Information Processing Method

Abstract

An information processing device 100 according to the present invention is equipped with: a conversion unit 121 which maps correct data, from among training data which includes status data representing the status of a prescribed person, and correct data which represents the physical condition of the prescribed person when the status data is obtained, onto a non-linear space; and a generation unit 122 for generating an estimation model which is used in order to estimate the physical condition of a target person, by training with the status data as the explanatory variable and the mapped correct data as the objective variable.

Description

[Technical field]

[0001] The present invention relates to an information processing method, an information processing device, and a program for estimating a person's physical condition. [Background technology]

[0002] A known method of estimating a person's stress involves learning and generating a model for calculating a stress value in advance, and inputting data from the person that is the cause of estimating stress, such as biometric data, into the model to calculate the stress value. In this case, the model is generated by learning the feature amount of the person's biometric data as explanatory variables, and the person's stress value obtained by a questionnaire or the like as a target variable. For example, Patent Document 1 describes a method of estimating stress using a model. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2021 / 090402 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the method of generating a model as described above, the learning data includes a large amount of data from typical people, that is, people who are neither high stress nor low stress, and there is a risk that data with stress values ​​extremely biased toward the median will be collected. Therefore, if a model is generated by learning data biased toward the median, the problem occurs that the accuracy of stress estimation for people with high or low stress decreases. As a result, it is difficult to estimate stress with high accuracy. In addition to stress, it is also difficult to estimate physical conditions such as a person's physical and mental fatigue and inner condition with high accuracy.

[0005] Therefore, an object of the present invention is to provide an information processing method capable of solving the above-mentioned problem that it is difficult to estimate physical condition with higher accuracy. [Means for solving the problem]

[0006] An information processing method according to one aspect of the present invention includes: Among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired, the correct answer data is mapped onto a nonlinear space; generating an estimation model to be used for estimating the physical condition of a target person by learning the state data as an explanatory variable and the mapped correct answer data as a target variable; The structure is as follows.

[0007] Further, an information processing device according to one aspect of the present invention includes: A conversion unit that maps the correct answer data into a nonlinear space from among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired; a generation unit that generates an estimation model to be used for estimating the physical condition of a target person by learning the state data as an explanatory variable and the mapped ground truth data as a target variable; Equipped with The structure is as follows.

[0008] In addition, a program according to one aspect of the present invention includes: In the information processing device, Among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired, the correct answer data is mapped onto a nonlinear space; generating an estimation model to be used for estimating the physical condition of a target person by learning the state data as an explanatory variable and the mapped correct answer data as a target variable; Execute the process, The structure is as follows. Effect of the Invention

[0009] By being configured as described above, the present invention can estimate a physical condition with higher accuracy. [Brief description of the drawings]

[0010] [Figure 1] 1 is a block diagram showing a configuration of a stress estimation device according to a first embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram showing an example of a test for acquiring subjective data on a person's stress to be input to the stress estimation device disclosed in FIG. 1. [Diagram 3] FIG. 2 is a diagram showing an example of a function used when generating a model in the stress estimation device disclosed in FIG. 1. [Figure 4] 2 is a flowchart showing an operation of the stress estimation device disclosed in FIG. 1. [Diagram 5] 2 is a flowchart showing an operation of the stress estimation device disclosed in FIG. 1. [Figure 6] FIG. 11 is a block diagram showing a hardware configuration of an information processing device according to a second embodiment of the present invention. [Figure 7] FIG. 11 is a block diagram showing a configuration of an information processing device according to a second embodiment of the present invention. [Figure 8] 10 is a flowchart showing the operation of an information processing device according to a second embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] <Embodiment 1> A first embodiment of the present invention will be described with reference to Figures 1 to 5. Figures 1 to 3 are diagrams for explaining the configuration of a stress estimation device, and Figures 4 to 5 are diagrams for explaining the processing operation of the stress estimation device.

[0012] [composition] The stress estimation device 10 (information processing device) of the present invention is used to estimate the stress of a person. For example, the stress estimation device 10 is used to calculate a stress value under a predetermined situation in which a person is performing work in a workplace to which the person belongs. However, the stress estimation device 10 of the present invention may calculate the stress of a person under any situation. In addition, the present invention is not limited to estimating stress, and can also be applied to estimating a person's physical condition such as physical and mental fatigue and inner condition. In other words, the stress value given in this embodiment is an example of a value of the physical condition of a person to be estimated, and other examples of the value of the physical condition may be any value, such as a fatigue level indicating the degree of fatigue or some index value indicating the condition.

[0013] The stress estimation device 10 is composed of one or more information processing devices each including a calculation device and a storage device. As shown in FIG. 1, the stress estimation device 10 includes a data acquisition unit 11, a preprocessing unit 12, a learning unit 13, a calculation unit 14, and an output unit 15. The functions of the data acquisition unit 11, the preprocessing unit 12, the learning unit 13, the calculation unit 14, and the output unit 15 can be realized by the calculation device executing a program for realizing each function stored in the storage device. The stress estimation device 10 also includes a person information storage unit 16 and a model storage unit 17. The person information storage unit 16 and the model storage unit 17 are each composed of a storage device. Each component will be described in detail below.

[0014] The data acquisition unit 11 acquires data used to estimate a person's stress. First, the data acquisition unit 11 acquires learning data for generating a stress estimation model used to calculate a stress value by machine learning. The learning data is data on a large number of arbitrary people (predetermined people) and is stored in the person information storage unit 16 as described below.

[0015] Specifically, the data acquisition unit 11 acquires, as learning data, status data representing a state of a person under a predetermined situation, such as a situation in which the person is performing work at a workplace. The status data is, for example, biometric data, which is various information emitted from the body of the person. The biometric data is, for example, a heart rate, an eye opening degree, and the like. For example, as shown in FIG. 1, the data acquisition unit 11 acquires, as status data, biometric data such as the heart rate of the person U via a measurement device such as a wearable terminal W worn by the person U, and an eye opening degree extracted from a face image of the person U captured by a camera (not shown). At this time, the data acquisition unit 11 measures the biometric data by distinguishing between the person and the period, and acquires the biometric data as status data. However, the data acquisition unit 11 may acquire any biometric data using any measurement device as the status data of the person, and may acquire other data representing the state of the person, not limited to biometric data.

[0016] Furthermore, the data acquisition unit 11 acquires, as the learning data, correct answer data representing the stress state of a specific person when the biometric data was acquired from the person as described above. For example, the correct answer data is data based on subjective data regarding the person's stress. For this reason, the data acquisition unit 11 asks the person a question set in advance via the input device 20 during work, at the end of work, or after a certain period of time has passed, and acquires and tallys up answers from person U, thereby acquiring data based on the subjective data regarding the stress of person U as correct answer data. At this time, the data acquisition unit 11 acquires the correct answer data by distinguishing between people and periods.

[0017] As an example, the data acquisition unit 11 acquires a "PSS score" obtained by aggregating "Perceived Stress Scale (PSS)" as correct answer data that is data based on subjective data regarding a person's stress. Here, the PSS is, as shown in FIG. 2 as an example, a set of 14 questions that ask the user how they feel about what is happening, and five levels of answers are prepared. A score of 0 to 4 points is given to each of the five levels of answers, and the sum of the scores of the answers to the 10 questions that are the subject of calculation among the 14 questions is calculated as the PSS score. Therefore, the PSS score has a value of 0 to 40 points. As an example, the data acquisition unit 11, as shown in FIG. 1, displays PSS questions as shown in FIG. 2 from an input device 20 such as an information processing terminal operated by the person U, acquires answers inputted from the person U to the input device 20, and acquires the PSS score by aggregating the answers. However, the data acquiring unit 11 is not limited to acquiring data based on subjective data regarding a person's stress as the correct answer data, and may acquire data representing the stress state acquired by another method as the correct answer data.

[0018] Then, the data acquiring unit 11 associates the status data and correct answer data acquired from a specific person as described above with each person and each period, and stores them as learning data in the person information storage unit 16. For example, when acquiring status data and correct answer data from a person every month, the data acquiring unit 11 associates the status data and correct answer data for each month with each person and stores them.

[0019] The preprocessing unit 12 (conversion unit) performs preprocessing of the learning data, which is obtained as described above and stored in the person information storage unit 16, before performing machine learning using the learning data consisting of the state data and the correct answer data. First, the preprocessing unit 12 maps the correct answer data into a nonlinear space for a data set consisting of the state data and the correct answer data. At this time, it is preferable that the preprocessing unit 12 maps the correct answer data into the nonlinear space using a monotonically increasing function with a rate of change of 1 or more so that the magnitude relationship of the correct answer data does not change before and after the mapping. Furthermore, it is preferable that the preprocessing unit 12 maps the correct answer data into the nonlinear space using a function that can define an inverse function so that the stress value calculated using the model can be inversely mapped into a linear space as described later. For this reason, in this embodiment, the correct answer data is mapped into the nonlinear space using any one of the cubic functions shown by symbols M1 and M2 in FIG. 3, the exponential function shown by symbol M3, and the quadratic function shown by symbol M4. Note that the "t" in each function is "t=x-med", where "x" represents the PSS score, which is the correct data, and "med" is the median of the entire distribution of the correct data. In this way, by mapping the correct data into a nonlinear space using each function as shown in FIG. 3, even if the correct data is biased toward the median, as shown in the graph on the right side of FIG. 3, it is possible to lower the weight of values ​​close to the median and increase the weight of values ​​far from the median. Note that, of the functions shown in FIG. 3, it is preferable to use a quadratic function that can equalize the weights, but any function may be used, and any function not listed in FIG. 3 may be used.

[0020] Next, the preprocessing unit 12 extracts multiple types of features from the biometric data, which is state data, from the learning data for each person and each period to generate feature data. For example, the preprocessing unit 12 extracts multiple types of feature data, such as the average value, variance / standard deviation, maximum value, minimum value, and quartiles in the time domain, the average value, variance, and quartiles of the histogram in the frequency domain, and even the peak of the power spectrum, from the biometric data acquired along a time series. Then, the preprocessing unit 12 associates the mapped correct answer data with each type of feature data to generate a dataset of the feature data and the mapped correct answer data. That is, the preprocessing unit 12 generates a dataset of each feature data and the mapped correct answer data for each period for each person. As an example, in the case where learning data consisting of three months' worth of state data and the mapped correct answer data is acquired for one person, if five types of feature data are extracted from the state data, five datasets are generated for one month, and 15 datasets are generated for three months. Then, such datasets are generated for the number of people. Note that the preprocessing unit 12 may generate multiple types of feature data from the state data before mapping the correct answer data to the nonlinear space as described above. In this case, the preprocessing unit 12 associates correct answer data with each type of feature data to generate a data set of the feature data and the correct answer data, and then maps the correct answer data in each data set to the nonlinear space to generate a data set of the feature data and the mapped correct answer data.

[0021] Next, the preprocessing unit 12 selects the type of feature data to be used for machine learning as learning data from among the feature data extracted from the state data as described above. At this time, the preprocessing unit 12 calculates a correlation value representing the degree of correlation between the feature data and the mapped correct answer data for each type of feature data, and selects the type of feature data based on the correlation value. For example, the preprocessing unit 12 calculates a correlation value representing the degree to which the data distribution of the mapped correct answer data and the data distribution of the feature data are similar to each other or correlated with each other based on a preset criterion for the mapped correct answer data and feature data that form all data sets belonging to a certain type. Then, the preprocessing unit 12 calculates a correlation value for each type. In other words, when there are five types of the above-mentioned feature data, the preprocessing unit 12 calculates the correlation value for each of the five types, and selects the feature data of the top N types with the highest correlation values ​​or the feature data of the type whose correlation value exceeds a preset threshold value. Note that the preprocessing unit 12 may select the type of feature data based on any criterion. For example, when the pre-processing unit 12 extracts feature data from the state data before mapping the super-correct data and generates a dataset of the feature data and the super-correct data, it may select feature data from the distribution of the super-correct data and the feature data before mapping, and then map the super-correct data into a non-linear space in the same manner as described above.

[0022] The learning unit 13 (generation unit) performs machine learning using a data set of the selected feature amount data and the correct answer data mapped to the nonlinear space as learning data, as described above, to generate a stress estimation model (estimation model). Specifically, the learning unit 13 performs machine learning using all data sets including the selected type of feature amount data as learning data, with the feature amount data as explanatory variables and the mapped correct answer data as objective variables. As a result, the generated stress estimation model can be a model configured to input the biometric data of person U and calculate a value of 0-40 as a value mapped to the nonlinear space, similar to the above-mentioned PSS score. Then, the learning unit 13 stores the stress estimation model generated by machine learning in the model storage unit 17.

[0023] Next, the functions of each part when estimating the stress of a target person (target person) using the stress estimation model after generating the stress estimation model as described above will be described. For example, assume that the target person U belongs to a certain workplace and is in a certain situation, such as working at the workplace, and the stress is estimated.

[0024] The data acquisition unit 11 acquires state data (target person measurement data) used to estimate the stress of a target person U. Specifically, the data acquisition unit 11 measures and acquires state data, which is biometric data, from the target person U when the target person U is in a situation where the stress is actually estimated, such as working in a workplace. For example, the data acquisition unit 11 acquires the heart rate of the target person U as biometric data through a measurement device such as a wearable terminal W worn by the target person U, as described above, or acquires the degree of eye opening extracted from a face image of the target person U captured by a camera (not shown) as biometric data. At this time, the data acquisition unit 11 acquires biometric data at a preset timing, for example, at regular time intervals, from the target person U in a situation where the target person U is working in a workplace.

[0025] As described above, the calculation unit 14 calculates feature amount data of the biometric data every time the calculation unit 14 acquires the biometric data from the person U. At this time, the calculation unit 14 extracts, from the biometric data, feature amount data (target person feature amount data) of the same type as the type of feature amount data selected by the pre-processing unit 12 as described above. Note that the calculation unit 14 may calculate all types of feature amount data set in advance from the biometric data, and extract the feature amount data of the same type as the selected type from the feature amount data.

[0026] Then, the calculation unit 14 reads out the stress estimation model stored in the model storage unit 17, and inputs the same type of feature amount data as the feature amount data selected by the preprocessing unit 12 to the stress estimation model, thereby calculating the stress value that is the output of the stress estimation model. That is, after the person U starts work, the calculation unit 14 calculates the stress value at each preset timing, such as a certain time interval for acquiring biometric data from the person U as described above. Note that the timing for calculating the stress value may be during work, or may be any interval, such as once a month.

[0027] Furthermore, the calculation unit 14 (inverse conversion unit) inversely maps the stress value calculated using the stress estimation model as described above into a linear space. At this time, the calculation unit 14 inversely maps the stress value calculated by the stress estimation model into a linear space using an inverse function of the function used when the pre-processing unit 12 maps the correct answer data into a nonlinear space before learning.

[0028] The output unit 15 outputs information based on the stress value calculated by the calculation unit 14 and inversely mapped to the linear space as described above. For example, each time a stress value is calculated, if the stress value exceeds a preset reference value for judging that stress is high, the output unit 15 outputs an alert to that effect on the display device 30 of the information processing device operated by the workplace manager who manages the person U. Alternatively, each time a stress value is calculated, the output unit 15 may always output the stress value itself, i.e., the time-series change in the stress value of the person U, or may output any data based on the stress value. In addition, the output unit 15 may output data based on the stress value to any person, such as to the person U as a target.

[0029] [Operation] Next, the operation of the above-mentioned stress estimation device 10 will be described mainly with reference to the flowcharts of Figures 4 to 5. First, the operation when generating a stress estimation model by machine learning will be described with reference to the flowchart of Figure 4.

[0030] The stress estimation device 10 acquires learning data for generating a stress calculation model used to calculate a stress value by machine learning. Specifically, the stress estimation device 10 acquires, as the learning data, state data representing a state of a person in a predetermined situation, such as a situation in which the person is performing work at a workplace, etc. For example, the stress estimation device 10 acquires, as state data, biological data such as the heart rate of the person U via a measuring device such as a wearable terminal W worn by the person U (step S1).

[0031] Next, the stress estimation device 10 acquires, as learning data, correct answer data that represents the stress state of a specific person when the biometric data was acquired from the person as described above (step S2). For example, the correct answer data is data based on subjective data related to the person's stress, and as an example, a "PSS score" that is an aggregate of the "Perceived Stress Scale (PSS)" is acquired.

[0032] Next, the stress estimation device 10 performs pre-processing of the learning data including the state data and the correct answer data acquired as described above before performing machine learning using the learning data. First, as pre-processing, the stress estimation device 10 maps the correct answer data included in the data set including the state data and the correct answer data onto a non-linear space (step S3). At this time, the stress estimation device 10 maps the correct answer data onto the non-linear space using, for example, a monotonically increasing function with a rate of change of 1 or more and a function for which an inverse function can be defined, as shown in FIG. 3.

[0033] Next, the stress estimation device 10 extracts a plurality of types of features from the biological data, which is the condition data, of the learning data to generate feature data. Then, the stress estimation device 10 associates each of the feature data with the mapped correct answer data to generate a data set of each feature data and the mapped correct answer data.

[0034] Furthermore, the stress estimation device 10 selects a type of feature amount data to be used as learning data for machine learning from the feature amount data extracted from the state data (step S4). At this time, the stress estimation device 10 calculates a correlation value representing a degree of correlation between the feature amount data and the correct answer data for each type of feature amount data, and selects the type of feature amount data based on the correlation value.

[0035] Then, the stress estimation device 10 performs machine learning using the data set of the selected feature amount data and the ground truth data mapped to the nonlinear space as learning data to generate a stress estimation model (step S5). Specifically, the stress estimation device 10 performs machine learning using all data sets including the selected type of feature amount data as learning data, with the feature amount data as explanatory variables and the mapped ground truth data as objective variables. In this way, the stress estimation device 10 generates a stress estimation model that uses the biometric data of person U as input and calculates a stress value consisting of a value of 0-40 as a value mapped to a nonlinear space, similar to the above-mentioned PSS score.

[0036] Next, the operation of estimating the stress of a target person using the generated stress calculation model will be described with reference to the flowchart in Fig. 5. Note that the scene in which the stress of a target person is estimated is assumed to be a case in which the person is in a specific situation, such as working in the workplace.

[0037] First, the stress estimation device 10 acquires state data used to estimate stress of a target person U. Specifically, the data acquisition unit 11 measures and acquires state data, which is biometric data, from the target person U when the target person U is in a situation where the person U is actually performing work in a workplace, which is a place where stress is to be estimated (step S11).

[0038] Next, the stress estimation device 10 extracts feature amount data from the biological data (step S12). At this time, the stress estimation device 10 extracts feature amount data of the same type as the type of feature amount data selected in the learning data when generating the model as described above.

[0039] Then, the stress estimation device 10 inputs the extracted feature data to the stress estimation model, and calculates the stress value, which is the output (step S13). Furthermore, the stress estimation device 10 inversely maps the calculated stress value to a linear space (step S14). At this time, the stress estimation device 10 inversely maps the stress value calculated by the stress estimation model to the linear space using an inverse function of the function used when mapping the correct answer data to the nonlinear space before learning.

[0040] After that, the stress estimation device 10 outputs information based on the stress value inversely mapped to the linear space (step S15). For example, the stress estimation device 10 outputs stress information of the target person U, such as the stress value itself or an alert based on the stress value.

[0041] As described above, according to the above-mentioned stress estimation device 10, correct answer data, which is a stress value among the learning data, is mapped to a nonlinear space, and then learning is performed to generate a model. As a result, even if data in which the stress value is extremely biased toward the median is collected as learning data, a model in which the stress value is suppressed from being biased toward the median can be generated. As a result, stress can be estimated with higher accuracy.

[0042] As described above, the present invention is not limited to estimating a person's stress, but can also be applied to estimating a person's physical condition, such as physical and mental fatigue and inner condition. In this case, by using the above-mentioned correct answer data as data representing the physical condition, for example, data such as a score based on subjective data regarding the person's fatigue and condition, it is possible to estimate a value representing the person's physical condition in the same manner as described above.

[0043] <Embodiment 2> Next, a second embodiment of the present invention will be described with reference to Fig. 6 to Fig. 8. Fig. 6 to Fig. 7 are block diagrams showing the configuration of an information processing device in the second embodiment, and Fig. 8 is a flowchart showing the operation of the information processing device. Note that this embodiment shows an outline of the configuration of the stress estimation device and stress estimation method described in the above-mentioned embodiment.

[0044] First, a hardware configuration of the information processing device 100 in this embodiment will be described with reference to Fig. 6. The information processing device 100 is configured as a general information processing device, and is equipped with the following hardware configuration, as an example. ·CPU(Central Processing Unit)101(Arithmetic unit) ROM (Read Only Memory) 102 (storage device) RAM (Random Access Memory) 103 (storage device) · Programs 104 loaded into RAM 103 A storage device 105 for storing a program group 104 A drive device 106 that reads and writes data from a storage medium 110 external to the information processing device A communication interface 107 for connecting to a communication network 111 outside the information processing device Input / output interface 108 for inputting and outputting data Bus 109 connecting each component

[0045] The information processing device 100 can be equipped with the conversion unit 121 and the generation unit 122 shown in Fig. 7 by the CPU 101 acquiring the program group 104 and executing it. The program group 104 is stored in the storage device 105 or the ROM 102 in advance, for example, and the CPU 101 loads the program group 104 into the RAM 103 and executes it as necessary. The program group 104 may be supplied to the CPU 101 via the communication network 111, or may be stored in the storage medium 110 in advance, and the drive device 106 reads out the program and supplies it to the CPU 101. However, the conversion unit 121 and the generation unit 122 described above may be constructed of dedicated electronic circuits for realizing such means.

[0046] 6 shows an example of the hardware configuration of the information processing device 100, and the hardware configuration of the information processing device is not limited to the above-mentioned case. For example, the information processing device may be configured with a part of the above-mentioned configuration, such as not having the drive device 106.

[0047] The information processing device 100 executes the information processing method shown in the flowchart of FIG. 8 by the functions of the conversion unit 121 and the generation unit 122 constructed by the program as described above.

[0048] As shown in FIG. 8, the information processing device 100 Among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired, the correct answer data is mapped onto a nonlinear space (step S101); A prediction model is generated for predicting the physical condition of the target person by learning the state data as explanatory variables and the mapped correct answer data as objective variables (step S102). The following process is executed.

[0049] With the above-mentioned configuration, the present invention generates an estimation model by mapping the correct answer data representing physical condition from among the training data to a nonlinear space and then performing training. As a result, even if data representing physical condition is extremely biased toward the median as training data, a model can be generated in which the values ​​representing physical condition are suppressed from being biased toward the median, and physical condition can be estimated with higher accuracy.

[0050] <Additional Notes> A part or all of the above-described embodiments can be described as follows: The following is an outline of the configurations of the information processing method, the information processing device, and the program according to the present invention. However, the present invention is not limited to the following configurations. (Appendix 1) Among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired, the correct answer data is mapped onto a nonlinear space; generating an estimation model to be used for estimating the physical condition of a target person by learning the state data as an explanatory variable and the mapped correct answer data as a target variable; Information processing methods. (Appendix 2) 13. The information processing method according to claim 1, calculating a value representing a physical condition of a target person by inputting target person condition data representing a condition of the target person obtained from the target person into the estimation model; The calculated values ​​representing the physical condition are inversely mapped to a linear space. Information processing methods. (Appendix 3) 3. The information processing method according to claim 1, further comprising: Mapping the ground truth data into a nonlinear space using a monotonically increasing function with a rate of change of 1 or more; Information processing methods. (Appendix 4) An information processing method according to any one of Supplementary Note 1 to 3, Mapping the ground truth data into a nonlinear space using a function whose inverse function can be defined; Information processing methods. (Appendix 5) An information processing method according to any one of Supplementary Note 1 to 4, the condition data is made up of a plurality of feature amount data extracted from measurement data measured on a predetermined person, generating the estimation model by learning the feature amount data selected based on the correct answer data from among the plurality of feature amount data as an explanatory variable; Information processing methods. (Appendix 6) 6. The information processing method according to claim 5, calculating a degree of correlation between the feature amount data and the correct answer data for each type of the feature amount data, and generating the estimation model by learning the feature amount data selected based on the calculated degree of correlation as an explanatory variable; Information processing methods. (Appendix 7) 7. The information processing method according to claim 5, further comprising: calculating a value representing a physical condition by inputting target person feature amount data extracted from target person measurement data measured on the target person, which is of the same type as the selected feature amount data, into the estimation model; The calculated values ​​representing the physical condition are inversely mapped to a linear space. Information processing methods. (Appendix 8) A conversion unit that maps the correct answer data into a nonlinear space from among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired; a generation unit that generates an estimation model to be used for estimating a physical condition of a target person by learning the state data as an explanatory variable and the mapped ground truth data as a target variable; An information processing device comprising: (Appendix 9) 9. The information processing device according to claim 8, a calculation unit that calculates a value representing a physical condition of a target person by inputting target person condition data representing a condition of the target person obtained from the target person into the estimation model; an inverse transformation unit that inversely maps the calculated value representing the physical condition into a linear space; An information processing device comprising: (Appendix 10) The information processing device according to claim 8 or 9, The conversion unit maps the correct answer data into a nonlinear space using a monotonically increasing function with a rate of change of 1 or more. Information processing device. (Appendix 11) An information processing device according to any one of claims 8 to 10, The conversion unit maps the correct answer data into a nonlinear space using a function capable of defining an inverse function. Information processing device. (Appendix 12) 12. The information processing device according to any one of claims 8 to 11, the condition data is made up of a plurality of feature amount data extracted from measurement data measured on a predetermined person, the generation unit generates the estimation model by learning, as an explanatory variable, the feature amount data selected based on the correct answer data from among the plurality of feature amount data. Information processing device. (Appendix 13) 13. The information processing device according to claim 12, the generation unit calculates a degree of correlation between the feature amount data and the ground truth data for each type of the feature amount data, and generates the estimation model by learning the feature amount data selected based on the calculated degree of correlation as an explanatory variable. Information processing device. (Appendix 14) 14. The information processing device according to claim 12, a calculation unit that calculates a value representing a physical condition by inputting target person feature amount data extracted from target person measurement data measured from the target person, which is of the same type as the selected feature amount data, into the estimation model; an inverse transformation unit that inversely maps the calculated value representing the physical condition into a linear space; An information processing device comprising: (Appendix 15) In the information processing device, Among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired, the correct answer data is mapped onto a nonlinear space; A physical condition estimation model is generated for estimating the physical condition of a target person by learning the state data as an explanatory variable and the mapped correct answer data as a target variable. A computer-readable storage medium that stores a program for executing processing. (Appendix 16) A computer-readable storage medium storing the program according to claim 15, calculating a value representing a physical condition of a target person by inputting target person condition data representing a condition of the target person obtained from the target person into the estimation model; The calculated values ​​representing the physical condition are inversely mapped to a linear space. A computer-readable storage medium that stores a program for executing processing. [Explanation of symbols]

[0051] 10 Stress Estimation Device 11 Data Acquisition Section 12 Pretreatment section 13 Learning Department 14 Calculation section 15 Output section 16 Person information storage section 17 Model memory section 20 Input Devices 30 Display device 100 Information processing device 101 CPU 102 ROM 103 RAM 104 Programs 105 Storage device 106 Drive device 107 Communication Interface 108 Input / Output Interface 109 Bus 110 Storage medium 111 Communication Network 121 Conversion unit 122 Generation part

Claims

1. Among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired, the correct answer data is mapped onto a nonlinear space; generating an estimation model to be used for estimating the physical condition of a target person by learning the state data as an explanatory variable and the mapped correct answer data as a target variable; calculating a value representing a physical condition of a target person by inputting target person condition data representing a condition of the target person obtained from the target person into the estimation model; The calculated values ​​representing the physical condition are inversely mapped to a linear space. Information processing methods.

2. Among training data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time when the status data was acquired, when mapping the correct answer data onto a nonlinear space, the correct answer data is mapped onto the nonlinear space using a monotonically increasing function with a rate of change of 1 or more; generating an estimation model to be used for estimating the physical condition of a target person by learning the state data as an explanatory variable and the mapped correct answer data as a target variable; Information processing methods.

3. Among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired, when mapping the correct answer data onto a nonlinear space, the correct answer data is mapped onto the nonlinear space using a function capable of defining an inverse function; generating an estimation model to be used for estimating the physical condition of a target person by learning the state data as an explanatory variable and the mapped correct answer data as a target variable; Information processing methods.

4. Among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired, the correct answer data is mapped onto a nonlinear space; generating an estimation model to be used for estimating the physical condition of a target person by learning the state data as an explanatory variable and the mapped correct answer data as a target variable; the condition data is made up of a plurality of feature amount data extracted from measurement data measured on a predetermined person, generating the estimation model by learning the feature amount data selected based on the correct answer data from among the plurality of feature amount data as an explanatory variable; calculating a value representing a physical condition by inputting target person feature amount data extracted from target person measurement data measured from the target person, which is of the same type as the selected feature amount data, into the estimation model; The calculated values ​​representing the physical condition are inversely mapped to a linear space. Information processing methods.

5. 4. An information processing method according to claim 1, further comprising: the condition data is made up of a plurality of feature amount data extracted from measurement data measured on a predetermined person, generating the estimation model by learning the feature amount data selected based on the correct answer data from among the plurality of feature amount data as an explanatory variable; Information processing methods.

6. 6. An information processing method according to claim 5, calculating a degree of correlation between the feature amount data and the correct answer data for each type of the feature amount data, and generating the estimation model by learning the feature amount data selected based on the calculated degree of correlation as an explanatory variable; Information processing methods.

7. A conversion unit that maps correct answer data into a nonlinear space from learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired; a generation unit that generates an estimation model to be used for estimating a physical condition of a target person by learning the state data as an explanatory variable and the mapped ground truth data as a target variable; a calculation unit that calculates a value representing a physical condition by inputting target person condition data representing a condition of the target person obtained from the target person to the estimation model; an inverse transformation unit that inversely maps the calculated value representing the physical condition into a linear space; An information processing device comprising:

8. a conversion unit that, when mapping the correct answer data into a nonlinear space, maps the correct answer data, among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired, to the nonlinear space using a monotonically increasing function with a rate of change of 1 or more; a generation unit that generates an estimation model to be used for estimating a physical condition of a target person by learning the state data as an explanatory variable and the mapped ground truth data as a target variable; An information processing device comprising:

9. a conversion unit that, when mapping the correct answer data into a nonlinear space, maps the correct answer data into a nonlinear space using a function that can define an inverse function, among learning data including status data that represents a status of a specific person and correct answer data that represents the physical condition of the specific person at the time the status data was acquired; a generation unit that generates an estimation model to be used for estimating a physical condition of a target person by learning the state data as an explanatory variable and the mapped ground truth data as a target variable; An information processing device comprising:

10. In the information processing device, Among learning data including status data representing a status of a specific person and correct answer data representing the physical condition of the specific person at the time the status data was acquired, the correct answer data is mapped onto a nonlinear space; generating an estimation model to be used for estimating the physical condition of a target person by learning the state data as an explanatory variable and the mapped correct answer data as a target variable; calculating a value representing a physical condition of a target person by inputting target person condition data representing a condition of the target person obtained from the target person into the estimation model; The calculated values ​​representing the physical condition are inversely mapped to a linear space. A program for executing a process.

Images