JP6712909B2 - Cognitive change prediction device and program - Google Patents

Description

The present invention, "Working Memory" Ya "Selective Attention", etc., related equipment Contact and program for predicting changes in cognitive information processing capability of the user (hereinafter referred to as cognitive performance).

In recent years, various techniques for estimating the cognitive ability of a user during work have been proposed, for example, for labor management of employees. For example, in Non-Patent Document 1, as in Advanced Trail Making Test (hereinafter referred to as ATMT), a work in progress is interrupted to execute a task for about several minutes, and the user's recognition based on the result of the task. Techniques for estimating capacity are described. Further, in Non-Patent Document 2, an electrocardiographic sensor is attached to a user, an R wave peak interval of the electrocardiogram (hereinafter referred to as RRI) is measured, and a plurality of feature amounts calculated from this RRI are used at the present time. A method for estimating the cognitive ability of humans is described.

K. Mizuno et al.: Utility of an Advanced Trail Making Test as a Neuropsychological Tool for an Objective Evaluation of Work Efficiency During Mental Fatigue, Fatigue Science for Human Health, pp.47-54, 2008. Tsunoda et al., “Incremental Cognitive Ability Estimation Method Using Heart Rate Variability”, Proc. of IPSJ Groupware and Network Service Workshop, pp. 1-10, 2015.

However, the technologies described in Non-Patent Documents 1 and 2 all estimate the user's cognitive ability at the present time. Therefore, when trying to use the estimation results for labor management, it is possible to take a rest when the cognitive ability declines, but it is impossible to take a rest before the cognitive ability declines. It was Therefore, reduction in productivity cannot be prevented, and improvement in labor management has been demanded.

The present invention has been made in view of the above circumstances, the place of interest, and our cognitive change projection equipment so as to be able to predict changes in cognitive ability allows the user management better It is intended to provide a program.

In order to solve the above problems, a first aspect of the present invention acquires data representing a heartbeat interval of a prediction target user, and based on the acquired data representing a heartbeat interval, a heartbeat fluctuation of a time section t. The feature amount and the feature amount of the heartbeat variability in the immediately preceding time interval t-1 are calculated, and the amount of change is calculated based on the calculated feature amount of each heartbeat variability. Then, based on the calculated feature amount of each heartbeat variability and its change amount and the learning data stored in advance , the change in the cognitive ability of the prediction target user in the subsequent time interval t+1, The prediction is made as an increase or decrease from the cognitive ability in the time section t, and information indicating the prediction result is output.

A second aspect of the present invention is, as means for calculating the feature amount of the heartbeat variability, means for calculating a feature amount in a time domain based on the acquired data indicating the heartbeat interval, and the obtained heartbeat. And means for calculating a feature amount in the frequency domain based on the data representing the interval. Then, by the means for calculating the feature amount in the time domain, the average value of the heartbeat intervals, the standard deviation of the heartbeat intervals (SDNN), and the ratio (pNN50) in which the difference between adjacent heartbeat intervals exceeds 50 msec, the The low-frequency component (0.04 to 0.15) when the heartbeat interval is spectrally analyzed by means for calculating at least one of the root mean square difference of the heartbeat intervals (RMSSD) and the feature amount in the frequency domain is calculated. Hz) total power (LF), total power (HF) of high frequency components (0.15 to 0.40 Hz) when spectrum analysis of heartbeat interval, total power LF of the above low frequency components and total power HF of the above high frequency components And at least one of the ratio (LF/HF) and is calculated.

A third aspect of the invention, the means for calculating a feature amount of variation of the above SL HRV, the feature amount of heart rate variability in the time interval t, the heart rate variability in the previous time interval t-1 a change amount or rate of change between the feature quantity, the amount of change in the correlation coefficient between the feature quantity of heart rate variability in the time interval t, the feature amount of heart rate variability in the previous time interval t-1 Alternatively, one of the change rate and the change rate is calculated.

According to a fourth aspect of the present invention, when calculating the change amount of the feature amount of the heartbeat variability, the feature amount of the heartbeat variability in the time period t and the feature of the heartbeat variability in the immediately preceding time period t-1. The principal component is analyzed for each quantity, and the T ² statistic indicating the change of the principal component and the Q statistic indicating the residual of dimension compression are calculated based on the analysis result of the principal component. It was done like this.

According to a sixth aspect of the present invention, as a means for predicting a change in the cognitive ability, a classifier representing a change in the cognitive ability is created based on the learning data, and the calculated feature amount of the heartbeat variability and its By inputting the change amount into the classifier, a regression model is created based on the means for predicting the change in the cognitive ability and the learning data, and the calculated feature amount of the heartbeat variability and the change amount thereof. One of the means for predicting the change amount of the cognitive ability by inputting into the regression model is provided.

According to the first aspect of the present invention, a change in the cognitive ability of the prediction target user, for example, in the time section t+1, which is one time after the current time section t, is predicted as an increase or decrease from the cognitive ability in the time section t. The result is displayed. Therefore, for example, it is possible for the prediction target user to take a rest before the cognitive ability deteriorates, or the administrator can take measures such as giving a rest to the prediction target user at an appropriate timing. As a result, it is possible to maintain high work efficiency and quality of the user who is executing the task.

According to the second aspect of the present invention, the feature amount in the time domain and the feature amount in the frequency domain are each calculated as the feature amount of the heartbeat fluctuation based on the heartbeat interval data. Therefore, it is possible to reliably obtain the feature amount that is related to the cognitive ability of the user.

According to a third aspect of the present invention, as the feature quantity variation of heart rate variability, and characteristics of heart rate variability in the time interval t, between the HRV feature value of the previous in time interval t-1 a change amount or rate of change, one of a change amount or rate of change of the correlation coefficient between each HRV feature amount is calculated. For this reason, it is possible to relatively easily calculate the amount of change in the characteristic amount of the heartbeat variability on a time interval basis.

According to the fourth aspect of the present invention, as the amount of change in the feature amount of heartbeat variability, there are T ² statistic indicating the change of the main component of the feature amount of heartbeat variability and Q statistic indicating the residual of dimensional compression. It is calculated. Therefore, the amount of change in the feature amount of heartbeat variability can be calculated more accurately.

In the sixth aspect of the present invention, the change in the cognitive ability of the user can be predicted as the presence or absence of the change in the cognitive ability or the change amount of the cognitive ability.

That is, according to each aspect of the invention, it is possible to provide a cognitive change projection equipment Contact and programs so as to allow predict changes in cognitive ability allows the user of the management more properly.

FIG. 3 is a block diagram showing a functional configuration of a cognitive ability change prediction device used for explaining the principle of the present invention. 3 is a flowchart showing a processing procedure and processing contents of the cognitive ability change prediction device shown in FIG. 1. The block diagram which shows the structure of the cognitive ability change prediction apparatus which concerns on the 1st Embodiment of this invention. 4 is a flowchart showing a processing procedure and processing contents of learning data creation processing by the cognitive ability change prediction device shown in FIG. 3. 4 is a flowchart showing a processing procedure and processing contents of cognitive ability change prediction processing by the cognitive ability change prediction device shown in FIG. 3. The flowchart which shows the process procedure and process content of the learning data creation process by the cognitive ability change prediction apparatus which concerns on the 2nd Embodiment of this invention. The flowchart which shows the process sequence and process content of the cognitive ability change prediction process by the cognitive ability change prediction apparatus which concerns on the 2nd Embodiment of this invention. The block diagram which shows the structure of the cognitive ability change prediction apparatus which concerns on the 3rd Embodiment of this invention.

[principle]
FIG. 1 is a diagram for use in explaining the principle of the present invention, in which 1 is a prediction processing device, 2 is data input means, 3 is result output means, and 4 is learning data storage means.
The data input unit 2 inputs the heartbeat interval data of the target user to the prediction processing device 1, and is composed of, for example, a wearable heartbeat sensor, a pulse wave sensor, or a database in which heartbeat interval data is accumulated. The result output means 3 outputs the prediction result of the cognitive ability obtained by the prediction processing device 1, and is composed of, for example, an application program used in a smartphone.

In the learning data accumulating means 4, the pre-measured characteristic amounts of the plurality of prediction target users, which indicate heart rate variability during work, and the information regarding the information processing ability of the users during work are stored at the same time. Learning data associated with each other is stored. More specifically, a value indicating a change in cognitive ability of the user (hereinafter referred to as a cognitive ability change value), an HRV feature amount, and information indicating an amount of change in the HRV feature amount are stored as learning data.

Examples of the cognitive ability change value include, for example, an ATMT (Advanced Trail Making Test) performance and its change amount. As the HRV feature amount, for example, the average value of RRI, the standard deviation of RRI, or the power spectrum of each high frequency band (0.15 Hz to 0.40 Hz) and low frequency band (0.04 Hz to 0.15 Hz) when spectrum analysis of RRI is performed. The ratio is LF/HF. As the variation amount of the HRV feature amount, T ² statistic indicating the variation of the principal component score calculated after applying the principal component analysis to a plurality of HRV feature amounts and Q indicating the variation of the correlation between the feature amounts. There are statistics.

The prediction processing device 1 has a feature amount calculation unit 11, a feature amount change calculation unit 12, and a cognitive ability change prediction unit 13 as processing functions for implementing the present invention.
The feature amount calculation means 11 calculates the HRV feature amount for each predetermined time interval from the heartbeat interval data input from the data input means 2. The feature amount change calculation means 12 calculates the change amount of the HRV feature amount (hereinafter referred to as HRV change amount) based on the HRV feature amount calculated for each time section by the feature amount calculation means 11. The cognitive ability change prediction unit 13 is stored in the learning data storage unit 4 based on the HRV feature amount calculated by the feature amount calculation unit 11 and the HRV change amount calculated by the feature amount change calculation unit 12. The change in cognitive ability is predicted by referring to the learned data.

With the above configuration, the prediction processing device 1 operates as follows. FIG. 2 is a flowchart showing the processing procedure and processing contents.
In step S11, the prediction processing device 1 first receives the heartbeat interval data of the target user from the data input unit 2 in time series, and calculates the HRV feature amount for each predetermined time interval from the heartbeat interval data. Then, in step S12, the HRV change amount is calculated based on the HRV feature amount calculated for each time section in step S11. Next, in step S13, based on the HRV feature amount calculated in step S11 and the HRV change amount calculated in step S12, referring to the learning data stored in the learning data storage means 4, the target user's recognition Predict the amount of change in ability. Then, the information indicating the predicted amount of change in cognitive ability is output to the result output means 3 and presented to the target user or the manager thereof.

Therefore, according to the present invention, since the prediction result of the change amount of the cognitive ability of the target user is presented, for example, more appropriate labor management such as giving a rest to the user before the cognitive ability of the user is deteriorated is performed. It becomes possible.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[First Embodiment]
(Constitution)
FIG. 3 is a block diagram showing the configuration of a system including the cognitive ability change prediction device according to the first embodiment of the present invention.
This system has the cognitive ability change prediction device 10A as a core, and additionally includes pulse wave sensors 21 and 22, an ATMT input unit 23, a display unit 30, and a learning database (learning DB) 40.

The pulse wave sensor 21 is attached to the user who measures the learning data. Then, the pulse wave of the learning data measurement target user is measured, data representing a heartbeat interval is calculated based on the measured data of the pulse wave, and the calculated data is used, for example, via a wireless interface to predict the cognitive ability change device 10A. It has a function to send to.

The pulse wave sensor 22 is attached to a user who is a target of prediction of a change in cognitive ability. Then, the pulse wave of the user who is the target of prediction of the change in cognitive ability is measured, data representing the heartbeat interval is calculated based on the measured data of the pulse wave, and the calculated data is predicted through, for example, a wireless interface. It has a function of transmitting to the device 10A.

As the wireless interface used by each of the pulse wave sensors 21 and 22, for example, a wireless interface adopting a short-range wireless data communication standard such as Bluetooth (registered trademark) is used. Note that the pulse wave sensors 21 and 22 may not be individually prepared, and one pulse wave sensor may be shared for learning data creation and cognitive ability change prediction.

The ATMT input unit 23 includes a mobile terminal (for example, a notebook personal computer, a tablet terminal, or a smart phone) having an interface for inputting data representing the execution result of the ATMT task. Then, the data representing the execution result of the input ATMT task is transmitted to the cognitive performance change prediction device 10A via, for example, a wireless interface.

The display unit 30 includes a mobile terminal (for example, a wearable terminal, a smartphone, or a tablet terminal) that includes an application program for displaying the cognitive ability change prediction result. Then, the information representing the prediction result of the cognitive ability change sent from the cognitive ability change prediction device 10A via the wireless interface is received, and the information is displayed on the display.

The learning DB 40 is provided in, for example, a server computer or a cloud computer on the Web, and receives the learning data created by the cognitive ability change prediction device 10A via the network and stores it. In addition, in response to the reference request sent from the cognitive ability change prediction device 10A, the corresponding learning data is transmitted to the request source cognitive performance change prediction device 10A. The learning DB 40 may be provided in the cognitive ability change prediction device 10A.

The cognitive ability change prediction device 10A includes a personal computer or a server computer including a processor such as a CPU (Central Processing Unit), a storage unit, and various external interfaces such as a wireless interface. Then, as the processing functions necessary for implementing the present embodiment, the feature amount calculation units 101 and 104, the feature amount change calculation units 102 and 105, the learning data creation unit 103, and the cognitive ability change prediction unit 106 are provided. Have Each of these processing functions is realized by causing the processor to execute the program stored in the storage unit.

Among the above processing functions, the feature amount calculation unit 101, the feature amount change calculation unit 102, and the learning data creation unit 103 are used to create learning data.

The feature amount calculation unit 101 receives the heartbeat interval data of the learning data measurement target user, which is transmitted in time series from the pulse wave sensor 21, and based on the received heartbeat interval data, for each predetermined time interval. HRV feature quantity is calculated.

Based on the calculated HRV feature amount, the feature amount change calculation unit 102 calculates the T ² statistic and the Q statistic of the temporally adjacent sections as data representing the change amount of the HRV feature amount. Then, the calculated T ² statistic and Q statistic are passed to the learning data creation unit 103 together with the HRV feature amount calculated by the feature amount calculation unit 101. In addition, the feature amount change calculation unit 102 calculates the amount of change or rate of change between the feature amounts of each heartbeat variability calculated in the plurality of different time intervals, and also calculates the phase between the feature amounts of each heartbeat variability. You may make it calculate the variation|change_quantity or rate of change of the number of relations.

The learning data creation unit 103 uses the HRV feature amount, the T ² statistic amount, and the Q statistic amount, which are passed from the feature amount change calculation unit 102, as explanatory variables, and the ATMT error input by the learning data measurement target user in the ATMT input unit 23. Learning data in which the data representing the increase/decrease in the rate is the explained variable is created, and the created learning data is transmitted to and stored in the learning DB 40.

On the other hand, among the above processing functions, the feature amount calculation unit 104, the feature amount change calculation unit 105, and the cognitive ability change prediction unit 106 are used to predict the change in the cognitive ability of the prediction target user.

The feature amount calculation unit 104 receives the heartbeat interval data of the prediction target user, which is transmitted in time series from the pulse wave sensor 22, and based on the received heartbeat interval data, the HRV feature amount for each predetermined time section. To calculate.

The feature amount change calculation unit 105 calculates the T ² statistic and the Q statistic of the temporally adjacent sections as the change amount of the HRV feature amount based on the calculated HRV feature amount. Then, the calculated T ² statistic and Q statistic are output to the cognitive ability change prediction unit 106 together with the HRV feature amount calculated by the feature amount calculation unit 104. In addition, the feature amount change calculation unit 105 calculates the amount of change or the rate of change between the feature amounts of each heartbeat variability calculated in the plurality of different time intervals, and also calculates the phase between the feature amounts of each heartbeat variability. You may make it calculate the variation|change_quantity or rate of change of the number of relations.

The cognitive ability change prediction unit 106 uses the T ² statistic and the Q statistic calculated as the HRV change amount by the feature amount change calculation unit 105, and the learning data stored in the learning DB 40 to target the target user. Predict changes in cognitive ability. Then, the display data of the prediction result is generated and transmitted to the display unit 30.

(motion)
Next, the operation of the cognitive ability change prediction device 10A configured as described above will be described. Here, a case will be described in which the predicted increase/decrease in the change in cognitive ability is predicted using only two values, "whether it has increased" and "has decreased".

(1) Creation of Learning Data Prior to the process of predicting the change in cognitive ability of the prediction target user, the cognitive performance change prediction device 10A executes the process of creating learning data as follows. FIG. 4 is a flowchart showing the processing procedure and processing contents.

The pulse wave sensor 21 is attached to the user whose learning data is to be measured. The pulse wave sensor 21 measures the pulse wave of the measurement target user, and calculates the heartbeat interval data based on the waveform data of the measured pulse wave. Then, the calculated heartbeat interval data is transmitted to the cognitive performance change prediction device 10A via the wireless interface.

Further, in parallel with the operation of measuring the heartbeat interval data, the user to be measured with the learning data performs the task of ATMT at intervals of 10 minutes, and inputs the error rate and execution time of ATMT as data representing his/her own cognitive ability. Input in the section 23. The ATMT input unit 23 transmits each time the ATMT error rate and the implementation time are input to the cognitive performance change prediction device 10A via the wireless interface.

Under the control of the feature amount calculation unit 101, the cognitive performance change prediction apparatus 10A first receives and takes in the heartbeat interval data transmitted in time series from the pulse wave sensor 21 in step S21, and sequentially stores the data in a storage unit (not shown). save. In step S22, the ATMT error rate and the execution time transmitted from the ATMT input unit 23 are received, the received ATMT error rate and the execution time are received by the learning data creation unit 103, and the execution time is received by the feature amount calculation unit 101. To enter.

Then, in step S23, the feature amount calculation unit 101 synchronizes the heartbeat interval data stored in the storage unit with a time interval of 10 minutes (interval t (t=1, 2,... , N), and the HRV feature amount is calculated while sliding the time window of 5 minutes with the slide width of 6 seconds for each section t, and the calculated HRV feature amount is calculated by the feature amount change calculation unit 102. Pass to.

The calculated HRV feature amount includes a time domain feature amount and a frequency domain feature amount.
The feature quantity in the time domain is, for example,
(1) Average heartbeat interval
(2) Standard deviation of heartbeat interval (SDNN)
(3) Ratio of adjacent heartbeat intervals exceeding 50 msec (pNN50)
(4) Sum of root mean squares of differences between adjacent heartbeat intervals (RMSSD)
There is.

On the other hand, for the feature quantity in the frequency domain, for example,
(5) Total power (LF) of low frequency components (0.04 to 0.15Hz) when spectrum analysis of heartbeat interval
(6) Total power (HF) of high frequency components (0.15 to 0.40Hz) when spectrum analysis of heartbeat interval
(7) Ratio of total power LF of low frequency components to total power HF of high frequency components (LF/HF)
There is.

The feature amount calculation unit 101 calculates, as the HRV feature amount, for example, SDNN of the time domain feature amounts and LF/HF of the frequency domain feature amounts. Other characteristic amounts may be calculated for each of the characteristic amounts, or a plurality of characteristic amounts may be calculated simultaneously.

Next, under the control of the feature amount change calculation unit 102, the cognitive ability change prediction device 10A, in step S24, based on the HRV feature amount passed from the feature amount calculation unit 101, a section t (t>1). HRV change amount is calculated for each. The HRV change amount consists of T ² statistic and Q statistic. Then, the calculated HRV change amount is passed to the learning data creation unit 103 together with the HRV feature amount calculated by the feature amount calculation unit 101.

The calculation method of the above T ² statistic and Q statistic is as follows.
Step1; Principal component analysis is applied to the vector of HRV feature quantity ^X ∈ ^RA×B (A=number of windows, B=number of feature quantity) as the number of principal component R in the section t−1, and the principal component score vector S∈R The load matrix V _R of ^A×B and the main component is calculated.

Step2; For the HRV feature vector X'ε ^{RA x B} in the section t, the principal component score vector S'ε ^{RA x B} is calculated using the load matrix V _R calculated in step 1.

Step3; From each principal component score S _r in the principal component score vector S′ and each principal component score S′ _r in the principal component score vector S′, the T ² statistic of the interval t is
Calculate by Note that σ _Sr is the standard deviation of the principal component score S _r .

Step4; 'from and the main component of the load matrix V _R, the vector X of the HRV feature quantity' principal component score vector S reconfiguration feature vector X ` 'reconstituted from main component,
X'=S'V ^T _R
Calculate by

Step5; The Q statistic is a value indicating the residual of dimensional compression, that is, the sum of squared errors of the reconstructed feature vector X′′ and the feature vector X′,
Q=Σ(X'-X') ²
Calculate by

Finally, under the control of the learning data creation unit 103, the cognitive performance change prediction apparatus 10A, in step S25, for each section t (1<t<n), the HRV feature amount in sections t 1 and t−1 and the section t 1. The HRV variation (T ² statistic and Q statistic) in the interval t calculated based on -1 is used as the explanatory variable, and the increase or decrease of the ATMT error rate in the interval t+1 and the interval t is used as the explained variable. Learning data is created and the learning data is stored in the learning DB 40.

Note that the above-described learning data creation process is performed for each of these users when there are a plurality of users who are targets for prediction of changes in cognitive ability.

In the above description, the case where learning data is created for one specific target user has been described as an example. However, learning data is similarly created and stored in the learning DB 40 for a plurality of users who are assumed to be targets of prediction of changes in cognitive ability.

(2) User's Cognitive Ability Change Prediction Process When the creation of the learning data is completed, the cognitive ability change prediction device 10A executes a process for predicting a change in the cognitive ability of the prediction target user as follows. FIG. 5 is a flowchart showing the processing procedure and processing contents.

The pulse wave sensor 22 is attached to the target user. The pulse wave sensor 22 measures the pulse wave of the target user and calculates heartbeat interval data based on the waveform data of the measured pulse wave. Then, the calculated heartbeat interval data is transmitted to the cognitive performance change prediction device 10A via the wireless interface.

Under the control of the feature amount calculation unit 104, the cognitive ability change prediction device 10A first receives and takes in the heartbeat interval data transmitted in time series from the pulse wave sensor 22 in step S31, and temporarily stores the data in a storage unit (not shown). save. Then, first, for the heartbeat interval data received from the pulse wave sensor 22 in the latest 10 minutes from the present time point, the HRV characteristic amount is calculated while sliding the time window of 5 minutes with the slide width of 6 seconds. The HRV feature amount calculated at this time is the same as the values (1) to (7) calculated by the feature amount calculation unit 101 at the time of creating the learning data described above. Then, the calculated latest HRV feature amount for 10 minutes is passed to the feature amount change calculation unit 105. After that, the calculation process of the HRV feature amount is repeated every 10 minutes.

Next, the cognitive ability change prediction device 10A calculates the HRV change amount as follows in step S32 under the control of the feature amount change calculation unit 105. That is, the HRV feature amount passed from the feature amount calculation unit 104 is sequentially stored in the storage unit. Then, first, the HRV feature amount in the section from the present time to 10 minutes ago is read from the storage unit, and the HRV change amount is calculated based on the HRV feature amount measured and calculated in the section 10 minutes to 20 minutes ago. To do. Specifically, from the above-mentioned present time, from the load matrix V _R and the principal component score vector S′ when the principal component analysis is applied to the HRV feature amount measured and calculated in the section from 10 minutes before to 20 minutes before, The T ² statistic and the Q statistic are calculated using the HRV features up to 10 minutes ago.

Instead of calculating the change amount or change rate between the feature amounts of each heartbeat variability, the change amount or change rate of the correlation coefficient between the feature amounts of each heartbeat change may be calculated.

Under the control of the cognitive performance change prediction unit 106, the cognitive performance change prediction device 10A finally reads data representing increase or decrease in the ATMT error rate from the learning DB 40 in step S33, and uses the SVM (Support Vector Machine) for prediction. Create a regression model classifier. By inputting the HRV feature amount calculated by the feature amount change calculation unit 105 and the change amount (T ² statistic amount and Q statistic amount) to the classifier, the ATMT error rate increases or decreases, that is, “increases”. "Or" or "will decrease"? Then, display data indicating the prediction result is generated and transmitted to the display unit 30. As a result, the display unit 30 displays a message, for example, "it is predicted that the cognitive ability will decrease in the next 10 minutes compared to the immediately preceding 10 minutes".

(effect)
As described above in detail, in the first embodiment, in the learning data creation unit 103, the intervals t 1, t−1 calculated by the feature amount calculation unit 101 for each interval t (1<t<n) in advance. , And the HRV change amount (T ² statistic and Q statistic) in the section t calculated based on the section t−1 calculated by the feature change calculation unit 102 as an explanatory variable, and Learning data having binary data, which is input by the ATMT input unit 23 and represents the increase or decrease of the ATMT error rate between the sections t+1 and t, as an explained variable, is created and stored in the learning DB 40. Then, in this state, the feature amount calculation unit 104 calculates the HRV feature amount for each predetermined time based on the heartbeat interval data of the prediction target user, and the feature amount change calculation unit 105 calculates the HRV feature amount based on the HRV feature amount. A T ² statistic and a Q statistic showing the HRV change amount are calculated. Then, the cognitive ability change prediction unit 106 reads data representing the increase or decrease of the ATMT error rate from the learning DB 40 to create a classifier, and calculates the above-mentioned HRV feature amount and its change amount (T ² statistic and Q statistic). By inputting (amount) into the classifier, increase/decrease in the ATMT error rate is predicted, and the result is displayed on the display unit 30.

Therefore, by predicting the change in the cognitive ability of the prediction target user and displaying the result, for example, the prediction target user takes a voluntary rest before the cognitive ability declines, or the administrator sets the prediction target user as the prediction target user. It is possible to take measures such as giving rest at appropriate timing. As a result, it is possible to maintain high work efficiency and quality of a user who is performing a task using, for example, a personal computer.

[Second Embodiment]
In the second embodiment of the present invention, the predicted increase/decrease in the change in cognitive ability is predicted as an increase/decrease amount that is a continuous value rather than a binary value.

(1) Creation of Learning Data FIG. 6 is a flowchart showing a procedure and processing contents of learning data creation processing by the cognitive ability change prediction device 10A. In the figure, the steps having the same processing contents as those in FIG. 4 will be described with the same reference numerals.

The second embodiment differs from the first embodiment in that the amount of change in the ATMT error rate is stored in the learning DB 40 as the explained variable. That is, the learning data creation unit 103 of the cognitive ability change prediction device 10A creates the learning data, in step S26, for each interval t (1<t<n), the HRV feature amount in the interval t 1, t-1. And the HRV change amount (T ² statistic and Q statistic) in the section t calculated based on the section t−1 as an explanatory variable, and the change amount of the ATMT error rate in the section t+1 and the section t, In other words, the learning data is created with the ATMT error rate itself as the explained variable. Then, the created learning data is stored in the learning DB 40.

(2) User's Cognitive Ability Change Prediction Process When the creation of the learning data is completed, the cognitive ability change prediction device 10A executes a process for predicting a change in the cognitive ability of the prediction target user as follows. FIG. 7 is a flowchart showing the processing procedure and processing contents. In the figure, steps having the same processing contents as those in FIG. 5 will be described with the same reference numerals.

In step S34, the cognitive performance change prediction unit 106 of the cognitive performance change prediction device 10A first reads data representing the amount of change in the ATMT error rate from the learning DB 40 and creates an estimation regression model using an SVR (Support Vector Regression) model. .. Then, the change amount of the ATMT error rate is predicted by inputting the HRV feature amount calculated by the feature amount change calculation unit 105 and the change amount (T ² statistic and Q statistic) into the estimation regression model. Then, the display data indicating the prediction result is generated and transmitted to the display unit 30. The regression model used for estimation is not limited to SVR, and other multiple regression models such as PLS (Partial Least Square) regression model may be used.

By executing the processing as described above, it is possible to predict the increase or decrease of the ATMT error rate and the increase or decrease thereof. Then, by generating display data indicating the prediction result and transmitting the display data to the display unit 30, the display unit 30 displays, for example, that “the cognitive ability is predicted to decrease by 20% for the next 10 minutes compared to the immediately preceding 10 minutes”. A message can be displayed. Therefore, it is possible to instruct the prediction target user not only whether or not to rest, but also for how long the rest should be performed and the type of proper rest (nap, gymnastics, snacks, etc.). Become.

[Third Embodiment]
The third embodiment of the present invention estimates the change in cognitive ability of the target user based on the HRV feature amount actually measured and calculated in the prediction target period after the process of predicting the change in cognitive ability of the target user. , Based on the estimation result of the change in cognitive ability, new learning data is created and added to the learning DB.

FIG. 8 is a block diagram showing the configuration of the cognitive ability change prediction device according to the third embodiment of the present invention. In the figure, the same parts as those in FIG. 3 are designated by the same reference numerals, and detailed description thereof will be omitted.

The cognitive ability change prediction device 10B of the present embodiment further includes a cognitive ability change estimation unit 107. The cognitive ability change estimation unit 107 reads the HRV feature amount actually calculated by the feature amount calculation unit 104 during the prediction target period after the target user's cognitive ability change prediction process. Then, the cognitive ability of the target user in the prediction target period is estimated based on the calculated HRV feature amount, and the estimation result is passed to the learning data creating unit 103 to create new learning data.

With such a configuration, in the cognitive performance change prediction device 10B, when the cognitive performance change prediction unit 106 finishes the prediction process of the cognitive performance change of the target user for the next 10 minutes, for example, the cognitive performance change estimation unit 107 then The actual cognitive ability is estimated in the next 10 minutes.

The process of estimating the cognitive ability is performed as follows. That is, first, the feature amount calculation unit calculates the HRV feature amount for each of the latest 10-minute heartbeat interval data received from the pulse wave sensor 22 while sliding a 5-minute time window with a slide width of 6 seconds. Next, the cognitive ability change estimation unit 107 estimates the actual cognitive ability change of the target user in the 10 minutes based on the calculated HRV feature amount. Specifically, the change amount of the ATMT error rate is estimated based on the estimation result of the change in cognitive ability. Note that Non-Patent Document 2 describes in detail how to estimate the change in cognitive ability using the HRV feature amount.

The cognitive ability change estimation unit 107 calculates the estimation result of the change amount of the ATMT error rate, which is calculated by the feature amount calculation unit 104 and the feature amount change calculation unit 105 in the process of predicting the change in cognitive ability. The amount and the HRV change amount (T ² statistic and Q statistic) are passed to the learning data creation unit 103. The learning data creation unit 103 uses the HRV feature amount and the HRV change amount (T ² statistic amount and Q statistic amount) passed from the cognitive ability change estimation unit 107 as explanatory variables, and estimates the change amount of the ATMT error rate. New learning data with the result as the explained variable is created, and the created learning data is written in the learning DB 40.

As described above, in the third embodiment, after the target user's cognitive ability change prediction processing, the target user's cognitive ability change is calculated based on the HRV feature amount actually measured and calculated in the prediction target period. The learning data is estimated and new learning data is created based on the estimation result of the change in cognitive ability and is added to the learning DB. Therefore, the learning data based on the estimation result of the change in cognitive ability of the target user in the latest period can be stored, and thereby, the change in cognitive ability thereafter can be predicted more accurately.

[Other Embodiments]
In addition to the error rate of ATMT, the number of correct answers and the number of errors of ATMT may be used as an index of cognitive ability. You may use.

Further, in each of the above-described embodiments, the case in which the cognitive ability change prediction device is provided in the server computer or the like has been described as an example, but the device may be provided in a mobile terminal such as a smartphone or a tablet type terminal owned by the target user.
In addition, the configuration of the cognitive ability change prediction device, the procedure of the prediction processing, the processing content, and the like can be variously modified and implemented without departing from the scope of the present invention.

In short, the present invention is not limited to the above-mentioned embodiments as they are, and can be embodied by modifying the constituent elements within a range not departing from the gist of the invention in an implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in each of the above embodiments. For example, some constituent elements may be deleted from all the constituent elements shown in each embodiment. Furthermore, the constituent elements of different embodiments may be combined appropriately.

DESCRIPTION OF SYMBOLS 1... Processing device, 2... Data input means, 3... Result output means, 4... Learning data storage means, 11... Feature amount calculation means, 12... Feature amount change calculation means, 13... Cognitive ability change prediction means, 21, 22... Pulse wave sensor, 23... ATMT input unit, 30... Display unit, 40... Learning database, 101, 104... Feature amount calculating unit, 102, 105... Feature amount change calculating unit, 103... Learning data creating unit, 106... Cognitive ability change prediction unit, 107... Cognitive ability change estimation unit.

Claims (6)

Means for acquiring data representing the heartbeat interval of the prediction target user,
Means for calculating the characteristic amount of the heartbeat fluctuation in the time interval t and the characteristic amount of the heartbeat fluctuation in the immediately preceding time interval t-1, based on the acquired data representing the heartbeat interval,
Based on the calculated feature amount of each heartbeat variability, means for calculating the amount of change in the feature amount of each heartbeat variability,
Based on the calculated feature amount of each heartbeat variability and its change amount, and the learning data stored in advance, the change in the cognitive ability of the prediction target user in the subsequent time interval t+1 is calculated as the time. Means for predicting increase/decrease from cognitive ability in section t,
A unit for outputting information representing a prediction result of a change in cognitive ability of the prediction target user. The means for calculating the feature amount of the heartbeat variability,
A means for calculating a characteristic amount in the time domain based on the data representing the obtained heartbeat interval;
A means for calculating a feature amount in the frequency domain based on the acquired data representing the heartbeat interval,
The means for calculating the feature amount in the time domain includes an average value of heartbeat intervals, a standard deviation of heartbeat intervals (SDNN), a ratio (pNN50) in which the difference between adjacent heartbeat intervals exceeds 50 msec, and an adjacent heartbeat interval. And calculating at least one of the sum of root mean square (RMSSD) of the differences,
The means for calculating the feature amount in the frequency domain is the total power (LF) of the low frequency component (0.04 to 0.15 Hz) when the heartbeat interval is spectrally analyzed and the high frequency component (0.15 to 0.40 Hz) total power (HF) and at least one of a ratio (LF/HF) of the total power LF of the low-frequency component and the total power HF of the high-frequency component. The apparatus for predicting change in cognitive ability according to claim 1. The means for calculating the change amount of the feature amount of the heartbeat variability is the change amount or the change between the feature amount of the heartbeat change in the time period t and the feature amount of the heartbeat change in the immediately preceding time period t-1. Rate, the change amount or change rate of the correlation coefficient between the feature amount of the heartbeat variability in the time section t and the feature amount of the heartbeat change in the immediately preceding time section t-1. The cognitive ability change prediction device according to claim 1, wherein The means for calculating the amount of change in the feature amount of the heartbeat variability is configured to analyze the main component for the feature amount of the heartbeat variability in the time period t and the feature amount of the heartbeat variability in the immediately preceding time period t-1. And calculating a T ² statistic indicating a change in the principal component and a Q statistic indicating a residual of dimensional compression, respectively, based on the analysis result of the principal component. The described device for predicting changes in cognitive ability. The means for predicting a change in the cognitive ability,
A classifier that represents a change in cognitive ability is created based on the learning data, and the calculated feature amount of each heartbeat variability and its change amount are input to the classifier to predict the change in the cognitive ability. Means to do
A means for predicting a change amount of the cognitive ability by creating a regression model based on the learning data and inputting the calculated feature amount of each heartbeat variability and its change amount into the regression model. The cognitive ability change prediction apparatus according to claim 1, further comprising: A program that causes a computer to function as each of the means of the cognitive ability change prediction device according to any one of claims 1 to 5.