degree of stress), and supplies the generated display signal S2 to the display device 3. Note that the stress estimated by the information processing device 1 in this embodiment is assumed to be chronic stress, which is stress from a long-term (chronic) perspective on a scale of several days to a week or month.

[0018] The input device 2 is an interface that accepts user input (manual input) of information regarding each presumed subject. The user who inputs information using the input device 2 may be the presumed subject himself / herself, or may be a person who manages or supervises the activities of the presumed subject. The input device 2 may be, for example, various user input interfaces such as a touch panel, a button, a keyboard, a mouse, or a voice input device. The input device 2 supplies an input signal S1 generated based on the user's input to the information processing device 1. The display device 3 displays predetermined information based on a display signal S2 supplied from the information processing device 1. The display device 3 is, for example, a display or a projector.

[0019] The sensor 5 measures the biosignals of the estimated subject and supplies the measured biosignals to the information processing device 1 as a sensor signal S3. In this case, the sensor signal S3 may be any biosignal (including vital information) of the estimated subject, such as the heart rate, brain waves, sweat rate, hormone secretion, cerebral blood flow, blood pressure, body temperature, electromyography, respiratory rate, pulse wave, and acceleration. The sensor 5 may also be a device that analyzes blood collected from the estimated subject and outputs a sensor signal S3 indicating the analysis result. The sensor 5 may also be a wearable terminal worn by the estimated subject, a camera that captures the estimated subject or a microphone that generates an audio signal of the estimated subject's speech, or a terminal such as a personal computer or smartphone operated by the estimated subject. In the latter case, the sensor 5 may supply information equivalent to the amount of operation of the personal computer, smartphone, or the like to the information processing device 1 as the sensor signal S3. The sensor 5 may also output location information output by a GPS receiver or the like incorporated in the wearable terminal or the like as the sensor signal S3. The sensor signal S3 is used to generate features (also called "observation features") that represent observed features of the observed subject.

[0020] The storage device 4 is a memory that stores various information necessary for estimating a stress state, etc. The storage device 4 may be an external storage device such as a hard disk connected to or built into the information processing device 1, or may be a storage medium such as a flash memory. The storage device 4 may also be a server device that performs data communication with the information processing device 1. The storage device 4 may also be composed of multiple devices.

[0021] The storage device 4 functionally comprises an attribute information storage unit 40, an observation data storage unit 41, a training data storage unit 42, and a learning parameter storage unit 43.

[0022] The attribute information storage unit 40 stores attribute information on the attributes of the subject. Here, the "attributes" include, for example, the subject's personality, stress resistance, sex, occupation, age, cognitive tendency, or a combination of these. The attribute information may be generated by the information processing device 1 and stored in the storage device 4, or may be generated in advance by a device other than the information processing device 1 and stored in the storage device 4. The attribute information may include information generated based on the subject's response to a questionnaire. For example, the Big 5 personality test is a questionnaire that measures the subject's personality. The attribute information is stored in the attribute information storage unit 40 in association with the subject's identification information.

[0023] The observation data storage unit 41 stores the observation data generated based on the sensor signal S3 acquired by the information processing device 1 from the sensor 5. In this embodiment, the observation data is information in which, for example, an observation feature amount, environmental information indicating the environment such as the date and time or place where the observation was performed, activity information indicating the subject's activity state when the observation was performed (for example, physical exercise intensity, mental activity intensity such as mental workload, state such as sitting / walking / running, state such as awake / sleeping), and identification information of the subject are associated with each other. For convenience of explanation, the observation data storage unit 41 is assumed to store the observation data of the estimated subject, and the observation data of the sample subject is assumed to be stored in the training data storage unit 42.

[0024] The observed feature is any feature representing the feature of data observed from the subject, and may be a feature based on a biological feature such as sweating, acceleration, skin temperature, pulse wave, etc., or may be a feature based on a behavioral feature based on the subject's behavior such as the amount of operation of a device. Here, the process of converting the sensor signal S3 into the observed feature may be executed by the information processing device 1, or may be executed by a device other than the information processing device 1. In this case, the observed feature may be generated from the sensor signal S3 based on any method of calculating the feature from a biological signal or any other arbitrary feature calculation method. The environmental information is generated by the information processing device 1 or another device based on, for example, date and time information, location information, temperature and humidity information, carbon dioxide concentration information, illuminance information, environmental sound information, etc. included in the sensor signal S3, and the activity information is generated by the information processing device 1 or another device based on, for example, location information, acceleration, etc. included in the sensor signal S3.

[0025] The training data storage unit 42 stores training data used for learning the stress estimation model. The training data is data generated for a plurality of sample subjects, and includes a plurality of pairs of observed data of the sample subjects and correct stress values ​​based on responses to a questionnaire by the sample subjects. In this embodiment, a PSS (Perceived Stress Scale) value is used as the correct stress value. The PSS value is calculated from the response results of a PSS questionnaire that can measure dynamic stress that changes over time.

[0026] The learning parameter storage unit 43 stores parameters learned by the information processing device 1. The parameters stored in the learning parameter storage unit 43 include parameters necessary for configuring a stress estimation model. The stress estimation model is a model that is learned to output a stress estimation value of a subject when a set of specific observed features (feature vector) of the subject is input. Here, the stress estimation model may be any machine learning model (including a statistical model) such as a neural network or a support vector machine. In addition, as described later, the stress estimation model is learned for each classification based on the attributes of the subject. In this case, each stress estimation model may have an architecture suitable for each classification. The learning parameter storage unit 43 stores information on parameters necessary for configuring these stress estimation models. For example, when the stress estimation model is a model based on a neural network such as a convolutional neural network, the learning parameter storage unit 43 stores information on various parameters such as a layer structure, a neuron structure of each layer, the number of filters and filter size in each layer, and the weight of each element of each filter.

[0027] The configuration of the stress estimation system 100 shown in FIG. 1 is an example, and various changes may be made to the configuration. For example, the input device 2 and the display device 3 may be configured as one unit. In this case, the input device 2 and the display device 3 may be configured as a tablet terminal that is integrated with or separate from the information processing device 1. The input device 2 and the sensor 5 may be configured as one unit. The information processing device 1 may be configured from a plurality of devices. In this case, the plurality of devices that configure the information processing device 1 transmit and receive information required to execute pre-assigned processing between these plurality of devices. In this case, the information processing device 1 functions as an information processing system.

[0028] (2) Hardware configuration of information processing device 2 shows a hardware configuration of the information processing device 1. The information processing device 1 includes, as hardware, a processor 11, a memory 12, and an interface 13. The processor 11, the memory 12, and the interface 13 are connected via a data bus 90.

[0029] The processor 11 functions as a controller (arithmetic device) that performs overall control of the information processing device 1 by executing a program stored in the memory 12. The processor 11 is, for example, a processor such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or a TPU (Tensor Processing Unit). The processor 11 may be composed of multiple processors. The processor 11 is an example of a computer.

[0030] The memory 12 is composed of various types of volatile and non-volatile memories such as a RAM (Random Access Memory), a ROM (Read Only Memory), and a flash memory. The memory 12 also stores programs for executing processes executed by the information processing device 1. Note that a part of the information stored in the memory 12 may be stored in one or more external storage devices capable of communicating with the information processing device 1, or may be stored in a storage medium that is detachable from the information processing device 1.

[0031] The interface 13 is an interface for electrically connecting the information processing device 1 to other devices. These interfaces may be wireless interfaces such as a network adapter for wirelessly transmitting and receiving data to and from other devices, or may be hardware interfaces for connecting to other devices via cables or the like.

[0032] The hardware configuration of the information processing device 1 is not limited to the configuration shown in Fig. 2. For example, the information processing device 1 may include at least one of an input device 2 or a display device 3. The information processing device 1 may also be connected to or have a built-in sound output device such as a speaker.

[0033] (3) Learning Phase Next, the process in the learning phase executed by the information processing device 1 will be described. In summary, the information processing device 1 learns an estimation model for each classification based on the attributes of the subjects. At this time, the information processing device 1 divides the observation feature included in the training data into clusters (groups) having biases in stress tendency and biological information based on the attributes of the corresponding sample subjects, environmental information of the observation feature, and the like. Then, the information processing device 1 selects the observation feature (also called "stress estimation feature") to be used as an input for the stress estimation model based on the correlation between the cluster of the divided observation feature and the correct stress value, and the like. In this way, the information processing device 1 selects the observation feature highly correlated with the stress data as the stress estimation feature, and learns a stress estimation model specialized for each cluster having biases in stress tendency and biological information. As a result, the information processing device 1 acquires a stress estimation model capable of performing stress estimation with high accuracy for unknown data not used in learning.

[0034] (3-1) Functional block FIG. 3 is an example of a functional block of the information processing device 1. In the learning phase, the processor 11 of the information processing device 1 functionally includes a first division unit 14, "N" (N is an integer of 2 or more) second division units 15 (151 to 15N), "M" (M is an integer of 2 or more) feature quantity selection units 16 (1611 to 16NM), and N estimation model learning units 17 (171 to 17N). In FIG. 3, the blocks where data is exchanged are connected by solid lines, but the combination of blocks where data is exchanged is not limited to FIG. 3. The same applies to the diagrams of other functional blocks described later. In addition, the training data storage unit 42 functionally includes an observation data storage unit 421 and a stress data storage unit 422. In addition, the learning parameter storage unit 43 functionally includes a first estimation model information storage unit 431 to an N-th estimation model information storage unit 43N that respectively store parameters of the N stress estimation models to be learned.

[0035] The first division unit 14 extracts observation features for learning from the observation data storage unit 421, and performs a first division to divide the extracted observation features into N pieces based on at least one of the corresponding attribute information and environmental information (e.g., date and time information). As a result, a cluster of N observation features having a bias in stress, biometric features, etc. is formed. The first division unit 14 extracts attribute information from the attribute information storage unit 40, and extracts environmental information from the observation data storage unit 421.

[0036] Here, the division based on attribute information is, for example, division based on personality, sex, occupation, race, age, height, weight, muscle mass, lifestyle habits, exercise habits, or a combination of these. Furthermore, the division based on environmental information is division based on season at the time of examination, division based on time period, division based on location (outdoors or indoors, etc.), or a combination of these. Then, the first division unit 14 supplies the clusters of observed features divided based on the first division to each second division unit 15 (151 to 15N) previously associated with each cluster classification.

[0037] The first division is not limited to a mode in which one observed feature is exclusively sorted into one cluster, but may be a mode in which one observed feature is overlapped and sorted into two or more clusters. For example, when the first division is performed based on the observed season, there may be clusters in which overlapping periods are assigned, such as a "cluster corresponding to April to June" and a "cluster corresponding to June to September." This also applies to the second division described later.

[0038] In addition, the first division unit 14 preferably performs a process of generating pseudo data by expanding the training data. The generation of pseudo data will be described in detail in the section "(3-3) Expanding Training Data".

[0039] The second division unit 15 (151 to 15N) performs a second division in which the observation feature for each cluster supplied from the first division unit 14 is divided into M clusters (sub-clusters) based on the observation target of the observation feature or the subject's activity state at the time of observation. In this way, the first division unit 14 further divides the observation feature that should be treated differently in stress estimation. Then, each of the second division units 151 to 15N supplies the sub-clusters of the M observation feature divided based on the second division to the feature selection unit 16 (1611 to 16NM).

[0040] Here, the "observation target" refers to the observation target of the raw data used when the observation feature is calculated, and corresponds to various biological features such as sweating, acceleration, skin temperature, and pulse wave. Therefore, in the case of an observation feature based on a biological feature, "division based on the observation target" refers to division into an observation feature related to sweating, an observation feature related to acceleration, an observation feature related to skin temperature, an observation feature related to pulse wave, and the like. Furthermore, "division based on the activity state" refers to division according to the level of exercise intensity at the time of observation of the subject (e.g., stationary state, walking state, running state). Note that information indicating the observation target and activity state corresponding to each observation feature is stored, for example, in the observation data storage unit 421 in association with the observation feature.

[0041] The feature selection unit 16 (1611 to 16NM) selects stress estimation features, which are observation features to be input to the stress estimation model, from the sub-clusters of observation features divided into N×M pieces based on the first and second divisions, based on correlation with the correct stress data. Here, the feature selection unit 16 selects "R" (R is an integer equal to or greater than 0) types of observation features as stress estimation features. Details of the processing of the feature selection unit 16 will be described later. Note that, instead of uniformly providing M feature selection units 16 for each cluster in the first division, an appropriate number of feature selection units 16 may be provided for each cluster in the first division. Similarly, the value of R may also be different for each feature selection unit 16.

[0042] The estimation model learning units 17 (171 to 17N) learn the stress estimation model prepared for each cluster divided by the first division unit 14, based on the stress estimation feature selected by the feature selection unit 16 and the stress data referenced from the stress data storage unit 422. In this case, each estimation model learning unit 17 acquires a plurality of pairs in which the M×R stress estimation feature supplied from the M feature selection units 16 are input data to the stress estimation model, and the corresponding stress data referenced from the stress data storage unit 422 are correct data. Then, each estimation model learning unit 17 learns the corresponding stress estimation model, based on the plurality of pairs of the above-mentioned input data and correct data.

[0043] In learning the stress estimation model, the estimation model learning unit 17, for example, sequentially extracts pairs of the above-mentioned input data and correct answer data, and updates the parameters of the stress estimation model. In this case, the parameters of the stress estimation model are determined so that the error (loss) between the estimation result output by the stress estimation model when input data is input and the stress value (here, the PSS value) which is the correct answer data is minimized. The algorithm for determining the above-mentioned parameters so as to minimize the loss may be any learning algorithm used in machine learning, such as the gradient descent method or the backpropagation method. Then, each estimation model learning unit 17 stores the learned parameters of each stress estimation model in the first estimation model information storage unit 431 to the Nth estimation model information storage unit 43N, respectively.

[0044] Each of the components of the first division unit 14, the second division unit 15, the feature quantity selection unit 16, and the estimation model learning unit 17 described in FIG. 3 can be realized, for example, by the processor 11 executing a program. Also, each component may be realized by recording a necessary program in an arbitrary non-volatile storage medium and installing it as necessary. At least a part of each of these components may be realized by any combination of hardware, firmware, and software, without being limited to being realized by software by a program. Also, at least a part of each of these components may be realized by using an integrated circuit programmable by a user, such as a field-programmable gate array (FPGA) or a microcontroller. In this case, the integrated circuit may be used to realize a program composed of each of the above components. Also, at least a part of each of the components may be configured by an application specific standard produce (ASSP), an application specific integrated circuit (ASIC), or a quantum processor (quantum computer control chip). In this way, each of the components may be realized by various hardware. The above is the same in other embodiments described later. Furthermore, each of these components may be realized by the cooperation of multiple computers, for example, using cloud computing technology.

[0045] (3-2) Details of feature selection section Next, the details of the process executed by the feature quantity selection unit 16 (1611 to 16NM) will be described. Fig. 4 is an example of a functional block of a feature quantity selection unit 16nm (where "n" and "m" are integers that satisfy 1 <= n <= N and 1 <= m <= M). The feature quantity selection unit 16nm functionally includes a group generation unit 50, a correlation calculation unit 51, a ranking unit 52, and a selection unit 53.

[0046] The feature selection unit 16nm selects the observed feature “F p,q” is obtained from the stress data storage unit 422, and the observed feature quantity F p,q The correct stress value (PSS value) corresponding to "S p " is obtained. Here, "p" indicates the index of the sample subject (1≦p≦P, P is an integer equal to or greater than 2), and "q" indicates the index of the type of observed feature (1≦q≦Q, Q is an integer satisfying "Q≧R"). Note that there are generally many types of observed feature (e.g., tens of thousands). For example, in the case of a feature related to sweating, various indices related to sweating such as maximum, minimum, median, average, and other arbitrary statistics are applicable.

[0047] The group generation unit 50 randomly selects a predetermined number of observed features F p,q is extracted L times (L is an integer equal to or greater than 1), and the extracted observed feature F p,q In this case, for example, if there are 100 sample subjects, the group generation unit 50 randomly extracts 50 sample subjects L times and calculates the observed feature value F p,q Then, the group generating unit 50 forms the observed feature F p,q The groups are supplied to the correlation calculation units 511 to 51L, respectively.

[0048] The correlation calculation unit 51 (511 to 51L) calculates the observed feature quantity F p,q Based on the group, the observed feature F p,q and stress value S p The correlation coefficient with the observed feature F p,q The correlation coefficient may be any one of Pearson's product moment correlation coefficient, Spearman's rank correlation coefficient, and Kendall's rank correlation coefficient, or a combination of multiple correlation coefficients such as an average. In other words, the correlation calculation unit 51 calculates the correlation coefficient for each group generated by the group generation unit 50 and for the observed feature quantity F p,q For each type q of p,q and stress value S p Calculate the correlation with

[0049] The ranking unit 52 calculates the observed feature quantity F based on the calculation results of the L correlation calculation units 511 to 51L. p,q In this case, the ranking unit 52 ranks the types q of the observed feature quantity F p,q For each type q, a score (also called a "correlation score") based on the calculation results of the L correlation calculation units 511 to 51L is calculated, and the higher the correlation score, the higher the ranking is considered to be. In this case, the ranking unit 52 calculates the correlation score based on a statistical value such as an average of correlation between groups and a sign inversion degree, which will be described later. A method of calculating the correlation score will be described later.

[0050] The selection unit 53 selects the observed feature quantities F corresponding to the top R types in the ranking formed by the ranking unit 52. p,q as the stress estimation feature. In this case, the selection unit 53 stores information indicating the type of the observation feature selected as the stress estimation feature (also referred to as "feature selection information Ifs") in the learning parameter storage unit 43. As will be described later, the feature selection information Ifs is used in a process of selecting a stress estimation feature to be input to a stress estimation model from the observation feature obtained in the estimation phase.

[0051] Here, a specific example of a method for calculating correlation scores by the ranking unit 52 will be described. Fig. 5 shows a histogram that tallies correlations for type q, which is a target for calculating correlation scores, based on the calculation results of the correlation calculation units 511 to 51L. Note that, although a histogram is shown here for convenience of explanation, the generation of a histogram is not an essential process for calculating correlation scores.

[0052] In this case, first, the correlation calculation unit 51 calculates the average (here, 0.15) of the correlations (correlation coefficients) calculated by the correlation calculation units 511 to 51L for the type q of the object based on the calculation results of the correlation calculation units 511 to 51L. Furthermore, the correlation calculation unit 51 calculates the proportion of minority signs when the positive and negative signs of the calculated correlations are tallied as the sign inversion degree, as the sign inversion degree. In the example of FIG. 5, since the positive signs are the majority, the correlation calculation unit 51 recognizes the proportion of negative signs (0.3) as the sign inversion degree. The sign inversion degree is in the range from 0 to 0.5. Then, the correlation calculation unit 51 determines the correlation score for the type q of the object to be a value obtained by multiplying the absolute value of the average of the correlations by a value obtained by subtracting the sign inversion degree from 1 (i.e., a range from 0.5 to 1) as a weight, for example, as follows: Correlation score = |average of correlations| × (1 - degree of sign reversal) In the example of FIG. 5, the correlation score of type q is 0.105 (=|0.15|×0.7).

[0053] The method of calculating the correlation score is not limited to the above formula, and any formula or lookup table may be used that defines the correlation score so that it has a positive correlation with the average of correlations and a negative correlation with the degree of sign reversal.

[0054] The feature quantity selection unit 16nm has a functional configuration as shown in FIG. 4, and thus can suitably select an observed feature quantity that has a stable correlation with the stress value regardless of individual differences as a stress estimation feature quantity.

[0055] (3-3) Expanding training data Next, a method of training data augmentation (Data Augmentation) will be described. In general, samples with low stress values ​​and samples with high stress values ​​tend to be insufficient. On the other hand, chronic stress does not change rapidly over time. Taking the above into consideration, the first division unit 14 generates a correct stress value by interpolation.

[0056] FIG. 6 is a diagram clearly showing the timing of a stress questionnaire for measuring the PSS value during a measurement period of observation data of a certain sample subject. In FIG. 6, the measurement of the observation data of the sample subject is started at date and time "t1", and the first stress questionnaire is conducted. Also, at dates and times "t2" to "t4", the second to fourth stress questionnaires are conducted, respectively. The stress questionnaire is conducted, for example, once a month. Then, the PSS value (actually measured PSS value) of the target sample subject is measured by each stress questionnaire. Note that each actual PSS value measured at time t1 to time t4 actually corresponds to the stress value during the period covered by the questionnaire, and corresponds to the stress value for one month if the questionnaire conducted is a questionnaire conducted at monthly intervals, for example.

[0057] In this case, the first division unit 14 generates PSS values ​​(interpolated PSS values) that are interpolated (for example, linearly interpolated) using the actual PSS values ​​at regular intervals during the period between measurements of the actual PSS values, as pseudo data. Here, the first division unit 14 generates two interpolated PSS values ​​for each of the period from date and time t2 to date and time t3 and the period from date and time t3 to date and time t4. The first division unit 14 considers these interpolated PSS values ​​to be correct data for the stress value for a predetermined period according to the interval between surveys, etc. Here, the observation data is continuously measured regardless of the interval between surveys, etc., and the observation data corresponding to these interpolated PSS values ​​is generated between times t1 and t4. Note that, for the period from time t1 to time t2, since there is no observation data corresponding to the interpolated PSS value at the interpolation point close to time t1, only the interpolated PSS value at the interpolation point close to time t2 is generated.

[0058] In this way, the first division unit 14 generates the interpolated PSS value, taking into consideration that chronic stress does not change suddenly. This increases the amount of actual training data, and makes it possible to preferably secure the training data necessary for learning the stress estimation model.

[0059] (3-4) Processing flow FIG. 7 is an example of a flowchart showing the procedure of the learning process executed by the information processing device 1 in the learning phase in the first embodiment.

[0060] First, the information processing device 1 acquires training data used for learning the stress estimation model, and attribute information and / or environmental information of sample subjects corresponding to the training data (step S11). In this case, for example, the information processing device 1 acquires the training data and the environmental information from the training data storage unit 42, and acquires the attribute information from the attribute information storage unit 40.

[0061] Then, the first division unit 14 of the information processing device 1 performs a first division to divide the observed features of the training data based on at least one of the attributes of the sample subjects indicated by the attribute information or the environment at the time of measurement indicated by the environmental information such as date and time information (e.g., season, etc.) (step S12). As a result, the first division unit 14 divides the observed features so as to generate N clusters.

[0062] Next, the second division unit 15 of the information processing device 1 divides the observed features based on a second division according to the observation target of the observed features and the activity state of the corresponding sample subject at the time of observation (step S13). In this case, for example, the second division unit 15 generates sub-clusters of M observed features for each cluster of N observed features based on the second division according to the type of observed biometric feature, the exercise intensity of the sample subject, etc.

[0063] Next, the feature selection unit 16 of the information processing device 1 randomly generates groups for each sub-cluster, and calculates the correlation between each type of observed feature and the stress data included in the training data in the generated groups (step S14). Furthermore, the feature selection unit 16 ranks the types of observed feature according to the correlation and the degree of sign reversal for each sub-cluster, and selects the top R types of observed feature as stress estimation features (step S15).

[0064] Then, the estimation model learning unit 17 of the information processing device 1 learns a stress estimation model for each cluster divided by the first division based on the stress estimation feature and the stress data included in the training data indicating the corresponding correct stress value (step S16). Then, the information processing device 1 outputs, as a learning result, feature selection information Ifs related to the stress estimation feature selected in step S15 and parameters of the stress estimation model learned in step S16. Specifically, the information processing device 1 stores the feature selection information Ifs and the parameters of the stress estimation model in the learning parameter storage unit 43. This allows the information processing device 1 to store information required in the estimation phase in the storage device 4.

[0065] (4) Estimation Phase Next, a description will be given of the processing in the estimation phase executed by the information processing device 1. The information processing device 1 estimates the stress value of the estimation subject based on the stress estimation model learned in the learning phase.

[0066] 8 is an example of a functional block of the information processing device 1 in the estimation phase. In the estimation phase, the processor 11 of the information processing device 1 functionally has a classification unit 34, N feature quantity selection units 36 (361 to 36N), and N stress estimation units 37 (371 to 37N). Here, the first estimation model information storage unit 431 to the Nth estimation model information storage unit 43N included in the learning parameter storage unit 43 store parameters of N stress estimation models that have already been learned in the learning phase.

[0067] The classification unit 34 extracts the observation feature of the estimation subject from the observation data storage unit 41, and classifies the extracted observation feature according to the stress estimation model (first estimation model to Nth estimation model) to be used based on at least one of the corresponding attribute information or environmental information. The classification method of the observation feature by the classification unit 34 is the same as the cluster generation method performed in the first division by the first division unit 14. Therefore, the observation feature of the estimation subject corresponding to the attribute information or environmental information of the same classification as the cluster supplied by the first division unit 14 to the second division unit 15n (n is an arbitrary integer from 1 to N) is supplied to the feature selection unit 36n. That is, the second division unit 15n and the feature selection unit 36n handle the observation feature of the same classification.

[0068] The classification unit 34 is not limited to a classification mode in which the observed feature amount of the estimation subject is assigned to any one of the stress estimation models, and may classify the observed feature amount so as to assign the observed feature amount to two or more stress estimation models in a duplicated manner.

[0069] The feature quantity selection units 36 (361 to 36N) select stress estimation feature quantities from the observation feature quantities supplied from the classification unit 34, based on the feature quantity selection information Ifs stored in the learning parameter storage unit 43. In this case, the feature quantity selection units 36n (n is an arbitrary integer from 1 to N) each extract, from the observation feature quantities supplied by the classification unit 34, an observation feature quantity of the same type as the type of stress estimation feature quantity indicated by the feature quantity selection information Ifs generated by the feature quantity selection units 16n1 to 16nM, as a stress estimation feature quantity. Then, the feature quantity selection units 36n supply the extracted stress estimation feature quantities to the corresponding stress estimation units 37n.

[0070] The stress estimation unit 37 (371 to 37N) estimates the stress value of the estimation subject based on the stress estimation model. In this case, the stress estimation unit 37n (n is any integer from 1 to N) configures the corresponding nth estimation model by referring to the corresponding nth estimation model information storage unit 43n. Then, the stress estimation unit 37n inputs the stress estimation feature supplied from the corresponding feature selection unit 36n to the configured nth estimation model, thereby acquiring the stress value of the estimation subject output by the nth estimation model. Then, the stress estimation unit 37n supplies the stress value output by the nth estimation model to the output control unit 38.

[0071] The output control unit 38 performs output based on the estimated stress value (estimated stress value) of the estimation subject. For example, the output control unit 38 generates a display signal S2 for displaying information related to the estimated stress value, and supplies the display signal S2 to the display device 3, thereby causing the display device 3 to display the information related to the estimated stress value. Here, when the output control unit 38 acquires stress values ​​from a plurality of stress estimation units 37, it causes the display device 3 to display a statistical value representing the average, median, maximum, or other representative statistical value of the acquired plurality of stress values ​​as the estimated stress value of the estimation subject.

[0072] Note that instead of or in addition to controlling the display of the stress estimated value itself, the output control unit 38 may control the display of information regarding the level of stress determined based on a comparison between the stress estimated value and a predetermined threshold value, and / or information regarding advice according to the level. Note that the viewer of the display device 3 in this case may be, for example, the person to be estimated, or a person who manages or supervises the person to be estimated. Also, the output control unit 38 may output the information regarding the stress estimated value as sound by a sound output device not shown.

[0073] 9 is an example of a flowchart showing the procedure of the stress estimation process executed in the estimation phase by the information processing device 1. The timing of performing the stress estimation process may be a timing requested by the user based on the input signal S1, or may be a predetermined timing.

[0074] First, the information processing device 1 acquires the observed feature amount of the subject to be estimated, and the attribute information and / or the environmental information of the subject to be estimated (step S21). In this case, for example, the information processing device 1 acquires the observed feature amount and the environmental information from the observed data storage unit 41, and acquires the attribute information from the attribute information storage unit 40.

[0075] Next, the classification unit 34 of the information processing device 1 classifies the observed feature based on at least one of the attributes of the estimation subject indicated by the attribute information or the environment (e.g., season, etc.) at the time of measurement indicated by the environmental information such as date and time information (step S22). As a result, the classification unit 34 classifies the observed feature into at least one of the first estimation model to the Nth estimation model. Note that the observed feature may be classified into multiple stress estimation models in a duplicated manner.

[0076] Then, the feature quantity selection unit 36 ​​of the information processing device 1 selects the observed feature quantity to be input to the target stress estimation model (step S23). In this case, when the observed feature quantity is supplied from the classification unit 34, the feature quantity selection unit 36 ​​refers to the corresponding feature quantity selection information Ifs and selects the stress estimation feature quantity that is the observed feature quantity to be input to the stress estimation model.

[0077] Then, the stress estimation unit 37 of the information processing device 1 calculates a stress estimation value based on the target stress estimation model (step S24). In this case, when the stress estimation feature is supplied from the feature selection unit 36, the stress estimation unit 37 configures a corresponding stress estimation model by referring to the learning parameter storage unit 43, and inputs the stress estimation feature to the configured stress estimation model to calculate a stress estimation value. Note that, when a plurality of stress estimation models are used, the information processing device 1 determines a stress estimation value that integrates the estimation results of these stress estimation models. Then, the output control unit 38 of the information processing device 1 outputs information related to the stress estimation value (step S25).

[0078] (5) Variations The stress estimation model may be provided for each sub-cluster formed by the first and second divisions, instead of for each cluster formed by the first division.

[0079] In this case, in the learning phase, the information processing device 1 provides stress estimation models corresponding to the N×M feature quantity selection units 1611-16NM, and learns these stress estimation models using the stress estimation feature quantities output by the corresponding feature quantity selection units 16 as input data and the stress values ​​indicated by the corresponding stress data as correct answer data. Also, in the estimation phase, there are N×M feature quantity selection units 36, similar to the feature quantity selection units 16 in the learning phase, and each stress estimation unit 37 inputs the stress estimation feature quantities output by the corresponding M feature quantity selection units 36 to the corresponding M stress estimation models. As a result, the stress estimation unit 37 acquires M stress values ​​and determines a stress estimation value by integrating the M stress values.

[0080] In this way, even in this modified example, the information processing device 1 can accurately estimate the stress state of the person to be estimated from observed features not used in learning, based on the stress estimation model learned for each cluster that has a bias in stress tendency.

[0081] Moreover, the stress estimated by the information processing device 1 is not limited to chronic stress, and may be short-term stress, which is stress lasting for a relatively short period of time (several minutes to a day).

[0082] <Second embodiment> 10 shows a schematic configuration of a stress estimation system 100A in the second embodiment. The stress estimation system 100A in the second embodiment has a stress estimation device 1A that performs processing in the estimation phase of the information processing device 1 in the first embodiment, a learning device 1B that performs processing in the learning phase of the information processing device 1 in the first embodiment, and a terminal device 8 and a sensor 5 used by a person to be estimated. Hereinafter, the same components as those in the first embodiment are appropriately denoted by the same reference numerals, and the description thereof will be omitted.

[0083] 10, the stress estimation system 100A mainly includes a stress estimation device 1A functioning as a server, a storage device 4, and a terminal device 8 functioning as a client. The stress estimation device 1A and the terminal device 8 perform data communication via a network 7.

[0084] The learning device 1B has the same hardware configuration as the hardware configuration of the information processing device 1 shown in Fig. 2, and the processor 11 of the learning device 1B has the functional blocks shown in Fig. 3. Then, based on the information stored in the storage device 4, the learning device 1B performs learning processes such as updating parameters of the stress estimation model and generating feature quantity selection information Ifs.

[0085] The terminal device 8 is a terminal used by a user who is to be an estimation subject, and has an input function, a display function, and a communication function, and functions as the input device 2 and the display device 3 shown in FIG. 1. The terminal device 8 may be, for example, a personal computer, a tablet terminal such as a smartphone, or a PDA (Personal Digital Assistant). The terminal device 8 is electrically connected to a sensor 5 such as a wearable sensor worn by a user, and transmits a biological signal of the estimation subject output by the sensor 5 (i.e., information corresponding to the sensor signal S3 in FIG. 1) to the stress estimation device 1A via the network 7. The terminal device 8 also accepts user input related to a questionnaire answer, and transmits information generated by the user input (information corresponding to the input signal S1 in FIG. 1) to the stress estimation device 1A.

[0086] The stress estimation device 1A has the same hardware configuration as the information processing device 1 shown in Fig. 2, and the processor 11 of the stress estimation device 1A has the functional blocks shown in Fig. 8. The stress estimation device 1A receives information corresponding to the input signal S1 and the sensor signal S3 in Fig. 1 from the terminal device 8 via the network 7, and stores the received information in the storage device 4. The stress estimation device 1A then refers to the parameters of the stress estimation model learned by the learning device 1B and the feature quantity selection information Ifs, and executes stress estimation processing for the estimation subject. The stress estimation device 1A also transmits an output signal for outputting the stress estimation result to the terminal device 8 via the network 7, based on a display request from the terminal device 8.

[0087] In this way, in the stress estimation system 100A in the second embodiment, the learning phase and the estimation phase are executed by different devices, and learning of the stress estimation model and stress estimation using the stress estimation model can be performed in the same manner as in the first embodiment. Also, in the second embodiment, the stress estimation device 1A estimates the stress state of the estimation subject based on the biological signal of the estimation subject received from a terminal used by the estimation subject, and can suitably present the estimation result to the estimation subject on the terminal.

[0088] <Third embodiment> 11 is a block diagram of a learning device 1BX in the third embodiment. The information processing device 1X mainly includes a first division means 14X, a second division means 15X, a feature selection means 16X, and a learning means 17X. The learning device 1BX may be composed of multiple devices.

[0089] The first division means 14X performs a first division that divides the observed feature of the subject based on at least one of the subject's attributes or environment. The first division means 14X can be, for example, the first division section 14 in the first embodiment (including modified examples, the same applies below) or the second embodiment. The second division means 15X performs a second division that divides the observed feature based on at least one of the observed target of the observed feature or the subject's activity state. The second division means 15X can be, for example, the second division section 15 in the first embodiment or the second embodiment.

[0090] The feature selection means 16X selects a stress estimation feature, which is a feature used for stress estimation, from the observed feature divided based on the first division and the second division. The feature selection means 16X can be, for example, the feature selection unit 16 in the first or second embodiment. The learning means 17X learns a stress estimation model for at least each group divided by the first division, based on the stress estimation feature and the correct stress value corresponding to the stress estimation feature. The learning means 17X can be, for example, the estimation model learning unit 17 in the first or second embodiment.

[0091] FIG. 12 is an example of a flowchart executed by the learning device 1BX in the third embodiment. First, the first division means 14X performs a first division to divide the observation feature of the subject based on at least one of the subject's attributes or environment (step S31). The second division means 15X performs a second division to divide the observation feature based on at least one of the observation target of the observation feature or the subject's activity state (step S32). Furthermore, the feature selection means 16X selects a stress estimation feature, which is a feature used for stress estimation, from the observation feature divided based on the first division and the second division (step S33). Then, the learning means 17X performs learning of a stress estimation model at least for each group divided by the first division based on the stress estimation feature and the correct stress value corresponding to the stress estimation feature (step S34).

[0092] According to the third embodiment, the learning device 1BX learns a stress estimation model for each group having a bias in stress tendency, and can learn a stress estimation model capable of performing stress estimation with high accuracy.

[0093] In each of the above-described embodiments, the program can be stored using various types of non-transitory computer readable media and can be supplied to a computer, such as a processor. The non-transitory computer readable medium includes various types of tangible storage media. Examples of the non-transitory computer readable medium include magnetic storage media (e.g., flexible disks, magnetic tapes, hard disk drives), magneto-optical storage media (e.g., magneto-optical disks), CD-ROM (Read Only Memory), CD-R, CD-R / W, and semiconductor memory (e.g., mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, and RAM (Random Access Memory)). The program may also be supplied to the computer by various types of transitory computer readable media. Examples of the transitory computer readable medium include electric signals, optical signals, and electromagnetic waves. The transitory computer readable medium can supply the program to the computer via a wired communication path such as an electric wire and an optical fiber, or a wireless communication path.

[0094] In addition, some or all of the above embodiments may be described as, but are not limited to, the following supplementary notes.

[0095] [Appendix 1] A first division means for performing a first division of dividing an observed feature amount of a subject based on at least one of an attribute or an environment of the subject; A second division means for performing a second division to divide the observation feature based on at least one of an observation target of the observation feature or an activity state of the target person; a feature selection means for selecting a stress estimation feature, which is a feature used for stress estimation, from the observation feature divided based on the first division and the second division; A learning means for learning a stress estimation model for at least each group divided by the first division based on the stress estimation feature amount and a correct stress value corresponding to the stress estimation feature amount; A learning device having the above configuration. [Appendix 2] The learning device according to claim 1, wherein the feature selection means selects the stress estimation feature based on a correlation between the observed feature divided based on the first division and the second division and the stress value. [Appendix 3] The learning device according to claim 2, wherein the feature selection means forms a plurality of groups by random sampling from the observed features divided based on the first division and the second division, and selects the stress estimation feature based on an aggregate result of the correlation calculated for each of the groups across the plurality of groups. [Appendix 4] The learning device according to claim 3, wherein the feature selection means calculates, as the aggregation result, a statistical value of the correlation among the plurality of groups and a ratio of positive and negative signs of the correlation among the plurality of groups. [Appendix 5] The learning device according to claim 1, wherein the learning means learns the stress estimation model for each group divided by the first division and the second division. [Appendix 6] The learning device according to any one of appendix 1 to 5, wherein the learning means outputs, as learning results, feature selection information which is information relating to the stress estimation feature selected by the feature selection means, and parameters of the stress estimation model learned by the learning means. [Appendix 7] The learning device according to any one of appendix 1 to 6, wherein the first division means generates pseudo data indicating a stress value when the correct stress value is not measured by interpolating the correct stress value. [Appendix 8] The learning device according to any one of claims 1 to 7, wherein the first division means performs the first division such that a part of the observed features overlaps and belongs to two or more groups. [Appendix 9] A classification means for classifying an observed feature amount of a subject to be estimated based on at least one of an attribute or an environment of the subject to be estimated; a feature selection means for selecting, from the observed feature values ​​based on the classification, a stress estimation feature value to be used for stress estimation; a stress estimation means for selecting a stress estimation model based on the classification and inputting the stress estimation feature amount into the selected stress estimation model, thereby estimating a stress value of the estimation subject; A stress estimation device having the above structure. [Appendix 10] The stress estimation device according to claim 9, wherein the stress estimation means estimates the stress value based on a stress estimation model learned by a learning device according to any one of claims 1 to 8. [Appendix 11] The stress estimation device according to claim 9 or 10, wherein the feature selection means selects, as the stress estimation feature, an observed feature of the same type as the stress estimation feature selected by the learning device according to any one of claims 1 to 8. [Appendix 12] The classification means classifies the observed feature quantities into a plurality of categories, The stress estimation device according to any one of claims 9 to 11, wherein the stress estimation means estimates the stress value by integrating estimation results output by a plurality of stress estimation models selected based on the plurality of classifications. [Appendix 13] The computer performing a first division of the observed feature of the subject based on at least one of an attribute or an environment of the subject; performing a second division of the observed feature based on at least one of an observed object of the observed feature or an activity state of the subject; selecting stress estimation features, which are features to be used for stress estimation, from the observation features divided based on the first division and the second division; learning a stress estimation model for at least each group divided by the first division based on the stress estimation feature amount and a correct stress value corresponding to the stress estimation feature amount; How to learn. [Appendix 14] The computer classifying the observed feature amount of the subject to be estimated based on at least one of the attributes or the environment of the subject to be estimated; selecting, from the observed features based on the classification, stress estimation features to be used for stress estimation; selecting a stress estimation model based on the classification, and inputting the stress estimation feature amount into the selected stress estimation model, thereby estimating a stress value of the estimation subject; Stress estimation methods. [Appendix 15] performing a first division of the observed feature of the subject based on at least one of an attribute or an environment of the subject; performing a second division of the observed feature based on at least one of an observed object of the observed feature or an activity state of the subject; selecting stress estimation features, which are features to be used for stress estimation, from the observation features divided based on the first division and the second division; A storage medium storing a program that causes a computer to execute a process of learning a stress estimation model for at least each group divided by the first division based on the stress estimation feature and a correct stress value corresponding to the stress estimation feature. [Appendix 16] classifying the observed feature amount of the subject to be estimated based on at least one of the attributes or the environment of the subject to be estimated; selecting, from the observed features based on the classification, stress estimation features to be used for stress estimation; A storage medium storing a program that causes a computer to execute a process of estimating a stress value of the person to be estimated by selecting a stress estimation model based on the classification and inputting the stress estimation feature into the selected stress estimation model.

[0096] Although the present invention has been described above with reference to the embodiments, the present invention is not limited to the above-mentioned embodiments. Various modifications that can be understood by a person skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention. In other words, the present invention naturally includes various modifications and corrections that a person skilled in the art could make in accordance with the entire disclosure, including the claims, and the technical ideas. In addition, the disclosures of the above cited patent documents and the like are incorporated herein by reference. [Explanation of symbols]

[0097] 1. Information processing device 1A Stress Estimation Device 1B, 1BX Learning Device 2 Input Devices 3 Display device 4 Storage device 5 Sensors 8 Terminal Equipment 100, 100A Stress Estimation System

Claims

1. A first division means for performing a first division of dividing an observed feature amount of a subject into clusters based on at least one of an attribute or an environment of the subject; A second division means for performing a second division in which each of the clusters is divided into sub-clusters based on at least one of an observed target of the observation feature or an activity state of the subject; a feature selection means for selecting, from each sub-cluster of the observation feature, a type of observation feature that has a top predetermined number of correlation scores based on a correlation coefficient between each of the sub-clusters and stress data corresponding to the observation feature of each of the sub-clusters, as a stress estimation feature that is a feature to be used for stress estimation; a learning means for learning the stress estimation model for at least each of the clusters divided by the first division, using the stress estimation feature as input data for a stress estimation model and a set of the stress estimation feature as correct answer data as training data; A learning device having the above configuration.

2. 2. The learning device according to claim 1 , wherein the feature selection means forms a plurality of groups by randomly sampling from each of the sub-clusters, and calculates the correlation score based on a summary result of the correlation calculated for each of the groups across all of the plurality of groups.

3. 3. The learning device according to claim 2, wherein the feature selection means calculates, as the aggregation result, a statistical value of the correlation in the plurality of groups and a ratio of positive and negative signs of the correlation in the plurality of groups, and calculates the correlation score based on the statistical value of the correlation and the ratio.

4. The learning device according to claim 1 , wherein the learning means performs learning of the stress estimation model for each of the sub-clusters.

5. The learning device according to any one of claims 1 to 4, wherein the learning means outputs, as learning results, feature selection information which is information relating to the stress estimation feature selected by the feature selection means, and parameters of the stress estimation model learned by the learning means.

6. The learning device according to any one of claims 1 to 5, wherein the first division means generates pseudo data indicating a stress value when the correct stress value is not measured by interpolating the correct stress value.

7. The learning device according to claim 1 , wherein the first division means performs the first division such that a part of the observed features overlaps and belongs to two or more groups.

8. A classification means for classifying an observed feature amount of a subject to be estimated based on at least one of an attribute or an environment of the subject to be estimated; a feature selection means for selecting the observation feature corresponding to the type corresponding to the classification as a stress estimation feature to be used for stress estimation, by referring to feature selection information in which the type of the observation feature used as input data of training data used in machine learning of the stress estimation model to be used for the stress estimation is associated with the classification; a stress estimation means for selecting a stress estimation model based on the classification and inputting the stress estimation feature amount into the selected stress estimation model, thereby estimating a stress value of the estimation subject; A stress estimation device having the above structure.

9. The stress estimation device according to claim 8, wherein the stress estimation means estimates the stress value based on a stress estimation model learned by a learning device according to any one of claims 1 to 7.

10. The stress estimation device according to claim 8 or 9, wherein the feature selection means selects, as the stress estimation feature, an observed feature of the same type as a stress estimation feature selected by the learning device according to any one of claims 1 to 7.

11. The classification means classifies the observed feature quantities into a plurality of categories, The stress estimation device according to any one of claims 8 to 10, wherein the stress estimation means estimates the stress value by integrating estimation results output by a plurality of stress estimation models selected based on the plurality of classifications.

12. The computer performing a first division in which the observed feature of the subject is divided into clusters based on at least one of an attribute or an environment of the subject; performing a second division in which each of the clusters is divided into sub-clusters based on at least one of the observed object of the observation feature or the activity state of the subject; selecting, from each sub-cluster of the observation feature, a type of observation feature that has a top predetermined number of correlation scores based on a correlation coefficient between each of the sub-clusters and stress data corresponding to the observation feature of each of the sub-clusters as a stress estimation feature that is a feature to be used for stress estimation; the stress estimation feature is used as input data for a stress estimation model, and a set of the stress data corresponding to the stress estimation feature as correct answer data is used as training data, and the stress estimation model is trained for at least each of the clusters divided by the first division. How to learn.

13. The computer classifying the observed feature amount of the subject to be estimated based on at least one of the attributes or the environment of the subject to be estimated; referring to feature selection information that associates types of observation features used as input data of training data used in machine learning of the stress estimation model used in the stress estimation with the classification, and selecting the observation feature corresponding to the type corresponding to the classification as a stress estimation feature that is a feature to be used in stress estimation; selecting a stress estimation model based on the classification, and inputting the stress estimation feature amount into the selected stress estimation model, thereby estimating a stress value of the estimation subject; Stress estimation methods.

14. performing a first division in which the observed feature of the subject is divided into clusters based on at least one of an attribute or an environment of the subject; performing a second division in which each of the clusters is divided into sub-clusters based on at least one of the observed object of the observation feature or the activity state of the subject; selecting, from each sub-cluster of the observation feature, a type of observation feature that has a top predetermined number of correlation scores based on a correlation coefficient between each of the sub-clusters and stress data corresponding to the observation feature of each of the sub-clusters as a stress estimation feature that is a feature to be used for stress estimation; A program that causes a computer to execute a process of learning the stress estimation model, at least for each of the clusters divided by the first division, using the stress estimation feature as input data to a stress estimation model and a set of stress data corresponding to the stress estimation feature as correct answer data.

15. classifying the observed feature amount of the subject to be estimated based on at least one of the attributes or the environment of the subject to be estimated; referring to feature selection information that associates types of observation features used as input data of training data used in machine learning of the stress estimation model used in the stress estimation with the classification, and selecting the observation feature corresponding to the type corresponding to the classification as a stress estimation feature that is a feature to be used in stress estimation; A program that causes a computer to execute a process of estimating a stress value of the estimation subject by selecting a stress estimation model based on the classification and inputting the stress estimation feature into the selected stress estimation model.

Images