WO2022038776A1 - Stress inference device, inference method, program, and storage medium - Google Patents

Stress inference device, inference method, program, and storage medium Download PDF

Info

Publication number
WO2022038776A1
WO2022038776A1 PCT/JP2020/031662 JP2020031662W WO2022038776A1 WO 2022038776 A1 WO2022038776 A1 WO 2022038776A1 JP 2020031662 W JP2020031662 W JP 2020031662W WO 2022038776 A1 WO2022038776 A1 WO 2022038776A1
Authority
WO
WIPO (PCT)
Prior art keywords
feature amount
attribute
biological signal
pattern
stress
Prior art date
Application number
PCT/JP2020/031662
Other languages
French (fr)
Japanese (ja)
Inventor
嘉樹 中島
旭美 梅松
剛範 辻川
Original Assignee
日本電気株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日本電気株式会社 filed Critical 日本電気株式会社
Priority to JP2022543248A priority Critical patent/JPWO2022038776A1/ja
Priority to PCT/JP2020/031662 priority patent/WO2022038776A1/en
Publication of WO2022038776A1 publication Critical patent/WO2022038776A1/en

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state

Definitions

  • the present invention relates to stress estimation using biological signals.
  • a biological signal which is a signal that reflects the biological information of the subject (sweat amount, skin surface temperature, body movement, etc.), is measured over a long period of time from a wearable terminal worn by the subject on a daily basis, and is measured. Long-term stress (chronic stress) of a person is monitored. Then, a technique aimed at promoting stress reduction based on such monitoring results has been proposed.
  • Non-Patent Document 1 For example, in Non-Patent Document 1, from the 30-day data of 20 people, the activity states of the three bodies of sitting, walking, and running are described by Activity Magnitude (RMS (Rooted Mean Square) with 3-axis acceleration) that is common to all. A technique for identifying from the moving average of changes in)) is disclosed. Further, Non-Patent Document 2 discloses a technique for automatically deriving and applying a threshold value for distinguishing three activity states of sitting, walking, and running from an individual's Activity Magnitude histogram for each individual. Non-Patent Document 3 discloses a technique for estimating the cognitive stress scale of a subject with a certain accuracy.
  • Activity Magnitude RMS (Rooted Mean Square) with 3-axis acceleration
  • Non-Patent Document 4 during learning, the gradient for the label is learned, and at the same time, the gradient for classifying the domain is multiplied by a negative coefficient, and the domain classification is made to learn "hostile".
  • a method for obtaining a domain-invariant model is disclosed.
  • Non-Patent Document 5 discloses a method of selecting features by Random Forest or Decision Tree.
  • Non-Patent Document 1 and Non-Patent Document 2 attempt to estimate the stress level of a person to be measured by using feature quantities such as statistics of biological signals such as sweating, heartbeat, body movement, and body temperature of the person to be measured.
  • feature quantities such as statistics of biological signals such as sweating, heartbeat, body movement, and body temperature of the person to be measured.
  • the relationship between the feature amount used for stress estimation such as sweating, heart rate, body movement, and body temperature and the stress level depends on the life pattern and employment pattern of the subject, or the bias of gender and age. different. Therefore, unless we build a database that evenly incorporates these conditions, we cannot create a model that can be applied to everyone. However, creating such an unbiased database is costly.
  • An object of the present disclosure is to provide a stress estimation device, an estimation method, a program and a storage medium capable of suitably estimating stress in view of the above-mentioned problems.
  • One aspect of the stress estimation device is a biometric signal acquisition means for acquiring a biometric signal of a person to be measured, a feature amount calculation means for calculating a biometric signal feature amount which is a feature amount of the biometric signal related to stress, and the subject to be measured.
  • An attribute indicating at least one of a person's attribute or pattern, an attribute for acquiring pattern data, an attribute that is an attribute that does not depend on the attribute, a pattern data from the biological signal feature amount, and an attribute that does not depend on the pattern.
  • It is a stress estimation device including an attribute / pattern common feature amount selection means for selecting a pattern common feature amount and a stress estimation means for estimating the stress of the person to be measured based on the attribute / pattern common feature amount.
  • One aspect of the estimation method is to acquire the biological signal of the person to be measured by a computer, calculate the characteristic amount of the biological signal which is the characteristic amount of the biological signal related to stress, and at least one of the attributes or patterns of the person to be measured.
  • the attribute / pattern common feature amount that is the attribute / pattern common feature amount that does not depend on the attribute / pattern indicated by the attribute / pattern data is selected from the biometric signal feature amount, and the attribute / pattern common feature amount is selected.
  • One aspect of the program is to acquire the biometric signal of the person to be measured, calculate the biometric signal feature amount which is the feature amount of the biometric signal related to stress, and show at least one of the attributes or patterns of the person to be measured.
  • Pattern data is acquired, an attribute / pattern common feature amount that is a feature amount that does not depend on the attribute / pattern indicated by the attribute / pattern data is selected from the biometric signal feature amount, and based on the attribute / pattern common feature amount, the said It is a program that causes a computer to execute a process of estimating the stress of the person to be measured. This program is stored in the storage medium.
  • stress can be suitably estimated regardless of the attributes and patterns of the person to be measured.
  • the block configuration of the stress estimation system in the first embodiment is shown.
  • the hardware configuration of the stress estimation device in the first embodiment is shown.
  • This is an example of a flowchart showing a processing procedure executed by the stress estimation device of the first embodiment. It is a block diagram which shows the specific structure of the stress estimation system in 2nd Embodiment.
  • the block configuration of the stress estimation device in the second embodiment is shown.
  • a schematic diagram of the process of creating a stress estimation model is shown in the second embodiment.
  • the graph regarding the stress estimation result of the sales position in the second embodiment is shown.
  • It is a block diagram of the stress estimation apparatus in 3rd Embodiment.
  • It is an example of the flowchart which shows the processing procedure of the stress estimation apparatus in 3rd Embodiment.
  • FIG. 1 shows the configuration of the stress estimation system 150 in this embodiment.
  • the stress estimation system 150 according to the present embodiment includes a stress estimation device 100 and a wearable terminal 200.
  • the stress estimation device 100 is composed of one or a plurality of computers.
  • the stress estimation device 100 can perform wired or wireless data communication with a part of the body of the person to be measured, for example, a wearable terminal 200 worn on an arm.
  • the wearable terminal 200 measures the biological signal "Sb" of the person to be measured, and supplies the measured biological signal Sb to the stress estimation device 100.
  • the biological signal Sb is, for example, a signal indicating the sweating amount, skin surface temperature, body movement, pulse rate, heart rate, respiratory rate, etc. of the subject.
  • the biological signal Sb is not limited to the above-mentioned example, and may be any information that can estimate the mental state such as stress of the person to be measured, such as information reflecting the autonomic nerve activity of the person to be measured.
  • the configuration of the stress estimation system 150 shown in FIG. 1 is an example, and various changes may be made.
  • the stress estimation system 150 may further have a mobile device terminal such as a smartphone owned by the subject.
  • the stress estimation device 100 and the wearable terminal 200 may perform data communication via the portable device terminal.
  • the stress estimation device 100 includes the biological signal acquisition unit 101, the biological signal storage unit 102, the biological signal feature amount calculation unit 103, the attribute / pattern data acquisition unit 104, and the attribute / pattern common feature amount selection.
  • a unit 105 and a stress estimation unit 106 are included.
  • the biological signal acquisition unit 101 acquires the biological signal Sb from the wearable terminal 200 by data communication.
  • the wearable terminal 200 may immediately supply the generated biological signal Sb to the biological signal acquisition unit 101, accumulate the generated biological signal Sb, and collect the accumulated biological signal Sb at a predetermined timing. It may be supplied to the acquisition unit 101.
  • the biological signal storage unit 102 stores the biological signal Sb acquired by the biological signal acquisition unit 101.
  • the biological signal feature amount calculation unit 103 calculates the feature amount used for stress estimation (also referred to as “biological signal feature amount Fb”) from the biological signal Sb stored in the biological signal storage unit 102.
  • the biological signal feature amount Fb is an arbitrary feature amount (including a statistic) of the biological signal Sb used for estimating stress, and is, for example, a biological signal as disclosed in Non-Patent Document 1 and Non-Patent Document 2.
  • the attribute / pattern data acquisition unit 104 acquires data indicating at least one of the attributes or behavior patterns of the person to be measured (also referred to as "attribute / pattern data Da").
  • the attribute / pattern data Da is, for example, information indicating an arbitrary behavior pattern such as a life pattern or an employment pattern of the person to be measured, and / or a job type, sex, age of the person to be measured, which reflects a bias of gender or age.
  • Information indicating attributes such as.
  • the attribute / pattern data acquisition unit 104 may acquire the attribute / pattern data Da by various methods.
  • the attribute / pattern data acquisition unit 104 uses the attribute / pattern data acquisition unit 104 based on the input information input by the person to be measured by any user interface (including a voice input device) to the wearable terminal 200 or the stress estimation device 100. Generate pattern data Da.
  • the attribute / pattern data acquisition unit 104 generates the attribute / pattern data Da by estimating the attribute / pattern of the subject from the biological signal Sb of the subject.
  • the attribute / pattern data Da of the measured person is acquired from the storage device that stores the attribute / pattern data Da of the measured person in advance.
  • the above-mentioned storage device may be a memory in the stress estimation device 100, or may be an external device (for example, a server that manages a database related to attributes / patterns) different from the stress estimation device 100.
  • the attribute / pattern common feature amount selection unit 105 does not differ from the biometric signal feature amount Fb calculated by the biometric signal feature amount calculation unit 103 depending on the attribute / pattern indicated by the attribute / pattern data Da, that is, does not depend on the attribute / pattern.
  • Select the feature amount also referred to as "attribute / pattern common feature amount Fc"
  • the attribute / pattern common feature amount selection unit 105 excludes the feature amount that can distinguish the attribute or the pattern (also referred to as “hostile feature amount Fa”) from the biometric signal feature amount Fb, and the remaining biometric signal features.
  • the quantity Fb is specified as the attribute / pattern common feature quantity Fc.
  • the attribute / pattern common feature amount selection unit 105 selects, as a specific method, the selection of the common feature amount of the attribute / pattern, that is, the hostile feature amount Fa capable of distinguishing the attribute / pattern, and is hostile.
  • the process of eliminating the target feature amount Fa (hereinafter referred to as "hostile feature amount selection") is executed.
  • hostile feature selection and target stress estimation are performed alternately to obtain the optimum feature set as input and output to the model of hostile feature selection.
  • the model of hostile feature selection is trained in advance.
  • the parameters of the model of hostile feature amount selection obtained by learning are stored in a memory or the like that can be referred to by the attribute / pattern common feature amount selection unit 105.
  • the attribute / pattern common feature amount selection unit 105 constitutes a model of hostile feature amount selection by referring to the parameter, and selects the hostile feature amount Fa from the biological signal feature amount Fb by the model.
  • the stress estimation unit 106 estimates the stress of the subject based on the attribute / pattern common feature amount Fc selected by the attribute / pattern common feature amount selection unit 105. In this case, the stress estimation unit 106 calculates a stress score estimated from the attribute / pattern common feature amount Fc, for example, based on a model obtained by machine learning (also referred to as a “stress estimation model”). The parameters of the stress estimation model obtained by learning are stored in a memory or the like that can be referred to by the stress estimation unit 106. After the stress is estimated, the score indicating the estimated stress is presented to the subject and used for the purpose of promoting the reduction of stress.
  • FIG. 2 shows the hardware configuration of the stress estimation device 100.
  • the stress estimation device 100 includes, for example, a processor 11, a memory 12, and an interface 13 as hardware.
  • the processor 11, the memory 12, and the interface 13 are connected via the data bus 19.
  • the processor 11 executes a predetermined process by executing the program stored in the memory 12.
  • the processor 11 is a processor such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and a quantum processor. Then, for example, by executing the program stored in the memory 12, the processor 11 executes the biological signal acquisition unit 101, the biological signal feature amount calculation unit 103, the attribute / pattern data acquisition unit 104, and the attribute / pattern shown in FIG. It functions as a pattern common feature amount selection unit 105 and a stress estimation unit 106.
  • Each component of the biological signal acquisition unit 101, the biological signal feature amount calculation unit 103, the attribute / pattern data acquisition unit 104, the attribute / pattern common feature amount selection unit 105, and the stress estimation unit 106 is realized by software by a program.
  • the controller is not limited to this, and may be a controller realized by a combination of any one of hardware, firmware, and software. Further, each of these components may be realized by using a user-programmable integrated circuit such as an FPGA (field-programmable gate array) or a microcomputer. In this case, this integrated circuit may be used to realize a program composed of each of the above components.
  • the memory 12 is composed of various volatile memories such as RAM (Random Access Memory) and ROM (Read Only Memory) and non-volatile memory. Further, the memory 12 stores a program for executing the process executed by the stress estimation device 100. For example, the memory 12 functions as a biological signal storage unit 102. The program executed by the stress estimation device 100 may be stored in a storage medium other than the memory 12. Similarly, the biological signal storage unit 102 may be stored in an external device other than the stress estimation device 100.
  • the interface 13 is an interface for electrically connecting the stress estimation device 100 and another device.
  • the interface 13 may be a communication interface for the stress estimation device 100 to perform data communication with an external device such as a wearable terminal 200 by wire or wirelessly.
  • the interface 13 may be a hardware interface compliant with USB (Universal Serial Bus), SATA (Serial AT Atchment), or the like.
  • various devices other than the wearable terminal 200 may be electrically connected to the stress estimation device 100 via the interface 13.
  • an input device that accepts user input related to attribute / pattern data Da
  • a display device that outputs an estimation result by the stress estimation unit 106
  • a sound output device, or the like exchanges data with the stress estimation device 100 via the interface 13. May be good.
  • the hardware configuration of the stress estimation device 100 is not limited to the configuration shown in FIG.
  • the stress estimation device 100 may be a tablet terminal or the like incorporating at least one of an input device, a display device, and a sound output device.
  • FIG. 3 is an example of a flowchart showing the operation of the stress estimation device 100.
  • the biological signal acquisition unit 101 of the stress estimation device 100 acquires the biological signal Sb transmitted from the wearable terminal 200 (step S11). Then, the biological signal acquisition unit 101 stores the acquired biological signal Sb in the biological signal storage unit 102 (step S12).
  • the biological signal feature amount calculation unit 103 calculates the biological signal feature amount Fb based on the biological signal Sb stored in the biological signal storage unit 102 (step S13). Further, the attribute / pattern data acquisition unit 104 acquires the attribute / pattern data Da in parallel with the processing in which steps S11 to S13 are executed, or at the timing before and after these (step S14).
  • the biological signal feature amount Fb calculated in step S13 is a feature amount capable of distinguishing the attribute / pattern indicated by the attribute / pattern data Da based on the attribute / pattern data Da. It is determined whether or not it corresponds to a certain hostile feature amount Fa (step S15). Then, when the attribute / pattern common feature amount selection unit 105 corresponds to the hostile feature amount Fa capable of distinguishing the attribute / pattern (step S15; Yes), the target biometric signal feature amount Fb is an attribute. -It is determined that the pattern does not correspond to the common feature amount Fc, and the process proceeds to step S17 without performing stress estimation in step S16.
  • the attribute / pattern common feature amount selection unit 105 sets the target biological signal feature amount Fb as the attribute / pattern common feature amount Fc. Consider it as. Then, in this case, the stress estimation unit 106 estimates the stress of the person to be measured based on the attribute / pattern common feature amount Fc (step S16).
  • the stress estimation device 100 determines whether or not the stress estimation process should be completed (step S17). For example, the stress estimation device 100 cannot acquire the biological signal Sb, detects a user input to end the stress estimation, or satisfies other predetermined termination conditions of the stress estimation process. , Judge that the stress estimation process should be completed. Then, when the stress estimation device 100 determines that the stress estimation process should be completed (step S17; Yes), the stress estimation device 100 ends the process of the flowchart. On the other hand, when the stress estimation device 100 determines that the stress estimation process should not be completed (step S17; No), the stress estimation device 100 returns the process to step S11.
  • the attribute / pattern common which is a feature amount that does not differ depending on the attribute / pattern.
  • Stress estimation can be performed using only the feature Fc. Therefore, the stress of the person to be measured can be estimated with high accuracy without depending on the attributes and patterns of the person to be measured.
  • Non-Patent Document 4 is a method for obtaining a domain-invariant model, which can handle a biased database, but since it is based on deep learning on the premise of a large amount of data, it is difficult to obtain a large amount of data for long-term stress. Not applicable to such database analysis.
  • the stress estimation system is created by a biased database without a large amount of training data, the stress of the person to be measured is suitably estimated without depending on the attributes and patterns. can do.
  • FIG. 4 shows the configuration of the stress estimation system 150A according to the second embodiment.
  • the stress estimation system 150A has a plurality of wearable terminals 400 and a computer 600.
  • the computer 600 and each wearable terminal 400 communicate with each other via the communication means 500 (501, 502, 503, 504).
  • the wearable terminal 400 acquires the biological signal "Sb" of the person to be measured 300.
  • the biological signal Sb of the person to be measured 300 may be a skin surface electrical activity (Electrodermal Activity) that reflects the sweating of the person to be measured.
  • Another example of the biological signal Sb may be a signal that reflects various biological information such as body temperature, pulse wave, heartbeat, voice, brain wave, respiration, myoelectricity, electrocardiogram, and body movement.
  • the biological signal Sb may be a signal that reflects arbitrary biological information affected by the mental activity of the subject.
  • the wearable terminal 400 is a terminal that can be worn by the person to be measured 300, and measures at least one of the biological signals that reflect the biological information mentioned above. For example, the wearable terminal 400 acquires skin conductivity at a constant sampling rate and stores it in the built-in memory.
  • the wearable terminal 400 may be in various forms such as a wristband type, a badge type, an employee ID card type, an earphone type, and a shirt type.
  • the communication means 500 (501, 502, 503, 504) transmits the biological signal Sb (which may include an acceleration signal) acquired by the wearable terminal 400 to the computer 600.
  • the wearable terminal 400 connects to the smartphone 502 by short-range communication 501 such as Bluetooth (registered trademark), and transmits the biological signal Sb to the smartphone 502.
  • the smartphone 502 transmits the biological signal Sb to the Internet 504 by packet communication 503 by the installed application.
  • the biological signal Sb is uploaded to the computer 600 connected to the Internet 504.
  • the computer 600 corresponds to the stress estimation device 100 of the first embodiment, and has, for example, a hardware configuration as shown in FIG.
  • FIG. 5 shows the functional configuration of the computer 600 in the second embodiment.
  • the computer 600 includes a communication interface (I / F) 601, a biometric signal acquisition unit 602, a biometric signal storage unit 603, an attribute / pattern data acquisition unit 604, and an attribute / pattern data storage unit 605.
  • Biosignal feature amount calculation unit 606 biometric signal feature amount storage unit 607, attribute / pattern common feature amount selection unit 608, attribute / pattern common feature amount storage unit 609, stress estimation unit 610, and stress estimation results. It includes a storage unit 611 and a stress estimation result output unit 612.
  • the biological signal acquisition unit 602 stores the biological signal Sb obtained from the communication interface 601 in the biological signal storage unit 603.
  • the attribute / pattern data acquisition unit 604 acquires the attribute / pattern data “Da”.
  • the attribute / pattern data Da indicates the job type (sales position / technical position) of the person to be measured.
  • the acquired attribute / pattern data Da is stored in the attribute / pattern data storage unit 605.
  • the attribute / pattern data acquisition unit 604 may acquire the attribute / pattern data Da indicating the attribute / pattern of the person to be measured from the wearable terminal 400 or other external device, and the attribute / pattern data Da of the person to be measured may be acquired from the biological signal Sb.
  • the pattern may be estimated to generate the attribute / pattern data Da.
  • the biological signal feature amount calculation unit 606 calculates the biological signal feature amount "Fb", which is a characteristic amount related to stress, based on the biological signal Sb extracted from the biological signal storage unit 603.
  • the biological signal feature amount Fb is a time-series histogram of sweating and body movement, a power spectral density, and a statistic (mean value, median value, dispersion value).
  • the calculated biological signal feature amount Fb is stored in the biological signal feature amount storage unit 607.
  • the attribute / pattern common feature amount selection unit 608 eliminates the hostile feature amount “Fa”, which is a feature amount that distinguishes attributes / patterns, so that the attribute / pattern is a feature amount that does not differ depending on the attribute / pattern.
  • the pattern common feature amount "Fc" is selected from the biometric signal feature amount Fb.
  • the attribute / pattern common feature amount selection unit 608 selects the hostile feature amount using a random forest, and excludes the selected hostile feature amount Fa from the biological signal feature amount Fb. Then, the attribute / pattern common feature amount selection unit 608 selects the remaining biological signal feature amount Fb that does not depend on the attribute / pattern as the attribute / pattern common feature amount Fc.
  • the attribute / pattern common feature amount selection unit 608 classifies data having different attributes / patterns into classes by a random forest method.
  • the attribute / pattern common feature amount selection unit 608 performs two-class classification for identifying the technical position and the sales position. In this classification, the attribute / pattern common feature amount selection unit 608 evaluates the importance of each biological signal feature amount Fb. Then, the attribute / pattern common feature amount selection unit 608 considers the biological signal feature amount Fb having an importance equal to or higher than a predetermined threshold value as the hostile feature amount Fa, and excludes it from the feature amount set used for stress estimation.
  • the parameters required for hostile feature selection using the random forest are stored in the memory of the computer 600 or an external storage device.
  • the attribute / pattern common feature amount selection unit 608 stores the attribute / pattern common feature amount Fc common to the attributes / patterns selected through the hostile feature amount selection in the attribute / pattern common feature amount storage unit 609.
  • the stress estimation unit 610 estimates stress based on the attribute / pattern common feature amount Fc stored in the attribute / pattern common feature amount storage unit 609.
  • the stress estimation unit 610 uses the score of the 10-item version of PSS (Perceived Stress Scale) (hereinafter referred to as “PSS10”) as the correct answer value of stress, and this PSS10.
  • PSS10 Perceived Stress Scale
  • the score of PSS10 calculated from the PSS questionnaire conducted at the end of the experimental period (for example, 4 weeks) for the subject is used as the teacher data (correct answer value), and is based on the biological signal obtained from the subject.
  • the stress estimation unit 610 uses a stress estimation model obtained by learning a machine learning model such as an SVM model based on this training data, and estimates the score of PSS10 from the attribute / pattern common feature amount Fc. Then, the stress estimation unit 610 generates the estimated PSS10 score as the stress estimation result.
  • a machine learning model such as an SVM model based on this training data
  • the stress estimation unit 610 generates the estimated PSS10 score as the stress estimation result.
  • the parameters of the stress estimation model obtained by learning the machine learning model described above are stored in the memory of the computer 600 or an external storage device so that the stress estimation unit 610 can refer to them.
  • the estimated stress estimation result is stored in the stress estimation result storage unit 611.
  • the stress estimation result output unit 612 outputs the stress estimation result stored in the stress estimation result storage unit 611.
  • the output method include, but are not limited to, screen output and print output.
  • the timing of output may be output at all times or at the request of the person to be measured.
  • the stress estimation result output unit 612 uses the communication means 500 to store the stress estimation result stored in the stress estimation result storage unit 611 through the communication interface 601 to the wearable terminal 400 or the smartphone. Send to 502. After that, the wearable terminal 400 or the smartphone 502 outputs the stress estimation result on the accompanying screen. Thereby, the computer 600 can preferably present the stress estimation result to the person to be measured 300.
  • FIG. 6 shows a schematic diagram of the process of creating a stress estimation model.
  • the computer 600 will be described as executing the process of creating the stress estimation model.
  • the process of creating the stress estimation model may be executed by a device other than the computer 600.
  • the computer 600 generates the biological signal feature amount Fb of the sales staff and the biological signal feature amount Fb of the technical staff from the biological signal Sb of the sales staff and the biological signal Sb of the technical staff, and discriminates both attributes / patterns. Search for possible hostile features Fa using the random forest method.
  • the "sales staff data" in FIG. 6 corresponds to the biological signal Sb of the sales staff or the biological signal feature amount Fb of the sales staff based on the biological signal Sb, and the "technical staff data" of FIG. It corresponds to the biological signal Sb or the biological signal feature amount Fb of a technical worker based on the biological signal Sb.
  • the computer 600 creates a stress estimation model based on the attribute / pattern common feature amount Fc in which the searched hostile feature amount Fa is excluded from each biological signal feature amount Fb of the sales position and the technical position.
  • the computer 600 is the result of the attribute / pattern common feature amount Fc excluding the hostile feature amount Fa from the biological signal feature amount Fb generated from the biological signal Sb of the technical worker, and the result of the stress questionnaire of the technical worker.
  • the stress estimation model is trained based on the PSS10 score (that is, the teacher data of the objective variable at the time of stress estimation).
  • the computer 600 estimates the PSS10 score of the sales position (that is, the result of the stress questionnaire of the sales position) from the attribute / pattern common feature amount Fc of the sales position by the stress estimation model learned based on the data related to the technical position. do.
  • Support Vector Regression SVR is used as an example.
  • FIG. 7 shows a graph regarding the stress estimation results of sales positions by SVR.
  • the horizontal axis shows the number of hostile features Fa selected by the hostile feature selection by random forest (that is, the number of features to be excluded), and the vertical axis shows the estimated score of PSS10 by SVR and the correct answer. It is an error from the actual score of PSS10.
  • the features that can distinguish sales and technical positions well that is, the features that are highly dependent on the attributes and patterns of job types, are hostile in order. It is selected as the feature amount Fa.
  • the features used in the stress estimation model decrease, so stress is stressed only by features that do not depend on attributes / patterns (in this case, occupation).
  • estimation models There is an increasing tendency for estimation models to be formed. Therefore, even with the stress estimation model learned only from the data of the technical staff, there is a strong tendency that the stress of the sales staff can be estimated accurately. This can be understood from the situation in which the error indicated by the vertical axis gradually decreases as the number of features increases.
  • the features used in the stress estimation model are eliminated by that amount, and the number of training samples becomes too small.
  • the performance of the stress estimation model deteriorates, and the error indicated by the vertical axis increases.
  • the error on the vertical axis increases the “number of features to be excluded” with the number “Nopt” as the boundary when the “number of features to be excluded” indicated by the horizontal axis minimizes the error on the vertical axis. Increases with.
  • the relationship between the estimation error corresponding to the vertical axis of FIG. 7 and the parameter of hostile feature amount selection is obtained in the training data. Then, the computer 600 constructs a stress estimation model by excluding the features selected by the parameters when the estimation error takes the minimum value. By using such a stress estimation model, the computer 600 can estimate the stress of the subject with high accuracy.
  • FIG. 8 is a block diagram of the stress estimation device 100X according to the third embodiment.
  • the stress estimation device 100X mainly includes a biological signal acquisition means 101X, a biological signal feature amount calculation means 103X, an attribute / pattern data acquisition means 104X, an attribute / pattern common feature amount selection means 105X, and a stress estimation means 106X.
  • the stress estimation device 100X may be composed of a plurality of devices.
  • the biological signal acquisition means 101X acquires the biological signal of the person to be measured.
  • the biological signal acquisition unit 101X can be the biological signal acquisition unit 101 of the first embodiment or the biological signal acquisition unit 602 of the second embodiment.
  • the biological signal acquisition means 101X may be the biological signal feature amount calculation unit 103 that acquires the biological signal Sb from the biological signal storage unit 102 in the first embodiment.
  • the biological signal feature amount calculation means 103X calculates the biological signal feature amount, which is the feature amount of the biological signal related to stress.
  • the biological signal feature amount calculation means 103X can be the biological signal feature amount calculation unit 103 of the first embodiment or the biological signal feature amount calculation unit 606 of the second embodiment.
  • the attribute / pattern data acquisition means 104X acquires attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured.
  • the attribute / pattern data acquisition unit 104X can be the attribute / pattern data acquisition unit 104 of the first embodiment or the attribute / pattern data acquisition unit 604 of the second embodiment.
  • the attribute / pattern common feature amount selection means 105X selects an attribute / pattern common feature amount which is a feature amount independent of the attribute / pattern indicated by the attribute / pattern data from the biological signal feature amount.
  • the attribute / pattern common feature amount selection means 105X can be the attribute / pattern common feature amount selection unit 105 of the first embodiment or the attribute / pattern common feature amount selection unit 608 of the second embodiment.
  • the stress estimation means 106X estimates the stress of the person to be measured based on the common feature amount of the attribute / pattern.
  • the stress estimation means 106X can be the stress estimation unit 106 of the first embodiment or the stress estimation unit 610 of the second embodiment.
  • FIG. 9 is an example of a flowchart executed by the stress estimation device 100X in the third embodiment.
  • the biological signal acquisition means 101X acquires the biological signal of the person to be measured (step S21).
  • the biological signal feature amount calculation means 103X calculates the biological signal feature amount, which is the feature amount of the biological signal related to stress (step S22).
  • the attribute / pattern data acquisition unit 104X acquires attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured (step S23).
  • the attribute / pattern common feature amount selection means 105X selects an attribute / pattern common feature amount which is a feature amount independent of the attribute / pattern indicated by the attribute / pattern data from the biological signal feature amount (step S24).
  • the stress estimation means 106X estimates the stress of the subject based on the attribute / pattern common feature amount (step S25).
  • the stress estimation device 100X according to the third embodiment can suitably estimate the stress of the person to be measured regardless of the attributes and patterns of the person to be measured.
  • Non-temporary computer-readable media include various types of tangible storage medium.
  • Examples of non-temporary computer-readable media include magnetic storage media (eg flexible disks, magnetic tapes, hard disk drives), optomagnetic storage media (eg optomagnetic disks), CD-ROMs (ReadOnlyMemory), CD-Rs, Includes CD-R / W, semiconductor memory (eg, mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (RandomAccessMemory)).
  • the program may also be supplied to the computer by various types of transient computer readable medium.
  • Examples of temporary computer readable media include electrical, optical, and electromagnetic waves.
  • the temporary 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.
  • a biological signal acquisition means for acquiring the biological signal of the person to be measured, and A biological signal feature amount calculating means for calculating a biological signal feature amount which is a feature amount of the biological signal related to stress, and a biological signal feature amount calculating means.
  • An attribute / pattern data acquisition means for acquiring attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured, and An attribute / pattern common feature amount selection means for selecting an attribute / pattern common feature amount which is a feature amount independent of the attribute / pattern indicated by the attribute / pattern data from the biological signal feature amount.
  • a stress estimation means for estimating the stress of the person to be measured based on the attribute / pattern common feature amount, and A stress estimator equipped with.
  • the attribute / pattern common feature amount selection means selects the attribute / pattern common feature amount by excluding the hostile feature amount that distinguishes the attribute / pattern from the biometric signal feature amount, according to Appendix 1. Stress estimation device.
  • the attribute / pattern common feature amount selection means calculates the importance of the biological signal feature amount in the classification of the attribute / pattern, and selects the hostile feature amount based on the importance.
  • Appendix 4 The stress estimation device according to Appendix 2 or 3, wherein the attribute / pattern common feature amount selection means selects the hostile feature amount based on a random forest.
  • [Appendix 8] By computer Acquires the biological signal of the person to be measured and The biological signal feature amount, which is the feature amount of the biological signal related to stress, is calculated. Acquire attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured. From the biological signal feature amount, select an attribute / pattern common feature amount that is a feature amount that does not depend on the attribute / pattern indicated by the attribute / pattern data. The stress of the person to be measured is estimated based on the common feature amount of the attribute / pattern. Estimating method.
  • [Appendix 9] Acquires the biological signal of the person to be measured and The biological signal feature amount, which is the feature amount of the biological signal related to stress, is calculated.
  • Acquire attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured. From the biological signal feature amount, select an attribute / pattern common feature amount that is a feature amount that does not depend on the attribute / pattern indicated by the attribute / pattern data.
  • a program that causes a computer to execute a process of estimating the stress of the person to be measured based on the common feature amount of the attribute / pattern.
  • Appendix 10 A storage medium in which the program described in Appendix 9 is stored.
  • Stress estimation device 101 Biological signal acquisition unit 102 Biological signal storage unit 103 Biological signal feature amount calculation unit 104 Attribute / pattern data acquisition unit 105 Attribute / pattern common feature amount selection unit 106 Stress estimation unit 150 Stress estimation system 200 Wearable terminal 300 Measurer 400 Wearable terminal 500 Communication means 501 Bluetooth communication 502 Smartphone 503 Mobile data communication or Wifi communication 504 Internet 600 Computer 601 Communication interface 602 Biological signal acquisition unit 603 Biological signal storage unit 604 Attribute / pattern data acquisition unit 605 Attribute / pattern data storage Unit 606 Biological signal feature amount calculation unit 607 Biological signal feature amount storage unit 608 Attribute / pattern common feature amount selection unit 609 Attribute / pattern common feature amount storage unit 610 Stress estimation unit 611 Stress estimation result storage unit 612 Stress estimation result output unit

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Hospice & Palliative Care (AREA)
  • Pathology (AREA)
  • Developmental Disabilities (AREA)
  • Psychiatry (AREA)
  • Psychology (AREA)
  • Social Psychology (AREA)
  • Physics & Mathematics (AREA)
  • Child & Adolescent Psychology (AREA)
  • Biophysics (AREA)
  • Educational Technology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

A stress inference device (100X) mainly comprises a biological signal acquisition means (101X), a biological signal feature amount calculation means (103X), an attribute/pattern data acquisition means (104X), an attribute/pattern common feature amount selection means (105X), and a stress inference means (106X). The biological signal acquisition means (101X) acquires a biological signal of a measurement subject. The biological signal feature amount calculation means (103X) calculates biological signal feature amounts that are feature amounts of a biological signal related to stress. The attribute/pattern data acquisition means (104X) acquires attribute/pattern data that indicates an attribute and/or a pattern of the measurement subject. The attribute/pattern common feature amount selection means (105X) selects, from the biological signal feature amounts, an attribute/pattern common feature amount, which is a feature amount that does not depend on the attribute/pattern indicated by the attribute/pattern data. The stress inference means (106X) infers the stress of the measurement subject on the basis of the attribute/pattern common feature amount.

Description

ストレス推定装置、推定方法、プログラム及び記憶媒体Stress estimator, estimation method, program and storage medium
　本発明は、生体信号を用いたストレス推定に関する。 The present invention relates to stress estimation using biological signals.
　近年、長期のストレッサーへの暴露などにより交感神経が活発になった状態が長く続き、自律神経が失調することにより精神の健康を害することが問題となっている。このため、被測定者に日常的に装着させたウェアラブル端末から被測定者の生体情報（発汗量・皮膚表面温・体動など）を反映する信号である生体信号を長期にわたって測定し、被測定者の長期ストレス（慢性ストレス）をモニタリングすることが行われている。そして、このようなモニタリング結果に基づいてストレスの低減を促すことを目的とした技術が提案されている。 In recent years, the state in which the sympathetic nerve has become active due to long-term exposure to stressors has continued for a long time, and it has become a problem that mental health is impaired due to ataxia of the autonomic nerve. For this reason, a biological signal, which is a signal that reflects the biological information of the subject (sweat amount, skin surface temperature, body movement, etc.), is measured over a long period of time from a wearable terminal worn by the subject on a daily basis, and is measured. Long-term stress (chronic stress) of a person is monitored. Then, a technique aimed at promoting stress reduction based on such monitoring results has been proposed.
　例えば、非特許文献１には、２０名の人々の３０日のデータから、座位・歩行・走行の３つの身体の活動状態を、全員に共通のActivity Magnitude（３軸加速度のRMS（Rooted Mean Square）の変化の移動平均）から識別する技術が開示されている。また、非特許文献２には、座位・歩行・走行の３つの活動状態を区別する閾値を個人のActivity Magnitudeのヒストグラムから個人ごとに自動的に導出し、適用する技術が開示されている。非特許文献３には、被測定者の認知ストレススケールを一定の精度で推定する技術が開示されている。非特許文献４には、学習の際に、ラベルのための勾配を学習させると同時に、ドメインをクラス分類する勾配にマイナスの係数をかけ、ドメイン分類に「敵対的」な学習をさせることで、ドメイン不変のモデルを得る手法が開示されている。非特許文献5では、ランダムフォレスト（Random Forest）やDecision Treeによる特徴量選択の手法が開示されている。 For example, in Non-Patent Document 1, from the 30-day data of 20 people, the activity states of the three bodies of sitting, walking, and running are described by Activity Magnitude (RMS (Rooted Mean Square) with 3-axis acceleration) that is common to all. A technique for identifying from the moving average of changes in)) is disclosed. Further, Non-Patent Document 2 discloses a technique for automatically deriving and applying a threshold value for distinguishing three activity states of sitting, walking, and running from an individual's Activity Magnitude histogram for each individual. Non-Patent Document 3 discloses a technique for estimating the cognitive stress scale of a subject with a certain accuracy. In Non-Patent Document 4, during learning, the gradient for the label is learned, and at the same time, the gradient for classifying the domain is multiplied by a negative coefficient, and the domain classification is made to learn "hostile". A method for obtaining a domain-invariant model is disclosed. Non-Patent Document 5 discloses a method of selecting features by Random Forest or Decision Tree.
　非特許文献１や非特許文献２は、被測定者の発汗、心拍、体動、体温等の生体信号の統計量等の特徴量を用いて、被測定者のストレスレベルを推定しようとしている。しかし、特に慢性ストレスの推定の場合、被測定者の生活パターンや就業パターン、または性別や年齢の偏りによって、発汗、心拍、体動、体温等のストレス推定に用いる特徴量とストレスレベルの関係は異なる。よって、これらの条件をまんべんなく取り入れたデータベースを構築しなければ、全ての人に適用できるモデルは作成できない。しかしながら、こうした偏りのないデータベースを作成することにはコストがかかる。 Non-Patent Document 1 and Non-Patent Document 2 attempt to estimate the stress level of a person to be measured by using feature quantities such as statistics of biological signals such as sweating, heartbeat, body movement, and body temperature of the person to be measured. However, especially in the case of estimating chronic stress, the relationship between the feature amount used for stress estimation such as sweating, heart rate, body movement, and body temperature and the stress level depends on the life pattern and employment pattern of the subject, or the bias of gender and age. different. Therefore, unless we build a database that evenly incorporates these conditions, we cannot create a model that can be applied to everyone. However, creating such an unbiased database is costly.
　本開示の目的は、上述した課題を鑑み、ストレスを好適に推定可能なストレス推定装置、推定方法、プログラム及び記憶媒体を提供することである。 An object of the present disclosure is to provide a stress estimation device, an estimation method, a program and a storage medium capable of suitably estimating stress in view of the above-mentioned problems.
　ストレス推定装置の一の態様は、被測定者の生体信号を取得する生体信号取得手段と、ストレスに関する前記生体信号の特徴量である生体信号特徴量を算出する特徴量計算手段と、前記被測定者の属性又はパターンの少なくとも一方を示す属性・パターンデータを取得する属性・パターンデータ取得手段と、前記生体信号特徴量から前記属性・パターンデータが示す属性・パターンに依存しない特徴量である属性・パターン共通特徴量を選択する属性・パターン共通特徴量選択手段と、前記属性・パターン共通特徴量に基づき、前記被測定者のストレスを推定するストレス推定手段と、を備えるストレス推定装置である。 One aspect of the stress estimation device is a biometric signal acquisition means for acquiring a biometric signal of a person to be measured, a feature amount calculation means for calculating a biometric signal feature amount which is a feature amount of the biometric signal related to stress, and the subject to be measured. An attribute indicating at least one of a person's attribute or pattern, an attribute for acquiring pattern data, an attribute that is an attribute that does not depend on the attribute, a pattern data from the biological signal feature amount, and an attribute that does not depend on the pattern. It is a stress estimation device including an attribute / pattern common feature amount selection means for selecting a pattern common feature amount and a stress estimation means for estimating the stress of the person to be measured based on the attribute / pattern common feature amount.
　推定方法の一の態様は、コンピュータにより、被測定者の生体信号を取得し、ストレスに関する前記生体信号の特徴量である生体信号特徴量を算出し、前記被測定者の属性又はパターンの少なくとも一方を示す属性・パターンデータを取得し、前記生体信号特徴量から前記属性・パターンデータが示す属性・パターンに依存しない特徴量である属性・パターン共通特徴量を選択し、前記属性・パターン共通特徴量に基づき、前記被測定者のストレスを推定する、推定方法である。 One aspect of the estimation method is to acquire the biological signal of the person to be measured by a computer, calculate the characteristic amount of the biological signal which is the characteristic amount of the biological signal related to stress, and at least one of the attributes or patterns of the person to be measured. The attribute / pattern common feature amount that is the attribute / pattern common feature amount that does not depend on the attribute / pattern indicated by the attribute / pattern data is selected from the biometric signal feature amount, and the attribute / pattern common feature amount is selected. This is an estimation method for estimating the stress of the subject based on the above.
　プログラムの一の態様は、被測定者の生体信号を取得し、ストレスに関する前記生体信号の特徴量である生体信号特徴量を算出し、前記被測定者の属性又はパターンの少なくとも一方を示す属性・パターンデータを取得し、前記生体信号特徴量から前記属性・パターンデータが示す属性・パターンに依存しない特徴量である属性・パターン共通特徴量を選択し、前記属性・パターン共通特徴量に基づき、前記被測定者のストレスを推定する処理をコンピュータに実行させるプログラムである。このプログラムは記憶媒体に格納される。 One aspect of the program is to acquire the biometric signal of the person to be measured, calculate the biometric signal feature amount which is the feature amount of the biometric signal related to stress, and show at least one of the attributes or patterns of the person to be measured. Pattern data is acquired, an attribute / pattern common feature amount that is a feature amount that does not depend on the attribute / pattern indicated by the attribute / pattern data is selected from the biometric signal feature amount, and based on the attribute / pattern common feature amount, the said It is a program that causes a computer to execute a process of estimating the stress of the person to be measured. This program is stored in the storage medium.
　本開示によれば、被測定者の属性・パターンによらずに好適にストレスを推定することができる。 According to the present disclosure, stress can be suitably estimated regardless of the attributes and patterns of the person to be measured.
第１実施形態におけるストレス推定システムのブロック構成を示す。The block configuration of the stress estimation system in the first embodiment is shown. 第１実施形態におけるストレス推定装置のハードウェア構成を示す。The hardware configuration of the stress estimation device in the first embodiment is shown. 第１実施形態のストレス推定装置が実行する処理手順を示すフローチャートの一例である。This is an example of a flowchart showing a processing procedure executed by the stress estimation device of the first embodiment. 第２実施形態におけるストレス推定システムの具体的構成を示すブロック図である。It is a block diagram which shows the specific structure of the stress estimation system in 2nd Embodiment. 第２実施形態におけるストレス推定装置のブロック構成を示す。The block configuration of the stress estimation device in the second embodiment is shown. 第２実施形態においてストレス推定モデルの作成プロセスの模式図を示す。A schematic diagram of the process of creating a stress estimation model is shown in the second embodiment. 第２実施形態において営業職のストレス推定結果に関するグラフを示す。The graph regarding the stress estimation result of the sales position in the second embodiment is shown. 第３実施形態におけるストレス推定装置のブロック図である。It is a block diagram of the stress estimation apparatus in 3rd Embodiment. 第３実施形態におけるストレス推定装置の処理手順を示すフローチャートの一例である。It is an example of the flowchart which shows the processing procedure of the stress estimation apparatus in 3rd Embodiment.
　以下、図面を参照しながら、ストレス推定装置、推定方法、プログラム及び記憶媒体の実施形態について説明する。 Hereinafter, embodiments of the stress estimation device, estimation method, program, and storage medium will be described with reference to the drawings.
　＜第１実施形態＞
　［構成の説明］
　以下、第１実施形態におけるストレス推定システムの構成について、図１及び図２を参照して説明する。 <First Embodiment>
[Description of configuration]
Hereinafter, the configuration of the stress estimation system according to the first embodiment will be described with reference to FIGS. 1 and 2.
　図１は、本実施形態におけるストレス推定システム１５０の構成を示す。図１を参照すると、本実施形態に係るストレス推定システム１５０は、ストレス推定装置１００と、ウェアラブル端末２００とを有する。 FIG. 1 shows the configuration of the stress estimation system 150 in this embodiment. Referring to FIG. 1, the stress estimation system 150 according to the present embodiment includes a stress estimation device 100 and a wearable terminal 200.
　ストレス推定装置１００は、１又は複数のコンピュータによって構成される。ストレス推定装置１００は、本実施形態では、被測定者の身体の一部、例えば腕に装着されたウェアラブル端末２００と、有線または無線によるデータ通信が可能である。 The stress estimation device 100 is composed of one or a plurality of computers. In the present embodiment, the stress estimation device 100 can perform wired or wireless data communication with a part of the body of the person to be measured, for example, a wearable terminal 200 worn on an arm.
　ウェアラブル端末２００は、被測定者の生体信号「Ｓｂ」を計測し、計測した生体信号Ｓｂをストレス推定装置１００に供給する。生体信号Ｓｂは、例えば、被測定者の発汗量、皮膚表面温、体動、脈拍数、心拍数、呼吸数等を示す信号である。なお、生体信号Ｓｂは、上述した例に限られず、被測定者の自律神経活動を反映する情報等、被測定者のストレス等の精神状態を推定し得る情報であればよい。 The wearable terminal 200 measures the biological signal "Sb" of the person to be measured, and supplies the measured biological signal Sb to the stress estimation device 100. The biological signal Sb is, for example, a signal indicating the sweating amount, skin surface temperature, body movement, pulse rate, heart rate, respiratory rate, etc. of the subject. The biological signal Sb is not limited to the above-mentioned example, and may be any information that can estimate the mental state such as stress of the person to be measured, such as information reflecting the autonomic nerve activity of the person to be measured.
　なお、図１に示すストレス推定システム１５０の構成は、一例であり、種々の変更が行われてもよい。例えば、ストレス推定システム１５０は、被測定者が所有するスマートフォンなどの携帯機器端末をさらに有してもよい。この場合、ストレス推定装置１００とウェアラブル端末２００とは、携帯機器端末を介してデータ通信を行ってもよい。 The configuration of the stress estimation system 150 shown in FIG. 1 is an example, and various changes may be made. For example, the stress estimation system 150 may further have a mobile device terminal such as a smartphone owned by the subject. In this case, the stress estimation device 100 and the wearable terminal 200 may perform data communication via the portable device terminal.
　ストレス推定装置１００は、機能的には、生体信号取得部１０１と、生体信号記憶部１０２と、生体信号特徴量計算部１０３と、属性・パターンデータ取得部１０４と、属性・パターン共通特徴量選定部１０５と、ストレス推定部１０６と、を含む。 Functionally, the stress estimation device 100 includes the biological signal acquisition unit 101, the biological signal storage unit 102, the biological signal feature amount calculation unit 103, the attribute / pattern data acquisition unit 104, and the attribute / pattern common feature amount selection. A unit 105 and a stress estimation unit 106 are included.
　生体信号取得部１０１は、生体信号Ｓｂを、ウェアラブル端末２００からデータ通信により取得する。なお、ウェアラブル端末２００は、生成した生体信号Ｓｂを即時に生体信号取得部１０１に供給してもよく、生成した生体信号Ｓｂを蓄積し、蓄積した生体信号Ｓｂを所定のタイミングによりまとめて生体信号取得部１０１に供給してもよい。 The biological signal acquisition unit 101 acquires the biological signal Sb from the wearable terminal 200 by data communication. The wearable terminal 200 may immediately supply the generated biological signal Sb to the biological signal acquisition unit 101, accumulate the generated biological signal Sb, and collect the accumulated biological signal Sb at a predetermined timing. It may be supplied to the acquisition unit 101.
　生体信号記憶部１０２は、生体信号取得部１０１が取得した生体信号Ｓｂを記憶する。 The biological signal storage unit 102 stores the biological signal Sb acquired by the biological signal acquisition unit 101.
　生体信号特徴量計算部１０３は、生体信号記憶部１０２に記憶された生体信号Ｓｂからストレス推定に用いる特徴量（「生体信号特徴量Ｆｂ」とも呼ぶ。）を算出する。生体信号特徴量Ｆｂは、ストレスの推定に用いられる生体信号Ｓｂの任意の特徴量（統計量を含む）であり、例えば、非特許文献１、非特許文献２に開示されているように、生体信号Ｓｂの平均値、分散値、時系列ヒストグラム、パワースペクトル密度ヒストグラム等である。 The biological signal feature amount calculation unit 103 calculates the feature amount used for stress estimation (also referred to as “biological signal feature amount Fb”) from the biological signal Sb stored in the biological signal storage unit 102. The biological signal feature amount Fb is an arbitrary feature amount (including a statistic) of the biological signal Sb used for estimating stress, and is, for example, a biological signal as disclosed in Non-Patent Document 1 and Non-Patent Document 2. The average value, the dispersion value, the time series histogram, the power spectral density histogram, and the like of the signal Sb.
　属性・パターンデータ取得部１０４は、被測定者の属性または行動パターンの少なくとも一方を示すデータ（「属性・パターンデータＤａ」とも呼ぶ。）を取得する。属性・パターンデータＤａは、例えば、被測定者の生活パターンや就業パターンなどの任意の行動パターンを示す情報、又は／及び、性別や年齢の偏りを反映する、被測定者の職種、性別、年齢などの属性を示す情報である。 The attribute / pattern data acquisition unit 104 acquires data indicating at least one of the attributes or behavior patterns of the person to be measured (also referred to as "attribute / pattern data Da"). The attribute / pattern data Da is, for example, information indicating an arbitrary behavior pattern such as a life pattern or an employment pattern of the person to be measured, and / or a job type, sex, age of the person to be measured, which reflects a bias of gender or age. Information indicating attributes such as.
　ここで、属性・パターンデータ取得部１０４は、属性・パターンデータＤａを種々の方法により取得してもよい。第１の例では、属性・パターンデータ取得部１０４は、被測定者がウェアラブル端末２００またはストレス推定装置１００に対する任意のユーザインターフェース（音声入力装置を含む）により入力された入力情報に基づき、属性・パターンデータＤａを生成する。第２の例では、属性・パターンデータ取得部１０４は、被験者の生体信号Ｓｂから被測定者の属性・パターンを推定することで、属性・パターンデータＤａを生成する。第３の例では、被測定者の属性・パターンデータＤａを予め記憶する記憶装置から、被測定者の属性・パターンデータＤａを取得する。上述の記憶装置は、ストレス推定装置１００内のメモリであってもよく、ストレス推定装置１００とは別の外部装置（例えば属性・パターンに関するデータベースを管理するサーバ等）であってもよい。 Here, the attribute / pattern data acquisition unit 104 may acquire the attribute / pattern data Da by various methods. In the first example, the attribute / pattern data acquisition unit 104 uses the attribute / pattern data acquisition unit 104 based on the input information input by the person to be measured by any user interface (including a voice input device) to the wearable terminal 200 or the stress estimation device 100. Generate pattern data Da. In the second example, the attribute / pattern data acquisition unit 104 generates the attribute / pattern data Da by estimating the attribute / pattern of the subject from the biological signal Sb of the subject. In the third example, the attribute / pattern data Da of the measured person is acquired from the storage device that stores the attribute / pattern data Da of the measured person in advance. The above-mentioned storage device may be a memory in the stress estimation device 100, or may be an external device (for example, a server that manages a database related to attributes / patterns) different from the stress estimation device 100.
　属性・パターン共通特徴量選択部１０５は、生体信号特徴量計算部１０３が算出した生体信号特徴量Ｆｂから、属性・パターンデータＤａが示す属性・パターンによって違いのない、即ち属性・パターンに依存しない特徴量（「属性・パターン共通特徴量Ｆｃ」とも呼ぶ。）を選択する。この場合、属性・パターン共通特徴量選択部１０５は、属性またはパターンを区別し得る特徴量（「敵対的特徴量Ｆａ」とも呼ぶ。）を生体信号特徴量Ｆｂから排除し、残りの生体信号特徴量Ｆｂを属性・パターン共通特徴量Ｆｃとして特定する。この場合、属性・パターン共通特徴量選択部１０５は、具体的な手法としては、属性・パターンの共通の特徴量の選択、すなわち属性・パターンを区別し得る敵対的特徴量Ｆａを選択し、敵対的特徴量Ｆａを排除するプロセス（以降、「敵対的特徴量選択」と呼ぶ。）を実行する。この場合、例えば、訓練データにおいて、敵対的特徴量選択と、目的であるストレス推定とを交互に行い、敵対的特徴量選択のモデルへの入力及び出力として最適な特徴量セットを得ることで、敵対的特徴量選択のモデルの学習を予め行っておく。学習により得られた敵対的特徴量選択のモデルのパラメータは、属性・パターン共通特徴量選択部１０５により参照可能なメモリ等に記憶される。そして、属性・パターン共通特徴量選択部１０５は、当該パラメータを参照することで敵対的特徴量選択のモデルを構成し、当該モデルにより生体信号特徴量Ｆｂから敵対的特徴量Ｆａを選択する。 The attribute / pattern common feature amount selection unit 105 does not differ from the biometric signal feature amount Fb calculated by the biometric signal feature amount calculation unit 103 depending on the attribute / pattern indicated by the attribute / pattern data Da, that is, does not depend on the attribute / pattern. Select the feature amount (also referred to as "attribute / pattern common feature amount Fc"). In this case, the attribute / pattern common feature amount selection unit 105 excludes the feature amount that can distinguish the attribute or the pattern (also referred to as “hostile feature amount Fa”) from the biometric signal feature amount Fb, and the remaining biometric signal features. The quantity Fb is specified as the attribute / pattern common feature quantity Fc. In this case, the attribute / pattern common feature amount selection unit 105 selects, as a specific method, the selection of the common feature amount of the attribute / pattern, that is, the hostile feature amount Fa capable of distinguishing the attribute / pattern, and is hostile. The process of eliminating the target feature amount Fa (hereinafter referred to as "hostile feature amount selection") is executed. In this case, for example, in the training data, hostile feature selection and target stress estimation are performed alternately to obtain the optimum feature set as input and output to the model of hostile feature selection. The model of hostile feature selection is trained in advance. The parameters of the model of hostile feature amount selection obtained by learning are stored in a memory or the like that can be referred to by the attribute / pattern common feature amount selection unit 105. Then, the attribute / pattern common feature amount selection unit 105 constitutes a model of hostile feature amount selection by referring to the parameter, and selects the hostile feature amount Fa from the biological signal feature amount Fb by the model.
　ストレス推定部１０６は、属性・パターン共通特徴量選択部１０５によって選択された属性・パターン共通特徴量Ｆｃに基づき、被測定者のストレスを推定する。この場合、ストレス推定部１０６は、例えば、機械学習により得られるモデル（「ストレス推定モデル」とも呼ぶ。）に基づき、属性・パターン共通特徴量Ｆｃから推定されるストレスのスコアを算出する。学習により得られたストレス推定モデルのパラメータは、ストレス推定部１０６が参照可能なメモリ等に記憶される。ストレスの推定後、推定されたストレスを示すスコアは、被測定者に提示され、ストレスの低減を促す目的等に使用される。 The stress estimation unit 106 estimates the stress of the subject based on the attribute / pattern common feature amount Fc selected by the attribute / pattern common feature amount selection unit 105. In this case, the stress estimation unit 106 calculates a stress score estimated from the attribute / pattern common feature amount Fc, for example, based on a model obtained by machine learning (also referred to as a “stress estimation model”). The parameters of the stress estimation model obtained by learning are stored in a memory or the like that can be referred to by the stress estimation unit 106. After the stress is estimated, the score indicating the estimated stress is presented to the subject and used for the purpose of promoting the reduction of stress.
　図２は、ストレス推定装置１００のハードウェア構成を示す。ストレス推定装置１００は、例えば、ハードウェアとして、プロセッサ１１と、メモリ１２と、インターフェース１３とを含む。プロセッサ１１、メモリ１２及びインターフェース１３は、データバス１９を介して接続されている。 FIG. 2 shows the hardware configuration of the stress estimation device 100. The stress estimation device 100 includes, for example, a processor 11, a memory 12, and an interface 13 as hardware. The processor 11, the memory 12, and the interface 13 are connected via the data bus 19.
　プロセッサ１１は、メモリ１２に記憶されているプログラムを実行することにより、所定の処理を実行する。プロセッサ１１は、ＣＰＵ（Ｃｅｎｔｒａｌ　Ｐｒｏｃｅｓｓｉｎｇ　Ｕｎｉｔ）、ＧＰＵ（Ｇｒａｐｈｉｃｓ　Ｐｒｏｃｅｓｓｉｎｇ　Ｕｎｉｔ）、量子プロセッサなどのプロセッサである。そして、プロセッサ１１は、例えば、メモリ１２に格納されたプログラムを実行することで、図１に示される生体信号取得部１０１、生体信号特徴量計算部１０３、属性・パターンデータ取得部１０４、属性・パターン共通特徴量選択部１０５及びストレス推定部１０６として機能する。 The processor 11 executes a predetermined process by executing the program stored in the memory 12. The processor 11 is a processor such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and a quantum processor. Then, for example, by executing the program stored in the memory 12, the processor 11 executes the biological signal acquisition unit 101, the biological signal feature amount calculation unit 103, the attribute / pattern data acquisition unit 104, and the attribute / pattern shown in FIG. It functions as a pattern common feature amount selection unit 105 and a stress estimation unit 106.
　なお、生体信号取得部１０１、生体信号特徴量計算部１０３、属性・パターンデータ取得部１０４、属性・パターン共通特徴量選択部１０５及びストレス推定部１０６の各構成要素は、プログラムによるソフトウェアで実現することに限ることなく、ハードウェア、ファームウェア、及びソフトウェアのうちのいずれかの組合せ等により実現されるコントローラであってもよい。また、これらの各構成要素は、例えばＦＰＧＡ（field-programmable gate array）又はマイコン等の、ユーザがプログラミング可能な集積回路を用いて実現してもよい。この場合、この集積回路を用いて、上記の各構成要素から構成されるプログラムを実現してもよい。 Each component of the biological signal acquisition unit 101, the biological signal feature amount calculation unit 103, the attribute / pattern data acquisition unit 104, the attribute / pattern common feature amount selection unit 105, and the stress estimation unit 106 is realized by software by a program. The controller is not limited to this, and may be a controller realized by a combination of any one of hardware, firmware, and software. Further, each of these components may be realized by using a user-programmable integrated circuit such as an FPGA (field-programmable gate array) or a microcomputer. In this case, this integrated circuit may be used to realize a program composed of each of the above components.
　メモリ１２は、ＲＡＭ（Ｒａｎｄｏｍ　Ａｃｃｅｓｓ　Ｍｅｍｏｒｙ）、ＲＯＭ（Ｒｅａｄ　Ｏｎｌｙ　Ｍｅｍｏｒｙ）などの各種の揮発性メモリ及び不揮発性メモリにより構成される。また、メモリ１２には、ストレス推定装置１００が実行する処理を実行するためのプログラムが記憶される。例えば、メモリ１２は、生体信号記憶部１０２として機能する。なお、ストレス推定装置１００が実行するプログラムは、メモリ１２以外の記憶媒体に記憶されてもよい。同様に、生体信号記憶部１０２は、ストレス推定装置１００以外の外部装置に記憶されてもよい。 The memory 12 is composed of various volatile memories such as RAM (Random Access Memory) and ROM (Read Only Memory) and non-volatile memory. Further, the memory 12 stores a program for executing the process executed by the stress estimation device 100. For example, the memory 12 functions as a biological signal storage unit 102. The program executed by the stress estimation device 100 may be stored in a storage medium other than the memory 12. Similarly, the biological signal storage unit 102 may be stored in an external device other than the stress estimation device 100.
　インターフェース１３は、ストレス推定装置１００と他の装置とを電気的に接続するためのインターフェースである。例えば、インターフェース１３は、ストレス推定装置１００がウェアラブル端末２００などの外部装置と有線又は無線によりデータ通信を行うための通信インターフェースであってもよい。他の例では、インターフェース１３は、ＵＳＢ（Ｕｎｉｖｅｒｓａｌ　Ｓｅｒｉａｌ　Ｂｕｓ）、ＳＡＴＡ（Ｓｅｒｉａｌ　ＡＴ　Ａｔｔａｃｈｍｅｎｔ）などに準拠したハードウェアインターフェースであってもよい。また、インターフェース１３を介してウェアラブル端末２００以外の種々の装置がストレス推定装置１００と電気的に接続してもよい。例えば、属性・パターンデータＤａに関するユーザ入力を受け付ける入力装置、ストレス推定部１０６による推定結果を出力する表示装置又は音出力装置等が、インターフェース１３を介してストレス推定装置１００とデータの授受を行ってもよい。 The interface 13 is an interface for electrically connecting the stress estimation device 100 and another device. For example, the interface 13 may be a communication interface for the stress estimation device 100 to perform data communication with an external device such as a wearable terminal 200 by wire or wirelessly. In another example, the interface 13 may be a hardware interface compliant with USB (Universal Serial Bus), SATA (Serial AT Atchment), or the like. Further, various devices other than the wearable terminal 200 may be electrically connected to the stress estimation device 100 via the interface 13. For example, an input device that accepts user input related to attribute / pattern data Da, a display device that outputs an estimation result by the stress estimation unit 106, a sound output device, or the like exchanges data with the stress estimation device 100 via the interface 13. May be good.
　なお、ストレス推定装置１００のハードウェア構成は、図２に示す構成に限定されない。例えば、ストレス推定装置１００は、入力装置、表示装置、音出力装置の少なくともいずれかを内蔵するタブレット端末等であってもよい。 The hardware configuration of the stress estimation device 100 is not limited to the configuration shown in FIG. For example, the stress estimation device 100 may be a tablet terminal or the like incorporating at least one of an input device, a display device, and a sound output device.
　［動作の説明］
　次に、第１実施形態におけるストレス推定装置１００の動作について、図３を参照して説明する。図３は、ストレス推定装置１００の動作を示すフローチャートの一例である。 [Explanation of operation]
Next, the operation of the stress estimation device 100 in the first embodiment will be described with reference to FIG. FIG. 3 is an example of a flowchart showing the operation of the stress estimation device 100.
　まず、ストレス推定装置１００の生体信号取得部１０１は、ウェアラブル端末２００から送信された生体信号Ｓｂを取得する（ステップＳ１１）。そして、生体信号取得部１０１は、取得した生体信号Ｓｂを、生体信号記憶部１０２に記憶する（ステップＳ１２）。 First, the biological signal acquisition unit 101 of the stress estimation device 100 acquires the biological signal Sb transmitted from the wearable terminal 200 (step S11). Then, the biological signal acquisition unit 101 stores the acquired biological signal Sb in the biological signal storage unit 102 (step S12).
　生体信号特徴量計算部１０３は、生体信号記憶部１０２に記憶された生体信号Ｓｂに基づき、生体信号特徴量Ｆｂを算出する（ステップＳ１３）。また、属性・パターンデータ取得部１０４は、ステップＳ１１～ステップＳ１３が実行される処理と並行して又はこれらの前後のタイミングにおいて、属性・パターンデータＤａを取得する（ステップＳ１４）。 The biological signal feature amount calculation unit 103 calculates the biological signal feature amount Fb based on the biological signal Sb stored in the biological signal storage unit 102 (step S13). Further, the attribute / pattern data acquisition unit 104 acquires the attribute / pattern data Da in parallel with the processing in which steps S11 to S13 are executed, or at the timing before and after these (step S14).
　そして、属性・パターン共通特徴量選定部１０５は、属性・パターンデータＤａに基づき、ステップＳ１３で算出された生体信号特徴量Ｆｂが、属性・パターンデータＤａが示す属性・パターンを区別できる特徴量である敵対的特徴量Ｆａに該当するか否か判定する（ステップＳ１５）。そして、属性・パターン共通特徴量選定部１０５は、生体信号特徴量Ｆｂが属性・パターンを区別できる敵対的特徴量Ｆａに該当する場合（ステップＳ１５；Ｙｅｓ）、対象の生体信号特徴量Ｆｂは属性・パターン共通特徴量Ｆｃに該当しないと判定し、ステップＳ１６でのストレス推定を行うことなくステップＳ１７へ処理を進める。 Then, in the attribute / pattern common feature amount selection unit 105, the biological signal feature amount Fb calculated in step S13 is a feature amount capable of distinguishing the attribute / pattern indicated by the attribute / pattern data Da based on the attribute / pattern data Da. It is determined whether or not it corresponds to a certain hostile feature amount Fa (step S15). Then, when the attribute / pattern common feature amount selection unit 105 corresponds to the hostile feature amount Fa capable of distinguishing the attribute / pattern (step S15; Yes), the target biometric signal feature amount Fb is an attribute. -It is determined that the pattern does not correspond to the common feature amount Fc, and the process proceeds to step S17 without performing stress estimation in step S16.
　一方、属性・パターン共通特徴量選定部１０５は、生体信号特徴量Ｆｂが敵対的特徴量Ｆａに該当しない場合（ステップＳ１５；Ｎｏ）、対象の生体信号特徴量Ｆｂを属性・パターン共通特徴量Ｆｃとみなす。そして、ストレス推定部１０６は、この場合、属性・パターン共通特徴量Ｆｃに基づき、被測定者のストレスを推定する（ステップＳ１６）。 On the other hand, when the biological signal feature amount Fb does not correspond to the hostile feature amount Fa (step S15; No), the attribute / pattern common feature amount selection unit 105 sets the target biological signal feature amount Fb as the attribute / pattern common feature amount Fc. Consider it as. Then, in this case, the stress estimation unit 106 estimates the stress of the person to be measured based on the attribute / pattern common feature amount Fc (step S16).
　その後、ストレス推定装置１００は、ストレス推定処理を終了すべきか否か判定する（ステップＳ１７）。例えば、ストレス推定装置１００は、生体信号Ｓｂを取得できなくなった場合、ストレス推定を終了する旨のユーザ入力を検知した場合、又は、予め定めたその他のストレス推定処理の終了条件が満たされた場合、ストレス推定処理を終了すべきと判定する。そして、ストレス推定装置１００は、ストレス推定処理を終了すべきと判定した場合（ステップＳ１７；Ｙｅｓ）、フローチャートの処理を終了する。一方、ストレス推定装置１００は、ストレス推定処理を終了すべきでないと判定した場合（ステップＳ１７；Ｎｏ）、ステップＳ１１へ処理を戻す。 After that, the stress estimation device 100 determines whether or not the stress estimation process should be completed (step S17). For example, the stress estimation device 100 cannot acquire the biological signal Sb, detects a user input to end the stress estimation, or satisfies other predetermined termination conditions of the stress estimation process. , Judge that the stress estimation process should be completed. Then, when the stress estimation device 100 determines that the stress estimation process should be completed (step S17; Yes), the stress estimation device 100 ends the process of the flowchart. On the other hand, when the stress estimation device 100 determines that the stress estimation process should not be completed (step S17; No), the stress estimation device 100 returns the process to step S11.
　［効果の説明］
　次に、第１実施形態の効果について補足説明する。 [Explanation of effect]
Next, the effect of the first embodiment will be supplementarily described.
　第１実施形態では、属性・パターンを区別する特徴量である敵対的特徴量Ｆａを敵対的特徴量選択の手法で排除することで、属性・パターンによって違いのない特徴量である属性・パターン共通特徴量Ｆｃだけを用いてストレス推定を行うことができる。そのため、被測定者の属性・パターンに依存することなく、高精度に被測定者のストレスを推定することができる。 In the first embodiment, by eliminating the hostile feature amount Fa, which is a feature amount that distinguishes attributes / patterns, by a method of selecting a hostile feature amount, the attribute / pattern common, which is a feature amount that does not differ depending on the attribute / pattern. Stress estimation can be performed using only the feature Fc. Therefore, the stress of the person to be measured can be estimated with high accuracy without depending on the attributes and patterns of the person to be measured.
　例えば、非特許文献１では学生、非特許文献２ではＩＴ企業のオフィス勤務者等、被験者の生活パターンや年齢層に偏りのあるデータベースしか取得できない場合が多い。また、非特許文献４では、ドメイン不変のモデルを得る手法であり、偏りのあるデータベースに対応可能だが、大量データが前提の深層学習を基盤にしている為、大量のデータが得にくい長期ストレスのようなデータベースの分析には適用できない。これに対し、第１実施形態では、大量の訓練データがなく、偏りのあるデータベースにより作成されたストレス推定システムであっても、属性・パターンに依存することなく被測定者のストレスを好適に推定することができる。 For example, in non-patent document 1, students, in non-patent document 2, office workers of IT companies, etc., in many cases, only databases with biased subject life patterns and age groups can be obtained. In addition, Non-Patent Document 4 is a method for obtaining a domain-invariant model, which can handle a biased database, but since it is based on deep learning on the premise of a large amount of data, it is difficult to obtain a large amount of data for long-term stress. Not applicable to such database analysis. On the other hand, in the first embodiment, even if the stress estimation system is created by a biased database without a large amount of training data, the stress of the person to be measured is suitably estimated without depending on the attributes and patterns. can do.
　＜第２実施形態＞
　次に、図４～図７を参照し、第１実施形態をより具体化した実施形態である第２実施形態について説明する。 <Second Embodiment>
Next, with reference to FIGS. 4 to 7, a second embodiment, which is a more specific embodiment of the first embodiment, will be described.
　図４は、第２実施形態に係るストレス推定システム１５０Ａの構成を示す。ストレス推定システム１５０Ａは、複数のウェアラブル端末４００と、コンピュータ６００とを有している。コンピュータ６００と、各ウェアラブル端末４００とは、通信手段５００（５０１、５０２、５０３、５０４）を介し、通信を行う。 FIG. 4 shows the configuration of the stress estimation system 150A according to the second embodiment. The stress estimation system 150A has a plurality of wearable terminals 400 and a computer 600. The computer 600 and each wearable terminal 400 communicate with each other via the communication means 500 (501, 502, 503, 504).
　ウェアラブル端末４００は、被測定者３００の生体信号「Ｓｂ」を取得する。被測定者３００の生体信号Ｓｂは、被測定者の発汗を反映する皮膚表面電気活動（Electrodermal Activity）であってもよい。生体信号Ｓｂの他の例は、体温、脈波、心拍、音声、脳波、呼吸、筋電、心電、体動等の種々の生体情報を反映する信号であってもよい。このように、生体信号Ｓｂは、被測定者の精神活動の影響を受ける任意の生体情報を反映した信号であってもよい。 The wearable terminal 400 acquires the biological signal "Sb" of the person to be measured 300. The biological signal Sb of the person to be measured 300 may be a skin surface electrical activity (Electrodermal Activity) that reflects the sweating of the person to be measured. Another example of the biological signal Sb may be a signal that reflects various biological information such as body temperature, pulse wave, heartbeat, voice, brain wave, respiration, myoelectricity, electrocardiogram, and body movement. As described above, the biological signal Sb may be a signal that reflects arbitrary biological information affected by the mental activity of the subject.
　ウェアラブル端末４００は、被測定者３００が着用可能な端末であって、前記に挙げた生体情報を反映する生体信号のうち少なくともいずれかを測定する。例えば、ウェアラブル端末４００は、皮膚導電性を一定のサンプリングレートで取得し、内蔵メモリに保存する。ウェアラブル端末４００は、リストバンドタイプ、バッジタイプ、社員証タイプ、イヤホンタイプ、シャツタイプ等の種々の形態であってもよい。 The wearable terminal 400 is a terminal that can be worn by the person to be measured 300, and measures at least one of the biological signals that reflect the biological information mentioned above. For example, the wearable terminal 400 acquires skin conductivity at a constant sampling rate and stores it in the built-in memory. The wearable terminal 400 may be in various forms such as a wristband type, a badge type, an employee ID card type, an earphone type, and a shirt type.
　通信手段５００（５０１、５０２、５０３、５０４）は、ウェアラブル端末４００で取得した生体信号Ｓｂ（加速度信号を含んでもよい）を、コンピュータ６００に送信する。具体的には、例えば、ウェアラブル端末４００は、Ｂｌｕｅｔｏｏｔｈ（登録商標）などの近距離通信５０１によりスマートフォン５０２に接続し、生体信号Ｓｂをスマートフォン５０２に送信する。その後、スマートフォン５０２は、インストールされたアプリケーションにより、生体信号Ｓｂをパケット通信５０３によってインターネット５０４に送信する。これにより、インターネット５０４に接続したコンピュータ６００に生体信号Ｓｂがアップロードされる。 The communication means 500 (501, 502, 503, 504) transmits the biological signal Sb (which may include an acceleration signal) acquired by the wearable terminal 400 to the computer 600. Specifically, for example, the wearable terminal 400 connects to the smartphone 502 by short-range communication 501 such as Bluetooth (registered trademark), and transmits the biological signal Sb to the smartphone 502. After that, the smartphone 502 transmits the biological signal Sb to the Internet 504 by packet communication 503 by the installed application. As a result, the biological signal Sb is uploaded to the computer 600 connected to the Internet 504.
　コンピュータ６００は、第１実施形態のストレス推定装置１００に相当し、例えば、図２に示されるようなハードウェア構成を有している。 The computer 600 corresponds to the stress estimation device 100 of the first embodiment, and has, for example, a hardware configuration as shown in FIG.
　図５は、第２実施形態におけるコンピュータ６００の機能的な構成を示す。コンピュータ６００は、機能的には、通信インターフェース（Ｉ／Ｆ）６０１と、生体信号取得部６０２と、生体信号記憶部６０３と、属性・パターンデータ取得部６０４と、属性・パターンデータ記憶部６０５と、生体信号特徴量計算部６０６と、生体信号特徴量記憶部６０７と、属性・パターン共通特徴量選択部６０８と、属性・パターン共通特徴量記憶部６０９と、ストレス推定部６１０と、ストレス推定結果記憶部６１１と、ストレス推定結果出力部６１２とを含む。 FIG. 5 shows the functional configuration of the computer 600 in the second embodiment. Functionally, the computer 600 includes a communication interface (I / F) 601, a biometric signal acquisition unit 602, a biometric signal storage unit 603, an attribute / pattern data acquisition unit 604, and an attribute / pattern data storage unit 605. , Biosignal feature amount calculation unit 606, biometric signal feature amount storage unit 607, attribute / pattern common feature amount selection unit 608, attribute / pattern common feature amount storage unit 609, stress estimation unit 610, and stress estimation results. It includes a storage unit 611 and a stress estimation result output unit 612.
　まず、生体信号取得部６０２は、通信インターフェース６０１から得られた生体信号Ｓｂを、生体信号記憶部６０３に記憶する。次に、属性・パターンデータ取得部６０４は、属性・パターンデータ「Ｄａ」を取得する。第２実施形態では、属性・パターンデータＤａは、被測定者の職種（営業職・技術職）を示す。取得された属性・パターンデータＤａは、属性・パターンデータ記憶部６０５に記憶される。なお、属性・パターンデータ取得部６０４は、ウェアラブル端末４００又はその他の外部装置から被測定者の属性・パターンを示す属性・パターンデータＤａを取得してもよく、生体信号Ｓｂから被測定者の属性・パターンを推定して属性・パターンデータＤａを生成してもよい。 First, the biological signal acquisition unit 602 stores the biological signal Sb obtained from the communication interface 601 in the biological signal storage unit 603. Next, the attribute / pattern data acquisition unit 604 acquires the attribute / pattern data “Da”. In the second embodiment, the attribute / pattern data Da indicates the job type (sales position / technical position) of the person to be measured. The acquired attribute / pattern data Da is stored in the attribute / pattern data storage unit 605. The attribute / pattern data acquisition unit 604 may acquire the attribute / pattern data Da indicating the attribute / pattern of the person to be measured from the wearable terminal 400 or other external device, and the attribute / pattern data Da of the person to be measured may be acquired from the biological signal Sb. -The pattern may be estimated to generate the attribute / pattern data Da.
　次に、生体信号特徴量計算部６０６は、生体信号記憶部６０３から抽出した生体信号Ｓｂに基づき、ストレスに関する特徴量である生体信号特徴量「Ｆｂ」を算出する。本実施形態において、生体信号特徴量Ｆｂは、発汗及び体動の時系列ヒストグラム、パワースペクトル密度、統計量（平均値、中央値、分散値）である。算出された生体信号特徴量Ｆｂは、生体信号特徴量記憶部６０７に記憶される。 Next, the biological signal feature amount calculation unit 606 calculates the biological signal feature amount "Fb", which is a characteristic amount related to stress, based on the biological signal Sb extracted from the biological signal storage unit 603. In the present embodiment, the biological signal feature amount Fb is a time-series histogram of sweating and body movement, a power spectral density, and a statistic (mean value, median value, dispersion value). The calculated biological signal feature amount Fb is stored in the biological signal feature amount storage unit 607.
　次に、属性・パターン共通特徴量選択部６０８は、属性・パターンを区別する特徴量である敵対的特徴量「Ｆａ」を排除することで、属性・パターンによって違いのない特徴量である属性・パターン共通特徴量「Ｆｃ」を生体信号特徴量Ｆｂから選択する。 Next, the attribute / pattern common feature amount selection unit 608 eliminates the hostile feature amount “Fa”, which is a feature amount that distinguishes attributes / patterns, so that the attribute / pattern is a feature amount that does not differ depending on the attribute / pattern. The pattern common feature amount "Fc" is selected from the biometric signal feature amount Fb.
　本実施形態では、属性・パターン共通特徴量選択部６０８は、ランダムフォレストを用いた敵対的特徴量選択を行い、選択した敵対的特徴量Ｆａを生体信号特徴量Ｆｂから排除する。そして、属性・パターン共通特徴量選択部６０８は、属性・パターンに依存しない残りの生体信号特徴量Ｆｂを属性・パターン共通特徴量Ｆｃとして選択する。 In the present embodiment, the attribute / pattern common feature amount selection unit 608 selects the hostile feature amount using a random forest, and excludes the selected hostile feature amount Fa from the biological signal feature amount Fb. Then, the attribute / pattern common feature amount selection unit 608 selects the remaining biological signal feature amount Fb that does not depend on the attribute / pattern as the attribute / pattern common feature amount Fc.
　具体的には、まず、属性・パターン共通特徴量選択部６０８は、ランダムフォレストの手法により、属性・パターンの異なるデータをクラス分類する。第２実施形態では、属性・パターン共通特徴量選択部６０８は、技術職と営業職を識別する２クラス分類を行う。このクラス分類において、属性・パターン共通特徴量選択部６０８は、各生体信号特徴量Ｆｂの重要度を評価する。そして、属性・パターン共通特徴量選択部６０８は、所定の閾値以上の重要度となる生体信号特徴量Ｆｂを、敵対的特徴量Ｆａとみなし、ストレス推定に用いる特徴量セットから排除する。なお、ランダムフォレストを用いた敵対的特徴量選択に必要なパラメータは、コンピュータ６００のメモリ又は外部記憶装置に記憶されている。 Specifically, first, the attribute / pattern common feature amount selection unit 608 classifies data having different attributes / patterns into classes by a random forest method. In the second embodiment, the attribute / pattern common feature amount selection unit 608 performs two-class classification for identifying the technical position and the sales position. In this classification, the attribute / pattern common feature amount selection unit 608 evaluates the importance of each biological signal feature amount Fb. Then, the attribute / pattern common feature amount selection unit 608 considers the biological signal feature amount Fb having an importance equal to or higher than a predetermined threshold value as the hostile feature amount Fa, and excludes it from the feature amount set used for stress estimation. The parameters required for hostile feature selection using the random forest are stored in the memory of the computer 600 or an external storage device.
　属性・パターン共通特徴量選択部６０８は、敵対的特徴量選択を経て選定した属性・パターンに共通の属性・パターン共通特徴量Ｆｃを、属性・パターン共通特徴量記憶部６０９に記憶する。 The attribute / pattern common feature amount selection unit 608 stores the attribute / pattern common feature amount Fc common to the attributes / patterns selected through the hostile feature amount selection in the attribute / pattern common feature amount storage unit 609.
　次に、ストレス推定部６１０は、属性・パターン共通特徴量記憶部６０９に記憶された属性・パターン共通特徴量Ｆｃに基づき、ストレスを推定する。 Next, the stress estimation unit 610 estimates stress based on the attribute / pattern common feature amount Fc stored in the attribute / pattern common feature amount storage unit 609.
　具体的には、第２実施形態では、ストレス推定部６１０は、ストレスの正解値として、ＰＳＳ（Perceived Stress Scale)の１０項目版（以降、「ＰＳＳ１０」と呼ぶ。）のスコアを用い、このＰＳＳ１０のスコアを回帰分析によって推定するストレス推定モデルを作成する。この際、被測定者に対して実験期間（例えば４週間）の最後に実施したＰＳＳアンケートから算出したＰＳＳ１０のスコアを教師データ（正解値）とし、当該被測定者から得られた生体信号に基づく属性・パターン共通特徴量Ｆｃを対応する入力データとする訓練データを用意する。そして、ストレス推定部６１０は、この訓練データに基づきＳＶＭモデル等の機械学習モデルを学習することで得られたストレス推定モデルを用い、属性・パターン共通特徴量ＦｃからＰＳＳ１０のスコアを推定する。そして、ストレス推定部６１０は、推定したＰＳＳ１０のスコアを、ストレス推定結果として生成する。上述の学習のより詳細な説明については、図６及び図７を参照して後述する。なお、上述の機械学習モデルの学習により得られたストレス推定モデルのパラメータは、ストレス推定部６１０が参照できるように、コンピュータ６００のメモリ又は外部記憶装置に記憶される。推定したストレス推定結果は、ストレス推定結果記憶部６１１に記憶される。 Specifically, in the second embodiment, the stress estimation unit 610 uses the score of the 10-item version of PSS (Perceived Stress Scale) (hereinafter referred to as “PSS10”) as the correct answer value of stress, and this PSS10. Create a stress estimation model that estimates the score of. At this time, the score of PSS10 calculated from the PSS questionnaire conducted at the end of the experimental period (for example, 4 weeks) for the subject is used as the teacher data (correct answer value), and is based on the biological signal obtained from the subject. Prepare training data using the attribute / pattern common feature quantity Fc as the corresponding input data. Then, the stress estimation unit 610 uses a stress estimation model obtained by learning a machine learning model such as an SVM model based on this training data, and estimates the score of PSS10 from the attribute / pattern common feature amount Fc. Then, the stress estimation unit 610 generates the estimated PSS10 score as the stress estimation result. A more detailed explanation of the above-mentioned learning will be described later with reference to FIGS. 6 and 7. The parameters of the stress estimation model obtained by learning the machine learning model described above are stored in the memory of the computer 600 or an external storage device so that the stress estimation unit 610 can refer to them. The estimated stress estimation result is stored in the stress estimation result storage unit 611.
　次に、被測定者３００の要求に応じて、ストレス推定結果出力部６１２は、ストレス推定結果記憶部６１１に記憶されたストレス推定の結果を出力する。 Next, in response to the request of the person to be measured 300, the stress estimation result output unit 612 outputs the stress estimation result stored in the stress estimation result storage unit 611.
　出力方法は、具体的には、例えば、画面出力、印刷出力などが挙げられるが、これに限らない。出力するタイミングは、常時、または被測定者の要求によって、出力することが挙げられる。具体的には、画面出力の場合、ストレス推定結果出力部６１２は、ストレス推定結果記憶部６１１に記憶されたストレス推定結果を、通信インターフェース６０１を通じて、通信手段５００を用いて、ウェアラブル端末４００またはスマートフォン５０２に送信する。その後、ウェアラブル端末４００またはスマートフォン５０２は、付随する画面においてストレス推定結果を出力する。これにより、コンピュータ６００は、ストレス推定結果を好適に被測定者３００に提示することができる。 Specific examples of the output method include, but are not limited to, screen output and print output. The timing of output may be output at all times or at the request of the person to be measured. Specifically, in the case of screen output, the stress estimation result output unit 612 uses the communication means 500 to store the stress estimation result stored in the stress estimation result storage unit 611 through the communication interface 601 to the wearable terminal 400 or the smartphone. Send to 502. After that, the wearable terminal 400 or the smartphone 502 outputs the stress estimation result on the accompanying screen. Thereby, the computer 600 can preferably present the stress estimation result to the person to be measured 300.
　ここで、上述したストレス推定モデルの作成プロセス（学習）について、図６及び図７を用いて補足説明する。 Here, the above-mentioned stress estimation model creation process (learning) will be supplementarily explained with reference to FIGS. 6 and 7.
　図６は、ストレス推定モデルの作成プロセスの模式図を示す。以後では、一例として、コンピュータ６００がストレス推定モデルの作成プロセスを実行するものとして説明を行う。なお、ストレス推定モデルの作成プロセスは、コンピュータ６００以外の装置が実行してもよい。 FIG. 6 shows a schematic diagram of the process of creating a stress estimation model. Hereinafter, as an example, the computer 600 will be described as executing the process of creating the stress estimation model. The process of creating the stress estimation model may be executed by a device other than the computer 600.
　まず、コンピュータ６００は、営業職の生体信号Ｓｂ及び技術職の生体信号Ｓｂから、営業職の生体信号特徴量Ｆｂと、技術職の生体信号特徴量Ｆｂとを生成し、両属性・パターンを識別し得る敵対的特徴量Ｆａをランダムフォレストの手法で探索する。なお、図６の「営業職データ」は、営業職の生体信号Ｓｂ又は当該生体信号Ｓｂに基づく営業職の生体信号特徴量Ｆｂに相当し、図６の「技術職データ」は、技術職の生体信号Ｓｂ又は当該生体信号Ｓｂに基づく技術職の生体信号特徴量Ｆｂに相当する。 First, the computer 600 generates the biological signal feature amount Fb of the sales staff and the biological signal feature amount Fb of the technical staff from the biological signal Sb of the sales staff and the biological signal Sb of the technical staff, and discriminates both attributes / patterns. Search for possible hostile features Fa using the random forest method. The "sales staff data" in FIG. 6 corresponds to the biological signal Sb of the sales staff or the biological signal feature amount Fb of the sales staff based on the biological signal Sb, and the "technical staff data" of FIG. It corresponds to the biological signal Sb or the biological signal feature amount Fb of a technical worker based on the biological signal Sb.
　そして、コンピュータ６００は、探索した敵対的特徴量Ｆａを営業職及び技術職の各生体信号特徴量Ｆｂから排除した属性・パターン共通特徴量Ｆｃに基づき、ストレス推定モデルを作成する。この場合、まず、コンピュータ６００は、技術職の生体信号Ｓｂから生成した生体信号特徴量Ｆｂから敵対的特徴量Ｆａを排除した属性・パターン共通特徴量Ｆｃと、技術職のストレスアンケートの結果であるＰＳＳ１０のスコア（即ちストレス推定の際の目的変数の教師データ）とに基づき、ストレス推定モデルの学習を行う。そして、技術職に関するデータに基づき学習したストレス推定モデルによって、コンピュータ６００は、営業職の属性・パターン共通特徴量Ｆｃから、営業職のＰＳＳ１０のスコア（即ち、営業職のストレスアンケートの結果）を推定する。この推定では、一例として、Support Vector Regression（ＳＶＲ）を用いるものとする。 Then, the computer 600 creates a stress estimation model based on the attribute / pattern common feature amount Fc in which the searched hostile feature amount Fa is excluded from each biological signal feature amount Fb of the sales position and the technical position. In this case, first, the computer 600 is the result of the attribute / pattern common feature amount Fc excluding the hostile feature amount Fa from the biological signal feature amount Fb generated from the biological signal Sb of the technical worker, and the result of the stress questionnaire of the technical worker. The stress estimation model is trained based on the PSS10 score (that is, the teacher data of the objective variable at the time of stress estimation). Then, the computer 600 estimates the PSS10 score of the sales position (that is, the result of the stress questionnaire of the sales position) from the attribute / pattern common feature amount Fc of the sales position by the stress estimation model learned based on the data related to the technical position. do. In this estimation, Support Vector Regression (SVR) is used as an example.
　図７は、ＳＶＲによる営業職のストレス推定結果に関するグラフを示す。グラフにおいて、横軸はランダムフォレストによる敵対的特徴量選択によって選ばれた敵対的特徴量Ｆａの数（即ち排除する特徴量の数）を示し、縦軸はＳＶＲによるＰＳＳ１０の推定スコアと、正解となるＰＳＳ１０の実スコアとの誤差である。 FIG. 7 shows a graph regarding the stress estimation results of sales positions by SVR. In the graph, the horizontal axis shows the number of hostile features Fa selected by the hostile feature selection by random forest (that is, the number of features to be excluded), and the vertical axis shows the estimated score of PSS10 by SVR and the correct answer. It is an error from the actual score of PSS10.
　ここで、横軸に関連する、ランダムフォレストによる敵対的特徴量選択においては、営業職と技術職をよく識別できる特徴量、すなわち、職種という属性・パターンへの依存度が高い特徴量から順に敵対的特徴量Ｆａとして選んでいる。横軸が示す「排除する特徴量の数」が増加していくと、ストレス推定モデルに使用される特徴量が減少するので、属性・パターン（この場合は職種）に依存しない特徴量のみでストレス推定モデルが形成される傾向が強まる。よって、技術職のデータのみから学習したストレス推定モデルでも、営業職のストレスを精度よく推定できる傾向が強まる。このことは、特徴量数が増加するにつれて、縦軸が示す誤差が徐々に減っていく状況から理解できる。 Here, in the selection of hostile features by random forest related to the horizontal axis, the features that can distinguish sales and technical positions well, that is, the features that are highly dependent on the attributes and patterns of job types, are hostile in order. It is selected as the feature amount Fa. As the "number of features to be excluded" indicated by the horizontal axis increases, the features used in the stress estimation model decrease, so stress is stressed only by features that do not depend on attributes / patterns (in this case, occupation). There is an increasing tendency for estimation models to be formed. Therefore, even with the stress estimation model learned only from the data of the technical staff, there is a strong tendency that the stress of the sales staff can be estimated accurately. This can be understood from the situation in which the error indicated by the vertical axis gradually decreases as the number of features increases.
　その後、ランダムフォレストによって選択される敵対的特徴量Ｆａの数を更に増加させていくと、ストレス推定モデルに使用される特徴量がその分排除され、学習サンプル数が過少になることに起因してストレス推定モデルの性能が悪化し、縦軸が示す誤差が増加する。図７では、横軸が示す「排除する特徴量の数」が縦軸の誤差が最小となるときの数「Ｎｏｐｔ」を境として、縦軸の誤差が「排除する特徴量の数」の増加と共に増加する。 After that, when the number of hostile features Fa selected by the random forest is further increased, the features used in the stress estimation model are eliminated by that amount, and the number of training samples becomes too small. The performance of the stress estimation model deteriorates, and the error indicated by the vertical axis increases. In FIG. 7, the error on the vertical axis increases the “number of features to be excluded” with the number “Nopt” as the boundary when the “number of features to be excluded” indicated by the horizontal axis minimizes the error on the vertical axis. Increases with.
　以上を勘案し、第２実施形態では、訓練データにおいて、図７の縦軸に相当する推定誤差と敵対的特徴量選択のパラメータとの関係を求める。そして、コンピュータ６００は、推定誤差が最小値を取る時のパラメータによって選択された特徴量を排除することで、ストレス推定モデルを構成する。このようなストレス推定モデルを用いることで、コンピュータ６００は、被測定者のストレスを高精度に推定することができる。 In consideration of the above, in the second embodiment, the relationship between the estimation error corresponding to the vertical axis of FIG. 7 and the parameter of hostile feature amount selection is obtained in the training data. Then, the computer 600 constructs a stress estimation model by excluding the features selected by the parameters when the estimation error takes the minimum value. By using such a stress estimation model, the computer 600 can estimate the stress of the subject with high accuracy.
　＜第３実施形態＞
　図８は、第３実施形態におけるストレス推定装置１００Ｘのブロック図である。ストレス推定装置１００Ｘは、主に、生体信号取得手段１０１Ｘと、生体信号特徴量計算手段１０３Ｘと、属性・パターンデータ取得手段１０４Ｘと、属性・パターン共通特徴量選択手段１０５Ｘと、ストレス推定手段１０６Ｘとを有する。なお、ストレス推定装置１００Ｘは、複数の装置から構成されてもよい。 <Third Embodiment>
FIG. 8 is a block diagram of the stress estimation device 100X according to the third embodiment. The stress estimation device 100X mainly includes a biological signal acquisition means 101X, a biological signal feature amount calculation means 103X, an attribute / pattern data acquisition means 104X, an attribute / pattern common feature amount selection means 105X, and a stress estimation means 106X. Has. The stress estimation device 100X may be composed of a plurality of devices.
　生体信号取得手段１０１Ｘは、被測定者の生体信号を取得する。生体信号取得手段１０１Ｘは、第１実施形態の生体信号取得部１０１又は第２実施形態の生体信号取得部６０２とすることができる。他の例では、生体信号取得手段１０１Ｘは、第１実施形態において生体信号記憶部１０２から生体信号Ｓｂを取得する生体信号特徴量計算部１０３であってもよい。 The biological signal acquisition means 101X acquires the biological signal of the person to be measured. The biological signal acquisition unit 101X can be the biological signal acquisition unit 101 of the first embodiment or the biological signal acquisition unit 602 of the second embodiment. In another example, the biological signal acquisition means 101X may be the biological signal feature amount calculation unit 103 that acquires the biological signal Sb from the biological signal storage unit 102 in the first embodiment.
　生体信号特徴量計算手段１０３Ｘは、ストレスに関する生体信号の特徴量である生体信号特徴量を算出する。生体信号特徴量計算手段１０３Ｘは、第１実施形態の生体信号特徴量計算部１０３又は第２実施形態の生体信号特徴量計算部６０６とすることができる。 The biological signal feature amount calculation means 103X calculates the biological signal feature amount, which is the feature amount of the biological signal related to stress. The biological signal feature amount calculation means 103X can be the biological signal feature amount calculation unit 103 of the first embodiment or the biological signal feature amount calculation unit 606 of the second embodiment.
　属性・パターンデータ取得手段１０４Ｘは、被測定者の属性又はパターンの少なくとも一方を示す属性・パターンデータを取得する。属性・パターンデータ取得手段１０４Ｘは、第１実施形態の属性・パターンデータ取得部１０４又は第２実施形態の属性・パターンデータ取得部６０４とすることができる。 The attribute / pattern data acquisition means 104X acquires attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured. The attribute / pattern data acquisition unit 104X can be the attribute / pattern data acquisition unit 104 of the first embodiment or the attribute / pattern data acquisition unit 604 of the second embodiment.
　属性・パターン共通特徴量選択手段１０５Ｘは、生体信号特徴量から属性・パターンデータが示す属性・パターンに依存しない特徴量である属性・パターン共通特徴量を選択する。属性・パターン共通特徴量選択手段１０５Ｘは、第１実施形態の属性・パターン共通特徴量選定部１０５又は第２実施形態の属性・パターン共通特徴量選択部６０８とすることができる。 The attribute / pattern common feature amount selection means 105X selects an attribute / pattern common feature amount which is a feature amount independent of the attribute / pattern indicated by the attribute / pattern data from the biological signal feature amount. The attribute / pattern common feature amount selection means 105X can be the attribute / pattern common feature amount selection unit 105 of the first embodiment or the attribute / pattern common feature amount selection unit 608 of the second embodiment.
　ストレス推定手段１０６Ｘは、属性・パターン共通特徴量に基づき、被測定者のストレスを推定する。ストレス推定手段１０６Ｘは、第１実施形態のストレス推定部１０６又は第２実施形態のストレス推定部６１０とすることができる。 The stress estimation means 106X estimates the stress of the person to be measured based on the common feature amount of the attribute / pattern. The stress estimation means 106X can be the stress estimation unit 106 of the first embodiment or the stress estimation unit 610 of the second embodiment.
　図９は、第３実施形態においてストレス推定装置１００Ｘが実行するフローチャートの一例である。まず、生体信号取得手段１０１Ｘは、被測定者の生体信号を取得する（ステップＳ２１）。次に、生体信号特徴量計算手段１０３Ｘは、ストレスに関する生体信号の特徴量である生体信号特徴量を算出する（ステップＳ２２）。属性・パターンデータ取得手段１０４Ｘは、被測定者の属性又はパターンの少なくとも一方を示す属性・パターンデータを取得する（ステップＳ２３）。属性・パターン共通特徴量選択手段１０５Ｘは、生体信号特徴量から属性・パターンデータが示す属性・パターンに依存しない特徴量である属性・パターン共通特徴量を選択する（ステップＳ２４）。ストレス推定手段１０６Ｘは、属性・パターン共通特徴量に基づき、被測定者のストレスを推定する（ステップＳ２５）。 FIG. 9 is an example of a flowchart executed by the stress estimation device 100X in the third embodiment. First, the biological signal acquisition means 101X acquires the biological signal of the person to be measured (step S21). Next, the biological signal feature amount calculation means 103X calculates the biological signal feature amount, which is the feature amount of the biological signal related to stress (step S22). The attribute / pattern data acquisition unit 104X acquires attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured (step S23). The attribute / pattern common feature amount selection means 105X selects an attribute / pattern common feature amount which is a feature amount independent of the attribute / pattern indicated by the attribute / pattern data from the biological signal feature amount (step S24). The stress estimation means 106X estimates the stress of the subject based on the attribute / pattern common feature amount (step S25).
　第３実施形態に係るストレス推定装置１００Ｘは、被測定者の属性・パターンによらずに被測定者のストレスを好適に推定することができる。 The stress estimation device 100X according to the third embodiment can suitably estimate the stress of the person to be measured regardless of the attributes and patterns of the person to be measured.
　なお、上述した各実施形態において、プログラムは、様々なタイプの非一時的なコンピュータ可読媒体（non-transitory computer readable medium）を用いて格納され、コンピュータであるプロセッサ等に供給することができる。非一時的なコンピュータ可読媒体は、様々なタイプの実体のある記憶媒体（tangible storage medium）を含む。非一時的なコンピュータ可読媒体の例は、磁気記憶媒体（例えばフレキシブルディスク、磁気テープ、ハードディスクドライブ）、光磁気記憶媒体（例えば光磁気ディスク）、ＣＤ－ＲＯＭ（Read Only Memory）、ＣＤ－Ｒ、ＣＤ－Ｒ／Ｗ、半導体メモリ（例えば、マスクＲＯＭ、ＰＲＯＭ（Programmable ROM）、ＥＰＲＯＭ（Erasable PROM）、フラッシュＲＯＭ、ＲＡＭ（Random Access Memory））を含む。また、プログラムは、様々なタイプの一時的なコンピュータ可読媒体（transitory computer readable medium）によってコンピュータに供給されてもよい。一時的なコンピュータ可読媒体の例は、電気信号、光信号、及び電磁波を含む。一時的なコンピュータ可読媒体は、電線及び光ファイバ等の有線通信路、又は無線通信路を介して、プログラムをコンピュータに供給できる。 In each of the above-described embodiments, the program is stored using various types of non-transitory computer readable medium and can be supplied to a processor or the like which is a computer. Non-temporary computer-readable media include various types of tangible storage medium. Examples of non-temporary computer-readable media include magnetic storage media (eg flexible disks, magnetic tapes, hard disk drives), optomagnetic storage media (eg optomagnetic disks), CD-ROMs (ReadOnlyMemory), CD-Rs, Includes CD-R / W, semiconductor memory (eg, mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (RandomAccessMemory)). The program may also be supplied to the computer by various types of transient computer readable medium. Examples of temporary computer readable media include electrical, optical, and electromagnetic waves. The temporary 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.
　その他、上記の各実施形態の一部又は全部は、以下の付記のようにも記載され得るが以下には限られない。 Other than that, a part or all of each of the above embodiments may be described as in the following appendix, but is not limited to the following.
［付記１］
　被測定者の生体信号を取得する生体信号取得手段と、
　ストレスに関する前記生体信号の特徴量である生体信号特徴量を算出する生体信号特徴量計算手段と、
　前記被測定者の属性又はパターンの少なくとも一方を示す属性・パターンデータを取得する属性・パターンデータ取得手段と、
　前記生体信号特徴量から前記属性・パターンデータが示す属性・パターンに依存しない特徴量である属性・パターン共通特徴量を選択する属性・パターン共通特徴量選択手段と、
　前記属性・パターン共通特徴量に基づき、前記被測定者のストレスを推定するストレス推定手段と、
を備えるストレス推定装置。 [Appendix 1]
A biological signal acquisition means for acquiring the biological signal of the person to be measured, and
A biological signal feature amount calculating means for calculating a biological signal feature amount which is a feature amount of the biological signal related to stress, and a biological signal feature amount calculating means.
An attribute / pattern data acquisition means for acquiring attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured, and
An attribute / pattern common feature amount selection means for selecting an attribute / pattern common feature amount which is a feature amount independent of the attribute / pattern indicated by the attribute / pattern data from the biological signal feature amount.
A stress estimation means for estimating the stress of the person to be measured based on the attribute / pattern common feature amount, and
A stress estimator equipped with.
［付記２］
　前記属性・パターン共通特徴量選択手段は、前記属性・パターンを区別する敵対的特徴量を前記生体信号特徴量から排除することで、前記属性・パターン共通特徴量を選択する、付記１に記載のストレス推定装置。 [Appendix 2]
The attribute / pattern common feature amount selection means selects the attribute / pattern common feature amount by excluding the hostile feature amount that distinguishes the attribute / pattern from the biometric signal feature amount, according to Appendix 1. Stress estimation device.
［付記３］
　前記属性・パターン共通特徴量選択手段は、前記属性・パターンの分類上での前記生体信号特徴量に対する重要度を算出し、前記重要度に基づき、前記敵対的特徴量を選択する、付記２に記載のストレス推定装置。 [Appendix 3]
The attribute / pattern common feature amount selection means calculates the importance of the biological signal feature amount in the classification of the attribute / pattern, and selects the hostile feature amount based on the importance. The stress estimation device described.
［付記４］
　前記属性・パターン共通特徴量選択手段は、ランダムフォレストに基づき、前記敵対的特徴量を選択する、付記２または３に記載のストレス推定装置。 [Appendix 4]
The stress estimation device according to Appendix 2 or 3, wherein the attribute / pattern common feature amount selection means selects the hostile feature amount based on a random forest.
［付記５］
　前記ストレス推定手段は、回帰分析に基づくモデルを用いて、前記属性・パターン共通特徴量から前記ストレスの推定結果を示すスコアを算出する、付記１～４のいずれか一項に記載のストレス推定装置。 [Appendix 5]
The stress estimation device according to any one of Supplementary note 1 to 4, wherein the stress estimation means calculates a score indicating the stress estimation result from the attribute / pattern common feature amount using a model based on regression analysis. ..
［付記６］
　前記生体信号取得手段が取得した前記生体信号を記憶する生体信号記憶手段をさらに有し、
　前記特徴量計算手段は、前記生体信号記憶手段に記憶された前記生体信号から前記生体信号特徴量を算出する、付記１～５のいずれか一項に記載のストレス推定装置。 [Appendix 6]
Further having a biological signal storage means for storing the biological signal acquired by the biological signal acquisition means,
The stress estimation device according to any one of Supplementary note 1 to 5, wherein the feature amount calculating means calculates the biological signal feature amount from the biological signal stored in the biological signal storage means.
［付記７］
　前記生体信号取得手段は、前記被測定者により装着されたウェアラブル端末から前記生体信号を受信する、付記１～６のいずれか一項に記載のストレス推定装置。 [Appendix 7]
The stress estimation device according to any one of Supplementary note 1 to 6, wherein the biological signal acquisition means receives the biological signal from a wearable terminal worn by the person to be measured.
［付記８］
　コンピュータにより、
　被測定者の生体信号を取得し、
　ストレスに関する前記生体信号の特徴量である生体信号特徴量を算出し、
　前記被測定者の属性又はパターンの少なくとも一方を示す属性・パターンデータを取得し、
　前記生体信号特徴量から前記属性・パターンデータが示す属性・パターンに依存しない特徴量である属性・パターン共通特徴量を選択し、
　前記属性・パターン共通特徴量に基づき、前記被測定者のストレスを推定する、
推定方法。 [Appendix 8]
By computer
Acquires the biological signal of the person to be measured and
The biological signal feature amount, which is the feature amount of the biological signal related to stress, is calculated.
Acquire attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured.
From the biological signal feature amount, select an attribute / pattern common feature amount that is a feature amount that does not depend on the attribute / pattern indicated by the attribute / pattern data.
The stress of the person to be measured is estimated based on the common feature amount of the attribute / pattern.
Estimating method.
［付記９］
　被測定者の生体信号を取得し、
　ストレスに関する前記生体信号の特徴量である生体信号特徴量を算出し、
　前記被測定者の属性又はパターンの少なくとも一方を示す属性・パターンデータを取得し、
　前記生体信号特徴量から前記属性・パターンデータが示す属性・パターンに依存しない特徴量である属性・パターン共通特徴量を選択し、
　前記属性・パターン共通特徴量に基づき、前記被測定者のストレスを推定する処理をコンピュータに実行させるプログラム。 [Appendix 9]
Acquires the biological signal of the person to be measured and
The biological signal feature amount, which is the feature amount of the biological signal related to stress, is calculated.
Acquire attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured.
From the biological signal feature amount, select an attribute / pattern common feature amount that is a feature amount that does not depend on the attribute / pattern indicated by the attribute / pattern data.
A program that causes a computer to execute a process of estimating the stress of the person to be measured based on the common feature amount of the attribute / pattern.
［付記１０］
　付記９に記載のプログラムが格納された記憶媒体。 [Appendix 10]
A storage medium in which the program described in Appendix 9 is stored.
　以上、実施形態を参照して本願発明を説明したが、本願発明は上記実施形態に限定されるものではない。本願発明の構成や詳細には、本願発明のスコープ内で当業者が理解し得る様々な変更をすることができる。すなわち、本願発明は、請求の範囲を含む全開示、技術的思想にしたがって当業者であればなし得るであろう各種変形、修正を含むことは勿論である。また、引用した上記の特許文献等の各開示は、本書に引用をもって繰り込むものとする。 Although the invention of the present application has been described above with reference to the embodiment, the invention of the present application is not limited to the above embodiment. Various changes that can be understood by those skilled in the art can be made within the scope of the present invention in terms of the configuration and details of the present invention. That is, it goes without saying that the invention of the present application includes all disclosure including claims, various modifications and modifications that can be made by those skilled in the art in accordance with the technical idea. In addition, each disclosure of the above-mentioned patent documents cited shall be incorporated into this document by citation.
１００　ストレス推定装置
１０１　生体信号取得部
１０２　生体信号記憶部
１０３　生体信号特徴量計算部
１０４　属性・パターンデータ取得部
１０５　属性・パターン共通特徴量選択部
１０６　ストレス推定部
１５０　ストレス推定システム
２００　ウェアラブル端末
３００　被測定者
４００　ウェアラブル端末
５００　通信手段
５０１　Bluetooth通信
５０２　スマートフォン
５０３　モバイルデータ通信またはWifi通信
５０４　インターネット
６００　コンピュータ
６０１　通信インタフェース
６０２　生体信号取得部
６０３　生体信号記憶部
６０４　属性・パターンデータ取得部
６０５　属性・パターンデータ記憶部
６０６　生体信号特徴量計算部
６０７　生体信号特徴量記憶部
６０８　属性・パターン共通特徴量選択部
６０９　属性・パターン共通特徴量記憶部
６１０　ストレス推定部
６１１　ストレス推定結果記憶部
６１２　ストレス推定結果出力部 100 Stress estimation device 101 Biological signal acquisition unit 102 Biological signal storage unit 103 Biological signal feature amount calculation unit 104 Attribute / pattern data acquisition unit 105 Attribute / pattern common feature amount selection unit 106 Stress estimation unit 150 Stress estimation system 200 Wearable terminal 300 Measurer 400 Wearable terminal 500 Communication means 501 Bluetooth communication 502 Smartphone 503 Mobile data communication or Wifi communication 504 Internet 600 Computer 601 Communication interface 602 Biological signal acquisition unit 603 Biological signal storage unit 604 Attribute / pattern data acquisition unit 605 Attribute / pattern data storage Unit 606 Biological signal feature amount calculation unit 607 Biological signal feature amount storage unit 608 Attribute / pattern common feature amount selection unit 609 Attribute / pattern common feature amount storage unit 610 Stress estimation unit 611 Stress estimation result storage unit 612 Stress estimation result output unit

Claims (10)

  1. 　被測定者の生体信号を取得する生体信号取得手段と、
    　ストレスに関する前記生体信号の特徴量である生体信号特徴量を算出する生体信号特徴量計算手段と、
    　前記被測定者の属性又はパターンの少なくとも一方を示す属性・パターンデータを取得する属性・パターンデータ取得手段と、
    　前記生体信号特徴量から前記属性・パターンデータが示す属性・パターンに依存しない特徴量である属性・パターン共通特徴量を選択する属性・パターン共通特徴量選択手段と、
    　前記属性・パターン共通特徴量に基づき、前記被測定者のストレスを推定するストレス推定手段と、
    を備えるストレス推定装置。     A biological signal acquisition means for acquiring the biological signal of the person to be measured, and
    A biological signal feature amount calculating means for calculating a biological signal feature amount which is a feature amount of the biological signal related to stress, and a biological signal feature amount calculating means.
    An attribute / pattern data acquisition means for acquiring attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured, and
    An attribute / pattern common feature amount selection means for selecting an attribute / pattern common feature amount which is a feature amount independent of the attribute / pattern indicated by the attribute / pattern data from the biological signal feature amount.
    A stress estimation means for estimating the stress of the person to be measured based on the attribute / pattern common feature amount, and
    A stress estimator equipped with.
  2. 　前記属性・パターン共通特徴量選択手段は、前記属性・パターンを区別する敵対的特徴量を前記生体信号特徴量から排除することで、前記属性・パターン共通特徴量を選択する、請求項１に記載のストレス推定装置。 The attribute / pattern common feature amount selection means according to claim 1, wherein the attribute / pattern common feature amount is selected by excluding the hostile feature amount that distinguishes the attribute / pattern from the biometric signal feature amount. Stress estimation device.
  3. 　前記属性・パターン共通特徴量選択手段は、前記属性・パターンの分類上での前記生体信号特徴量に対する重要度を算出し、前記重要度に基づき、前記敵対的特徴量を選択する、請求項２に記載のストレス推定装置。 2. The attribute / pattern common feature amount selection means calculates the importance of the biological signal feature amount in the classification of the attribute / pattern, and selects the hostile feature amount based on the importance. The stress estimation device described in.
  4. 　前記属性・パターン共通特徴量選択手段は、ランダムフォレストに基づき、前記敵対的特徴量を選択する、請求項２または３に記載のストレス推定装置。 The stress estimation device according to claim 2 or 3, wherein the attribute / pattern common feature amount selection means selects the hostile feature amount based on a random forest.
  5. 　前記ストレス推定手段は、回帰分析に基づくモデルを用いて、前記属性・パターン共通特徴量から前記ストレスの推定結果を示すスコアを算出する、請求項１～４のいずれか一項に記載のストレス推定装置。 The stress estimation means according to any one of claims 1 to 4, wherein the stress estimation means calculates a score indicating the stress estimation result from the attribute / pattern common feature amount using a model based on regression analysis. Device.
  6. 　前記生体信号取得手段が取得した前記生体信号を記憶する生体信号記憶手段をさらに有し、
    　前記特徴量計算手段は、前記生体信号記憶手段に記憶された前記生体信号から前記生体信号特徴量を算出する、請求項１～５のいずれか一項に記載のストレス推定装置。     Further having a biological signal storage means for storing the biological signal acquired by the biological signal acquisition means,
    The stress estimation device according to any one of claims 1 to 5, wherein the feature amount calculating means calculates the biological signal feature amount from the biological signal stored in the biological signal storage means.
  7. 　前記生体信号取得手段は、前記被測定者により装着されたウェアラブル端末が生成した前記生体信号を受信する、請求項１～６のいずれか一項に記載のストレス推定装置。 The stress estimation device according to any one of claims 1 to 6, wherein the biological signal acquisition means receives the biological signal generated by the wearable terminal worn by the person to be measured.
  8. 　コンピュータにより、
    　被測定者の生体信号を取得し、
    　ストレスに関する前記生体信号の特徴量である生体信号特徴量を算出し、
    　前記被測定者の属性又はパターンの少なくとも一方を示す属性・パターンデータを取得し、
    　前記生体信号特徴量から前記属性・パターンデータが示す属性・パターンに依存しない特徴量である属性・パターン共通特徴量を選択し、
    　前記属性・パターン共通特徴量に基づき、前記被測定者のストレスを推定する、
    推定方法。     By computer
    Acquires the biological signal of the person to be measured and
    The biological signal feature amount, which is the feature amount of the biological signal related to stress, is calculated.
    Acquire attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured.
    From the biological signal feature amount, select an attribute / pattern common feature amount that is a feature amount that does not depend on the attribute / pattern indicated by the attribute / pattern data.
    The stress of the person to be measured is estimated based on the common feature amount of the attribute / pattern.
    Estimating method.
  9. 　被測定者の生体信号を取得し、
    　ストレスに関する前記生体信号の特徴量である生体信号特徴量を算出し、
    　前記被測定者の属性又はパターンの少なくとも一方を示す属性・パターンデータを取得し、
    　前記生体信号特徴量から前記属性・パターンデータが示す属性・パターンに依存しない特徴量である属性・パターン共通特徴量を選択し、
    　前記属性・パターン共通特徴量に基づき、前記被測定者のストレスを推定する処理をコンピュータに実行させるプログラム。     Acquires the biological signal of the person to be measured and
    The biological signal feature amount, which is the feature amount of the biological signal related to stress, is calculated.
    Acquire attribute / pattern data indicating at least one of the attributes or patterns of the person to be measured.
    From the biological signal feature amount, select an attribute / pattern common feature amount that is a feature amount that does not depend on the attribute / pattern indicated by the attribute / pattern data.
    A program that causes a computer to execute a process of estimating the stress of the person to be measured based on the common feature amount of the attribute / pattern.
  10. 　請求項９に記載のプログラムが格納された記憶媒体。 A storage medium in which the program according to claim 9 is stored.
PCT/JP2020/031662 2020-08-21 2020-08-21 Stress inference device, inference method, program, and storage medium WO2022038776A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
JP2022543248A JPWO2022038776A1 (en) 2020-08-21 2020-08-21
PCT/JP2020/031662 WO2022038776A1 (en) 2020-08-21 2020-08-21 Stress inference device, inference method, program, and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2020/031662 WO2022038776A1 (en) 2020-08-21 2020-08-21 Stress inference device, inference method, program, and storage medium

Publications (1)

Publication Number Publication Date
WO2022038776A1 true WO2022038776A1 (en) 2022-02-24

Family

ID=80322637

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2020/031662 WO2022038776A1 (en) 2020-08-21 2020-08-21 Stress inference device, inference method, program, and storage medium

Country Status (2)

Country Link
JP (1) JPWO2022038776A1 (en)
WO (1) WO2022038776A1 (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH04341243A (en) * 1991-05-17 1992-11-27 Mitsubishi Electric Corp Amenity evaluation system and amenity evaluation/ control system
JP2012061222A (en) * 2010-09-17 2012-03-29 Tokai Rika Co Ltd Driver condition estimating device
US20150265211A1 (en) * 2012-08-01 2015-09-24 Soma Analytics Ug (Haftungsbeschränkt) Device, method and application for establishing a current load level
JP2018149262A (en) * 2017-03-13 2018-09-27 株式会社疲労科学研究所 Autonomic nerve evaluation apparatus, autonomic nerve evaluation method, program, and recording medium
WO2018221750A1 (en) * 2017-06-02 2018-12-06 学校法人慶應義塾 Sleep determining device, sleep determining method, and sleep determining program
WO2019159252A1 (en) * 2018-02-14 2019-08-22 日本電気株式会社 Stress estimation device and stress estimation method using biosignal

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH04341243A (en) * 1991-05-17 1992-11-27 Mitsubishi Electric Corp Amenity evaluation system and amenity evaluation/ control system
JP2012061222A (en) * 2010-09-17 2012-03-29 Tokai Rika Co Ltd Driver condition estimating device
US20150265211A1 (en) * 2012-08-01 2015-09-24 Soma Analytics Ug (Haftungsbeschränkt) Device, method and application for establishing a current load level
JP2018149262A (en) * 2017-03-13 2018-09-27 株式会社疲労科学研究所 Autonomic nerve evaluation apparatus, autonomic nerve evaluation method, program, and recording medium
WO2018221750A1 (en) * 2017-06-02 2018-12-06 学校法人慶應義塾 Sleep determining device, sleep determining method, and sleep determining program
WO2019159252A1 (en) * 2018-02-14 2019-08-22 日本電気株式会社 Stress estimation device and stress estimation method using biosignal

Also Published As

Publication number Publication date
JPWO2022038776A1 (en) 2022-02-24

Similar Documents

Publication Publication Date Title
US20240079117A1 (en) Biometric characteristic application using audio/video analysis
US20230060732A1 (en) Method and system for identifying biometric characteristics using machine learning techniques
Bogomolov et al. Pervasive stress recognition for sustainable living
CN107714024B (en) Method, system and apparatus for monitoring cardiac activity
JP6435128B2 (en) Physiological parameter monitoring
US20200074380A1 (en) Work support device, work support method, and work support program
US20190117143A1 (en) Methods and Apparatus for Assessing Depression
US20200294670A1 (en) System and method for real-time estimation of emotional state of user
US10629225B2 (en) Information processing method, information processing device, and recording medium recording information processing program
Lawanot et al. Daily stress and mood recognition system using deep learning and fuzzy clustering for promoting better well-being
JP6975265B2 (en) Computing devices, non-transient computer-readable storage media, methods for removing artifacts in electroencephalogram (EEG) signals, and computer programs
JP7136341B2 (en) Stress estimation device, stress estimation method and program
US20190042978A1 (en) Computer system
WO2022038776A1 (en) Stress inference device, inference method, program, and storage medium
US20220206745A1 (en) Relationship analysis utilizing biofeedback information
Eggert et al. Recognizing mental stress in chess players using vital sign data
Djemai et al. A genetic algorithm-based support vector machine model for detection of hearing thresholds
JP7276586B2 (en) STRESS ESTIMATION DEVICE, STRESS ESTIMATION METHOD, AND PROGRAM
KR102290233B1 (en) System and Method for Analyzing Vocal Cord Condition
Ekiz et al. Long short-term memory network based unobtrusive workload monitoring with consumer grade smartwatches
James et al. An IoT-Based Health Monitoring System for Stress Detection in Human Beings
WO2022144978A1 (en) Information processing device, control method, and storage medium
US20220101072A1 (en) Detecting a User&#39;s Outlier Days Using Data Sensed by the User&#39;s Electronic Devices
WO2022113276A1 (en) Information processing device, control method, and storage medium
US20240057912A1 (en) Biological information processing device, biological information processing system, and biological information processing method

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20950341

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2022543248

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20950341

Country of ref document: EP

Kind code of ref document: A1