Driving quality estimation system, driving quality estimation server, on-vehicle device, and computer program

Abstract

This driving quality estimation system includes: a storage unit that acquires and stores a time series of sensor data output by a sensor pertaining to biological information from the driver of a vehicle; and a driving quality estimation unit that estimates the quality of driving by the driver on the basis of the sensor data from a predetermined period stored in the storage unit. The driving quality estimation unit includes a state number estimation unit for estimating the number of states of the driver on the basis of the sensor data over the predetermined period, and a determination unit for determining the quality of driving by the driver in accordance with the number of states estimated by the state number estimation unit.

Description

Driving quality estimation system, driving quality estimation server, in-vehicle device, and computer program

[0001] This disclosure relates to a driving quality estimation system, a driving quality estimation server, an in-vehicle device, and a computer program. This application claims priority to Japanese Application No. 2024-062792 filed on April 9, 2024, and incorporates by reference all of the contents of said Japanese application.

[0002] One of the main causes of traffic accidents is distracted driving due to fatigue, etc. Traffic accidents caused by distracted driving due to fatigue, etc. account for approximately one-quarter of all traffic accidents. In order to prevent such traffic accidents, it is necessary to accurately detect the driving quality of passengers (drivers), and if there is a change in the driving quality, it is necessary to have the driver take appropriate rest or vacation, or reduce the driver's workload as necessary.

[0003] However, fatigue may differ from the driver's perception. In other words, even if the driver does not feel tired, the body may actually be fatigued, which may result in a decline in driving quality. Therefore, whether or not a driver's driving quality is declining should not be left to the driver's own judgment, but rather an external, objective standard for judging the decline is required.

[0004] Japanese Patent Laid-Open No. 2006-124222 discloses a technology for classifying a driver into one of a number of groups based on biological information detected from the driver in real time. The technology states that the driver's condition can be estimated with high accuracy by selecting an optimal feature for each group.

[0005] JP 2013-164748 A

[0006] A driving quality estimation system according to a certain aspect of this disclosure includes a memory unit that acquires and stores a time series of sensor data output by a sensor related to biometric information of a vehicle occupant, and a driving quality estimation unit that estimates the quality of driving by the occupant based on the sensor data for a predetermined period stored in the memory unit, wherein the driving quality estimation unit includes a state number estimation unit that estimates the number of states of the occupant based on the sensor data for the predetermined period, and a judgment unit that judges the quality of driving by the occupant in accordance with the number of states estimated by the state number estimation unit.

[0007] This disclosure can be realized not only as a driving quality estimation system including such a characteristic processing unit, but also as a driving quality estimation method including such characteristic processing steps, as a computer program for causing a computer to execute such steps, as a semiconductor integrated circuit that realizes some or all of the devices constituting the driving quality estimation system, or as a larger system including the driving quality estimation system.

[0008] FIG. 1 is a block diagram showing a schematic configuration of a fatigue level estimation system, which is an example of a driving quality estimation system according to a first embodiment of the present disclosure. FIG. 2 is a schematic diagram for explaining the principle of driving quality determination in the first embodiment of the present disclosure. FIG. 3 is a schematic diagram for explaining the principle of driving quality estimation in a second embodiment of the present disclosure. FIG. 4 is a block diagram showing the configuration of an in-vehicle device according to the first embodiment. FIG. 5 is a block diagram showing the configuration of a fatigue level determination server shown in FIG. 1. FIG. 6 is a block diagram showing the functional configuration of the in-vehicle system shown in FIG. 1. FIG. 7 is a schematic diagram showing an example of installation of a seating sensor in a vehicle. FIG. 8 is a block diagram showing the functional configuration of the fatigue level determination server shown in FIG. 1. FIG. 9 is a schematic diagram showing an example of sensor placement in the fatigue level estimation system. FIG. 10 is a schematic diagram for explaining a method for estimating the number of states according to the first embodiment of the present disclosure. FIG. 11 is a flowchart showing the control structure of a computer program (hereinafter referred to as the program) executed by a fatigue level determination server, which is an example of the driving quality determination server according to the first embodiment of the present disclosure. FIG. 12 is a graph showing an example of data analysis results immediately after the start of driving. Fig. 13 is a graph showing an example of data analysis results of the seat sensor after a predetermined time has elapsed since the start of driving. Fig. 14 is a schematic diagram showing a method for estimating the number of valid states according to a second embodiment of the present disclosure. Fig. 15 is a block diagram showing the functional configuration of an in-vehicle system according to a third embodiment of the present disclosure.

[0009] [Problem to be Solved by the Present Disclosure] In the technology disclosed in Patent Document 1, examples of biometric information used include information on eye movement, line of sight, eyelid opening, heart rate, and body movement. To acquire this information, it is necessary to have the driver wear a sensor or to install a camera in the vehicle. However, it is not realistic for employers to have drivers wear sensors. Furthermore, with conventional technology, there is a problem in that the processing of sensor data needs to be changed when the type of sensor changes. It would be preferable to be able to estimate the quality of a driver's driving from sensor data regardless of the type of sensor, but the technology in Patent Document 1 cannot meet such a demand.

[0010] Therefore, an object of this disclosure is to provide a driving quality estimation system, a driving quality estimation server, an in-vehicle device, and a computer program that are independent of the type of sensor.

[0011] [Effects of the Present Disclosure] With this configuration, it is possible to provide a driving quality estimation system, a driving quality estimation server, an in-vehicle device, and a computer program that are independent of the type of sensor.

[0012] [Description of the embodiments of the present disclosure] In the following description and drawings, the same components are denoted by the same reference numerals. Therefore, detailed description thereof will not be repeated. Note that at least some of the embodiments described below may be combined in any manner.

[0013] (1) A driving quality estimation system according to a first aspect of this disclosure includes a storage unit that acquires and stores a time series of sensor data output by a sensor related to biometric information of a vehicle occupant, and a driving quality estimation unit that estimates the quality of driving by the occupant based on the sensor data for a predetermined period stored in the storage unit, wherein the driving quality estimation unit includes a state number estimation unit that estimates a number of states of the occupant based on the sensor data for the predetermined period, and a determination unit that determines the quality of driving by the occupant in accordance with the number of states estimated by the state number estimation unit. With this configuration, the quality of driving by the occupant can be determined based on the number of states estimated from the occupant's biometric information, regardless of the type of sensor. As a result, a driving quality estimation system that is independent of the type of sensor can be provided.

[0014] (2) In the above (1), the sensor may include a seating sensor disposed in a seat of the vehicle. With this configuration, the quality of driving by a passenger can be determined without requiring the passenger to wear a special sensor.

[0015] (3) In the above (2), the seat sensor may include a plurality of pressure sensors, and the time series of the sensor data may be represented by a plurality of vectors whose elements are the outputs of the plurality of pressure sensors at each measurement point. With this configuration, the quality of driving by a passenger can be determined by numerically processing the sensor data without requiring the passenger to wear a special sensor.

[0016] (4) In the above (3), the number-of-states estimation unit may include a plurality of clustering units that cluster the plurality of vectors over the predetermined period into a plurality of cluster groups each having a different number of clusters, and an estimation unit that estimates the number of states by evaluating the plurality of cluster groups obtained by the plurality of clustering units. With this configuration, the number of states of a passenger can be estimated from sensor data using an existing clustering method.

[0017] (5) In the above (4), each of the plurality of cluster numbers may be an integer greater than or equal to 2 and less than or equal to a predetermined upper limit. With this configuration, the number of states of the passenger can be estimated within a certain range.

[0018] (6) In the above (5), the upper limit may be equal to or greater than 3 and equal to or less than 16. With this configuration, the number of states of the passenger can be estimated within a certain limited range.

[0019] (7) In the above (4), the plurality of clustering units may include a plurality of K-means clustering units that cluster the sensor data into different numbers of clusters using a K-means clustering method. With this configuration, the number of states of a passenger can be efficiently estimated using an existing K-means clustering method.

[0020] (8) In the above (4), the estimation unit may include an evaluation value calculation unit that calculates a predetermined evaluation value for each of the plurality of cluster groups obtained by the plurality of clustering units, an eligible cluster group selection unit that selects, from the plurality of cluster groups, an eligible cluster group whose evaluation value calculated by the evaluation value calculation unit satisfies a predetermined condition, and a cluster number confirmation unit that estimates the number of states by confirming the number of clusters included in each of the eligible cluster groups selected by the eligible cluster group selection unit. With this configuration, a highly reliable clustering result can be used, and the quality of driving by a passenger can be determined with high reliability.

[0021] (9) In the above (4), the estimation unit may include an evaluation value calculation unit that calculates a predetermined evaluation value for each of the plurality of cluster groups obtained by the plurality of clustering units, and a qualified cluster group selection unit that selects, from the plurality of cluster groups, a qualified cluster group whose evaluation value calculated by the evaluation value calculation unit satisfies a predetermined condition, and the determination unit may determine whether the occupant is fatigued based on whether the number of qualified cluster groups selected by the cluster group selection unit exceeds the number of previous qualified cluster groups. With this configuration, it is possible to determine from the clustering result that the number of states of the occupant has changed and that driving quality has deteriorated.

[0022] (10) In the above (8), the determination unit may include a first determination unit that determines that the occupant is fatigued in response to the number of clusters confirmed by the cluster number confirmation unit being all greater than a threshold value, and a second determination unit that determines whether the occupant is fatigued in accordance with whether any of the numbers of clusters confirmed by the cluster number confirmation unit is equal to or less than the threshold value and whether the number of qualified cluster groups exceeds the number of previous qualified cluster groups. This configuration makes it possible to determine the occupant's condition with high accuracy. As a result, the quality of the occupant's driving can be estimated with high reliability.

[0023] (11) In the above (4), the number-of-states estimating unit may further include a contracting unit that contracts data of a data block made up of the plurality of vectors over the predetermined period prior to clustering by the plurality of clustering units. With this configuration, the amount of calculation required to estimate driving quality is reduced, and processing can be performed quickly.

[0024] (12) A second aspect of the present disclosure provides an in-vehicle device that includes any one of the driving quality estimation systems described above. As a result, the in-vehicle device can determine the driving quality of a passenger based on the number of states estimated from the passenger's biometric information, regardless of the type of sensor. As a result, an in-vehicle device that can estimate driving quality without depending on the type of sensor can be provided.

[0025] (13) A driving quality estimation server according to a third aspect of this disclosure includes: a storage unit that acquires, via communication, a time series of sensor data output by a sensor related to biometric information of a vehicle occupant from an on-board device mounted on the vehicle and stores the acquired data; a driving quality estimation unit that estimates the driving quality of the occupant based on the sensor data for a predetermined period stored in the storage unit; and a processing unit that executes a predetermined process to address the occupant's fatigue in response to a determination that the driving quality of the occupant estimated by the driving quality estimation unit satisfies a predetermined condition. The driving quality estimation unit includes a state number estimation unit that estimates a number of states of the sensor data for the predetermined period, and a determination unit that determines the driving quality of the occupant according to the number of states estimated by the state number estimation unit. With this configuration, the driving quality of each occupant can be determined based on the state number estimated from the biometric information of the occupant of each vehicle, regardless of the type of sensor. As a result, a driving quality estimation server that estimates the driving quality of each occupant can be provided, regardless of the type of sensor.

[0026] (14) A computer program according to a fourth aspect of this disclosure causes a computer to function as a storage unit that acquires and stores a time series of sensor data output by a sensor related to biometric information of a vehicle occupant, and a driving quality estimation unit that estimates the quality of driving by the occupant based on the sensor data for a predetermined period stored in the storage unit, wherein the driving quality estimation unit includes a state number estimation unit that estimates a number of states of the sensor data for the predetermined period, and a determination unit that determines the quality of driving by the occupant based on the number of states estimated by the state number estimation unit. With this configuration, the quality of driving by each occupant can be determined based on the number of states estimated from the biometric information of each vehicle occupant, regardless of the type of sensor. As a result, a computer program that estimates the quality of driving by each occupant can be provided, regardless of the type of sensor.

[0027] (15) A computer program according to a fifth aspect of the present disclosure causes a computer to function as: a storage unit that acquires, via communication, a time series of sensor data output by a sensor related to biometric information of a vehicle occupant from an on-board device mounted on the vehicle and stores the acquired data; a driving quality estimation unit that estimates a driving quality of the occupant based on the sensor data for a predetermined period stored in the storage unit; and a processing unit that executes a predetermined process to suppress a decline in the driving quality of the occupant in response to a determination that the driving quality of the occupant estimated by the driving quality estimation unit satisfies a predetermined condition, wherein the driving quality estimation unit includes: a state number estimation unit that estimates a number of states of the sensor data for the predetermined period; and a determination unit that determines the driving quality of the occupant in accordance with the state number estimated by the state number estimation unit. With this configuration, the state number of each occupant can be estimated based on biometric information of the occupant of each vehicle received from each external vehicle, regardless of the type of sensor, and the driving quality of each occupant can be determined based on the state number. As a result, a computer program can be provided that estimates the quality of driving by each occupant without relying on the type of sensor.

[0028] [Details of the embodiment of the present disclosure] 1 First embodiment 1.1 Configuration Fig. 1 shows a schematic configuration of a fatigue level estimation system 50 according to a first embodiment of the present disclosure, which estimates a driver's fatigue level as an index of the quality of driving by a vehicle occupant (driver). Referring to Fig. 1, the fatigue level estimation system 50 includes a vehicle 60, a fatigue level determination server 62, and a driver management system 64. The vehicle 60, the fatigue level determination server 62, and the driver management system 64 are capable of communicating with each other via a network 66.

[0029] The vehicle 60 is equipped with an in-vehicle system 68 for controlling various systems inside the vehicle 60 and for communicating with the driver management system 64. The in-vehicle system 68 includes an in-vehicle device 80 that has the function of collecting and storing the output of a seating sensor (not shown in FIG. 1 ) provided in the vehicle 60 and periodically transmitting the output to the fatigue level determination server 62.

[0030] The fatigue level determination server 62 receives and stores a time series of output from the seating sensor from the in-vehicle device 80 and estimates the driver's fatigue level at regular intervals. If the fatigue level determination server 62 determines that the driver is fatigued, it notifies the driver management system 64 of this fact. In other words, the fatigue level determination server 62 determines that the quality of driving has deteriorated due to the driver's fatigue and notifies the driver management system 64 of this fact.

[0031] In response to the notification from the fatigue level determination server 62, the driver management system 64 notifies the administrator that a fatigued driver is present. The administrator may encourage the problematic driver to take a break or arrange for a replacement driver to restore the driver's driving quality. A system may be constructed that performs these processes automatically without the intervention of an administrator.

[0032] In this first embodiment, a method for estimating the driver's fatigue level will be described below. In this disclosure, the driver's fatigue is estimated using the law of increasing entropy. According to the law of increasing entropy, the entropy of an orderly system is low, and the entropy of a disordered system is high. The entropy of an isolated system only changes in an increasing direction. If a driver is considered to be an isolated system, the entropy of the driver increases as the driver becomes fatigued. Therefore, it is believed that the driver's fatigue level can be known by knowing the change in the driver's entropy.

[0033] That is, in this disclosure, it is assumed that when a driver drives continuously, the entropy of the driver system increases. In other words, in this disclosure, it is assumed that the order of the driver system is lost. When the order of the driver system is lost, the quality of the driver's driving is also thought to decline. That is, fatigue accumulates in the driver, or the driver's attention becomes distracted, increasing the likelihood of careless driving. In this disclosure, based on this assumption, the entropy of the driver system is estimated, and when it is thought that the entropy has increased beyond a certain level, it is determined that the driver is in a fatigued state.

[0034] Here, the problem is how to estimate the entropy of the driver system. In this disclosure, the driver's entropy is estimated as follows. Generally, when the entropy is low, the number of microscopic states that the driver system can take is small, and as the entropy increases, the number of microscopic states that the driver system can take increases. Therefore, in this disclosure, the number of microscopic states that the driver system can take is estimated by some means, and the magnitude of the driver's entropy is estimated based on the result.

[0035] That is, referring to the left side of Fig. 2, when the number of micro states taken by the driver 100 is constant and small as shown by state group 102, it is determined that the driver 100 is not fatigued, as shown by icon 104. Conversely, referring to the right side of Fig. 2, when the number of micro states taken by the driver 100 is large as shown by state group 112, it is determined that the driver 100 is fatigued, i.e., the quality of driving by the driver 100 is deteriorating, as shown by icon 114.

[0036] 3, it may be determined that the driver 100 is fatigued when the number of micro states that the driver 100 is in increases, such as when the number of micro states that the driver 100 is in changes from only state group 102 to two states as shown by state group 120 and state group 122. The first embodiment described below uses the method shown in FIG.

[0037] That is, in this disclosure, the macro state of the driver 100, whether or not the driver is fatigued, is determined based on the number of micro states of the driver 100.

[0038] The problem with this approach is how to estimate the number of microscopic states the driver is in. In this disclosure, the number of such states is estimated by acquiring some kind of driver movement using a sensor. Various types of sensors are possible. Various biosensors used in conventional technologies can be used, such as those that acquire information on eye movement, line of sight, eyelid opening, heart rate, and body movement. In this embodiment, a seat sensor, which is a type of biosensor, is used in consideration of the burden on the driver. The seat sensor has the function of simultaneously measuring the pressure distribution caused by the driver's weight on various parts of the seat at multiple points at each measurement time and outputting the pressure distribution in the form of a vector of measurement values. Therefore, the seat sensor can obtain a time series of measurement vectors consisting of the measurement vectors at each measurement time. This time series of measurement vectors is considered to represent one characteristic of the driver's body movement.

[0039] - Hardware Configuration of In-Vehicle Device 80 Fig. 4 shows the hardware configuration of the in-vehicle device 80. Referring to Fig. 4, the in-vehicle device 80 is essentially a computer, and includes a control unit 150 that controls the entire in-vehicle device 80, an auxiliary storage device 158 that stores various data, an in-vehicle communication unit 154 that communicates with other in-vehicle devices via an in-vehicle network (not shown), and a communication unit 156 that communicates with an external wireless device (not shown) installed in the vehicle. The control unit 150, the auxiliary storage device 158, the in-vehicle communication unit 154, and the communication unit 156 are all connected to a communication bus 152, and data exchange between them is performed via the communication bus 152.

[0040] The control unit 150 includes a CPU (Central Processing Unit) 180, a ROM (Read-Only Memory) 182 that stores a boot-up program for the in-vehicle device 80, and a RAM (Random Access Memory) 184 that can be written to and read from as needed. The CPU 180 includes, for example, a CPU or an MPU (Micro Processing Unit) as a computing element (processor). The auxiliary storage device 158 includes, for example, a non-volatile memory such as a flash memory. The ROM 182 or the auxiliary storage device 158 stores software (programs) executed by the CPU 180 and various information (data).

[0041] The in-vehicle communication unit 154 communicates with other in-vehicle devices via an in-vehicle network in accordance with a predetermined communication protocol. The in-vehicle network may be any of a Controller Area Network (CAN), a Local Interconnect Network (LIN), a Media Oriented Systems Transport (MOST), a FlexRay, 10BASE-T1S, a 100BASE-T1, a 1000BASE-T1, a Clock Extension Peripheral Interface (CXPI), an Automotive Safety Restraints Bus (ASRB), and a CAN XL. In this embodiment, although not particularly limited, a CAN is used as the in-vehicle network.

[0042] Under the control of the control unit 150, the in-vehicle device 80 receives sensor data output by each in-vehicle sensor via the in-vehicle network and stores the data in the RAM 184. The in-vehicle device 80 periodically transmits the sensor data stored in the RAM 184 to the fatigue level determination server 62 shown in Fig. 1. The communication unit 156 provides communication with the fatigue level determination server 62 via an external wireless device not shown in Fig. 4.

[0043] 5 , the fatigue level assessment server 62 includes a control unit 200, an auxiliary storage device 204, and a communication unit 206. The control unit 200 includes a CPU 230, a GPU (Graphics Processing Unit) 232, a ROM 234, and a RAM 236. The control unit 200, the auxiliary storage device 204, and the communication unit 206 are all connected to a communication bus 202, and data exchange between them is performed via the communication bus 202.

[0044] The auxiliary storage device 204 includes a non-volatile storage device such as a flash memory or a hard disk drive. The auxiliary storage device 204 stores various information and programs executed by the CPU 230. The communication unit 206 provides communication with other devices, including the in-vehicle device 80, via the network 66.

[0045] The fatigue level determination server 62 communicates with the in-vehicle device 80 (see FIG. 1 ) via the network 66 and the communication unit 206. The fatigue level determination server 62 acquires data for estimating the fatigue level of the driver of the vehicle 60 in which the in-vehicle device 80 is installed from the in-vehicle device 80. The fatigue level determination server 62 estimates the fatigue level of the driver of the vehicle 60 based on the acquired data, and notifies an administrator of the driver's fatigue level when the estimated fatigue level exceeds a predetermined threshold. Note that instead of or in addition to notifying the driver's fatigue level to the driver management system 64, the fatigue level determination server 62 may also notify the in-vehicle device 80 of the vehicle 60 driven by the driver.

[0046] A program for causing the fatigue level determination server 62 to function as each functional unit according to this embodiment is stored and distributed in a predetermined storage medium such as a DVD (Digital Versatile Disc) or a USB (Universal Serial Bus) memory, and is then transferred from this medium to the auxiliary storage device 204. Alternatively, the program may be transmitted to the fatigue level determination server 62 from an external device via the network 66 and stored in the auxiliary storage device 204.

[0047] - Functional Configuration of In-Vehicle System 68 and In-Vehicle Device 80 Fig. 6 shows the functional configuration of an in-vehicle system 68 equipped with an in-vehicle device 80. Referring to Fig. 6, the in-vehicle system 68 includes, in addition to the in-vehicle device 80, a seat occupancy sensor 252 as a biometric sensor, and an in-vehicle navigation system 254. The seat occupancy sensor 252, the in-vehicle device 80, and the in-vehicle navigation system 254 are connected to each other via an in-vehicle network 250 so as to be able to communicate with each other. The in-vehicle device 80 can also communicate with various ECUs (Electronic Control Units) via the in-vehicle network 250.

[0048] 7, the seating sensor 252 is a body pressure (pressure) sensor provided on the seat surface of the driver's seat 350 inside the vehicle 60. The seating sensor 252 measures the seating pressure at each point, which indicates the distribution of the body pressure of the driver seated in the seat 350, at predetermined measurement intervals and outputs the measured data as sensor data. In this embodiment, the sensor data output by the seating sensor 252 during a given measurement is in a vector format with seating pressure data as elements.

[0049] 6 , in-vehicle navigation system 254 includes a route guidance unit 302 that executes processing for route guidance, a display device 304, an audio output unit 306, and a control unit 300 that controls these to assist the driver. Control unit 300 is connected to in-vehicle network 250.

[0050] The main functional components of the in-vehicle device 80 are as follows: That is, the functions realized by the in-vehicle device 80 include a communication function by the communication unit 156, a sensor data receiving unit 280 that receives sensor data from in-vehicle sensors such as the seat occupancy sensor 252 via the in-vehicle network 250, and a sensor data storage unit 282 that stores the sensor data received by the sensor data receiving unit 280. The functions realized by the in-vehicle device 80 further include a sensor data transmitting unit 284 that executes a process of summarizing the sensor data stored in the sensor data storage unit 282 for each predetermined period and transmitting the summaries to the fatigue level determination server 62 via the communication unit 156, and a warning notification unit 286 that notifies the driver of a warning, such as a driver fatigue warning received by the communication unit 156 from an external source, using the in-vehicle navigation system 254.

[0051] - Functional Configuration of Fatigue Level Determination Server 62 Fig. 8 shows the functional configuration of the fatigue level determination server 62. Referring to Fig. 8, the functions realized by the fatigue level determination server 62 include a timer 380, a fatigue level processing unit 382, ​​and a warning notification unit 384 in addition to the communication function of the communication unit 206.

[0052] The fatigue level processing unit 382 includes a sensor data acquisition unit 400 , a sensor data accumulation unit 402 , a state number estimation unit 404 , a state number increase determination unit 406 , a state number storage unit 410 , and a fatigue level determination unit 408 .

[0053] The sensor data acquisition unit 400 acquires sensor data from the in-vehicle device 80 of each vehicle 60. The sensor data accumulation unit 402 accumulates the sensor data acquired by the sensor data acquisition unit 400 for each vehicle (driver). The number-of-states estimation unit 404 reads sensor data for each vehicle (driver) for a predetermined period immediately preceding the period from the sensor data accumulation unit 402 at predetermined intervals in accordance with the timing of the timer 380, and estimates the number of micro states of the driver for that period. The number-of-states storage unit 410 stores the number of micro states estimated by the number-of-states estimation unit 404. The number-of-states increase determination unit 406 determines whether the number of micro states estimated by the number-of-states estimation unit 404 has increased from the number of micro states estimated by the previous processing, which is stored in the number-of-states storage unit 410. The fatigue level determination unit 408 determines whether a predetermined condition, that the fatigue level of the target driver has exceeded a threshold, is satisfied, based on the number of micro states estimated by the number of states estimation unit 404 and the determination result by the number of states increase determination unit 406. For a driver determined to satisfy this condition, the warning notification unit 384 executes a predetermined process to address the driver's fatigue, for example, a process of notifying a warning to the driver management system 64. The warning notification unit 384 may directly notify the in-vehicle device 80 of the warning. In this case, it is considered that the control unit 300 shown in FIG. 6 uses the display device 304 and the audio output unit 306 to warn the driver.

[0054] - Seat occupancy sensor 252 Fig. 9 shows an example of the arrangement of pressure sensors in the seat occupancy sensor 252. Fig. 9 is merely one example, and for example, the seat occupancy sensor 252 may be rectangular and the pressure sensors may be arranged in a matrix. Alternatively, the number of pressure sensors may be less or more.

[0055] As shown in FIG. 9, each pressure sensor of the seat occupancy sensor 252 is assigned a number to distinguish it from the others. The vector output by the seat occupancy sensor 252 includes elements in which the outputs of the pressure sensors are arranged in order according to the numbers. In the following description, the number of pressure sensors is assumed to be m. A plurality of vectors measured over a certain period of time (e.g., 30 minutes) form one data block. Note that the order in which the numbers are assigned is not limited to that shown in FIG. 9.

[0056] 10 is a diagram showing a clustering analysis process 440 for estimating the number of effective states in this embodiment. This process is executed in the number-of-states estimation unit 404 shown in FIG.

[0057] 10 , assume that seat occupancy sensor 252 outputs sensor data at n measurement times T1, T2, ..., Tn during the predetermined period described above. As described above, a measurement vector having m sensor outputs as elements is obtained at each measurement time. The elements of this measurement vector are P1, P2, ..., Tm. That is, for each predetermined period, a data block 450 consisting of n x m pieces of data is processed by number-of-states estimation unit 404 ( FIG. 8 ).

[0058] In this embodiment, the clustering analysis process 440 includes a K-means clustering unit 452 for elements belonging to the data block 450. In this embodiment, the K-means clustering method is used for clustering. This clustering is performed by the K-means clustering method while changing the number of clusters from 2 to L (L-1 ways). L is an integer greater than 2, for example, an integer greater than 6 and less than a predetermined upper limit (for example, 16). Of course, the upper limit value may be changed depending on the computing capacity of the fatigue level determination server 62, the number of vehicles to be processed by the fatigue level determination server 62, the number of pressure sensors in the seating sensor 252 ( FIG. 6 ), etc.

[0059] Clustering by the K-means clustering unit 452 produces L-1 different clustering results, with the number of clusters ranging from 2 to L. In this specification, a collection of multiple clusters obtained by each clustering is referred to as a cluster group obtained by that clustering. In other words, L-1 cluster groups are obtained by the K-means clustering unit 452. Note that the clustering process performed by the K-means clustering unit 452 in this manner can be executed in parallel. Furthermore, the process is purely numerical processing. Therefore, by having the GPU 232 (FIG. 5) execute these processes, efficient processing can be achieved.

[0060] The clustering analysis process 440 further includes a score calculation unit 454 that calculates a score, which is a predetermined evaluation value, for each of the L-1 cluster groups obtained by the K-means clustering unit 452. Any score that evaluates the results of clustering may be used as this score. For example, scores obtained by silhouette analysis, the elbow method, gap statistics, the Davies-Bouldin method, the Calinski-Harabasz method, etc. may be used. In this embodiment, a silhouette score obtained by silhouette analysis is used as the score. The method for calculating the silhouette score is publicly known, and its value falls between -1 and 1. The closer the silhouette score is to 1, the higher the clustering evaluation. The calculation formula for the silhouette score in this example is as follows, with the silhouette score being SCORE:

[0061] SCORE = (b - a) / max (a, b) where, in this example, b is the distance between the data belonging to cluster i and the center of cluster a, and a is the distance between this data and the cluster closest to cluster i among the other clusters. Various definitions of distance can be used. For example, the distance between data and a cluster can be the distance between the data and the center of the cluster. Alternatively, the distance between data x and data y, which is the closest data to data x among the data belonging to the target cluster, can be used, or conversely, the distance between data z, which is the farthest from data x, can be used. Furthermore, the average distance between data x and all data belonging to the other cluster can be used. Note that the scores shown in FIG. 10 are for reference only.

[0062] In this embodiment, for each cluster in a cluster group obtained by a certain clustering, the score calculation unit 454 calculates the silhouette score of the data belonging to that cluster. The score calculation unit 454 further calculates the score of that cluster group by averaging the silhouette scores across all cluster groups obtained as a result of the certain clustering. For example, in FIG. 10, the score of the cluster group with the number of clusters = 2 is SCORE1 = 0.705. The score of the cluster group with the number of clusters = 3 is SCORE2 = 0.598. The same applies below.

[0063] The clustering analysis process 440 further includes an average score calculation unit 456 that calculates the average (mean score) of the scores obtained for all cluster groups by the score calculation unit 454, a comparison process 462 that compares the score of each cluster group with the average score, and a qualified cluster group selection unit 458 that selects cluster groups determined to have silhouette scores higher than the average score as a result of the comparison process 462. In the example shown in FIG. 10 , two cluster groups, one with two clusters and the other with four clusters, are selected by the qualified cluster group selection unit 458. In this disclosure, the cluster groups selected by the qualified cluster group selection unit 458 are referred to as qualified cluster groups, in the sense that they are cluster groups that satisfy predetermined criteria. Note that, in calculating the average score in the average score calculation unit 456, the number of data belonging to each cluster may be weighted, or the average score may be calculated by simply dividing the sum of the scores of each cluster by the number of clusters.

[0064] The clustering analysis process 440 further includes a cluster number confirmation component 460 that confirms the number of eligible cluster groups selected by the eligible cluster group selection component 458 and the number of clusters within each cluster group.

[0065] FIG. 11 shows, in flowchart form, the control structure of a program for implementing fatigue level processing unit 382 of FIG. 8 in cooperation with computer hardware. This program is started periodically (e.g., every 30 minutes) by timer 380 of FIG. 8 for each target in-vehicle device. Referring to FIG. 11 , this program includes step 480 of acquiring sensor data for a period to be analyzed from the sensor data accumulated in sensor data accumulation unit 402 ( FIG. 8 ) as a data block consisting of sensor data from the target in-vehicle device for the immediately preceding 30 minutes. This program further includes step 482 of performing the clustering analysis outlined in FIG. 10 on the acquired data block.

[0066] The program further includes step 484, which functions as a first determination unit that branches the flow of control depending on whether the number of clusters contained in each of the eligible cluster groups obtained as a result of the clustering analysis process 440 in step 482 exceeds a threshold value.

[0067] In this disclosure, each cluster in each cluster group contains similar data that differs from the data in other clusters. These data represent the output of the seat occupancy sensor. Therefore, the data belonging to each cluster is considered to indicate that the driver's state is approximately the same at the time each data is acquired. In other words, the number of clusters is considered to represent the order of the driver's state. A small number of clusters indicates order and low entropy, while a large number of clusters indicates disorder and high entropy. When entropy is low, the driver is not fatigued, and when entropy is high, the driver is determined to be fatigued. The "threshold" described in step 484 is a reference state number for determining whether the driver is fatigued. It is desirable to use different values ​​for this threshold depending on the type of work the target driver performs, the type of vehicle driven, the vehicle's equipment, etc.

[0068] From the above explanation, if the determination in step 484 is positive, it corresponds to the state shown on the right side of Figure 2, and it can be determined that the driver is fatigued. If the determination in step 484 is negative, it means that there is at least a possibility that the driver is not fatigued.

[0069] 11 further includes step 486, which functions as a second determination unit that branches the flow of control depending on whether the number of eligible cluster groups exceeds the number of eligible cluster groups obtained by the previous process when the determination in step 484 is negative. The situation in which the number of eligible cluster groups has increased from the previous situation means that the situation has changed from the situation shown on the left side of FIG. 3 to the situation shown on the right side. In other words, it is unclear whether the driver's state is maintaining order, as indicated by state group 120, or is losing order, as indicated by state group 122. As a result, if the determination in step 486 is positive, it is safe to assume that the driver is fatigued. On the other hand, if the determination in step 486 is negative, it can be determined that the driver is not yet fatigued.

[0070] This program further includes step 488, which is executed when the determination in step 484 is positive, and when the determination in step 484 is negative and the determination in step 486 is positive, to determine that the driver's fatigue is accumulating and the quality of the driver's driving is deteriorating, and to notify the driver management system 64 (see FIG. 1) of a warning regarding the driver, and step 490, which is executed after step 488 and when the determination in step 486 is negative, to store the number of eligible cluster groups calculated in the current process in state number storage unit 410 shown in FIG. 8 as the state number of driver 352, and then terminate execution of this program. The value saved in state number storage unit 410 is referenced in step 486 the next time this program is executed.

[0071] 1.2 Operation The fatigue level estimation system 50, the structure of which has been described above, operates as follows: The in-vehicle device 80 and the fatigue level determination server 62 are assumed to be pre-programmed to perform predetermined operations.

[0072] 7, when the driver 352 sits in the seat 350, the seating sensor 252 starts outputting sensor data. The output indicates the pressure distribution caused by the driver 352, as shown in FIG. 9, and is output in vector format at predetermined time intervals. The vector format allows various numerical processing to be applied to the sensor data.

[0073] 6 , the sensor data output by the seat occupancy sensor 252 is received by the sensor data receiving unit 280 via the in-vehicle network 250. The sensor data receiving unit 280 stores the received sensor data in the sensor data storage unit 282, along with a data reception time stamp. The sensor data transmitting unit 284 periodically reads the sensor data from the sensor data storage unit 282, attaches identification information of the in-vehicle device 80 to the sensor data, and transmits the data to the fatigue level determination server 62 via the communication unit 156 and the exterior wireless device 82.

[0074] Referring to Figure 8, the sensor data acquisition unit 400 of the fatigue level determination server 62 receives this sensor data via the communication unit 206 and stores it in the sensor data storage unit 402 (Figure 8) together with the address of the exterior wireless device 82 (Figure 6) that sent the sensor data or the identification information of the in-vehicle device 80.

[0075] The seating sensor 252 (FIG. 6), the in-vehicle device 80 (FIGS. 4 and 6), and the fatigue level processing unit 382 of the fatigue level determination server 62 (FIG. 8) repeat this process.

[0076] Meanwhile, the fatigue level determination server 62 launches a program whose control structure is shown in Fig. 11 at regular intervals (e.g., every 30 minutes). For example, the number-of-states estimation unit 404 reads and acquires a data block consisting of sensor data from a specific vehicle for the 30 minutes immediately before the program is launched from the sensor data accumulation unit 402 (step 480 in Fig. 11). The number-of-states estimation unit 404 executes the clustering analysis process 440 shown in Fig. 10 on this data block. As a result, one or more eligible cluster groups are selected.

[0077] If the number of states contained in each of the selected eligible cluster groups exceeds the threshold (the determination in step 484 is positive), a warning is issued, indicating that the driver's fatigue is accumulating and the quality of driving is deteriorating (step 488). Even if the determination in step 484 is negative, if the number of eligible cluster groups selected this time is greater than the number of eligible cluster groups selected the previous time (the determination in step 486 is positive), a warning is issued in the same way, since the driver may be fatigued. If the number of selected eligible cluster groups has not increased, this process ends without issuing a warning.

[0078] The fatigue level determination server 62 repeats the above-described process at predetermined time intervals for each vehicle that is using the service of the fatigue level determination server 62 .

[0079] 12 and 13 show an example of the relationship between the number of clusters and the score during clustering in this first embodiment. Fig. 12 shows an example of the clustering result for sensor data up to 30 minutes after the driver starts driving. Fig. 13 shows an example of the clustering result for sensor data from one hour after the driver starts driving to another hour later (two hours after starting driving).

[0080] In Figures 12 and 13, the number of clusters during clustering is calculated from 2 to 15, and the horizontal axis represents the number of clusters, while the vertical axis represents the score of the cluster group when clustered to that number of clusters.The relationship between the number of clusters and the score is shown in graph 520 (Figure 12) and graph 530 (Figure 13), and the average score is shown in graph 522 (Figure 12) and graph 532 (Figure 13).

[0081] 12, in this example, the score is highest when the number of clusters is 2, followed by the scores decreasing in the order of 4 and 3. These three cases are the only ones that exceed the average score and are selected as eligible cluster groups, and the entropy is low. If the threshold is set to, for example, "4," the determination in step 484 in FIG. 11 will be negative. Since a comparison with the previous result is not possible, the determination in step 486 will be negative, and it can be determined that the driver is not yet fatigued.

[0082] In the example shown in Fig. 13, the cluster groups with cluster numbers of 2, 3, 4, 5, 11, 13, 14, and 15 have low scores overall and are selected as eligible cluster groups. If the threshold value were set to "4" as in Fig. 12, the determination in step 484 would be negative. However, in the case of Fig. 13, the probability that the determination in step 484 will be positive is high, and therefore it is determined that there is a high possibility that the driver is fatigued.

[0083] As described above, the fatigue level estimation system 50 according to this embodiment estimates the microscopic number of states of the driver using the output of a sensor, such as a seating sensor, that obtains biological information from the driver. The driver's fatigue level is determined based on the estimated number of states. As a result, by using an appropriate sensor, it is possible to determine a fatigue level that is not apparent from the driver's appearance and that the driver himself is not even aware of. A sensor that can collect necessary information without being worn directly by the driver, such as a seating sensor, is desirable. However, the fatigue level estimation system 50 according to this embodiment is not limited to a seating sensor and can also use other sensors. This is because the system targets the abstract state of the driver represented by the sensor data, rather than the value of the sensor data itself, and therefore introduces the concept of clustering sensor data. Therefore, the driver's fatigue can be flexibly determined based on sensor data, regardless of the type of sensor.

[0084] Second Embodiment In the first embodiment, the clustering analysis process 440 in FIG. 10 compares the score of each cluster group with the overall score, and selects cluster groups with higher scores as eligible cluster groups. However, this disclosure is not limited to such an embodiment. In this second embodiment, eligible cluster groups are selected by taking into account all the scores of each cluster within each cluster group.

[0085] 14, a clustering analysis process 550 in the second embodiment includes a K-means clustering unit 452 that performs processing on a data block 450 similar to that of the clustering analysis process 440 of Fig. 10. The clustering analysis process 550 differs from the clustering analysis process 440 in that it includes a score calculation unit 560 that calculates all the scores of each cluster in each cluster group, instead of the score calculation unit 454 of Fig. 10.

[0086] Score calculation unit 454 calculates only one score for each cluster group. However, score calculation unit 560 calculates a score for each cluster in each cluster group. That is, if there are two clusters, two scores are calculated; if there are three clusters, three scores are calculated; and so on.

[0087] The clustering analysis process 550 further includes an average score calculation unit 564 that calculates the average score of the scores calculated for each cluster group, instead of the average score calculation unit 456 of the clustering analysis process 440 shown in Fig. 10. In this average score calculation unit 564 as well, a weighted average of the scores may be calculated using the number of data belonging to each cluster, or a number obtained by simply dividing the sum of the scores of each cluster by the total number of clusters may be used.

[0088] 10, the clustering analysis process 550 further includes a comparison process 562 that compares the score of each cluster with the average score calculated by an average score calculation section 564. The clustering analysis process 550 further includes a qualified cluster group selection section 566 that selects, as a qualified cluster group, a cluster group in which all clusters have scores higher than the average score as a result of the comparison by the comparison process 562, and a cluster number confirmation section 568 that confirms the number of qualified cluster groups selected by the qualified cluster group selection section 566 and the number of clusters belonging to the qualified cluster group.

[0089] In the example shown in FIG. 14 , all of the scores of clusters belonging to a cluster group with two clusters exceed the average score. On the other hand, of the scores of clusters belonging to a cluster group with three clusters, two exceed the average score, but one is below the average score. Furthermore, for a cluster group with four clusters, three exceed the average score, but one is below the average score. If the same applies when the number of clusters is four or more, only cluster groups with two clusters are selected as eligible cluster groups by the eligible cluster group selection unit 566. In this case, the number of eligible cluster groups is one, and the number of clusters is two.

[0090] The second embodiment is realized by replacing the clustering analysis process 440 according to the first embodiment with this clustering analysis process 550.

[0091] The operation of the second embodiment is the same as that of the first embodiment, except for the clustering analysis process 550. The processing of the clustering analysis process 550 is also clear from the above description, and therefore will not be repeated here.

[0092] The second embodiment also provides the same effects as the first embodiment.

[0093] Third Embodiment In both the first and second embodiments, the output of the seat occupancy sensor is stored as a vector, and multiple vectors obtained over a given time period are processed as a single data block. However, this disclosure is not limited to such an embodiment. In this third embodiment, preprocessing is performed on the sensor data from the seat occupancy sensor to reduce the data for clustering.

[0094] Fig. 15 shows a functional block diagram of an in-vehicle system 600 according to the third embodiment. The in-vehicle system 600 differs from the in-vehicle system 68 shown in Fig. 6 in that, instead of the in-vehicle device 80 shown in Fig. 6, the in-vehicle system 600 includes an in-vehicle device 610 having a function of performing a predetermined reduction process on the sensor data received from the seating sensor 252 by the sensor data receiving unit 280, storing the data, and periodically transmitting the reduced sensor data to a fatigue level determination server.

[0095] In-vehicle device 610, in in-vehicle device 80 shown in Figure 6, further includes a sensor data reduction unit 620 that is provided at the output of sensor data receiving unit 280 and the input of sensor data storage unit 282 and has the function of performing a predetermined reduction process on the sensor data received by sensor data receiving unit 280 and storing the data in sensor data storage unit 282.

[0096] The data reduction performed by the sensor data reduction unit 620 may be as follows.

[0097] (1) The sensor surface of the seat occupancy sensor is divided into several parts, and the average pressure for each part is used as sensor data. This configuration reduces the number of components of the vector that is the sensor data at each measurement point.

[0098] (2) Quantize the output of each pressure sensor of the seat sensor. For example, if the output of each pressure sensor is a real value, quantize the value into, for example, 16 levels, 128 levels, or 256 levels, and use the quantized value as an element of the sensor data vector. This may be combined with (1).

[0099] By reducing the sensor data in this manner, the following effects can be obtained. First, the amount of data transmitted from the in-vehicle system 68 to the fatigue level assessment server 62 can be reduced. As a result, the time required to transmit data from the in-vehicle system 68 to the fatigue level assessment server 62 can be shortened. Second, when clustering is performed after data reduction, the amount of calculation required for clustering can be reduced. As a result, the calculation load on the fatigue level assessment server 62 can be reduced. Note that such data reduction processing and interpretation of data content by clustering have some commonality with various known data processing methods, such as image processing methods, and techniques used in such techniques can also be applied to the technology of this disclosure. Furthermore, in the above embodiment, the driver's fatigue level is estimated based on the output of the seat sensor as an indicator of the driver's driving quality. By using the seat sensor, there is an effect that information for estimating driving quality can be obtained without requiring the driver to wear a dedicated sensor. However, this disclosure is not limited to such an embodiment. As an indicator of the quality of driving by the driver, information on eye movement, line of sight, eyelid opening, heart rate, and body movement may be obtained by having the driver wear a dedicated sensor for each, or by installing a camera in the vehicle.

[0100] Each process (each function) in the above-described embodiments is realized by a processing circuit (circuitry) including one or more processors. The processing circuit may be configured by an integrated circuit that combines one or more memories, various analog circuits, and various digital circuits in addition to the one or more processors. The one or more memories store programs (instructions) that cause the one or more processors to execute each of the processes. The one or more processors may execute each of the processes according to the program read from the one or more memories, or according to a logic circuit designed in advance to execute each of the processes. The processor may be any of various processors suitable for computer control, such as a CPU, GPU, DSP (Digital Signal Processor), FPGA (Field-Programmable Gate Array), ASIC (Application Specific Integrated Circuit), and DPU (Data Processing Unit). The physically separated processors may cooperate with each other to execute the processes. For example, the processors installed in the physically separated computers may cooperate with each other via a network such as a LAN (Local Area Network), a WAN (Wide Area Network), or the Internet to execute the processes. The program may be installed into the memory from an external server device or the like via the network, or may be distributed in a state stored on a recording medium such as a CD-ROM (Compact Disc Read-Only Memory), a DVD-ROM (Digital Versatile Disc Read-Only Memory), or a semiconductor memory, and then installed into the memory from the recording medium.

[0101] The embodiments disclosed herein are merely examples, and the present disclosure is not limited to the above-described embodiments. The scope of the present disclosure is defined by the claims in the scope of the claims, taking into consideration the description of the detailed description of the invention, and includes all modifications within the meaning and scope equivalent to the wording described therein.

[0102] 50 Fatigue level estimation system 60 Vehicle 62 Fatigue level determination server 64 Driver management system 66 Network 68, 600 In-vehicle system 80, 610 In-vehicle device 82 Exterior wireless device 100, 352 Driver 102, 112, 120 Status group 104, 114 Icon 150, 200, 300 Control unit 152, 202 Communication bus 154 In-vehicle communication unit 156, 206 Communication unit 158, 204 Auxiliary storage device 180, 230 CPU 182, 234 ROM 184, 236 RAM 232 GPU 250 In-vehicle network 252 Seat occupancy sensor 254 In-vehicle navigation system 280 Sensor data receiving unit 282, 402 Sensor data storage unit 284 Sensor data transmitting unit 286, 384 Warning notification unit 302 Route guidance unit 304 Display device 306 Audio output unit 350 Seat 380 Timer 382 Fatigue level processing unit 400 Sensor data acquisition unit 404 State number estimation unit 406 State number increase determination unit 408 Fatigue level determination unit 410 State number storage unit 440, 550 Clustering analysis process 450 Data block 452 K-means clustering unit 454, 560 Score calculation unit 456, 564 Average score calculation unit 458, 566 Eligible cluster group selection unit 460, 568 Cluster number confirmation unit 462, 562 Comparison process 520, 522, 530, 532 Graph 620 Sensor data reduction unit

Claims

1. A driving quality estimation system comprising: a memory unit that acquires and stores a time series of sensor data output by a sensor related to the biometric information of a vehicle occupant; and a driving quality estimation unit that estimates the quality of driving by the occupant based on the sensor data for a predetermined period stored in the memory unit, wherein the driving quality estimation unit comprises: a state number estimation unit that estimates the number of states of the occupant based on the sensor data for the predetermined period; and a judgment unit that judges the quality of driving by the occupant in accordance with the number of states estimated by the state number estimation unit.

2. The driving quality estimation system according to claim 1, wherein the sensor includes an occupancy sensor disposed in a seat of the vehicle.

3. The driving quality estimation system of claim 2, wherein the seating sensor includes a plurality of pressure sensors, and the time series of the sensor data is represented by a plurality of vectors whose elements are the outputs of the plurality of pressure sensors at each measurement point.

4. The driving quality estimation system of claim 3, wherein the state number estimation unit includes: a plurality of clustering units that cluster the plurality of vectors over the predetermined period into a plurality of cluster groups, each of which has a different number of clusters; and an estimation unit that estimates the number of states by evaluating the plurality of cluster groups obtained by the plurality of clustering units.

5. The driving quality estimation system according to claim 4, wherein the number of clusters is an integer greater than or equal to 2 and less than or equal to a predetermined upper limit.

6. The driving quality estimation system according to claim 5, wherein the upper limit is 3 or more and 16 or less.

7. The driving quality estimation system according to claim 4, wherein the plurality of clustering units include a plurality of K-means clustering units that cluster the sensor data into different numbers of clusters using the K-means method.

8. A driving quality estimation system as described in claim 4, wherein the estimation unit includes: an evaluation value calculation unit that calculates a predetermined evaluation value for each of the plurality of cluster groups obtained by the plurality of clustering units; a qualified cluster group selection unit that selects, from the plurality of cluster groups, a qualified cluster group whose evaluation value calculated by the evaluation value calculation unit satisfies a predetermined condition; and a cluster number confirmation unit that estimates the number of states by confirming the number of clusters included in each of the qualified cluster groups selected by the qualified cluster group selection unit.

9. The driving quality estimation system of claim 4, wherein the estimation unit includes: an evaluation value calculation unit that calculates a predetermined evaluation value for each of the plurality of cluster groups obtained by the plurality of clustering units; and an eligible cluster group selection unit that selects, from the plurality of cluster groups, an eligible cluster group whose evaluation value calculated by the evaluation value calculation unit satisfies a predetermined condition; and the determination unit determines whether the passenger is fatigued based on whether the number of eligible cluster groups selected by the cluster group selection unit exceeds the number of previous eligible cluster groups.

10. The driving quality estimation system of claim 8, wherein the judgment unit includes: a first judgment unit that judges that the occupant is fatigued in response to the number of clusters confirmed by the cluster number confirmation unit being all greater than a threshold value; and a second judgment unit that judges whether the occupant is fatigued in accordance with whether any of the number of clusters confirmed by the cluster number confirmation unit is less than or equal to the threshold value and whether the number of eligible cluster groups exceeds the number of previous eligible cluster groups.

11. A driving quality estimation system as described in claim 4, wherein the state number estimation unit further includes a contraction unit that contracts data of a data block consisting of the plurality of vectors over the predetermined period prior to clustering by the plurality of clustering units.

12. An in-vehicle device equipped with the driving quality estimation system according to any one of claims 1 to 11.

13. A driving quality estimation server comprising: a memory unit that acquires via communication from an on-board device mounted on the vehicle and stores a time series of sensor data output by a sensor related to the biometric information of a vehicle occupant; a driving quality estimation unit that estimates the quality of driving by the occupant based on the sensor data for a predetermined period stored in the memory unit; and a processing unit that executes predetermined processing to deal with changes in the quality of driving by the occupant in response to the quality of driving by the occupant estimated by the driving quality estimation unit satisfying predetermined conditions, wherein the driving quality estimation unit comprises: a state number estimation unit that estimates the number of states of the sensor data for the predetermined period; and a judgment unit that judges the quality of driving by the occupant in accordance with the number of states estimated by the state number estimation unit.

14. A computer program that causes a computer to function as: a memory unit that acquires and stores time series of sensor data output by sensors related to the biometric information of a vehicle occupant; and a driving quality estimation unit that estimates the quality of driving by the occupant based on the sensor data for a predetermined period stored in the memory unit, wherein the driving quality estimation unit includes: a state number estimation unit that estimates the number of states of the sensor data for the predetermined period; and a judgment unit that judges the quality of driving by the occupant in accordance with the number of states estimated by the state number estimation unit.

15. A computer program that causes a computer to function as: a memory unit that acquires and stores time series of sensor data output by sensors related to the biometric information of a vehicle occupant from an on-board device installed in the vehicle via communication; a driving quality estimation unit that estimates the quality of driving by the occupant based on the sensor data for a predetermined period stored in the memory unit; and a processing unit that executes predetermined processing to deal with changes in the quality of driving by the occupant in response to the occupant's driving quality estimated by the driving quality estimation unit satisfying predetermined conditions, wherein the driving quality estimation unit includes: a state number estimation unit that estimates the number of states of the sensor data over the predetermined period; and a judgment unit that judges the occupant's driving quality in accordance with the number of states estimated by the state number estimation unit.

Images