Fatigue determination device, method, and program

Abstract

To improve the accuracy of fatigue assessment in subjects. [Solution] The calculation unit 36 ​​calculates, from the changes in the heart rate of a subject who has taken actions to recover after a temporary increase in heart rate, at least the RRI overshoot ratio, which is the ratio of the increase in RRI to the decrease in RRI, as a characteristic quantity of heart rate change. The determination unit 44 then determines the fatigue of the subject based on the characteristic quantity of heart rate change calculated by the calculation unit 36.

Description

Technical Field

[0001] The present disclosure relates to a fatigue determination device, a fatigue determination method, and a fatigue determination program.

Background Art

[0002] Patent Document 1 describes a technique for estimating the degree of fatigue of a vehicle driver based on the stabilization time, which is the time until the driver's biological information, which changes during the braking period from the start of deceleration to the stop of the vehicle, stabilizes after the braking period.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The technique described in Patent Document 1 assumes that biological information (such as blood pressure and heart rate) increases when the vehicle decelerates. However, there are individual differences (variations) in the degree of increase in blood pressure, heart rate, etc. when the vehicle decelerates. And the length of the stabilization time, which is the time until blood pressure, heart rate, etc. stabilize after the braking period of the vehicle, depends on the degree of increase in blood pressure, heart rate, etc. when the vehicle decelerates. For this reason, the technique described in Patent Document 1 is affected by individual differences (variations) in the degree of increase in blood pressure, heart rate, etc. when the vehicle decelerates, and the estimation result of the degree of fatigue fluctuates, so the estimation accuracy of the degree of fatigue is not sufficient.

[0005] The present disclosure has been made in consideration of the above facts, and an object thereof is to obtain a fatigue determination device, a fatigue determination method, and a fatigue determination program that can improve the determination accuracy of fatigue for a subject.

Means for Solving the Problems

[0006] The fatigue determination device according to the first embodiment includes a calculation unit that calculates, as a characteristic quantity of heart rate change, at least the RRI overshoot ratio, which is the ratio of the increase in RRI to the decrease in RRI (RR-Interval: the time interval between the peaks of adjacent R waves, which is the reciprocal of the heart rate), from the changes in the heart rate of a subject who has taken action to recover after a temporary increase in heart rate, and a determination unit that determines the fatigue of the subject based on the characteristic quantity of heart rate change calculated by the calculation unit.

[0007] In the second aspect, in the first aspect, the calculation unit determines the overshoot ratio of the RRI as follows: NNmin is the minimum value of the moving average of the RRI during the first period in which the subject's heart rate temporarily increased; NNmax is the maximum value of the moving average of the RRI during the second period in which the subject's heart rate recovers; and NNend is the moving average of the RRI during the third period following the second period. NNos=(NNmax-NNend) / (NNend-NNmin) …(1) The overshoot ratio NNos of the moving average of RRI is calculated using equation (1) above.

[0008] In a third embodiment, in the first or second embodiment, the determination unit determines the fatigue of the subject using a feature quantity that quantifies the variability of heart rate changes in addition to the overshoot ratio of the RRI.

[0009] The fourth aspect is that, in the third aspect, the feature quantity used to quantify the variability of the heart rate change is at least one of the following: the total power change of the RRI spectrum, which is the sum of the components in a predetermined frequency band (for example, 0.04 Hz to 0.4 Hz) of the RRI spectrum; SDNN (Standard Deviation of the NN intervals), which is the standard deviation of the RRI; and rMSSD (root Mean Square of Successive Differences), which is the root mean square of the difference between two consecutive RRI values.

[0010] The fifth embodiment further includes a learning unit that, in any of the first to fourth embodiments, learns a reference range which is the range of the heart rate change feature quantities when the subject is not fatigued or fatigued, or a threshold value for the heart rate change feature quantities for determining the subject's fatigue, based on the heart rate change feature quantities calculated by the calculation unit during the learning period, and the determination unit determines the subject's fatigue during the determination period by comparing the heart rate change feature quantities calculated by the calculation unit with the reference range or threshold value learned by the learning unit.

[0011] In the sixth embodiment, in the fifth embodiment, the learning unit learns the reference range or threshold for each subject, and the determination unit determines the fatigue of the subject by comparing the characteristic quantity of heart rate change calculated by the calculation unit during the determination period with the reference range or threshold corresponding to the current subject from among the reference ranges or thresholds learned by the learning unit for each subject.

[0012] In the seventh embodiment, in the fifth embodiment, the learning unit learns the reference range or threshold for each of several different time periods throughout the day, and the determination unit determines the fatigue of the subject by comparing the characteristic quantity of heart rate change calculated by the calculation unit during the determination period with the reference range or threshold for the time period that includes the current time, among the reference range or threshold learned by the learning unit for each of the several time periods.

[0013] The fatigue determination method according to the eighth aspect includes a process performed by a computer that includes calculating, as a characteristic quantity of heart rate change, at least the RRI overshoot ratio, which is the ratio of the increase in RRI to the decrease in RRI, from the changes in the heart rate of a subject who has taken action to recover after a temporary increase in heart rate, and determining the fatigue of the subject based on the calculated characteristic quantity of heart rate change.

[0014] The fatigue determination program according to the ninth embodiment causes a computer to perform a process that includes calculating, as a feature quantity of heart rate change, at least the RRI overshoot ratio, which is the ratio of the increase in RRI to the decrease in RRI, from the changes in the heart rate of a subject who has taken action to recover after a temporary increase in heart rate, and determining the fatigue of the subject based on the calculated feature quantity of heart rate change. [Effects of the Invention]

[0015] In the first embodiment, the RRI overshoot ratio, which is the ratio of the increase in RRI to the decrease in RRI, is calculated as a feature of heart rate change from the heart rate changes of a subject who has taken an action to recover after a temporary increase in heart rate. Here, for example, even if the degree of increase in heart rate due to the action taken by the subject was relatively large, that is, the decrease in RRI was relatively large, the RRI overshoot ratio will be normalized to some extent by the fact that the increase in RRI will also be relatively large. Then, in the first embodiment, the fatigue of the subject is determined based on the feature of heart rate change that includes at least this RRI overshoot ratio. As a result, the influence of the degree of increase in heart rate due to the action taken by the subject on the accuracy of fatigue determination for the subject is reduced, and the accuracy of fatigue determination for the subject can be improved.

[0016] In the second embodiment, the overshoot ratio of the RRI is calculated as the overshoot ratio of the moving average of the RRI, NNos, using equation (1) above. This allows the RRI overshoot ratio (overshoot ratio of the moving average of the RRI NNos) to be calculated by a simple process of identifying the parameters of the minimum value of the moving average of the RRI in the first period, the maximum value of the moving average of the RRI in the second period, and the moving average of the RRI in the third period, and substituting them into equation (1). Note that "NNend-NNmin" in equation (1) is an example of the "decrease in RRI" in the first embodiment, and "NNmax-NNend" in equation (1) is an example of the "increase in RRI" in the first embodiment.

[0017] In the third embodiment, fatigue in subjects is determined using a feature that quantifies the variability of heart rate changes, in addition to the RRI overshoot ratio. This makes it possible to further improve the accuracy of fatigue determination for subjects by using the feature that quantifies the variability of heart rate changes, even when the accuracy of fatigue determination using the RRI overshoot ratio is insufficient.

[0018] The total power change of the RRI spectrum, SDNN, and rMSSD are all features that quantify the variability of heart rate changes. In the fourth embodiment, the subject's fatigue level is determined using at least one of these features in addition to the RRI overshoot ratio, making it possible to further improve the accuracy of fatigue determination for the subject.

[0019] It is conceivable that there is variability in the changes in the heart rate of subjects who take actions to recover after their heart rate temporarily increases. In contrast, in the fifth embodiment, the learning unit learns, during the learning period, a reference range which is the range of the heart rate change features when the subject is not fatigued or fatigued, or a threshold for the heart rate change features to determine the subject's fatigue, based on the heart rate change features calculated by the calculation unit. The determination unit then determines the subject's fatigue by comparing the heart rate change features calculated by the calculation unit during the determination period with the reference range or threshold learned by the learning unit. This further improves the accuracy of fatigue determination for subjects compared to embodiments in which a reference range or threshold is not learned.

[0020] It is considered that there are individual differences in the transition of the heart rate of subjects who took actions to recover after the heart rate temporarily increased. In contrast, in the sixth aspect, the learning unit learns the reference range or threshold value for each subject. Then, the determination unit determines the fatigue of the subject by comparing the feature amount of the heart rate change calculated by the calculation unit during the determination period with the reference range or threshold value learned for each subject by the learning unit and corresponding to the current subject. Thereby, compared with the mode in which the reference range or threshold value is not learned for each subject, the determination accuracy of fatigue for the subject can be further improved.

[0021] It is considered that the transition of the heart rate of a subject who took an action to recover after the heart rate temporarily increased differs depending on the time zone in which the action was taken due to the influence of the circadian rhythm. In contrast, in the seventh aspect, the learning unit learns the reference range or threshold value for each of a plurality of different time zones within a day. Then, the determination unit determines the fatigue of the subject by comparing the feature amount of the heart rate change calculated by the calculation unit during the determination period with the reference range or threshold value learned for each of the plurality of time zones by the learning unit and corresponding to the time zone including the current time. Thereby, compared with the mode in which the reference range or threshold value is not learned for each of the plurality of time zones, the determination accuracy of fatigue for the subject can be further improved.

[0022] In the eighth aspect and the ninth aspect, similar to the first aspect, the determination accuracy of fatigue for the subject can be improved.

Brief Description of Drawings

[0023] <00​​​​​​​​​​​​​This is an explanatory diagram illustrating the calculation of the RRI overshoot ratio by the calculation unit. [Figure 6] This is an explanatory diagram for calculating the reference range for the RRI overshoot ratio and for determining fatigue based on that reference range. [Figure 7] This is a block diagram showing the schematic configuration of the fatigue determination system according to the second embodiment. [Figure 8] This is a flowchart showing the fatigue determination process according to the second embodiment. [Figure 9] This is an explanatory diagram for explaining the calculation of the total power change in the RRI spectrum. [Figure 10] This is a flowchart showing the fatigue determination process according to the third embodiment. [Figure 11] This is an explanatory diagram used to describe the details of the experiments conducted by the inventors. [Figure 12] This graph shows the trend of the moving average of RRI obtained in the above experiment. [Figure 13] This graph shows the progression of the total power of the RRI spectrum obtained in the above experiment. [Figure 14] This figure shows the calculation results for the overshoot ratio NNos of the moving average of the RRI and the total power change TPd of the RRI spectrum obtained in the above experiment. [Modes for carrying out the invention]

[0024] Hereinafter, an example of an embodiment of this disclosure will be described in detail with reference to the drawings.

[0025] [First Embodiment] Figure 1 shows a fatigue determination system 10A according to the first embodiment. The fatigue determination system 10A includes a touch sensor 28 provided on the steering wheel 12 of the vehicle and an on-board ECU 30 mounted on the vehicle.

[0026] The steering wheel 12 (see Figures 2 and 3), equipped with a touch sensor 28, is positioned on the front side of the vehicle, above the seat (driver's seat) where the subject (driver) sits. In Figures 2 and 3, the front of the vehicle is indicated by arrow FR, the top of the vehicle by arrow UP, and the right side of the vehicle in the vehicle width direction by arrow HR. The radial direction of the steering wheel 12 is indicated by arrow R, and the circumferential direction of the steering wheel 12 is indicated by arrow L.

[0027] As shown in Figure 2, the steering wheel 12 includes an annular rim portion 14 as a gripping portion, a boss portion 16 provided in the center, and a stay portion 18. The steering wheel 12 is provided with a metal core. The core consists of a rim core portion 20 of the rim portion 14 (see Figure 3), a boss core portion of the boss portion (not shown), and a stay core portion of the stay portion 18 (not shown), with the rim core portion 20 being formed in an annular (ring shape). In the steering wheel 12, the rim core portion 20 and the boss core portion are connected via the stay core portion, and the core forms a framework, integrating the rim portion 14, the boss portion 16, and the stay portion 18.

[0028] The vehicle is equipped with a steering shaft (not shown), which is positioned so that its axis aligns with the vehicle's longitudinal direction and is rotatably supported by the vehicle body. The steering wheel 12 has a boss core portion of the boss portion 16 fixed to the rear end of the steering shaft and is supported by the steering shaft, allowing it to rotate integrally with the steering shaft. Therefore, when the steering wheel 12 is rotated, the steering shaft rotates and the vehicle is steered.

[0029] As shown in Figure 3, the rim portion 14 has a substantially circular (or substantially elliptical) radial cross-section of the steering wheel 12, and a base body 22 formed in an annular shape from a resin material such as urethane as an insulating material is arranged inside the rim portion 14. The rim core portion 20 is housed in the base body 22 by insert formation, and the rim core portion 20 is covered by the base body 22.

[0030] Furthermore, decorative parts 24 are arranged on the outer circumference of the base body 22 as contact parts, and the entire circumference of the base body 22 in the radial cross-section of the steering wheel 12, and the entire circumference (entire area) of the steering wheel 12 in the circumferential direction are covered by the decorative parts 24. A resin material such as urethane is used for the decorative parts 24 as an insulating material, and the rim portion 14 of the steering wheel 12 is decorated with decorative parts 24. The decorative parts 24 may also be made of leather such as tanned leather.

[0031] Furthermore, a touch sensor 28, including a sensor electrode 32, is embedded between the base body 22 and the decorative part 24 within the rim portion 14 of the steering wheel 12. The sensor electrode 32 is formed in a roughly strip shape from a sheet-like or film-like conductive material. The touch sensor 28 may also include a shield electrode formed in a roughly strip shape from an insulating material, similar to the sensor electrode 32, and a known configuration can be applied, such as the sensor electrode 32 being placed on one side of the strip-shaped insulating material and the shield electrode being placed on the other side. In addition, the sensor electrode 32 of the touch sensor 28 (and the shield electrode as well) may be made of a conductive fabric in which a conductive material such as metal is attached to the surface of a woven fabric that is stretchable by weaving warp and weft threads.

[0032] As shown in Figure 2, the touch sensor 28 (sensor electrode 32) is positioned with its longitudinal direction aligned with the circumferential direction of the steering wheel 12, and is arranged on the rim portion 14 in a range of approximately half a turn in the circumferential direction of the steering wheel 12. Also, as shown in Figure 2, the touch sensor 28 is wrapped around approximately the entire circumference of the outer circumference of the base body 22, with each sensor electrode 32 facing radially outward from the rim portion 14 and its width aligned with the circumferential direction of the rim portion 14.

[0033] As a result, the touch sensor 28 is wrapped around the outer surface of the base body 22 over substantially the entire circumference of the steering wheel 12 and the rim portion 14, and is covered by the decorative portion 24. Furthermore, the two touch sensors 28 are electrically isolated, with one positioned on the right side of the vehicle and the other on the left side of the vehicle when the steering wheel 12 is in the straight-ahead steering position (the position shown in Figure 2). The two touch sensors 28 are each connected to the on-board ECU 30. Note that the processing of the two touch sensors 28 in the on-board ECU 30 is identical, and the following description will focus on the configuration corresponding to one touch sensor 28.

[0034] Although not shown in the diagram, the in-vehicle ECU 30 includes a CPU (Central Processing Unit), memory such as ROM (Read Only Memory) and RAM (Random Access Memory), storage such as HDD (Hard Disk Drive) and SSD (Solid State Drive), a communication interface (I / F) section, and an input / output interface (I / F) section, all of which are connected to each other via a bus for communication. The input / output interface section is connected to a touch sensor 28 via an A / D converter, as well as a display unit such as an in-vehicle display and an audio output unit such as an in-vehicle speaker. The in-vehicle ECU 30 is an example of a fatigue determination device according to this disclosure.

[0035] Furthermore, a fatigue determination program is stored in the ROM or storage of the in-vehicle ECU 30. The in-vehicle ECU 30 reads the fatigue determination program from the ROM or storage and loads it into memory. The CPU then executes the fatigue determination program loaded into memory, and the ECU 30 functions as the electrocardiogram waveform acquisition unit 34, calculation unit 36, heart rate data DB (DataBase) 38, learning unit 40, reference range table 42, determination unit 44, and output unit 46 shown in Figure 1, respectively, and performs the fatigue determination process described later. Note that the fatigue determination program described above is an example of a fatigue determination program related to this disclosure.

[0036] The electrocardiogram waveform acquisition unit 34 acquires the electrocardiogram waveform (heart rate change) of a subject who has taken an action to recover after a temporary increase in heart rate from the touch sensor 28 as digital electrocardiogram waveform data via an A / D converter (not shown). The calculation unit 36 ​​calculates the RRI overshoot ratio, which is the ratio of the increase in RRI to the decrease in RRI, as a characteristic quantity of heart rate change from the electrocardiogram waveform data acquired by the electrocardiogram waveform acquisition unit 34. The heart rate data DB 38 stores the RRI overshoot ratio calculated by the calculation unit 36.

[0037] During the learning period, the learning unit 40 learns a reference range, which is the range of the RRI overshoot ratio when the subject is not fatigued, based on the RRI overshoot ratio calculated by the calculation unit 36 ​​and stored in the heart rate data DB 38. The reference range table 42 stores the reference range of the RRI overshoot ratio when the subject is not fatigued, which has been learned by the learning unit 40.

[0038] During the determination period, the determination unit 44 compares the RRI overshoot ratio calculated by the calculation unit 36 ​​with the reference range of the RRI overshoot ratio when the subject is not fatigued, which has been learned by the learning unit 40 and stored in the reference range table 42, to determine whether or not the subject is fatigued.

[0039] The output unit 46 is connected to an in-vehicle UI (User Interface) unit 48, which includes a display unit such as an in-vehicle display and an audio output unit such as an in-vehicle speaker. The output unit 46 outputs the result of the determination unit 44's determination of whether or not the subject is fatigued via the in-vehicle UI unit 48. If the determination unit 44 determines that the subject is fatigued, the output unit 46 may also output a warning to the subject via the in-vehicle UI unit 48, prompting them to take a break.

[0040] Furthermore, if there is an administrator who manages the subject's physical condition, the output unit 46 may transmit the determination result of whether or not the subject is fatigued, made by the determination unit 44, to the administrator terminal 52 via the communication unit 50. In this embodiment, the communication unit 50 can be a DCM (Data Communication Module) installed in the vehicle. In this embodiment, the output unit 46 may be configured, for example, to output a warning from the administrator terminal 52 prompting the administrator to take action when the determination unit 44 determines that the subject is fatigued.

[0041] Next, as an example of the operation of the first embodiment, the fatigue determination process performed by the on-board ECU 30 when the vehicle's ignition switch is turned on and the steering wheel 12 is gripped by the subject will be explained with reference to Figure 4.

[0042] Before the subject turns on the vehicle's ignition switch and grips the steering wheel 12, the subject walks out of their home to the parking lot, unlocks the vehicle doors, and sits in the driver's seat. Therefore, when the vehicle's ignition switch is turned on and the steering wheel 12 is gripped, the subject's heart rate temporarily increases, and then recovers as the subject remains seated in the driver's seat. Thus, the subject's actions described above are an example of "actions that cause a temporary increase in heart rate followed by recovery" in this disclosure.

[0043] In step 102 of the fatigue determination process, the electrocardiogram waveform acquisition unit 34 acquires electrocardiogram waveform data representing the subject's heart rate progression from the touch sensor 28.

[0044] In step 104, the calculation unit 36 ​​first extracts the R wave from the electrocardiogram waveform data acquired by the electrocardiogram waveform acquisition unit 34 and calculates the moving average of RRI. Next, from the calculation result of the moving average of RRI, as shown in Figure 5, it identifies the minimum value NNmin of the moving average of RRI during the first period when the subject's heart rate temporarily increased, the maximum value NNmax of the moving average of RRI during the second period when the subject's heart rate recovered, and the moving average NNend of RRI during the third period following the second period. Then, it substitutes the identified NNmin, NNmax, and NNend into equation (1) above and calculates the overshoot ratio NNos of the moving average of RRI.

[0045] In step 108, the calculation unit 36 ​​stores the moving average overshoot ratio NNos of the RRI calculated in step 104 in the heart rate data DB 38. Thus, in the first embodiment, each time the subject turns on the vehicle's ignition switch and grips the steering wheel 12, the moving average overshoot ratio NNos of the RRI is calculated and stored in the heart rate data DB 38.

[0046] In step 114, the learning unit 40 determines whether or not to perform learning of the reference range for the overshoot ratio. The determination in step 114 is denied, for example, until the number of overshoot ratio data stored in the heart rate data DB 38 reaches a first predetermined value, in which case the process proceeds from step 114 to step 128. Alternatively, for example, after the number of overshoot ratio data stored in the heart rate data DB 38 reaches the first predetermined value, the determination in step 114 is affirmed and the process proceeds to step 116.

[0047] Furthermore, after the number of overshoot ratio data points stored in the heart rate data DB38 reaches a first predetermined value, the determination in step 114 may be affirmed each time the increment in the number of overshoot ratio data points stored since the last learning of the reference range for the overshoot ratio reaches a second predetermined value. Alternatively, if the above conditions are not met, the determination in step 114 may be denied.

[0048] In step 116, the learning unit 40 calculates a reference range for the RRI overshoot ratio, which is the range of the RRI overshoot ratio when the subject is not fatigued, from the overshoot ratio data stored in the heart rate data DB38. Specifically, for example, outlier detection is performed on the overshoot ratio data stored in the heart rate data DB38, and the range of the overshoot ratio is calculated from the data set after excluding the detected outliers. When a subject is driving a vehicle, their physical condition is generally not fatigued for the most part. Therefore, through the above outlier detection, the data of the overshoot ratio when the subject is fatigued is detected and excluded as outliers, and the range of the overshoot ratio when the subject is not fatigued (see also "Variation Range When Not Fatigued" shown in Figure 6) is obtained as the reference range for the overshoot ratio.

[0049] Alternatively, anomaly detection or unsupervised learning may be applied instead of outlier detection. Furthermore, during the learning period (the period during which the judgment in step 114 is affirmed), subjects may be asked to self-report whether or not they were fatigued. Based on the self-reported results, the overshoot ratio data may be separated into a non-fatigued data group and a fatigued data group, and a reference range for the overshoot ratio in the non-fatigued state or the fatigued state may be calculated from the separated data groups (supervised learning).

[0050] In step 122, the learning unit 40 stores the reference range of the RRI overshoot ratio for the subject in a non-fatigued state, calculated in step 116, in the reference range table 42.

[0051] In step 128, the determination unit 44 determines whether or not to perform a fatigue assessment on the subject. The determination in step 128 is, for example, denied if the reference range for the RRI overshoot ratio is not stored in the reference range table 42, and affirmed if the reference range for the RRI overshoot ratio is stored in the reference range table 42. If the determination in step 128 is affirmed, the process proceeds to step 130.

[0052] In step 130, the determination unit 44 reads the reference range for the overshoot ratio from the reference range table 42. Then, in step 136, the determination unit 44 compares the RRI overshoot ratio calculated in step 104 with the reference range for the RRI overshoot ratio read from the reference range table 42 in step 130 to determine whether or not the subject is fatigued.

[0053] As shown in Figure 6 as an example, in step 136, if the RRI overshoot ratio calculated in step 104 falls within the standard range for the RRI overshoot ratio, the subject can be judged as "not fatigued." On the other hand, if the RRI overshoot ratio calculated in step 104 falls outside the standard range for the RRI overshoot ratio, the subject can be judged as "fatigued."

[0054] After determining whether the subject is fatigued in step 136, the process proceeds to step 140. In step 140, the output unit 46 outputs the result of the subject's fatigue determination, and the fatigue determination process is completed.

[0055] Thus, in the first embodiment, the calculation unit 36 ​​calculates, as a feature of heart rate change, at least the RRI overshoot ratio, which is the ratio of the increase in RRI to the decrease in RRI, from the changes in the heart rate of a subject who has taken actions to recover after a temporary increase in heart rate. The determination unit 44 then determines the subject's fatigue based on the feature of heart rate change calculated by the calculation unit 36. This reduces the influence of the degree of increase in heart rate due to the subject's actions on the accuracy of fatigue determination for the subject, and improves the accuracy of fatigue determination for the subject.

[0056] Furthermore, in the first embodiment, the calculation unit 36 ​​calculates the RRI overshoot ratio NNos using equation (1) above, where NNmin is the minimum value of the moving average of RRI during the first period when the subject's heart rate temporarily increased, NNmax is the maximum value of the moving average of RRI during the second period when the subject's heart rate recovers, and NNend is the moving average of RRI during the third period following the second period. This allows the RRI overshoot ratio (the RRI moving average overshoot ratio NNos) to be calculated with a simple process.

[0057] Furthermore, in the first embodiment, the learning unit 40 learns a reference range, which is the range of the heart rate change features when the subject is not fatigued, based on the heart rate change features calculated by the calculation unit 36 ​​during the learning period. The determination unit 44 then determines the subject's fatigue by comparing the heart rate change features calculated by the calculation unit 36 ​​with the reference range learned by the learning unit 40 during the determination period. This further improves the accuracy of fatigue determination for the subject compared to an embodiment in which the reference range of heart rate change features is not learned.

[0058] [Second Embodiment] Next, a second embodiment of this disclosure will be described. Note that the same reference numerals are used for parts identical to those in the first embodiment, and their descriptions are omitted. While the first embodiment assumes that the same subject always sits in the driver's seat of the vehicle, the second embodiment is a configuration that corresponds to a case where any of multiple subjects sits in the driver's seat of the vehicle.

[0059] As shown in Figure 7, in the fatigue determination system 10B according to the second embodiment, a subject identification unit 54 is provided in the in-vehicle ECU 30. The subject identification unit 54 acquires biometric information of a subject seated in the driver's seat of a vehicle, such as a facial image or fingerprint information, and identifies the subject seated in the driver's seat of the vehicle based on the acquired biometric information. Alternatively, the subject identification unit 54 may identify the subject by communicating with a smartwatch or the like worn by the subject, instead of acquiring the subject's biometric information. The subject identification result by the subject identification unit 54 is input to the learning unit 40 and the determination unit 44.

[0060] Next, with reference to Figure 8, the fatigue determination process according to the second embodiment will be described. In the fatigue determination process according to the second embodiment, first in step 100, the subject identification unit 54 identifies the current subject sitting in the driver's seat of the vehicle.

[0061] Furthermore, in the fatigue determination process according to the second embodiment, the process of step 106 is performed instead of step 104 described in the first embodiment. In step 106, the calculation unit 36 ​​calculates the RRI overshoot ratio from the electrocardiogram waveform data acquired by the electrocardiogram waveform acquisition unit 34 in the same manner as in the first embodiment, and also calculates the total power change amount of the RRI spectrum as a feature quantity that quantifies the variability of heart rate changes.

[0062] Specifically, the calculation unit 36 ​​first extracts the R wave from the electrocardiogram waveform data and calculates the change in total power of the RRI spectrum (time series change), which is the sum of the components in a predetermined frequency band (e.g., 0.04 Hz to 0.4 Hz) of the RRI spectrum. When a subject takes action to recover after a temporary increase in heart rate, the total power of the RRI spectrum shows a change in which it increases initially and then decreases, converging to a certain value, as shown in Figure 9. From the change in total power of the RRI spectrum, the calculation unit 36 ​​extracts the maximum value TPmax and the convergence value TPend of the total power of the RRI spectrum. Then, it substitutes the extracted maximum value TPmax and convergence value TPend into the following equation (2) to calculate the change in total power of the RRI spectrum TPd. TPd = TPmax - TPend …(2)

[0063] In the next step 110, the calculation unit 36 ​​stores the RRI overshoot ratio and the total power change of the RRI spectrum calculated in step 106 in the heart rate data DB 38, associating them with the subject ID of the current subject identified in step 100.

[0064] Furthermore, in the fatigue determination process according to the second embodiment, the process of step 118 is performed instead of step 116 as described in the first embodiment. In step 118, the learning unit 40 calculates, for each subject, the reference range for the RRI overshoot ratio and the reference range for the total power change of the RRI spectrum from the data stored in the heart rate data DB 38. The reference range for the total power change of the RRI spectrum can also be calculated by applying outlier detection, etc., in the same way as the calculation of the reference range for the RRI overshoot ratio described in the first embodiment.

[0065] Then, in the next step 124, the learning unit 40 stores the reference range for the RRI overshoot ratio and the reference range for the total power change of the RRI spectrum, which were calculated for each subject in step 118, in the reference range table 42, associating them with the subject ID of each individual subject.

[0066] Furthermore, in the fatigue determination process according to the second embodiment, the process of step 132 is performed instead of step 130 described in the first embodiment. In step 132, the determination unit 44 reads the reference range of the RRI overshoot ratio and the reference range of the total power change of the RRI spectrum, which are associated with the subject ID of the current subject, from the reference range table 42.

[0067] Then, in step 138, the determination unit 44 compares the RRI overshoot ratio calculated in step 106 with the reference range of the RRI overshoot ratio read in step 132, and compares the total power change of the RRI spectrum calculated in step 106 with the reference range of the total power change of the RRI spectrum read in step 132 to determine whether or not the subject is currently fatigued.

[0068] In step 138, fatigue can be determined, for example, if fatigue is detected based on the RRI overshoot ratio and also based on the change in the total power of the RRI spectrum, the subject may be deemed fatigued. Alternatively, the subject may be deemed fatigued if fatigue is detected based on at least one of the RRI overshoot ratio and the change in the total power of the RRI spectrum. Furthermore, the presence or absence of fatigue in the subject may be determined by calculating a weighted average of the RRI overshoot ratio and the change in the total power of the RRI spectrum as an evaluation value for the subject's fatigue, and comparing the calculated evaluation value with a reference range or threshold for that evaluation value.

[0069] Thus, in the second embodiment, the determination unit 44 determines the subject's fatigue using, in addition to the RRI overshoot ratio, a feature that quantifies the variability of heart rate changes, specifically the total power change of the RRI spectrum, as a feature of heart rate change. This makes it possible to further improve the accuracy of fatigue determination for the subject.

[0070] In the second embodiment, the learning unit 40 learns a reference range for heart rate change features for each subject, and the determination unit 44 determines the fatigue of the current subject by comparing the heart rate change features calculated by the calculation unit 36 ​​during the determination period with the reference range corresponding to the current subject among the reference ranges for heart rate change features learned for each subject by the learning unit. This further improves the accuracy of fatigue determination for subjects compared to the first embodiment, in which the reference range for heart rate change features is not learned for each subject.

[0071] [Third Embodiment] Next, a third embodiment of this disclosure will be described. Since the third embodiment has the same configuration as the second embodiment, the same reference numerals are used for each part, and the description of the configuration will be omitted. Hereinafter, the fatigue determination process according to the third embodiment will be described with reference to Figure 8.

[0072] In the fatigue determination process according to the third embodiment, the process of step 112 is performed instead of step 110 described in the second embodiment. In step 112, the calculation unit 36 ​​stores the RRI overshoot ratio and the total power change of the RRI spectrum calculated in step 106 in the heart rate data DB 38, associating them with the subject ID of the current subject and the current time, including the current time. The time period may be divided into two parts, for example, AM / PM, or into three or more finer time periods.

[0073] Furthermore, in the fatigue determination process according to the third embodiment, the process of step 120 is performed instead of step 118 described in the second embodiment. In step 120, the learning unit 40 calculates, for each subject and for each time period, a reference range for the overshoot ratio of the RRI and a reference range for the total power change of the RRI spectrum, respectively, from the data stored in the heart rate data DB 38.

[0074] Furthermore, in the fatigue determination process according to the third embodiment, the process of step 126 is performed instead of step 124 described in the second embodiment. In step 126, the learning unit 40 stores the reference range of the RRI overshoot ratio and the reference range of the total power change of the RRI spectrum, which were calculated in step 120 for each subject and each time period, in the reference range table 42, associating them with the subject ID and individual time period of each subject.

[0075] Furthermore, in the fatigue determination process according to the third embodiment, the process of step 134 is performed instead of step 1#2 described in the second embodiment. In step 134, the determination unit 44 reads the subject ID of the current subject and the reference range of the RRI overshoot ratio and the reference range of the total power change of the RRI spectrum corresponding to the current time period from the reference range table 42.

[0076] Then, in the next step 138, similar to the second embodiment, the RRI overshoot ratio calculated in step 106 is compared with the reference range of the RRI overshoot ratio read in step 134, and the total power change of the RRI spectrum calculated in step 106 is compared with the reference range of the total power change of the RRI spectrum read in step 134 to determine whether or not the subject is currently fatigued.

[0077] Thus, in the third embodiment, the learning unit 40 learns reference ranges for heart rate change features for multiple different time periods throughout the day. The determination unit 44 then determines the subject's fatigue by comparing the heart rate change features calculated by the calculation unit 36 ​​during the determination period with the reference range for the current time period, which is among the reference ranges for heart rate change features learned by the learning unit 40 for multiple time periods. This further improves the accuracy of fatigue determination for the subject compared to the embodiment in which reference ranges for heart rate change features are not learned for multiple time periods.

[0078] In the above embodiment, the learning unit 40 learns a reference range for heart rate change features, and the determination unit 44 determines the subject's fatigue based on this reference range. However, this disclosure is not limited thereto. For example, the learning unit 40 may learn (calculate) a threshold for heart rate change features (a boundary value between the range of heart rate change features when fatigued and the range of heart rate change features when not fatigued) for determining the subject's fatigue, and the determination unit 44 may determine the subject's fatigue by comparing the heart rate change features calculated by the calculation unit 36 ​​with the threshold for heart rate change features learned by the learning unit 40.

[0079] Furthermore, in the above embodiment, a learning unit 40 learns a reference range or threshold for the heart rate change feature quantities, and a determination unit 44 determines the subject's fatigue based on the heart rate change feature quantities calculated by the calculation unit 36 ​​and the reference range or threshold learned by the learning unit 40. However, this disclosure is not limited thereto. For example, the learning unit 40 may be omitted, and the determination unit 44 may determine the subject's fatigue by comparing the heart rate change feature quantities calculated this time by the calculation unit 36 ​​with the heart rate change feature quantities calculated last time by the calculation unit 36. Although omitting the learning unit 40 may reduce the accuracy of the subject's fatigue determination, this disclosure also includes such embodiments within its scope.

[0080] Furthermore, although the above embodiment describes an embodiment in which the changes in the subject's heart rate are acquired by a touch sensor 28 provided on the steering wheel 12, this disclosure is not limited thereto. For example, the changes in the subject's heart rate may be acquired by a smartwatch with an electrocardiogram measurement function worn by the subject.

[0081] Furthermore, in the second and third embodiments, an example of applying the total power change of the RRI spectrum as a feature quantity to quantify the variability of heart rate changes was described, but this disclosure is not limited thereto. As a feature quantity to quantify the variability of heart rate changes, for example, SDNN, which is the standard deviation of RRI, or rMSSD, which is the root mean square of the difference between two consecutive RRI values, may be applied.

[0082] Furthermore, while the above embodiment describes a method for determining a subject's fatigue using the overshoot ratio NNos of the moving average of the RRI as a feature of heart rate change, this disclosure is not limited thereto. For example, the subject's fatigue may be determined using the ratio of the morning value to the afternoon value of the overshoot ratio NNos of the moving average of the RRI (e.g., NNos_am / NNos_pm).

[0083] Furthermore, although the above embodiment describes an embodiment in which an in-vehicle ECU 30 functions as a fatigue determination device according to the present disclosure, the present disclosure is not limited thereto, and it is also possible to use equipment such as an ergometer or treadmill as a fatigue determination device according to the present disclosure. In this case, by instructing the subject to exercise at a predetermined load for a predetermined time, the subject can be put into a state in which their heart rate temporarily increases and then recovers. It is also possible to adjust the exercise menu performed by the subject based on the fatigue determination result.

[0084] Furthermore, although the above describes a configuration in which the fatigue determination program is pre-stored (installed) in the ROM or storage of the in-vehicle ECU 30, the fatigue determination program according to this disclosure can also be provided in a form recorded on a non-temporary recording medium such as an HDD, SSD, or DVD. [Examples]

[0085] Next, we will describe the experiments conducted by the inventors of this application to verify the validity of this disclosure.

[0086] This experiment was conducted over two days with six subjects. As shown in Figure 11, on the first day, no exercise was performed to simulate a non-fatigued state. On the second day, to simulate a fatigued state, the subjects were subjected to one hour of exercise (cycling) on ​​an ergometer between 10:00 and 12:00, maintaining a constant heart rate. Heart rate recovery tests were also conducted at 9:00 and 15:00 on both days. In these heart rate recovery tests, after resting for one minute, the subjects were asked to climb stairs back and forth, followed by five minutes of rest, during which their heart rate was measured.

[0087] Figure 12 shows the moving average of the RRI calculated from the heart rate changes obtained in the 15:00 heart rate recovery test, with the data for day 1 (no exercise load, non-fatigued) and day 2 (with exercise load, fatigued) overlaid for each of the three subjects (data 1-3 shown in Figure 12). As is clear from the results shown in Figure 12, the degree of RRI overshoot differs between the non-fatigued and fatigued states, with the RRI overshoot being smaller in the fatigued state than in the non-fatigued state.

[0088] This revealed that by calculating and comparing the RRI overshoot ratio, which is the ratio of the increase in RRI to the decrease in RRI, from the changes in the heart rate of subjects who took action to recover after a temporary increase in heart rate (for example, light exercise such as going up and down stairs in a heart rate recovery test), it is possible to determine the fatigue level of the subjects.

[0089] Furthermore, Figure 13 shows the calculation of the total power of the RRI spectrum, which is the sum of the 0.04Hz to 0.4Hz components of the RRI, based on the heart rate changes obtained in the 15:00 heart rate recovery test. For each of the three subjects, the data for day 1 (non-fatigued) without exercise load and the data for day 2 (fatigued) with exercise load are superimposed (data 1-3 shown in Figure 13). As is clear from the results shown in Figure 13, although not as clear as the RRI overshoot, the change in the total power of the RRI spectrum also differs between non-fatigued and fatigued states, with the change in the total power of the RRI spectrum being smaller when fatigued than when non-fatigued.

[0090] Figure 14 shows the results of calculating the RRI moving average overshoot ratio NNos and the RRI spectral total power change TPd for each of the six subjects. In Figure 14, the smaller value for each parameter is shown in thicker lines for the data from day 1 (non-fatigued) with no exercise load and the data from day 2 (fatigued) with exercise load. For the RRI moving average overshoot ratio NNos, only for participant s6 the value was smaller in the non-fatigued state, indicating that fatigue cannot be properly determined for participant s6 using only the RRI moving average overshoot ratio NNos. On the other hand, for the RRI spectral total power change TPd, only for participant s2 the value was smaller in the non-fatigued state, indicating that fatigue can be properly determined for participant s6 by using the RRI spectral total power change TPd.

[0091] This revealed that, when determining fatigue in subjects who temporarily increased their heart rate and then recovered, using the total power of the RRI spectrum in addition to the RRI overshoot ratio may improve the accuracy of determining fatigue in subjects. [Explanation of Symbols]

[0092] 10A, 10B... Fatigue detection system, 28... Touch sensor, 30... In-vehicle ECU (fatigue detection device), 34... Electrocardiogram waveform acquisition unit, 36... Calculation unit, 38... Heart rate data DB, 40... Learning unit, 42... Reference range table, 44... Judgment unit, 46... Output unit, 54... Subject identification unit

Claims

1. A calculation unit calculates, from the changes in the heart rate of subjects who took actions to recover after a temporary increase in heart rate, at least the RRI overshoot ratio, which is the ratio of the increase in RRI to the decrease in RRI, as a characteristic of heart rate change. A determination unit that determines the fatigue of the subject based on the characteristic quantity of heart rate change calculated by the calculation unit, A fatigue determination device that includes [a specific component].

2. The calculation unit determines the overshoot ratio of the RRI as follows: NNmin is the minimum value of the moving average of RRI during the first period when the subject's heart rate temporarily increased; NNmax is the maximum value of the moving average of RRI during the second period when the subject's heart rate recovers; and NNend is the moving average of RRI during the third period following the second period. NNos=(NNmax-NNend) / (NNend-NNmin)...(1) The fatigue determination device according to claim 1, wherein the overshoot ratio NNos of the moving average of RRI is calculated using the above formula (1).

3. The fatigue determination device according to claim 1, wherein the determination unit determines the fatigue of the subject using a feature quantity that quantifies the variability of heart rate changes in addition to the overshoot ratio of the RRI.

4. The fatigue determination device according to claim 3, wherein the feature quantity used to quantify the variability of the heart rate change is at least one of the following: the total power change of the RRI spectrum, which is the sum of components in a predetermined frequency band of the RRI spectrum; SDNN, which is the standard deviation of RRI; and rMSSD, which is the root mean square of the difference between two consecutive RRI values.

5. The system further includes a learning unit that, during the learning period, learns a reference range which is the range of the heart rate change feature quantities when the subject is not fatigued or fatigued, or a threshold for the heart rate change feature quantities to determine the subject's fatigue, based on the heart rate change feature quantities calculated by the calculation unit. The fatigue determination device according to claim 1, wherein the determination unit determines the fatigue of the subject by comparing the characteristic quantity of the heart rate change calculated by the calculation unit with the reference range or threshold learned by the learning unit during the determination period.

6. The learning unit learns the reference range or threshold for each subject, The fatigue determination device according to claim 5, wherein the determination unit determines the fatigue of a subject by comparing the characteristic quantity of heart rate change calculated by the calculation unit during the determination period with the reference range or threshold learned for each subject by the learning unit, which corresponds to the current subject.

7. The learning unit learns the reference range or threshold for multiple different time periods throughout the day. The fatigue determination device according to claim 5, wherein the determination unit determines the fatigue of the subject by comparing the characteristic quantity of heart rate change calculated by the calculation unit during the determination period with the reference range or threshold learned by the learning unit for each of the multiple time periods, specifically for the time period including the current time.

8. From the heart rate changes of subjects who took actions to recover after a temporary increase in heart rate, at least the RRI overshoot ratio, which is the ratio of the increase in RRI to the decrease in RRI, was calculated as a characteristic of heart rate change. The fatigue of the subject is determined based on the calculated characteristic quantities of heart rate change. A fatigue determination method that involves performing a process including the following using a computer.

9. On the computer, From the heart rate changes of subjects who took actions to recover after a temporary increase in heart rate, at least the RRI overshoot ratio, which is the ratio of the increase in RRI to the decrease in RRI, was calculated as a characteristic of heart rate change. The fatigue of the subject is determined based on the calculated characteristic quantities of heart rate change. A fatigue determination program for executing processes that include the following.

Images