Concentration level determination device and concentration level determination method

Abstract

A concentration level determination device (1) comprises: a feature quantity calculation unit (11) for calculating a first feature quantity (X) indicating an inter-waveform relationship between a head motion waveform representing the motion of the head of a person (A) and an eyeball motion waveform representing the motion of the eyeballs of the person (A), and a second feature quantity (Y) indicating an inter-waveform relationship different from that of the first feature quantity (X); and a concentration level determination unit (12) for determining the concentration level (E) of the person (A), on the basis of the first feature quantity (X) and the second feature quantity (Y).

Description

Concentration level determination device and concentration level determination method

[0001] The present disclosure relates to a concentration level determination device and a concentration level determination method.

[0002] There are known technologies for assessing the concentration state of a vehicle driver or other operator (hereinafter referred to as a driver, etc.). For example, Patent Document 1 describes a concentration assessment device for assessing the concentration level of a vehicle driver, etc., on visual information. This concentration assessment device measures eye movement, head movement, and visual field fluctuation to detect reflex eye movements, including the vestibulo-ocular reflex and the optokinetic reflex. Here, the vestibulo-ocular reflex is a compensatory behavior that moves the eyes in response to head movement so as to prevent blurring of an image on the retina. The optokinetic reflex is a compensatory behavior that moves the eyes in response to movement of the entire visual field so as to prevent blurring of an image on the retina. The concentration assessment device then detects reflex eye movement data, including eye movement, head position, head posture, and visual field fluctuation, and estimates the concentration level of the driver, etc., based on the detected reflex eye movement detection results.

[0003] Patent No. 5030142

[0004] In the conventional technology described in Patent Document 1, there are differences in various characteristics between the data representing head movement and the data representing eye movement, which are used to estimate the concentration level of a driver, etc. For this reason, it is difficult to appropriately design an index of concentration level that is sensitive to the characteristics of interest, and there is a problem in that it is not possible to improve the accuracy of determining the concentration level.

[0005] The present disclosure is intended to solve the above-mentioned problem, and aims to provide a concentration level determination device that can improve the accuracy of determining a person's concentration level.

[0006] The concentration level determination device according to the present disclosure includes a feature calculation unit that calculates a first feature indicating the relationship between a head movement waveform representing the movement of a person's head and an eye movement waveform representing the movement of a person's eye, and a second feature indicating the relationship between the waveforms different from the first feature, and a concentration level determination unit that determines the concentration level of a person based on the first feature and the second feature.

[0007] According to the present disclosure, since the first feature amount and the second feature amount have mutually different relationships between the waveforms of the head movement waveform representing the movement of the person's head and the eye movement waveform representing the movement of the person's eye, it is possible to determine a person's concentration level from these relationships, taking into account the differences in various mutually different features between the waveforms. As a result, the concentration level determination device according to the present disclosure can improve the accuracy of determining a person's concentration level.

[0008] FIG. 1 is a block diagram showing the configuration of a concentration level determination device according to embodiment 1. FIG. 2 is a block diagram showing the configuration of a concentration level determination device according to embodiment 1 and various waveform data. FIG. 3 is a diagram showing a two-dimensional feature space in which feature points having first feature amounts and second feature amounts as elements are plotted. FIG. 4 is a flowchart showing a concentration level determination method according to embodiment 1. FIG. 5 is a diagram showing a two-dimensional feature space in which multiple feature points are plotted. FIGS. 6A and 6B are diagrams showing an example of the distribution of multiple feature points in the feature space. FIG. 6A is a graph showing the relationship between the evaluation value of the concentration level over time. FIG. 9A is a diagram showing the relationship between the radial transition amount and the circumferential transition amount of multiple feature points. FIG. 9B is a block diagram showing a hardware configuration for realizing the functions of the concentration level determination device according to embodiment 1.

[0009] Embodiment 1. Fig. 1 is a block diagram showing the configuration of a concentration level determination device 1 according to embodiment 1. Fig. 2 is a block diagram showing the configuration of the concentration level determination device 1 and various waveform data. In Figs. 1 and 2, the concentration level determination device 1 determines the concentration level E of, for example, a vehicle driver (hereinafter referred to as "person A"). The concentration level determination device 1 includes a feature amount calculation unit 11 and a concentration level determination unit 12.

[0010] The feature calculation unit 11 calculates a first feature X indicating the relationship between a head movement waveform representing the movement of person A's head and an eye movement waveform representing the movement of person A's eye, and a second feature Y indicating the relationship between the waveforms different from the first feature X. The feature calculation unit 11 includes a first low-pass filter 111, a second low-pass filter 112, a phase shifter 113, a first correlation calculator 114, and a second correlation calculator 115. Hereinafter, the low-pass filter will be abbreviated as "LPF." Note that the feature calculation unit 11 may not include the first LPF 111 and the second LPF 112, and may instead acquire waveform signals from the first LPF 111 and the second LPF 112 provided outside the concentration level determination device 1.

[0011] First LPF 111 is a filter that performs so-called "low-pass filtering," which passes waveform signals in a frequency band below a cutoff frequency. In concentration level assessment device 1, a head movement waveform signal is input to first LPF 111. The head movement waveform signal is a signal indicative of head movement waveform B, which indicates the movement of the head of person A. First LPF 111 performs low-pass filtering on the head movement waveform signal, thereby passing the head movement waveform signal from which noise has been removed. The head movement waveform signal from which noise has been removed is output from first LPF 111 to phase shifter 113 and first correlation calculator 114.

[0012] The head movement of person A includes the person A's head shaking or head up and down movement. The head movement waveform signal is detected by a measurement device provided separately from the concentration assessment device 1. The measurement device may be a measurement device equipped with a radio wave sensor. In this measurement device, changes in distance information to the head continuously detected by the radio wave sensor are measured as a waveform signal indicating head movement. The measurement device may also be a measurement device equipped with a camera. In this measurement device, changes in head position between captured images of person A captured by the camera are identified through image analysis and measured as a waveform signal indicating head movement. The head movement waveform signal is acquired as a vibration waveform signal following the head movement. Note that person A's head does not necessarily move with the same period, and the waveform following this movement does not necessarily form a clean periodic waveform. However, although the head movement waveform is a periodic waveform, the period often varies over time. Therefore, in the following description, it is assumed that the head movement waveform is a periodic waveform such as a sine wave.

[0013] The second LPF 112 is a filter that performs low-pass filtering, similar to the first LPF 111. In the concentration determination device 1, an eye movement waveform signal is input to the second LPF 112. The eye movement waveform signal is a signal indicative of eye movement waveform C that indicates the movement of the eyeballs of person A. The second LPF 112 performs low-pass filtering on the eye movement waveform signal, thereby passing the eye movement waveform signal from which noise has been removed. The eye movement waveform signal from which noise has been removed is output from the second LPF 112 to the first correlation calculator 114 and the second correlation calculator 115.

[0014] The eye movement waveform signal, like the head movement waveform signal, is detected by a measurement device provided separately from the concentration level determination device 1. The measurement device may be a measurement device equipped with an infrared camera. This measurement device, for example, measures changes in brightness of the eyeballs continuously photographed by the infrared camera as a waveform signal indicating the eyeball movement. The eye movement waveform signal is acquired as a vibration waveform signal following the eyeball movement. Person A's eyeball movement changes according to the movement of his or her head. Since the head movement waveform signal is a periodic waveform as described above, hereinafter, it is assumed that the eye movement waveform is a periodic waveform such as a sine wave. Note that the first LPF 111 and the second LPF 112 may be any filter capable of removing noise contained in the waveform signal, and may be, for example, a band-pass filter.

[0015] Phase shifter 113 shifts the phase of the input signal. For example, phase shifter 113 receives the head movement waveform signal low-pass filtered by first LPF 111 and shifts the phase of head movement waveform B indicated by the input head movement waveform signal by π / 2. Phase shifter 113 outputs head movement waveform D, which has been phase-shifted by π / 2, to second correlation calculator 115.

[0016] First correlation calculator 114 acquires from first LPF 111 a head movement waveform signal indicating head movement waveform B from which noise has been removed by low-pass filtering, and acquires from second LPF 112 an eye movement waveform signal indicating eye movement waveform C from which noise has been removed by low-pass filtering. First correlation calculator 114 performs correlation calculation processing by adding and multiplying head movement waveform B and eye movement waveform C, thereby calculating a first feature quantity X. The correlation calculation processing is processing to calculate a correlation coefficient that indicates the relationship between the two waveforms.

[0017] For example, first correlation calculator 114 performs addition to overlap head movement waveform B and eye movement waveform C on the time axis, and inverts one of the waveforms. Next, first correlation calculator 114 multiplies the overlapping portions of the two waveforms at each time and integrates the overlapping portions to calculate first feature quantity X, which is a correlation coefficient. First feature quantity X calculated by first correlation calculator 114 is a correlation coefficient that indicates the relationship between head movement waveform B and eye movement waveform C, and is output from first correlation calculator 114 to concentration level determination unit 12.

[0018] Second correlation calculator 115 acquires from phase shifter 113 a head movement waveform signal indicating head movement waveform D phase-shifted by π / 2, and acquires from second LPF 112 an eye movement waveform signal indicating eye movement waveform C from which noise has been removed by low-pass filtering. Similar to first correlation calculator 114, second correlation calculator 115 performs correlation calculation processing by adding and multiplying eye movement waveform C and head movement waveform D to calculate second feature amount Y. Second feature amount Y calculated by second correlation calculator 115 is a correlation coefficient indicating the relationship between waveforms different from first feature amount X in head movement waveform B and eye movement waveform C, and is output from second correlation calculator 115 to concentration level determination unit 12.

[0019] The concentration level determination unit 12 determines the concentration level E of person A based on the first feature amount X and the second feature amount Y. For example, the concentration level determination unit 12 calculates feature points using the first feature amount X and the second feature amount Y as elements, and determines the concentration level E of person A based on the transition of the feature points plotted in a two-dimensional feature space.

[0020] 3 is a diagram showing a two-dimensional feature space in which feature points having elements of a first feature amount X and a second feature amount Y are plotted. For example, the feature amount calculation unit 11 calculates the first feature amount X1 and the second feature amount Y1 using a head movement waveform B and an eye movement waveform C acquired at time t1, and calculates the first feature amount X2 and the second feature amount Y2 using a head movement waveform B and an eye movement waveform C acquired at time t2.

[0021] The concentration level determination unit 12 plots a feature point p1, whose elements are the first feature amount X1 and the second feature amount Y1, and a feature point p2, whose elements are the first feature amount X2 and the second feature amount Y2, in a circular area in the feature space indicated by the dashed line in Fig. 3. The above circular area is a circular area with a radius of "1" centered at the origin O in a two-dimensional feature space whose axes are the first feature amount X and the second feature amount Y. The concentration level determination unit 12 determines the concentration level E of person A based on the transition of feature points p1 and p2 in the above circular area. In the following description, feature points p1, p2, etc. will be collectively referred to as feature points p as appropriate.

[0022] Next, a specific example of the concentration level determination method according to the first embodiment will be described. FIG. 4 is a flowchart showing the concentration level determination method according to the first embodiment, illustrating a series of processes for determining a concentration level by the concentration level determination device 1. In FIG. 4, step ST1 shows a feature amount calculation process for calculating a first feature amount X and a second feature amount Y, and step ST2 shows a concentration level determination process using the first feature amount X and the second feature amount Y. Furthermore, a series of processes from step ST11 to step ST14 is a specific example of step ST1, and processes from step ST15 to step ST16 is a specific example of step ST2.

[0023] First, the first LPF 111 acquires a head movement waveform signal from the measurement device, and the second LPF 112 acquires an eye movement waveform signal from the measurement device (step ST11). Next, the first LPF 111 performs noise removal on the head movement waveform signal by low-pass filtering, and the second LPF 112 performs noise removal on the eye movement waveform signal by low-pass filtering (step ST12).

[0024] Next, phase shifter 113 generates head movement waveform D by shifting the phase of head movement waveform B indicated by the noise-removed head movement waveform signal by π / 2 (step ST13). The waveform signal indicating head movement waveform D is output from phase shifter 113 to second correlation calculator 115.

[0025] The first correlation calculator 114 calculates a correlation coefficient using the head movement waveform B and the eye movement waveform C to generate a first feature quantity X, which is a correlation coefficient. The second correlation calculator 115 calculates a correlation coefficient using the eye movement waveform C and the head movement waveform D to generate a second feature quantity Y, which is a correlation coefficient (step ST14). For example, the first correlation calculator 114 adds the head movement waveform B and the eye movement waveform C to overlap them on the time axis and inverts the eye movement waveform C. Next, the first correlation calculator 114 identifies a portion where the two waveforms overlap at each time and multiplies the head movement waveform B by the eye movement waveform C at the identified portion. The first correlation calculator 114 calculates a correlation coefficient as the first feature quantity X by integrating the portion where the two waveforms overlap at each time. The first feature quantity X is output from the first correlation calculator 114 to the concentration level determination unit 12. Similar to the first correlation calculator 114, the second correlation calculator 115 calculates a correlation coefficient as a second feature amount Y by adding and multiplying the eye movement waveform C and the head movement waveform D. The second feature amount Y is output from the second correlation calculator 115 to the concentration level determination unit 12. The series of processes up to this point constitutes the feature amount calculation process.

[0026] Next, the concentration level determination unit 12 generates a feature point p having elements of the first feature amount X acquired from the first correlation calculator 114 and the second feature amount Y acquired from the second correlation calculator 115, and plots the generated feature point p in a feature space (step ST15). FIG. 5 illustrates a two-dimensional feature space in which multiple feature points p are plotted. In FIG. 5 , the feature point p is plotted in a circular region with a radius of "1" and a center at the origin O in the two-dimensional feature space with the first feature amount X and the second feature amount Y as axes. For example, the concentration level determination unit 12 converts the first feature amount X and the second feature amount Y, which are elements of the feature point p, into coordinate values ​​of a circular region with a radius of "1" and a center at the origin O, thereby plotting the feature point p in the circular region. When ideal two waveforms with inverted phases are input, the feature point is theoretically plotted at a point (-1, 0) in the feature space. Therefore, it is expected that the feature points p will be concentrated when the concentration level E is high. However, depending on the influence of individual differences among person A, feature points p may be concentrated at a position slightly shifted from point (-1, 0). Hereinafter, point (-1, 0) when ideal two waveforms are input is taken as the reference point on the circumference of the circular region in feature space.

[0027] (Regarding feature points p(1-1) to p(1-5)) Feature points p(1-1), p(1-2), p(1-3), p(1-4), and p(1-5) are plotted in area E1 in the feature space. Feature point p(1-1) is a feature point p(X, Y) whose elements are the first feature amount X and the second feature amount Y, and which was generated using head movement waveform signal B(t(1-1)) indicating head movement waveform B at time t(1-1) and eye movement waveform signal C(t(1-1)) indicating eye movement waveform C at time t(1-1). Similarly, feature point p(1-2) is a feature point p(X, Y) having as its elements a first feature amount X and a second feature amount Y, and is generated using head movement waveform signal B(t(1-2)) indicating head movement waveform B at time t(1-2) and eye movement waveform signal C(t(1-2)) indicating eye movement waveform C at time t(1-2). Feature point p(1-3) is a feature point p(X, Y) having as its elements a first feature amount X and a second feature amount Y, and is generated using head movement waveform signal B(t(1-3)) indicating head movement waveform B at time t(1-3) and eye movement waveform signal C(t(1-3)) indicating eye movement waveform C at time t(1-3). Feature point p(1-4) is a feature point p(X, Y) having elements of a first feature amount X and a second feature amount Y, generated using a head movement waveform signal B(t(1-4)) indicating head movement waveform B at time t(1-4) and an eye movement waveform signal C(t(1-4)) indicating eye movement waveform C at time t(1-4). Feature point p(1-5) is a feature point p(X, Y) having elements of a first feature amount X and a second feature amount Y, generated using a head movement waveform signal B(t(1-5)) indicating head movement waveform B at time t(1-5) and an eye movement waveform signal C(t(1-5)) indicating eye movement waveform C at time t(1-5). Time progresses from time t(1-1) to time t(1-2), time t(1-3), time t(1-4), and time t(1-5).

[0028] As shown in FIG. 5 , when feature points p(1-1) to p(1-5) are plotted near point (-1,0), these feature points p exhibit small radial and circumferential transitions. For example, at time t(1-1), when feature point p(1-1) is plotted near point (-1,0), there is high symmetry between the waveforms of the time-series changes of head movement waveform B indicated by head movement waveform signal B(t(1-1)) and eye movement waveform C indicated by eye movement waveform signal C(t(1-1)). This is also true for feature points p(1-2) to p(1-5). When two symmetrical waveforms with inverted phases are input, theoretically, the value of X is calculated to be -1 and the value of Y is calculated to be 0. On the other hand, if the symmetry of the input waveform is lost, the position at which feature point p is plotted shifts from point (-1,0), i.e., shifts in the radial or circumferential direction, and ideal symmetry is lost. In this way, when two contrasting waveforms are input, ideally, the characteristic point p is plotted near the reference point (-1, 0), and the more the symmetry of the input waveform is lost, the more the characteristic point p shifts in the radial or circumferential direction.

[0029] (Regarding feature points p(2-1) to p(2-5)) Feature points p(2-1), p(2-2), p(2-3), p(2-4), and p(2-5) are plotted in area F1 in the feature space. Feature point p(2-1) is a feature point p(X, Y) whose elements are the first feature amount X and the second feature amount Y, and which is generated using head movement waveform signal B(t(2-1)) indicating head movement waveform B at time t(2-1) and eye movement waveform signal C(t(2-1)) indicating eye movement waveform C at time t(2-1). Similarly, feature point p(2-2) is a feature point p(X, Y) having as its elements a first feature amount X and a second feature amount Y, and is generated using head movement waveform signal B(t(2-2)) indicating head movement waveform B at time t(2-2) and eye movement waveform signal C(t(2-2)) indicating eye movement waveform C at time t(2-2). Feature point p(2-3) is a feature point p(X, Y) having as its elements a first feature amount X and a second feature amount Y, and is generated using head movement waveform signal B(t(2-3)) indicating head movement waveform B at time t(2-3) and eye movement waveform signal C(t(2-3)) indicating eye movement waveform C at time t(2-3). Feature point p(2-4) is a feature point p(X, Y) having as its elements a first feature amount X and a second feature amount Y, and is generated using head movement waveform signal B(t(2-4)) indicating head movement waveform B at time t(2-4) and eye movement waveform signal C(t(2-4)) indicating eye movement waveform C at time t(2-4). Feature point p(2-5) is a feature point p(X, Y) having as its elements a first feature amount X and a second feature amount Y, and is generated using head movement waveform signal B(t(2-5)) indicating head movement waveform B at time t(2-5) and eye movement waveform signal C(t(2-5)) indicating eye movement waveform C at time t(2-5). Time advances from time t(1-1) to time t(2-1), from time t(1-2) to time t(2-2), from time t(1-3) to time t(2-3), from time t(1-4) to time t(2-4), and from time t(1-5) to time t(2-5). Also, time advances from time t(2-1) to time t(2-2), time t(2-3), time t(2-4), and time t(2-5).

[0030] As shown by the bold arrows in FIG. 5 , feature points p(2-1) to p(2-5) are plotted at positions radially shifted from point (-1,0). For example, feature point p(2-1) is located at a position radially shifted from point (-1,0). This indicates that there is low symmetry between the waveforms of the time-series changes at time t(2-1) between the head movement waveform B indicated by head movement waveform signal B(t(2-1)) and the eye movement waveform C indicated by eye movement waveform signal C(t(2-1)), and that feature point p has shifted in the radial direction of the feature space. When two contrasting waveforms are input in this way, feature point p is ideally plotted near the reference point (-1,0), and the loss of symmetry may cause feature point p to shift in the radial direction. The same is true for feature points p(2-2) to p(2-5).

[0031] (Regarding feature points p(3-1) to p(3-5)) Feature points p(3-1), p(3-2), p(3-3), p(3-4), and p(3-5) are plotted in area G1 in the feature space. Feature point p(3-1) is a feature point p(X, Y) whose elements are a first feature amount X and a second feature amount Y, and which is generated using head movement waveform signal B(t(3-1)) indicating head movement waveform B at time t(3-1) and eye movement waveform signal C(t(3-1)) indicating eye movement waveform C at time t(3-1). Similarly, feature point p(3-2) is a feature point p(X, Y) having as its elements a first feature amount X and a second feature amount Y, and is generated using head movement waveform signal B(t(3-2)) indicating head movement waveform B at time t(3-2) and eye movement waveform signal C(t(3-2)) indicating eye movement waveform C at time t(3-2). Feature point p(3-3) is a feature point p(X, Y) having as its elements a first feature amount X and a second feature amount Y, and is generated using head movement waveform signal B(t(3-3)) indicating head movement waveform B at time t(3-3) and eye movement waveform signal C(t(3-3)) indicating eye movement waveform C at time t(3-3). Feature point p(3-4) is a feature point p(X, Y) having as its elements a first feature amount X and a second feature amount Y, and is generated using head movement waveform signal B(t(3-4)) indicating head movement waveform B at time t(3-4) and eye movement waveform signal C(t(3-4)) indicating eye movement waveform C at time t(3-4). Feature point p(3-5) is a feature point p(X, Y) having as its elements a first feature amount X and a second feature amount Y, and is generated using head movement waveform signal B(t(3-5)) indicating head movement waveform B at time t(3-5) and eye movement waveform signal C(t(3-5)) indicating eye movement waveform C at time t(3-5). Time advances from time t(1-1) to time t(3-1), from time t(1-2) to time t(3-2), from time t(1-3) to time t(3-3), from time t(1-4) to time t(3-4), and from time t(1-5) to time t(3-5). Also, time advances from time t(3-1) to time t(3-2), time t(3-3), time t(3-4), and time t(3-5).

[0032] As shown by the thin arrows in FIG. 5 , feature points p(3-1) to p(3-5) are plotted at positions shifted circumferentially from the point (−1,0). For example, feature point p(3-1) is located at a position shifted circumferentially from the point (−1,0). This indicates that there is low symmetry between the waveforms of the time-series changes between the head movement waveform B indicated by head movement waveform signal B(t(3-1)) and the eye movement waveform C indicated by eye movement waveform signal C(t(3-1)) at time t(3-1), and that feature point p has shifted circumferentially in the feature space. The same is true for feature points p(3-2) to p(3-5). When two contrasting waveforms are input in this way, feature point p would ideally be plotted near the reference point (−1,0), and the loss of symmetry may cause feature point p to shift circumferentially.

[0033] The concentration level determination unit 12 determines the concentration level E of person A based on the feature point p plotted in the feature space (step ST16). For example, the concentration level determination unit 12 determines the concentration level E of person A based on the coordinate transition amount of the feature point p in the two-dimensional feature space.

[0034] A specific example (1) of determining the concentration level E will be explained using Fig. 5. The concentration level determination unit 12 identifies the relationship between the head movement waveform B and the eye movement waveform C as follows, based on the distribution of feature points p(1-1) to p(1-5) plotted near point (-1, 0), which is area E1 shown in Fig. 5. Then, the concentration level determination unit 12 determines the concentration level E of person A based on the relationship between the identified waveforms.

[0035] The multiple feature points p plotted near the point (-1, 0) have substantially the same first feature X value, but substantially zero second feature Y value. The second feature Y is obtained by performing a correlation calculation process between the head movement waveform B and the eye movement waveform C that has been phase-shifted by π / 2, while the first feature X is obtained by performing a correlation calculation process between the head movement waveform B and the non-phase-shifted eye movement waveform C. Therefore, when feature points p are plotted near the point (-1, 0), where the first feature X is the main element, the relationship between the head movement waveform B and the eye movement waveform C indicated by these feature points p indicates that there is a high degree of symmetry between the waveforms of the time-series changes of these waveforms.

[0036] For example, at time t(1-1), there is a high degree of symmetry between the waveforms of the time series changes of the head movement waveform B indicated by the head movement waveform signal B(t(1-1)) and the eye movement waveform C indicated by the eye movement waveform signal C(t(1-1)). At time t(1-2), there is a high degree of symmetry between the waveforms of the time series changes of the head movement waveform B indicated by the head movement waveform signal B(t(1-2)) and the eye movement waveform C indicated by the eye movement waveform signal C(t(1-2)). At time t(1-3), there is a high degree of symmetry between the waveforms of the time series changes of the head movement waveform B indicated by the head movement waveform signal B(t(1-3)) and the eye movement waveform C indicated by the eye movement waveform signal C(t(1-3)). At time t(1-4), there is a high degree of symmetry between the waveforms of the time series changes of the head movement waveform B indicated by the head movement waveform signal B(t(1-4)) and the eye movement waveform C indicated by the eye movement waveform signal C(t(1-4)). At time t(1-5), there is a high degree of symmetry between the waveforms of the time series changes of the head movement waveform B indicated by the head movement waveform signal B(t(1-5)) and the eye movement waveform C indicated by the eye movement waveform signal C(t(1-5)).

[0037] When there is a high degree of symmetry between the head movement waveform B and the eye movement waveform C, the two waveforms have similar shapes, indicating that the head and eye movements of person A are linked. This indicates that person A, such as a driver, is likely concentrating on driving and moving his or her head and eyes to view various locations in front of the vehicle. Therefore, when feature points are distributed in area E1, the concentration level determination unit 12 determines that person A's concentration level E is high.

[0038] 5, that is, based on the distribution of feature points p(2-1) to p(2-5) that have transitioned from point (-1, 0) as a starting point to the vicinity of the origin O in the radial direction of the circular region of the feature space, the concentration level determination unit 12 identifies the relationship between the head movement waveform B and the eye movement waveform C as follows: Then, the concentration level determination unit 12 determines the concentration level E of person A based on the relationship between the identified waveforms.

[0039] At multiple feature points p that transition in the radial direction starting from point (-1, 0), such as in region F1, the first feature amount X has approximately the same value, but both the first feature amount X and the second feature amount Y are close to zero. In other words, the influence of the feature indicating the relationship between waveforms indicated by the first feature amount X and the feature indicating the relationship between waveforms indicated by the second feature amount Y is small. For this reason, there is low symmetry between the waveforms of the time-series changes in the head movement waveform B and the eye movement waveform C indicated by these feature points p.

[0040] For example, at time t(2-1), there is low symmetry between the waveforms of the time series changes of the head movement waveform B indicated by head movement waveform signal B(t(2-1)) and the eye movement waveform C indicated by eye movement waveform signal C(t(2-1)), and the feature point p has shifted in the radial direction to the vicinity of the origin O. At time t(2-2), there is low symmetry between the waveforms of the time series changes of the head movement waveform B indicated by head movement waveform signal B(t(2-2)) and the eye movement waveform C indicated by eye movement waveform signal C(t(2-2)), and the feature point p has shifted in the radial direction to the vicinity of the origin O. At time t(2-3), there is low symmetry between the waveforms of the time series changes of the head movement waveform B indicated by head movement waveform signal B(t(2-3)) and the eye movement waveform C indicated by eye movement waveform signal C(t(2-3)), and the feature point p has shifted in the radial direction to the vicinity of the origin O. At time t(2-4), there is low symmetry between the waveforms of the time series changes in the head movement waveform B indicated by head movement waveform signal B(t(2-4)) and the eye movement waveform C indicated by eye movement waveform signal C(t(2-4)), and the feature point p has shifted in the radial direction to the vicinity of the origin O. At time t(2-5), there is low symmetry between the waveforms of the time series changes in the head movement waveform B indicated by head movement waveform signal B(t(2-5)) and the eye movement waveform C indicated by eye movement waveform signal C(t(2-5)), and the feature point p has shifted in the radial direction to the vicinity of the origin O.

[0041] When the symmetry between the head movement waveform B and the eye movement waveform C is low and the feature point p has shifted radially to near the origin O, this indicates that both the head and eye movements of person A have become smaller. In this case, for example, a driver, etc., is likely to be in a state where the driver's head is not moving much, and the eye movement changes significantly, then slows down, and the driver continues to stare at something, i.e., the driver is likely to be looking away. Therefore, when the feature point p is distributed in region F1, the concentration level determination unit 12 determines that person A's concentration level E has decreased due to inattentiveness. The change in the feature point p in the radial direction of the circular region is affected by the degree of deviation between the waveforms, and the change in the feature point p in the circumferential direction of the circular region is affected by the degree of time difference between the waveforms. When the driver is looking away, the gaze changes significantly independent of the head movement, and the waveforms are likely to diverge, resulting in a radial transition of the feature point p.

[0042] 5, i.e., based on the distribution of feature points p(3-1) to p(3-5) that have transitioned in the circumferential direction of the circular region starting from point (-1, 0) until the first feature amount X becomes almost zero, the concentration level determination unit 12 identifies the relationship between the head movement waveform B and the eye movement waveform C as follows: The concentration level determination unit 12 determines the concentration level E of person A based on the relationship between the identified waveforms.

[0043] In multiple feature points p shifted in the circumferential direction starting from point (-1, 0), as in region G1, the first feature amount X is nearly zero and the second feature amount Y is large. Since the second feature amount Y is the correlation coefficient between the head movement waveform B and the eye movement waveform C phase-shifted by π / 2, when multiple feature points p are plotted in region G1, the influence of the feature indicating the relationship between the head movement waveform B and the eye movement waveform C indicated by these feature points p becomes greater, indicating that there is low symmetry between the waveforms of the time-series changes of these waveforms.

[0044] For example, at time t(3-1), there is low symmetry between the waveforms of the time series changes of the head movement waveform B indicated by head movement waveform signal B(t(3-1)) and the eye movement waveform C indicated by eye movement waveform signal C(t(3-1)), and the first feature amount X, which is an element of feature point p, transitions in the circumferential direction until it becomes almost zero. At time t(3-2), there is low symmetry between the waveforms of the time series changes of the head movement waveform B indicated by head movement waveform signal B(t(3-2)) and the eye movement waveform C indicated by eye movement waveform signal C(t(3-2)), and the first feature amount X, which is an element of feature point p, transitions in the circumferential direction until it becomes almost zero. At time t(3-3), there is low symmetry between the waveforms of the time series changes of the head movement waveform B indicated by head movement waveform signal B(t(3-3)) and the eye movement waveform C indicated by eye movement waveform signal C(t(3-3)), and the first feature amount X, which is an element of feature point p, transitions in the circumferential direction until it becomes almost zero. At time t(3-4), there is low symmetry between the waveforms of the time series changes of the head movement waveform B indicated by head movement waveform signal B(t(3-4)) and the eye movement waveform C indicated by eye movement waveform signal C(t(3-4)), and the first feature amount X, which is an element of feature point p, transitions in the circumferential direction until it becomes almost zero. At time t(3-5), there is low symmetry between the waveforms of the time series changes of the head movement waveform B indicated by the head movement waveform signal B(t(3-5)) and the eye movement waveform C indicated by the eye movement waveform signal C(t(3-5)), and the first feature amount X, which is an element of the feature point p, transitions in the circumferential direction until it becomes almost zero.

[0045] If the symmetry between the head movement waveform B and the eye movement waveform C is low and the first feature quantity X, which is an element of the feature point p, shifts in the circumferential direction until it reaches almost zero, this indicates that either the head or eye movement of person A is becoming smaller. In this case, person A, such as a driver, is likely in a state where he / she is not moving his / her head due to drowsiness, with his / her eyes moving while his / her gaze is unfocused, or in a state where he / she is moving his / her head with his / her eyes closed due to drowsiness. Therefore, if feature point p is distributed in region G1, the concentration level determination unit 12 determines that person A's concentration level E has decreased due to drowsiness. As described above, the change in feature point p in the radial direction in the circular region is affected by the degree of deviation in the shapes between the waveforms, and the change in feature point p in the circumferential direction in the circular region is affected by the degree of time difference between the waveforms. When person A, such as a driver, is looking straight ahead to some extent but is in a state of absentminded drowsiness, the eye reaction to head movement is likely to appear as a time difference, resulting in a circumferential transition of feature point p.

[0046] A specific example (2) of determining the concentration level E will be described using Figures 6A and 6B. Figure 6A shows a feature space containing a distribution of multiple feature points p plotted near point (-1, 0) and a distribution of multiple feature points p plotted at an angle θ in the circumferential direction starting from point (-1, 0). In Figure 6A, feature points p(1-1) to p(1-5) are plotted near point (-1, 0), as shown in Figure 5. On the other hand, feature points p(2-1) to p(2-5) are plotted at positions shifted at an angle θ in the circumferential direction with respect to the X-axis, unlike in Figure 5.

[0047] Fig. 6B shows a feature space having X' and Y' axes obtained by rotating the X and Y axes of the feature space in Fig. 6A by a predetermined angle. In Fig. 6B, by rotating the X and Y axes, feature points p(1-1) to p(1-5) become feature points p(1'-1) to p(1'-5), and feature points p(2-1) to p(2-5) become feature points p(2'-1) to p(2'-5). In the following description, feature points p(1'-1) to p(1'-5) and feature points p(2'-1) to p(2'-5) will be collectively referred to as feature point p' as appropriate.

[0048] The amount of circumferential movement of the feature point p′ shown in FIG. 6B can be calculated using the Y coordinate value of the feature point p on the XY coordinate system, the Y′ coordinate value of the feature point p′ on the X′Y′ coordinate system, the angle θ that the feature point p makes with the X axis on the XY coordinate system, and the angle θ′ that the feature point p′ makes with the X axis on the X′Y′ coordinate system. The radial shift amount of the feature point p′ shown in FIG. 6B can be calculated using the X coordinate value of the feature point p on the XY coordinate system, the X coordinate value of the feature point p′ on the X′Y′ coordinate system, the distance from the feature point p on the XY coordinate system to the origin O, the distance from the feature point p′ on the X′Y′ coordinate system to the origin O, the distance from the center of gravity of the multiple feature points p on the XY coordinate system (the center of gravity of feature points p(1-1) to p(1-5) or the center of gravity of feature points p(2-1) to p(2-5)) to the origin O, and the distance from the center of gravity of the multiple feature points p′ on the X′Y′ coordinate system (the center of gravity of feature points p(1′-1) to p(1′-5) or the center of gravity of feature points p(2′-1) to p(2′-5)) to the origin O. In a state of concentration, ideally, the plot of feature point p would be concentrated at the point (-1, 0), but due to individual differences, it is expected that even in a state of concentration, the plot may deviate slightly from the ideal position, in which case the reference position will change to a position different from the point (-1, 0). Therefore, by evaluating feature point p in a coordinate system with tilted axes as described above, it is possible to determine the concentration level E in accordance with the tendencies of each individual.

[0049] A specific example (3) of determining the concentration level E will be described with reference to FIG. 7 . The concentration level determination unit 12 determines whether the concentration level E of a person A, such as a driver, is high or low based on the evaluation value of the feature point p. The evaluation value includes the transition amount of the feature point p or the distribution of the feature point p in a feature space. For example, the evaluation value is a numerical value calculated by performing a product-sum operation (e.g., weighted addition) on the radial transition amount and the circumferential transition amount of the feature point p in a feature space that is an XY coordinate plane with the first feature amount X as the X axis and the second feature amount Y as the Y axis. Alternatively, the evaluation value may be the distribution of the feature points p in a circular region centered at the origin O of the XY coordinate plane with the first feature amount X and the second feature amount Y as axes.

[0050] The concentration level determination unit 12 compares the evaluation value with a preset threshold value TH, and determines that the concentration level E of the driver or the like, who is person A, is high if the evaluation value is smaller than the threshold value TH, and determines that the concentration level E of the driver or the like is low if the evaluation value is equal to or greater than the threshold value TH. If the concentration level E of the driver or the like is determined to be high, it is highly likely that the driver or the like is concentrating on driving, etc. On the other hand, if the concentration level E of the driver or the like is determined to be low, it is highly likely that the driver or the like is not concentrating on driving, etc., or is dozing off or looking away.

[0051] For example, the concentration level determination unit 12 determines whether or not the driver, etc., who is person A, is dozing off, which affects the concentration level E, based on the magnitude of the circumferential shift of the feature point p included in the evaluation value, i.e., the magnitude of the circumferential shift of the feature point p in the circular area on the XY coordinate plane, which is the feature space. Furthermore, the concentration level determination unit 12 determines whether or not the driver, etc., who is person A, is looking away, which affects the concentration level E, based on the magnitude of the radial shift of the feature point p included in the evaluation value, i.e., the magnitude of the radial shift of the feature point p in the circular area on the XY coordinate plane, which is the feature space.

[0052] 7 is a graph showing the relationship between the evaluation value of the concentration level and the passage of time. In FIG. 7, the horizontal axis represents the time during which the driver or the like has driven, the vertical axis represents the evaluation value of the driver or the like over time, and TH represents the evaluation threshold. The concentration level determination unit 12 determines that the concentration level E of the driver or the like is maintained at a high level because the evaluation value is smaller than the threshold TH until time ta. After time ta, the concentration level determination unit 12 determines that the concentration level E of the driver or the like has decreased because the evaluation value is equal to or greater than the threshold TH.

[0053] A specific example (4) of determining the concentration level E will be described with reference to Fig. 8. The concentration level determination unit 12 may determine the level of the concentration level E of person A based on the relationship between the radial and circumferential transition amounts of a feature point p in a circular region centered at the origin O of an XY coordinate plane, which is a feature space and has the first feature amount X and the second feature amount Y as its axes. Note that the feature point p is a point whose elements are the first feature amount X and the second feature amount Y.

[0054] 8 is a diagram showing the relationship between the radial and circumferential shift amounts of the feature point p, with the vertical axis representing the radial shift amount of the feature point p and the horizontal axis representing the circumferential shift amount of the feature point p. In FIG. 8, region H1 is a region including the origin O' and its vicinity where the vertical and horizontal axes intersect at right angles, and is, for example, a fan-shaped region centered on the origin O'. Region H2 is a region located farther from the origin O' than region H1. Region H3 is a region located farther from the origin O' than region H2.

[0055] Points q(1-1) to q(1-5) at times t(1-1) to t(1-5) are points whose elements are the radial and circumferential transition amounts of feature point p. That is, points q(1-1) to q(1-5) are identified by the radial and circumferential transition amounts of feature point p on the XY coordinate plane, which is the feature space. Points q(2-1) to q(2-5) at times t(2-1) to t(2-5) are points whose elements are the radial and circumferential transition amounts of feature point p. That is, points q(2-1) to q(2-5) are identified by the radial and circumferential transition amounts of feature point p on the XY coordinate plane, which is the feature space. Points q(3-1) to q(3-5) at times t(3-1) to t(3-5) are points whose elements are the radial and circumferential transition amounts of feature point p. That is, points q(3-1) to q(3-5) are specified by the radial and circumferential transition amounts of feature point p on the XY coordinate plane, which is the feature space.

[0056] If points q(1-1) to q(1-5) are distributed in area H1 near origin O' at times t(1-1) to t(1-5), this indicates that the amount of radial transition of feature point p is small and the amount of circumferential transition is also small. In this case, there is high symmetry between the waveforms of the time-series changes in head movement waveform B and eye movement waveform C, so concentration level determination unit 12 determines that person A, such as a driver, has a high concentration level E.

[0057] Furthermore, if points q(2-1) to q(2-5) at times t(2-1) to t(2-5) are distributed in a portion along the horizontal axis in region H2, this indicates that the radial transition amount of feature point p is small and the circumferential transition amount is large. In this case, the symmetry between the waveforms of the time-series changes in head movement waveform B and eye movement waveform C is low, and concentration level determination unit 12 determines that the concentration level E of the driver or the like is low due to drowsiness.

[0058] Furthermore, if points q(3-1) to q(3-5) at times t(3-1) to t(3-5) are distributed in a portion along the horizontal axis in region H3, this indicates that the radial transition amount of feature point p is large, or that both the radial transition amount and the circumferential transition amount are large. In this case, the symmetry between the waveforms of the time-series changes of head movement waveform B and eye movement waveform C is low, and concentration level determination unit 12 determines that the concentration level E of the driver, etc. is low due to inattentiveness.

[0059] As described above, the concentration level determination unit 12 determines the concentration level E of person A based on the first feature amount X and the second feature amount Y, which respectively indicate the relationship between the different waveforms of person A's head movement waveform B and eye movement waveform C. In other words, the concentration level determination unit 12 only needs to be able to determine the concentration level from the relationship between these different waveforms, based on the first feature amount X and the second feature amount Y, which indicate the relationship between the different waveforms of person A's head movement waveform B and eye movement waveform C, while taking into account differences in various features contained between the waveforms. For example, as described with reference to FIGS. 5 and 6 , the concentration level determination unit 12 may plot a feature point p, whose elements are the first feature amount X and the second feature amount Y, in a two-dimensional feature space and determine person A's concentration level E based on the transition of the feature point p in the feature space. In particular, as described with reference to FIG. 5 , the concentration level determination unit 12 may determine whether person A is looking away, which affects person A's concentration level E, based on the amount of change in feature point p in the radial direction of a circular region centered on the origin O in a two-dimensional feature space whose axes are the first feature amount X and the second feature amount Y. Furthermore, as described with reference to FIG. 5 , the concentration level determination unit 12 may determine whether person A is dozing, which affects person A's concentration level E, based on the amount of change in feature point p in the circumferential direction of the circular region. Furthermore, as described with reference to FIG. 7 , the concentration level determination unit 12 may determine person A's concentration level E based on the result of evaluating feature point p by comparing an evaluation value including the amount of change in feature point p or the distribution of feature points p in the feature space with a predetermined evaluation threshold. This enables the concentration level determination device 1 to improve the accuracy of determining person A's concentration level E.

[0060] Next, we will explain the hardware configuration that realizes the functions of the concentration level determination device 1. The functions of the feature amount calculation unit 11 and the concentration level determination unit 12 provided in the concentration level determination device 1 are realized by processing circuits. That is, the concentration level determination device 1 includes a processing circuit that executes the processes of steps ST1 and ST2 in Fig. 4. The processing circuit may be dedicated hardware, or may be a CPU (Central Processing Unit) that executes a program stored in memory.

[0061] Fig. 9A is a block diagram showing a hardware configuration that realizes the functions of concentration level determination device 1. Fig. 9B is a block diagram showing a hardware configuration that executes software that realizes the functions of concentration level determination device 1. In Figs. 9A and 9B, input interface 100 relays waveform signals output from an external measurement device to concentration level determination device 1. Output interface 101 relays, for example, data indicating the determination result of concentration level E that is output from concentration level determination device 1 to an external device.

[0062] 9A , the processing circuit 102 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination thereof. The functions of the feature calculation unit 11 and the concentration determination unit 12 included in the concentration determination device 1 may be realized by separate processing circuits, or these functions may be realized together by a single processing circuit.

[0063] 9B , the functions of the feature calculation unit 11 and the concentration determination unit 12 included in the concentration determination device 1 are realized by software, firmware, or a combination of software and firmware. The software or firmware is written as a program and stored in the memory 104.

[0064] The processor 103 reads and executes programs stored in the memory 104 to realize the functions of the feature calculation unit 11 and the concentration determination unit 12 included in the concentration determination device 1. For example, the concentration determination device 1 includes a memory 104 for storing programs that, when executed by the processor 103, result in the processing of steps ST1 and ST2 in the flowchart shown in FIG. 4 being executed. These programs cause a computer to execute the procedures or methods of the processing performed by the feature calculation unit 11 and the concentration determination unit 12. The memory 104 may be a computer-readable storage medium that stores programs for causing a computer to function as the feature calculation unit 11 and the concentration determination unit 12.

[0065] The memory 104 may be, for example, a non-volatile or volatile semiconductor memory such as a RAM (Random Access Memory), a ROM (Read Only Memory), a flash memory, an EPROM (Erasable Programmable Read Only Memory), or an EEPROM (Electrically-EPROM), a magnetic disk, a flexible disk, an optical disk, a compact disk, a mini disk, or a DVD.

[0066] Furthermore, some of the functions of the feature calculation unit 11 and the concentration determination unit 12 included in the concentration determination device 1 may be realized by dedicated hardware, and some may be realized by software or firmware. For example, the function of the feature calculation unit 11 may be realized by a processing circuit 102 that is dedicated hardware, and the function of the concentration determination unit 12 may be realized by a processor 103 reading and executing a program stored in a memory 104. In this way, the processing circuit can realize the above functions by hardware, software, firmware, or a combination of these.

[0067] In the above description, the first feature value X is calculated by adding and multiplying a noise-removed head movement waveform B and a noise-removed eye movement waveform C to perform a correlation calculation process, and the second feature value Y is calculated by adding and multiplying a noise-removed eye movement waveform C and a head movement waveform D that has been phase-shifted by π / 2 after noise removal to perform a correlation calculation process. However, the feature values ​​handled by the concentration level assessment device 1 are not limited to this. That is, the first feature value X and the second feature value Y may each represent a relationship between the different waveforms of person A, namely, head movement waveform B and eye movement waveform C. For example, the first feature value X may be calculated as a feature value of a portion common to both head movement waveform B and eye movement waveform C. In order to represent a relationship different from that of the first feature value X, the second feature value Y may be calculated as a feature value of a portion common to the waveform after the change and the unchanged waveform, obtained by changing either head movement waveform B or eye movement waveform C by a certain amount. Furthermore, the feature quantities handled by the concentration level determination device 1 may be three or more types of feature quantities, including a first feature quantity X and a second feature quantity Y. In this case, it is sufficient that each of the three or more types of feature quantities indicates a relationship between different waveforms. Furthermore, the feature points may also be elements of three or more types of feature quantities, and the concentration level determination unit 12 may determine the concentration level E of person A based on the transition amounts or distribution of the feature points plotted in a three-dimensional or higher-dimensional feature space centered on these feature points. In this manner, the concentration level determination device 1 can also determine the concentration level E of person A with high accuracy.

[0068] As described above, the concentration level assessment device 1 according to the first embodiment includes a feature amount calculation unit 11 that calculates a first feature amount X indicating the relationship between the waveforms of person A's head movement waveform B and eye movement waveform C, and a second feature amount Y indicating the relationship between the waveforms different from the first feature amount X, and a concentration level assessment unit 12 that assesses person A's concentration level E based on the first feature amount X and the second feature amount Y. Because the first feature amount X and the second feature amount Y indicate the relationship between the different waveforms of person A's head movement waveform B and eye movement waveform C, it is possible to assess person A's concentration level E from the relationship between these waveforms, taking into account differences in various features contained between the waveforms. This enables the concentration level assessment device 1 to improve the accuracy of assessing person A's concentration level E.

[0069] In the concentration level assessment device 1 according to the first embodiment, the concentration level assessment unit 12 plots feature points p, each of which has a first feature amount X and a second feature amount Y as elements, in a two-dimensional feature space, and assesses the concentration level E of person A based on the transition of feature points p in the feature space. This enables the concentration level assessment device 1 to assess the concentration level E based on the transition of feature points that indicate the relationship between head movement waveform B and eye movement waveform C in the feature space, taking into account the differences in various features contained between these waveforms.

[0070] In the concentration level assessment device 1 according to the first embodiment, the concentration level assessment unit 12 assesses the level of concentration E of person A based on the result of evaluating the feature point p by comparing an evaluation value including the amount of transition of the feature point p or the distribution of the feature point p in the feature space with a predetermined evaluation threshold. This enables the concentration level assessment device 1 to assess the level of concentration E taking into account the differences in various features included between the head movement waveform B and the eye movement waveform C, based on the amount of transition of the feature point p included in the evaluation value.

[0071] In the concentration level determination device 1 according to the first embodiment, the concentration level determination unit 12 determines whether or not person A is dozing, which affects the concentration level E, based on the amount of shift of feature point p in the circumferential direction of a circular area centered at origin O in a two-dimensional feature space whose axes are the first feature amount X and the second feature amount Y. This enables the concentration level determination device 1 to determine whether or not person A is dozing, based on the movement of the head or eyes that corresponds to the magnitude of the amount of shift of feature point p in the circumferential direction.

[0072] In the concentration level assessment device 1 according to the first embodiment, the concentration level assessment unit 12 assesses whether person A is looking away, which affects the concentration level E, based on the amount of shift of feature point p in the radial direction of a circular region centered at origin O in a two-dimensional feature space whose axes are the first feature amount X and the second feature amount Y. This enables the concentration level assessment device 1 to assess whether person A is looking away, based on the movement of the head or eyeballs that corresponds to the magnitude of the amount of shift of feature point p in the radial direction.

[0073] In the concentration level assessment device 1 according to the first embodiment, the feature amount calculation unit 11 calculates a first feature amount X, which is the correlation coefficient between the noise-removed head movement waveform B and the noise-removed eye movement waveform C, and calculates a second feature amount Y, which is the correlation coefficient between the noise-removed eye movement waveform C and the head movement waveform D that has been phase-shifted by π / 2 after noise removal. This enables the concentration level assessment device 1 to calculate the first feature amount X and the second feature amount Y, which indicate different relationships between the head movement waveform B and the eye movement waveform C.

[0074] The concentration level determination method according to the first embodiment includes step ST1 in which a feature amount calculation unit 11 calculates a first feature amount X indicating the relationship between the waveforms of a head movement waveform B and an eye movement waveform C of person A, and a second feature amount Y indicating the relationship between the waveforms different from the first feature amount X, and step ST2 in which a concentration level determination unit 12 determines a concentration level E of person A based on the first feature amount X and the second feature amount Y. By performing this method, the concentration level determination device 1 can improve the accuracy of determining the concentration level E of person A.

[0075] Any of the components of the embodiments may be modified or omitted.

[0076] The concentration level determination device according to the present disclosure can be used to assist a driver of a vehicle, for example.

[0077] 1 Concentration level determination device, 11 feature amount calculation unit, 12 concentration level determination unit, 100 input interface, 101 output interface, 102 processing circuit, 103 processor, 104 memory, 111 first LPF, 112 second LPF, 113 phase shifter, 114 first correlation calculator, 115 second correlation calculator.

Claims

1. a feature calculation unit that calculates a first feature indicating a relationship between a head movement waveform indicating a movement of a person's head and an eye movement waveform indicating a movement of the person's eyeballs, and a second feature that is different from the first feature and indicates the relationship between the waveforms; a concentration level determination unit that determines a concentration level of the person based on the first feature amount and the second feature amount, The concentration level determination unit plots feature points, which are elements of the first feature amount and the second feature amount, in a two-dimensional feature space, and determines the concentration level of the person based on transitions of the feature points in the feature space. A concentration determination device characterized by:

2. The concentration level determination unit determines the concentration level of the person based on a result of evaluating the feature points by comparing an evaluation value including a transition amount of the feature points or a distribution of the feature points in the feature space with a predetermined evaluation threshold. The concentration determination device according to claim 1 .

3. The concentration level determination unit determines whether or not the person is dozing, which affects the concentration level, based on a transition amount of the feature points in a circumferential direction of a circular area centered at an origin in the two-dimensional feature space whose axes are the first feature amount and the second feature amount. The concentration determination device according to claim 1 .

4. The concentration level determination unit determines whether or not the person is looking away, which affects the concentration level of the person, based on a transition amount of the feature point in a radial direction of a circular area centered at an origin in the two-dimensional feature space whose axes are the first feature amount and the second feature amount. The concentration determination device according to claim 1 .

5. The feature calculation unit calculates the first feature, which is a correlation coefficient between the noise-removed head movement waveform and the noise-removed eye movement waveform, and calculates the second feature, which is a correlation coefficient between the noise-removed eye movement waveform and the head movement waveform phase-shifted by π / 2 after noise removal. The concentration determination device according to any one of claims 1 to 4.

6. A concentration level determination method executed by a concentration level determination device, a feature calculation unit calculating a first feature indicating a relationship between a head movement waveform indicating a movement of a person's head and an eye movement waveform indicating a movement of the person's eyeballs, and a second feature indicating the relationship between the waveforms, which is different from the first feature; a concentration level determination unit determining a concentration level of the person based on the first feature amount and the second feature amount; The concentration level determination unit plots feature points, which are elements of the first feature amount and the second feature amount, in a two-dimensional feature space, and determines the concentration level of the person based on transitions of the feature points in the feature space. A concentration determination method comprising: