Drowsing prediction device, drowsing prediction system, drowsing prediction program and drowsing prediction method

Abstract

To provide a drowsing prediction device, drowsing prediction system, drowsing prediction program, and drowsing prediction method capable of accurately determining a driver's sleep state and predicting the risk of drowsing during driving to create a safe and reasonable operation plan and preventing traffic accidents caused by drowsy driving in advance.SOLUTION: A drowsing prediction device includes a pulse rate data acquisition unit 41 that acquires pulse rate data during non-sleep and sleep; an average pulse rate drop rate calculation unit 42 that calculates a drop rate of the average pulse rate during sleep relative to the average pulse rate during non-sleep; a total pulse rate ratio calculation unit 43 that calculates a ratio of the total pulse rate during sleep to the total pulse rate during non-sleep; and a sleep state determination unit 44 that determines a sleep state on the basis of the drop rate and the ratio.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a doze prediction device, a doze prediction system, a doze prediction program, and a doze prediction method that are suitable for predicting doze based on a sleep state. [Background technology]

[0002] Conventionally, there have been proposed technologies for detecting whether a vehicle driver is dozing off, etc. For example, Japanese Patent Application Laid-Open Publication No. 2017-143889 discloses a dozing off detection system that determines the degree of eye opening from captured images of both eyes of a subject, and detects whether the subject is dozing off based on the ratio of the eye opening time to the eye closing time obtained from the transition of the eye opening degree (Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2017-143889 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, conventional drowsiness detection systems, including that of Patent Document 1, only detect after the fact that a driver has fallen asleep, and do not predict the risk of falling asleep in advance. Conventionally, operation plans for long-distance trucks and buses have often been created in a bureaucratic manner without considering the sleep state of each individual driver. For this reason, if unreasonable operation plans are made due to a driver shortage or the like, there is a risk that traffic accidents caused by drowsy driving will be triggered.

[0005] The present invention has been made to solve these problems, and aims to provide a drowsiness prediction device, drowsiness prediction system, drowsiness prediction program, and drowsiness prediction method that can accurately determine a driver's sleep state and predict the risk of falling asleep while driving, thereby creating a safe and reasonable operation plan and preventing traffic accidents caused by drowsy driving. [Means for solving the problem]

[0006] In order to solve the problem of creating a safe and reasonable operation plan and preventing traffic accidents caused by drowsy driving by accurately determining a driver's sleep state and predicting the risk of drowsing while driving, the drowsiness prediction device of the present invention has a pulse rate data acquisition unit that acquires pulse rate data during non-sleep and sleep, an average pulse rate decline rate calculation unit that calculates the decline rate of the average pulse rate during sleep to the average pulse rate during non-sleep based on the pulse rate data, a total pulse rate ratio calculation unit that calculates the ratio of the total pulse rate during sleep to the total pulse rate during non-sleep based on the pulse rate data, and a sleep state determination unit that determines the sleep state based on the decline rate calculated by the average pulse rate decline rate calculation unit and the ratio calculated by the total pulse rate ratio calculation unit.

[0007] In addition, as one aspect of the present invention, in order to solve the problem of easily determining the driver's sleep state based only on the driver's pulse rate data, the sleep state determination unit may determine the sleep state based on the following determination criteria. Good: D>Td1 and R <Tr1 Normal: D≦Td1 and Tr1≦R <Tr2 Bad: D≦Td2 and Tr1≦R <Tr2 Worst case: D≦Td2 and Tr2≦R Here, each symbol represents the following: D: The rate of decrease in the average pulse rate during sleep compared to the average pulse rate during non-sleep R: Ratio of total pulse rate during sleep to total pulse rate during non-sleep Td1, Td2: Threshold of the rate of decline (Td1>Td2) Tr1,Tr2: Ratio threshold (Tr1 <Tr2)

[0008] Furthermore, as one aspect of the present invention, in order to solve the problem of identifying with high accuracy the decline rate and the ratio that are characteristic of each sleep state classified by the apnea-hypopnea index and improving the reliability of the judgment criteria, a learning dataset consisting of heart rates during non-sleep and sleep and an apnea-hypopnea index (AHI) measured simultaneously with the measurement of the heart rate may be stratified into a plurality of groups corresponding to the sleep states according to the value of the apnea-hypopnea index, and a threshold value of the judgment criteria may be determined based on features extracted from the learning dataset of each group using machine learning.

[0009] The drowsiness prediction system of the present invention comprises a drowsiness prediction device having any of the above-mentioned aspects and a pulse rate measuring means for measuring the pulse rate data in order to solve the problem of creating a safe and reasonable driving plan and preventing traffic accidents caused by drowsiness while driving by accurately determining the driver's sleep state and predicting the risk of drowsiness while driving.

[0010] The drowsiness prediction program of the present invention solves the problem of creating a safe and reasonable operation plan and preventing traffic accidents caused by drowsiness while driving by accurately determining the driver's sleep state and predicting the risk of drowsiness while driving, and causes a computer to function as a pulse rate data acquisition unit that acquires pulse rate data during non-sleep and sleep, an average pulse rate decline rate calculation unit that calculates the decline rate of the average pulse rate during sleep to the average pulse rate during non-sleep based on the pulse rate data, a total pulse rate ratio calculation unit that calculates the ratio of the total pulse rate during sleep to the total pulse rate during non-sleep based on the pulse rate data, and a sleep state determination unit that determines the sleep state based on the decline rate calculated by the average pulse rate decline rate calculation unit and the ratio calculated by the total pulse rate ratio calculation unit.

[0011] The drowsiness prediction method of the present invention solves the problem of creating a safe and reasonable operation plan and preventing traffic accidents caused by drowsiness while driving by accurately determining a driver's sleep state and predicting the risk of drowsiness while driving, and includes the following steps: a pulse rate data acquisition step of acquiring pulse rate data during non-sleep and sleep; an average pulse rate decline rate calculation step of calculating a decline rate of the average pulse rate during sleep to the average pulse rate during non-sleep based on the pulse rate data; a total pulse rate ratio calculation step of calculating the ratio of the total pulse rate during sleep to the total pulse rate during non-sleep based on the pulse rate data; and a sleep state determination step of determining the sleep state based on the decline rate calculated in the average pulse rate decline rate calculation step and the ratio calculated in the total pulse rate ratio calculation step. Effect of the Invention

[0012] According to the present invention, by determining a driver's sleep state with high accuracy and predicting the risk of falling asleep while driving, a safe and reasonable driving plan can be created and traffic accidents caused by falling asleep while driving can be prevented. [Brief description of the drawings]

[0013] [Figure 1] 1 is a block diagram showing an embodiment of a dozing-off prediction system including a dozing-off prediction device according to the present invention; [Diagram 2] 4 is a flowchart showing a dozing-off prediction method executed by the dozing-off prediction device and the dozing-off prediction program of the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an embodiment of a dozing-off prediction device, a dozing-off prediction system, a dozing-off prediction program, and a dozing-off prediction method according to the present invention will be described with reference to the drawings.

[0015] As shown in FIG. 1, the drowsiness prediction system 10 of this embodiment is composed of a pulse rate measuring means 5 for measuring the driver's pulse rate, and a drowsiness prediction device 1 for predicting the risk of drowsiness based on the pulse rate data measured by the pulse rate measuring means 5.

[0016] The pulse rate measuring means 5 measures the pulse rate of the driver. In this embodiment, the pulse rate measuring means 5 is configured with a tracker or the like that is attached to the driver's wrist, and measures the pulse rate when not sleeping and the pulse rate when sleeping. In addition, in the case of a person without arrhythmia, the pulse rate can be treated as being the same as the heart rate.

[0017] The drowsiness prediction device 1 predicts the risk of a driver falling asleep at the wheel. In this embodiment, the drowsiness prediction device 1 is configured by a computer such as a personal computer, a tablet, or a smartphone, and as shown in Fig. 1, mainly includes a display means 2, a storage means 3, and a calculation processing means 4. Each component will be described below.

[0018] The display means 2 displays various display data. In this embodiment, the display means 2 is configured with a touch panel display or the like also having an input function, and displays an operation menu screen, a determination result to be described later, and the like, and also allows the user to input various selection instructions. When using a display means 2 without an input function, a separate input means such as a mouse or a keyboard may be provided.

[0019] The storage means 3 stores various data and functions as a working area when the arithmetic processing means 4 performs arithmetic processing. In this embodiment, the storage means 3 is composed of a solid state drive (SSD), a hard disk drive (HDD), a ROM (Read Only Memory), a RAM (Random Access Memory), a flash memory, etc., and has a program storage unit 31 and a judgment threshold storage unit 32 as shown in FIG.

[0020] A dozing-off prediction program 1a of the present embodiment is installed in the program storage unit 31. Then, the calculation processing means 4 executes the dozing-off prediction program 1a, thereby causing the computer as the dozing-off prediction device 1 to function as each of the components described below.

[0021] The use of the dozing prediction program 1a is not limited to the above configuration. For example, the dozing prediction program 1a may be stored in a non-transitory computer-readable recording medium such as an SD card, a CF card, a USB memory, an external HDD, an external SSD, a CD-ROM, a DVD-ROM, etc., and may be read and executed directly from the recording medium. The dozing prediction program 1a may also be used from an external server or the like using a cloud computing method or an ASP (Application Service Provider) method.

[0022] The judgment threshold storage unit 32 stores judgment thresholds for judging the sleep state of the driver. In this embodiment, when the judgment thresholds are determined, first, the heart rates of several hundred subjects (300 to 500 subjects) are measured during non-sleep and sleep by a Holter electrocardiogram test. In addition, simultaneously with the measurement of the heart rates, the apnea hypopnea index (AHI) is measured by a simple sleep apnea test using an apnoea monitor. Here, the apnea hypopnea index refers to the combined number of apnea (apnea) and hypopnea (hypopnea) per hour.

[0023] Next, the learning data set consisting of the heart rate and apnea / hypopnea index simultaneously measured for each subject is stratified into a plurality of groups according to the value of the apnea / hypopnea index. Here, in this embodiment, the sleep state is classified into the following four types according to the value of the apnea / hypopnea index. Good: AHI<5 Normal: 5≦AHI<15 Poor: 15≦AHI<40 Worst: 40 or less AHI

[0024] Then, feature values ​​are extracted from the learning data sets of each group corresponding to the above four types of sleep states using machine learning, and judgment threshold values ​​are determined based on the feature values. As the judgment threshold values, as described below, a threshold value related to the rate of decrease in the average pulse rate during sleep relative to the average pulse rate during non-sleep, and a threshold value related to the ratio of the total pulse rate during sleep to the total pulse rate during non-sleep are determined.

[0025] In this embodiment, the sleep states are classified into four types according to the value of the apnea / hypopnea index, but the present invention is not limited to this and may be classified into a plurality of types. The value of the apnea / hypopnea index for classifying the sleep states is not limited to the above and may be set appropriately.

[0026] The calculation processing means 4 is composed of a CPU (Central Processing Unit) and the like, and by executing the dozing prediction program 1a installed in the program storage unit 31, as shown in Fig. 1, functions as a pulse rate data acquisition unit 41, an average pulse rate decrease rate calculation unit 42, a total pulse rate ratio calculation unit 43, a sleep state determination unit 44, and a determination result output unit 45. Each component will be described in more detail below.

[0027] The pulse rate data acquisition unit 41 acquires the pulse rate data of the driver. In this embodiment, the pulse rate data acquisition unit 41 acquires pulse rate data when the driver is not sleeping and pulse rate data when the driver is sleeping from the pulse rate measurement means 5 as data corresponding to the driver's heart rate.

[0028] The average pulse rate decrease rate calculation unit 42 calculates the decrease rate of the average pulse rate during sleep relative to the average pulse rate during non-sleep. In this embodiment, the average pulse rate decrease rate calculation unit 42 calculates the decrease rate of the average pulse rate during sleep relative to the average pulse rate during non-sleep based on the pulse rate data acquired by the pulse rate data acquisition unit 41. For example, if the average pulse rate during non-sleep is 70 (beats / min) and the average pulse rate during sleep is 60 (beats / min), the decrease rate is approximately 14%. It is considered that the larger this decrease rate, the better the quality of sleep, and the smaller the decrease rate, the worse the quality of sleep.

[0029] The total pulse rate ratio calculation unit 43 calculates the ratio of the total pulse rate during sleep to the total pulse rate during non-sleep based on the pulse rate data acquired by the pulse rate data acquisition unit 41. It is considered that the smaller this ratio is, the better the quality of sleep is, and the larger this ratio is, the worse the quality of sleep is.

[0030] The sleep state determination unit 44 determines the sleep state based on the pulse rate of the driver. In this embodiment, the sleep state determination unit 44 determines the sleep state based on the decrease rate calculated by the average pulse rate decrease rate calculation unit 42 and the ratio calculated by the total pulse rate ratio calculation unit 43. Specifically, the sleep state determination unit 44 reads out the determination threshold value from the determination threshold value storage unit 32, and determines the sleep state of the driver by applying the calculated decrease rate and ratio to the following determination criteria.

[0031] Good: D>Td1 and R <Tr1 Normal: D≦Td1 and Tr1≦R <Tr2 Bad: D≦Td2 and Tr1≦R <Tr2 Worst case: D≦Td2 and Tr2≦R Here, each symbol represents the following: D: The rate of decrease in the average pulse rate during sleep compared to the average pulse rate during non-sleep R: Ratio of total pulse rate during sleep to total pulse rate during non-sleep Td1, Td2: Threshold of the rate of decline (Td1>Td2) Tr1,Tr2: Ratio threshold (Tr1 <Tr2)

[0032] If the sleep state is "good" as a result of the above-mentioned judgment, the risk of falling asleep while driving is predicted to be low. On the other hand, as the sleep state becomes worse from "normal" to "bad" to "worst", the risk of falling asleep while driving is predicted to increase.

[0033] The judgment result output unit 45 outputs the judgment result regarding the sleep state of the driver. In this embodiment, the judgment result output unit 45 acquires the judgment result judged by the sleep state judgment unit 44 and outputs it to the display means 2. In addition to the judgment result by the sleep state judgment unit 44, the judgment result output unit 45 may output driving advice based on the judgment result.

[0034] For example, if the result of the judgment is "good," the risk of falling asleep at the wheel is low, and therefore advice may be output to the effect that it is OK to drive a relatively long distance that day. On the other hand, if the result of the judgment is "poor," the risk of falling asleep at the wheel is high, and therefore advice may be output to refrain from driving that day, or to limit the distance to a shorter distance if driving is to be performed.

[0035] Next, the actions of the doze prediction device 1, the doze prediction system 10, the doze prediction program 1a, and the doze prediction method of the present embodiment will be described.

[0036] When predicting a driver's dozing off using the dozing off prediction device 1, dozing off prediction system 10, dozing off prediction program 1a, and dozing off prediction method of the present embodiment, first, the pulse rate data acquisition unit 41 acquires pulse rate data when the driver is not asleep and when the driver is asleep from the pulse rate measurement means 5 worn by the driver (step S1: pulse rate data acquisition step). This allows data corresponding to the driver's heart rates when the driver is not asleep and when the driver is asleep to be obtained.

[0037] Next, the average pulse rate decrease rate calculation unit 42 calculates the decrease rate of the average pulse rate during sleep relative to the average pulse rate during non-sleep based on the pulse rate data acquired in the pulse rate data acquisition step (step S1) (step S2: average pulse rate decrease rate calculation step). This makes it possible to grasp the degree to which the average pulse rate during sleep has decreased relative to the average pulse rate during non-sleep.

[0038] Next, the total pulse rate ratio calculation unit 43 calculates the ratio of the total pulse rate during sleep to the total pulse rate during non-sleep based on the pulse rate data acquired in the pulse rate data acquisition step (step S1) (step S3: total pulse rate ratio calculation step). This allows the ratio of the total pulse rate during sleep to the total pulse rate during non-sleep to be understood.

[0039] Next, the sleep state determination unit 44 determines the sleep state based on the decrease rate calculated in the average pulse rate decrease rate calculation step (step S2) and the ratio calculated in the total pulse rate ratio calculation step (step S3) (step S4: sleep state determination step). As a result, the decrease rate and the ratio are taken into consideration comprehensively, so that the sleep state of the driver can be determined with high accuracy and the risk of falling asleep while driving can be predicted.

[0040] In addition, in this embodiment, the sleep state is stratified according to the value of the apnea-hypopnea index, which is closely related to the quality of sleep, and the judgment criteria for judging the sleep state are determined using the decline rate and the ratio. Therefore, the sleep state of the driver can be easily judged based only on the pulse rate data of the driver, and the sleep state accurately reflects the quality of the driver's sleep.

[0041] Furthermore, in this embodiment, the threshold value used for the sleep state determination criterion is determined based on features extracted by machine learning from a learning dataset consisting of the heart rate and the apnea / hypopnea index, so that the decline rate and the ratio, which are characteristic of each sleep state classified by the apnea / hypopnea index, are identified with high accuracy, improving the reliability of the determination criterion.

[0042] Finally, the judgment result output unit 45 outputs the judgment result judged in the sleep state judgment step (step S4) to the display means 2 (step S5). This allows the manager of the transportation company to which the driver belongs to understand the driver's sleep state and the risk of falling asleep, and can use this as a reference when making a safe and reasonable operation plan.

[0043] The doze prediction device 1, the doze prediction system 10, the doze prediction program 1a, and the doze prediction method of the present embodiment as described above provide the following advantages. 1. It is possible to judge the driver's sleep state with a high degree of accuracy and predict the risk of falling asleep while driving. 2. It is possible to create a safe and reasonable driving plan and prevent traffic accidents caused by drowsy driving. 3. The driver's sleep state can be easily determined based solely on the driver's pulse rate data. 4. The rate of decline and the ratio characteristic of each sleep state classified by the apnea / hypopnea index can be identified with high accuracy, thereby improving the reliability of the judgment criteria.

[0044] The doze prediction device 1, the doze prediction system 10, the doze prediction program 1a and the doze prediction method according to the present invention are not limited to the above-described embodiments, and may be modified as appropriate.

[0045] For example, in the above-described embodiment of the dozing prediction system 10, the dozing prediction device 1 and the pulse rate measuring means 5 are configured separately, but the present invention is not limited to this configuration. If a wristwatch-type wearable computer (smart watch), which has become widespread in recent years, is used, the dozing prediction system 10 can be configured with the dozing prediction device 1 and the pulse rate measuring means 5 integrated together. [Explanation of symbols]

[0046] 1. Drowsiness prediction device 1a Drowsiness prediction program 2 Display means 3 Memory means 4. Processing means 5. Pulse rate measuring means 10. Drowsiness prediction system 31 Program memory section 32 Judgment threshold memory unit 41 Pulse rate data acquisition unit 42 Average pulse rate decrease rate calculation unit 43 Total pulse rate ratio calculation unit 44 Sleep state determination unit 45 Judgment result output section

Claims

1. a pulse rate data acquisition unit for acquiring pulse rate data during non-sleep and sleep; an average pulse rate decrease rate calculation unit that calculates a decrease rate of an average pulse rate during sleep relative to an average pulse rate during non-sleep based on the pulse rate data; a total pulse rate ratio calculation unit that calculates a ratio of a total pulse rate during sleep to a total pulse rate during non-sleep based on the pulse rate data; a sleep state determination unit that determines a sleep state based on the decrease rate calculated by the average pulse rate decrease rate calculation unit and the ratio calculated by the total pulse rate ratio calculation unit; A dozing off prediction device comprising:

2. The dozing prediction device according to claim 1 , wherein the sleep state determination unit determines the sleep state based on the following determination criteria: Good: D>Td1 and R<Tr1 Normal: D≦Td1 and Tr1≦R<Tr2 Bad: D≦Td2 and Tr1≦R<Tr2 Worst case: D≦Td2 and Tr2≦R Here, each symbol represents the following: D: The rate of decrease in the average pulse rate during sleep compared to the average pulse rate during non-sleep R: Ratio of total pulse rate during sleep to total pulse rate during non-sleep Td1, Td2: Threshold value of the decrease rate (Td1>Td2) Tr1, Tr2: ratio threshold (Tr1<Tr2)

3. 3. The drowsiness prediction device according to claim 2, wherein a learning data set consisting of heart rates during non-sleep and sleep and an apnea-hypopnea index (AHI) measured simultaneously with the measurement of the heart rates is stratified into a plurality of groups corresponding to the sleep states according to the values ​​of the apnea-hypopnea index, and the threshold value of the judgment criterion is determined based on features extracted from the learning data set of each group using machine learning.

4. 4. A dozing-off prediction system comprising: the dozing-off prediction device according to claim 1; and a pulse rate measuring means for measuring the pulse rate data.

5. a pulse rate data acquisition unit for acquiring pulse rate data during non-sleep and sleep; an average pulse rate decrease rate calculation unit that calculates a decrease rate of an average pulse rate during sleep relative to an average pulse rate during non-sleep based on the pulse rate data; a total pulse rate ratio calculation unit that calculates a ratio of a total pulse rate during sleep to a total pulse rate during non-sleep based on the pulse rate data; a sleep state determination unit that determines a sleep state based on the decrease rate calculated by the average pulse rate decrease rate calculation unit and the ratio calculated by the total pulse rate ratio calculation unit; A drowsy driving prediction program that enables computers to function in this way.

6. a pulse rate data acquiring step of acquiring pulse rate data during non-sleep and sleep; an average pulse rate decrease rate calculation step of calculating a decrease rate of an average pulse rate during sleep relative to an average pulse rate during non-sleep based on the pulse rate data; a total pulse rate ratio calculation step of calculating a ratio of a total pulse rate during sleep to a total pulse rate during non-sleep based on the pulse rate data; a sleep state determination step of determining a sleep state based on the decrease rate calculated in the average pulse rate decrease rate calculation step and the ratio calculated in the total pulse rate ratio calculation step; The method for predicting dozing off comprises the steps of:

Images