JP2022182563A - Driver's drowsiness estimation method - Google Patents

Abstract

To provide a driver's drowsiness estimation method which can accurately detect a sleepy or shallow sleepy state before a driver realizes.SOLUTION: A driver's drowsiness estimation method for estimating drowsiness of a driver of a vehicle comprises: a preparation step (S1) of preparing an in-vehicle camera 2 which detects a head behavior of the driver and a steering angle sensor 3 which detects a steering angle by a steering operation of the driver; a head roll angle calculation step (S3) of calculating a head roll angle based on the head behavior detected by the in-vehicle camera 2; a coherence calculation step (S4) of calculating a coherence value of the steering angle and the head roll angle detected by the steering angle sensor 3; and a drowsiness level determination step (S5, S6, S8, S9) of determining a preset drowsiness level 2 when the coherence value is 0.7 or more and less than 0.9 and determining the drowsiness level 3 at which the drowsiness is stronger than the drowsiness level 2 when the coherence value is less than 0.7.SELECTED DRAWING: Figure 7

Description

The present invention relates to a drowsiness estimation method for estimating drowsiness of a vehicle driver.

Conventionally, there have been proposed techniques for estimating the biological state of a vehicle driver, particularly drowsiness (arousal), based on a plurality of feature quantities detected from the driver's face image, biological information, and the like.
In such vehicle driving support technology, the biological condition of the driver is monitored in real time while the vehicle is being driven, and is measured in a non-constraining and non-invasive manner without interfering with the forward visibility of the vehicle or the operating behavior of the driving equipment. ing.

The drowsy driving warning device of Patent Document 1 includes a behavior sensor such as a steering angle sensor, an image drowsiness level determination unit that determines an image drowsiness level and reliability based on the image of the driver, and a vehicle behavior drowsiness level and a vehicle behavior drowsiness level based on the vehicle behavior. a vehicle behavior drowsiness level determination unit that determines reliability; a total drowsiness level determination unit that determines a total drowsiness level and a total reliability based on the image drowsiness level and reliability and the vehicle behavior drowsiness level and reliability; and a total drowsiness level and an alarm level setting unit for setting an alarm level based on the total reliability.

The image drowsiness level determination unit determines the drowsiness level using known evaluation criteria.
As shown in FIG. 8, the drowsiness level of the driver is classified into five levels from 1 to 5.
If the motion of the driver acquired from the captured image is fast and frequent, the movement is active and accompanied by body movement, it is determined to be drowsiness level 1, and if the lips are open and the movement of the line of sight is slow, the drowsiness level is determined. If it is determined to be 2 and blinks slowly and frequently, it is determined to be drowsiness level 3.
In addition, when there is intentional blinking, shaking the head, and frequent yawning, it is determined to be drowsiness level 4, and when actions such as closing the eyelids and tilting the neck forward and backward are detected, the highest drowsiness level is detected. 5 is determined.

Japanese Patent No. 6662352

Driving a vehicle is the action (output) after the information input from the driver's visual sense, vestibular sense, somatosensory, etc. has been processed, and is represented by the operation behavior of driving equipment. It consists of (cognitive) movement and low-level involuntary movements such as maintaining head posture.
Both the cerebral cortex, which controls higher voluntary functions, and the brainstem, which controls lower involuntary functions, are aggregates of nerve cells, and these nerve cells output according to predetermined firing frequencies (firing characteristics).
In normal times, the cortical system activity supervised by the cerebral cortex and the brainstem system activity supervised by the brainstem have interlocking behavior and rhythmic activity.
Therefore, when drowsiness is regarded as a kind of minor brain movement abnormality as a function of the brain as a whole, it is expected that drowsiness impairs the periodic linkage between cortical system activity and brainstem system activity.

The drowsy driving warning device of Patent Document 1 comprehensively determines the drowsiness level by using photographed facial movements (expressions) and the occurrence frequency of corrective steering.
However, in Patent Literature 1, cortical system activity and brainstem system activity are not evaluated from the viewpoint of interrelationship in the brain, so there is a possibility that sufficient determination accuracy cannot be ensured.
In the evaluation criteria of FIG. 8, for example, the determination of drowsiness level 4, which is a high drowsiness level, is easy to determine and has high accuracy because it determines a large physical and superficial action of the driver. The timing of completion of is late, and it interferes with post-processing such as alarms.
On the other hand, determination of drowsiness level 2, which is a low level of drowsiness, determines a small motion of the driver, and although the determination timing can be accelerated, actual determination is difficult and accuracy tends to be low.

Moreover, both the facial expression and the steering correction frequency are evaluated after they appear in a manner that can be visually recognized from the outside, in other words, after the drowsiness level of the driver has increased to some extent.
Therefore, it is not intended to extract periodic interlocking between cortical system activity and brainstem system activity, and to evaluate drowsiness levels chronologically in minute frequency bands that cannot be easily recognized from the outside.
That is, it is not easy to accurately detect the state of light drowsiness before the driver becomes aware of drowsiness.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method for estimating driver drowsiness, etc., capable of accurately detecting drowsiness before the driver becomes aware of it or light drowsiness.

A driver drowsiness estimation method according to claim 1 is a driver drowsiness estimation method for estimating drowsiness of a driver of a vehicle, comprising: head behavior detection means for detecting the head behavior of the driver; a preparation step of preparing steering angle detection means for detecting a steering angle by a steering angle operation; a head roll angle calculation step of calculating a head roll angle based on the head behavior detected by the head behavior detection means; a coherence calculation step of calculating the coherence between the steering angle detected by the steering angle detection means and the head roll angle; and a preset first drowsiness level when the coherence is greater than or equal to a first threshold and less than a second threshold. and a drowsiness level determination step of determining a second drowsiness level that is stronger than the first drowsiness level when the sleepiness level is less than the first threshold.

Since this method for estimating driver drowsiness includes a preparation step of preparing head behavior detection means for detecting the driver's head behavior and steering angle detection means for detecting the steering angle by the driver's steering operation, It is possible to reliably extract the driver's head behavior and the steering angle that reflects the driver's steering angle operation, which is a high-order voluntary movement.
Since it has a head roll angle calculation step of calculating a head roll angle based on the head behavior detected by the head behavior detection means, it reflects low-order involuntary movements that are performed involuntarily as brainstem activity. A driver's head roll angle can be calculated.
Since there is a coherence calculation step of calculating the coherence between the steering angle detected by the steering angle detection means and the head roll angle, the steering angle that is the result of cortical system activity and the head roll angle that is the result of brainstem system activity are calculated. can be computed in a very small frequency band.
When the coherence is greater than or equal to a first threshold and less than a second threshold, it is determined to be a preset first drowsiness level, and when it is less than the first threshold, it is determined to be a second drowsiness level that is stronger than the first drowsiness level. Since it has a drowsiness level determination step, drowsiness before the driver becomes aware of it or light drowsiness state can be determined based on the periodic interlocking between the cortical system activity and the brainstem system activity.

The invention of claim 2 is characterized in that, in the invention of claim 1, the coherence is calculated using the steering angle and the head roll angle with a frequency band of 0.2 Hz or less in the coherence calculation step. .
According to this configuration, it is possible to calculate the periodic interlocking relationship between the steering angle and the head roll angle in the frequency band where the synchrony between the cortical system activity and the brainstem system activity remarkably appears. can be improved.

The invention of claim 3 is characterized in that, in the invention of claim 1 or 2, in the coherence calculation step, the coherence is calculated when the steering angle is equal to or greater than a predetermined threshold.
According to this configuration, it is possible to calculate the periodic interlocking relationship between the steering angle and the head roll angle motion in a driving environment in which the synchrony between the cortical system activity and the brainstem system activity remarkably appears, and the accuracy of drowsiness level determination. can be further improved.

The invention of claim 4 is the invention of any one of claims 1 to 3, wherein in the preparation step, the driver is in an awake state and the frequency band is 0.06 Hz or more and 0.15 Hz or less. The first and second thresholds are preset based on coherence calculated using the steering angle and the head roll angle, and the preset first and second thresholds are stored in a storage medium. and
According to this configuration, wakefulness maintenance support can be executed in a state of drowsiness or light sleepiness before the driver becomes aware of it, and the effect of wakefulness maintenance support can be enhanced.

According to the method for estimating driver drowsiness of the present invention, drowsiness before the driver becomes aware of it or light drowsiness can be accurately detected by evaluating the periodic linkage between cortical system activity and brainstem system activity. can be done.

1 is a block diagram of a device for estimating driver drowsiness according to Example 1. FIG. It is a course explanatory drawing of a driving simulator. Time-series data of driving behavior, (a) vehicle speed, (b) accelerator pedal depression amount, (c) steering angle, and (d) drowsiness level. Time-series data of head behavior, where (a) is head yaw angle, (b) is head pitch angle, (c) is head roll angle, and (d) is drowsiness level. . 4 is a table showing coherence between a steering angle and a head roll angle for each frequency band; 4 is a graph showing the correlation between drowsiness level and coherence value. 10 is a flowchart showing drowsiness estimation processing; It is an example of sleepiness level evaluation criteria.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail based on the drawings.
The following description exemplifies the application of the present invention to a driver riding in a vehicle (driving simulator), and does not limit the present invention, its applications, or its uses.

Embodiment 1 of the present invention will be described below with reference to FIGS. 1 to 8. FIG.
The drowsiness estimating device 1 judges the drowsiness level or extremely light drowsiness state that the driver perceives during driving of a vehicle, through a sign (coherence) related to intracerebral functions.
This drowsiness estimation device 1 executes wakefulness maintenance control to recover the driver's wakefulness when an extremely light drowsiness state is determined. is executed.

As shown in FIG. 1, the drowsiness estimation device 1 includes an in-vehicle camera 2 (head behavior detection means), a steering angle sensor 3 (steering angle detection means), one or more displays 4, and an audio output device 5. , a steering control device 6, a brake control device 7, an accelerator control device 8, and an ECU (Electronic Control Unit) 10 are main components.

The in-vehicle camera 2 is installed, for example, inside the windshield (not shown) in order to obtain time-series data of the driver's head behavior every 100 mmsec. The in-vehicle camera 2 is an image sensor such as a CCD (Charge Coupled Device) or a CMOS (Complementary Metal Oxide Semiconductor) capable of photographing the interior conditions including the driver operating the steering wheel (not shown). An image captured by the in-vehicle camera 2 is transmitted to the ECU 10 via an in-vehicle network (not shown). The steering angle sensor 3 detects a steering angle, which is one of the driving behaviors, every 100 mmsec, and outputs the detected steering angle to the ECU 10 .

Each of the devices 4 to 8 receives command signals from the ECU 10 and performs prescribed operations.
The display 4 is a display that can be visually recognized by the driver, and includes a center display, a head-up display, and the like arranged on an instrument panel (not shown). The display 4 provides the driver with a visual stimulus within a range that does not interfere with the driving operation when the wakefulness maintenance control is executed when the light drowsiness state (for example, drowsiness level 2) is determined. For example, the brightness or brightness of the image of the driving lane in front of the vehicle may be increased (or decreased), or an alarm may be displayed to the extent that it does not interfere with driving.

The audio output device 5 provides auditory stimulation to the driver within a range that does not interfere with the driving operation during the execution of the wakefulness maintenance control. For example, BGM or the like set in advance at a low volume is played.
The display 4 and the audio output device 5 are mainly in operation during execution of wakefulness maintenance control.
When executing abnormal control in which a high drowsiness state (for example, drowsiness level 3) is determined, the display 4 warns the driver via visual information more strongly than when the wakefulness maintenance control is executed, and the audio output device 5 outputs sound. to warn the driver more strongly than when the wakefulness maintenance control is executed.

A steering control device 6 controls a steering angle for the vehicle to turn a curve. The brake control device 7 generates a braking force for decelerating the vehicle. The accelerator control device 8 generates a driving force for running the vehicle. For example, when a highly drowsy state is estimated and determined to be in an abnormal state, an abnormal output is performed, and in order to avoid danger, vehicle-led control is performed to reduce vehicle speed and make an emergency stop on the shoulder of the road.

Next, the ECU 10 will be explained.
The ECU 10 includes a central processing unit that executes various programs, memories (RAM and ROM), an input/output bus, and the like. As shown in FIG. 1, the ECU 10 includes a head roll angle calculator 11, a coherence calculator 12, a drowsiness level determiner 13, and the like.
Note that the following description includes a description of a drowsiness estimation method using the drowsiness estimation device 1 .

The head roll angle calculation unit 11 extracts the head behavior of the driver from the images continuously captured in time series by the in-vehicle camera 2 while driving, and determines the driving based on the head roll angle movement among the head behaviors. It calculates the head roll angle of the person. The head roll angle is defined as the tilt angle of the head in the horizontal direction around the central axis of rotation when the axis in the longitudinal direction of the vehicle body is the central axis of rotation.
A coherence calculator 12 calculates a coherence value between the steering angle detected by the steering angle sensor 3 and the head roll angle.

Here, in order to clarify the correlation between the steering angle and the driver's head roll angle during vehicle driving, a verification experiment was conducted using a driving simulator (hereinafter abbreviated as DS).
As shown in FIG. 2, DS is a course in which 12 straight lines and 12 curves are alternately formed continuously. The experimental conditions are a target vehicle speed of 80 km/h and a running time of about 60 minutes.
The curve has a radius of curvature that requires a steering angle greater than or equal to a decision threshold (eg, about 30°).
During the verification experiment, the vehicle behavior (vehicle speed, accelerator pedal depression amount, steering angle) corresponding to the driver's driving operation behavior, and the driver's head behavior (head yaw angle, head pitch angle, head roll angle) and the drowsiness level of the driver based on preset evaluation criteria (see FIG. 8) were detected in chronological order.

As shown in FIGS. 3(a) to 3(d), in the region of drowsiness level 4 or higher, large disturbances occur in all of the vehicle speed (a), accelerator depression amount (b), and steering angle (c). In the region near drowsiness level 2, there is almost no disturbance in the vehicle speed (a) and the accelerator depression amount (b), and only the steering angle (c) is disturbed. As a result, it was found that around drowsiness level 2, changes in the steering angle (c) among the cortical system activities supervised by the cerebral cortex during driving were more noticeable than other cortical system activities.

As shown in FIGS. 4(a) to 4(d), in the region of drowsiness level 4 or higher, all of the head yaw angle (a), head pitch angle (b), and head roll angle (c) Great turmoil occurs. In the region near drowsiness level 2, the head yaw angle (a) and the head pitch angle (b) do not change significantly, and only the head roll angle (c) changes significantly. As a result, it was found that around drowsiness level 2, changes in the head roll angle (c), which is part of brainstem system activities controlled by the brainstem, are more pronounced than other brainstem system activities during driving.

Based on the above analysis results, since the coherence value for each frequency corresponds to periodic interlocking, for the time-series data of the steering angle and head roll angle, the coherence value for each frequency is calculated using the coherence (relevance function). Calculated. Fast Fourier transform (FFT) or frequency analysis method was used for coherence calculation.

FIG. 5 shows the results of coherence analysis of the steering angle and head roll angle.
In the analysis results, the coherence value (correlation strength) between the steering angle and the head roll angle is expressed as a value between 0 and 1 for each frequency band. The higher the brightness of the region, the higher the coherence value, and the maximum value of the coherence value is 1.0.

As shown in FIG. 5, the higher the frequency band, the lower (darker) the coherence value tends to be, regardless of the operational status of straight lines, curves, etc. in DS.
On the other hand, in a frequency band equal to or lower than a predetermined minute frequency (for example, 0.2 Hz), there is a region with a high coherence value (bright) corresponding to each curve region.
As a result, when the driver is in an awake state (normal state), the steering angle, which is the result of cortical system activity, and the head roll angle, which is the result of brainstem system activity, have periodic interlocking in the minute frequency band. It has been found.

Further, as shown in FIG. 5, even in the minute frequency band, the higher the drowsiness level, that is, the longer the running time, the more the mutual synchronization between the steering angle and the head roll angle is lost. , the coherence values of both tend to be low.
As a result, the experimental data show that drowsiness impairs the periodic linkage between cortical and brainstem system activities, that is, drowsiness can be regarded as a kind of minor brain movement abnormality as a function of the entire brain. Confirmed by

Next, among the minute frequency bands in which periodic interlocking occurs, for the four frequency bands of 0.06 Hz, 0.12 Hz, 0.18 Hz, and 0.23 Hz, the 1st to 4th, 6th, 9th, and 11th The coherence values at the 1-th curve points P1 to P7 were obtained, and a graph showing the correlation was created. Except for drowsiness level 4 in the frequency bands of 0.06 Hz and 0.12 Hz, all frequency bands have the characteristic that the higher the drowsiness level, the lower the coherence value.

As shown in FIG. 6, at drowsiness level 1 (point P1) corresponding to the wakeful state, the coherence value in the minute frequency band of 0.2 Hz or less is 0.9 (second threshold) or more.
At drowsiness level 2 (point P3), the coherence value of any frequency band is approximately 07 (first threshold value) or more, and if limited to the coherence value of a minute frequency band of 0.2 Hz or less, all coherence values are 0.7. That's it. At drowsiness level 3 (point P5), the coherence value in the minute frequency band of 0.2 Hz or less is 0.7 or more. Become.

Based on the above, in the present embodiment, when the coherence value is 0.9 or more, it is determined to be drowsiness level 1, when it is 0.7 or more and less than 0.9, it is determined to be drowsiness level 2, and when it is less than 0.7 , drowsiness level 3. In other words, the first threshold of 0.7 and the second threshold of 0.9 correspond to the steering angle and the head roll angle with a frequency band of 0.2 Hz or less, particularly 0.06 Hz or more and 0.15 Hz or less. can be said to be a coherence value calculated using , and is stored in advance in a vehicle storage medium (for example, a hard disk, etc.). That is, the first and second thresholds are set using the coherence value in the frequency band of 0.2 Hz or less, and the drowsiness level is also determined using the coherence value in the frequency band of 0.2 Hz or less.

It returns to description of ECU10.
When the coherence value calculated by the coherence calculation unit 12 is 0.7 or more and less than 0.9, the drowsiness level determination unit 13 determines that the drowsiness level is 2, and when the coherence value is less than 0.7, the drowsiness level is determined. Drowsiness level 3, which is stronger than level 2, is determined.
The ECU 10 executes wakefulness maintenance control when the drowsiness level determination unit 13 determines drowsiness level 2, and executes abnormal control when the drowsiness level determination unit 13 determines drowsiness level 3. When sleepiness level 1 is determined, no control is executed.

Next, drowsiness estimation processing will be described based on the flowchart of FIG.
Note that Si (i=1, 2, . . . ) indicates steps for each process.
First, as shown in FIG. 7, the drowsiness estimation device 1 reads various information such as photographed data acquired by the in-vehicle camera 2, the steering angle detected by the steering angle sensor 3, and coherence values corresponding to the first and second thresholds. (S1), it shifts to S2. Moreover, S1 also serves as a preparatory step for preparing the in-vehicle camera 2 and the steering angle sensor 3 .

In S2, it is determined whether or not the steering angle of the steering wheel operated by the driver is greater than or equal to the determination threshold.
As a result of the determination in S2, if the steering angle is equal to or greater than the determination threshold value, the vehicle is traveling on a curve, and the periodic interlocking between cortical system activity and brainstem system activity can be used to determine the drowsiness level, so the process proceeds to S3.
As a result of the determination in S2, if the steering angle is less than the determination threshold value, the vehicle is running in a straight line, and the periodic linkage between the cortical system activity and the brainstem system activity is low. Return because it is difficult to use.
In S3, after the head roll angle calculator 11 calculates the driver's head roll angle, the coherence calculator 12 calculates a coherence value (S4).

In S5, it is determined whether or not the coherence value is less than 0.7, which is the first threshold.
As a result of the determination in S5, if the coherence value is less than 0.7, the drowsiness of the driver is strong, so drowsiness level 3 is determined (S6), and the process proceeds to S7.
In S7, after the ECU 10 executes the abnormality control, the process returns.
If the coherence value is 0.7 or more as a result of the determination in S5, the process proceeds to S8.

In S8, it is determined whether or not the coherence value is less than the second threshold of 0.9.
As a result of the determination in S8, if the coherence value is less than 0.9, the driver's drowsiness is lighter than drowsiness level 3, so drowsiness level 2 is determined (S9), and the process proceeds to S10.
In S10, the ECU 10 executes wakefulness maintenance control, and then returns.
If the coherence value is equal to or greater than 0.9 as a result of the determination in S8, the driver is not drowsy at all and the drowsiness level is 1, which corresponds to a normal state, so the process simply returns.

Next, the operation and effects of the driver drowsiness estimation method will be described.
According to this method for estimating driver drowsiness, there is a preparatory step (S1) of preparing an in-vehicle camera 2 for detecting the driver's head behavior and a steering angle sensor 3 for detecting the steering angle by the driver's steering operation. Therefore, it is possible to reliably extract the head behavior of the driver and the steering angle reflecting the rudder angle operation of the steering wheel by the driver, which is a high-order voluntary movement.
Since it has a head roll angle calculation step (S3) that calculates the head roll angle based on the head behavior detected by the in-vehicle camera 2, it reflects low-level involuntary movements that are performed involuntarily as brainstem activity. A driver's head roll angle can be calculated.
Since there is a coherence calculation step (S4) for calculating the coherence value between the steering angle detected by the steering angle sensor 3 and the head roll angle, the steering angle that is the result of cortical system activity and the head roll angle that is the result of brainstem system activity are included. can be computed in a small frequency band.
When the coherence value is greater than or equal to 0.7, which is the first threshold, and less than 0.9, which is the second threshold, it is determined to be a preset sleepiness level 2, and when it is less than the first threshold, sleepiness is stronger than sleepiness level 2. Since the drowsiness level determination step (S5, S6, S8, S9) for determining drowsiness level 3 is provided, the drowsiness or light drowsiness state before the driver becomes aware of it based on the periodic interlocking between the cortical system activity and the brainstem system activity. can be determined.

In the coherence calculation step (S4), since the coherence is calculated using the steering angle and the head roll angle whose frequency band is 0.2 Hz or less, the frequency at which the synchrony between the cortical system activity and the brainstem system activity appears remarkably. It is possible to calculate the periodic linkage between the steering angle and the head roll angle in the band, and improve the accuracy of the drowsiness level determination.

In the coherence calculation step (S4), when the steering angle is equal to or greater than a predetermined threshold value, the coherence value is calculated. It is possible to calculate the periodic coherence with motion, and to further improve the accuracy of drowsiness level determination.

In the preparation step (S1), the driver is in an awake state, and a first coherence value is calculated using a steering angle and a head roll angle with a frequency band of 0.06 Hz or more and 0.15 Hz or less. , and second threshold values are set in advance, and the preset first and second threshold values are stored in a storage medium. Therefore, in a state of drowsiness or light drowsiness before the driver becomes aware of it, it is possible to execute wakefulness maintenance support. It is possible to enhance the effect of support for maintaining wakefulness.

Next, a modified example in which the above embodiment is partially changed will be described.
1) In the above embodiment, an example in which drowsiness levels are divided into five stages was described. You can do it.

2) In the above embodiment, an example was described in which wakefulness maintenance control at sleepiness level 2 is performed when the first threshold value is greater than or equal to the second threshold value and less than the second threshold value, and abnormal control at sleepiness level 3 is performed when the value is less than the first threshold value. , three types of thresholds are set, and when the first threshold or more and less than the second threshold, the first wakefulness maintenance control of drowsiness level 2 is performed, and when the second threshold or more and less than the third threshold, the first wakefulness maintenance of drowsiness level 3 is performed. A second wakefulness maintenance control with a stronger stimulus than the control may be performed, and when the third threshold is equal to or higher, the sleepiness level 4 abnormality control may be performed.

3) In addition, those skilled in the art can implement various modifications to the above embodiment without departing from the scope of the present invention, and the present invention includes such modifications. be.

1 drowsiness estimation device S3 head roll angle calculation step S4 coherence calculation step

Claims (4)

A driver drowsiness estimation method for estimating drowsiness of a vehicle driver, comprising:
a preparation step of preparing head behavior detection means for detecting a driver's head behavior and steering angle detection means for detecting a steering angle by a driver's steering angle operation;
a head roll angle calculation step of calculating a head roll angle motion based on the head behavior detected by the head behavior detection means;
a coherence calculation step of calculating the coherence between the steering angle detected by the steering angle detection means and the head roll angle;
When the coherence is greater than or equal to a first threshold and less than a second threshold, it is determined to be a preset first drowsiness level, and when it is less than the first threshold, it is determined to be a second drowsiness level that is stronger than the first drowsiness level. a drowsiness level determination step to
A method for estimating drowsiness of a driver, comprising: 2. The method for estimating drowsiness of a driver according to claim 1, wherein in said coherence calculation step, coherence is calculated using said steering angle and head roll angle with a frequency band of 0.2 Hz or less. 3. The method for estimating drowsiness of a driver according to claim 1, wherein in the coherence calculation step, the coherence is calculated when the steering angle is equal to or greater than a predetermined threshold. In the preparation step, the driver is in an awake state, and the first coherence is calculated using the steering angle and head roll angle motion with a frequency band of 0.06 Hz or more and 0.15 Hz or less. , second threshold values are set in advance, and the preset first and second threshold values are stored in a storage medium. Method.

Images