FR3012029A1 - METHOD FOR DETECTING THE VIGILANCE DROP OF THE DRIVER OF A MOTOR VEHICLE - Google Patents

Abstract

A method for detecting the driver's vigilance decrease, comprising the steps of comparing (500; 600) the current vigilance level (NVC) of said driver with a detection threshold (SDS; SDP), transmission (800) of the order to send a warning message as soon as said current vigilance level (NVC) is below said detection threshold (SDS; SDP); said method comprising previously in said comparing step (500; 600), the steps of determining (100) the conductive profile by facial recognition and selection (200, 400) of the detection threshold (SDS; SDP) to be adopted, said step of selection consisting when said conductive profile corresponds to a profile already known and associated with a personalized detection threshold (SDP), to opt for said personalized detection threshold, and when these conditions are not met to opt for a standard detection threshold predefined (SDS), said personalized detection threshold (SDP) being calculated according to the nominal vigilance level of said driver determined at the end of a learning phase.

Description

FIELD OF THE INVENTION The method relates to the field of driver assistance devices intended for the driver of a motor vehicle, and in particular the devices for detecting the decline in alertness of this driver.

BACKGROUND OF THE INVENTION The converging results of numerous studies show that distraction and reduced vigilance at the wheel of the driver constitute one of the main causes of serious traffic accidents. This hypovigilance gradually sets in insidiously, so that drivers are often aware of it too late, while their perception of the road and their ability to react are already considerably altered.

In recent years, a number of devices have been developed to detect the occurrence of the decline of vigilance in the driver. Thus, embedded devices are known inside the vehicle comprising a camera oriented towards the road which is connected to a unit of analysis and control. This verifies that the vehicle correctly follows the route of its lane by comparing its position with respect to those marking lines of this lane. When the device estimates that the vehicle repeatedly deviates from the ideal trajectory, it emits an audible and / or visual warning message to the driver. Other devices for detecting hypovigilance are based on the analysis of the driver's face via a camera. The detection module analyzes the image provided by this camera and deduces a state of fatigue according to certain parameters such as the closing frequency of the eyelids, the duration of blinking, or the speed of movement of the pupils. As soon as this state of fatigue exceeds a certain threshold, the device sends an alert to the driver.

Japanese application JP 2009-208739 also discloses a system associating a first device for analyzing the face of the driver of the vehicle with the aid of a first camera, and a second device for monitoring the position of the vehicle in the track. traffic through a second camera.

An audible alert is activated when each of the two devices detects a level of drowsiness greater than a predetermined threshold. The purpose of combining these two detection means is to limit the number of false alerts generated by the system. Nevertheless, the latter are still too frequent, the chosen detection thresholds are not necessarily the most appropriate. OBJECT AND SUMMARY OF THE INVENTION The present invention aims at improving the efficiency of such a system for detecting driver hypovigilance.

It proposes for this purpose, a method of detecting the decline of vigilance of the driver of a motor vehicle, said method comprising the following steps: - comparison of the current level of vigilance of said driver at the wheel of said vehicle with a level detection threshold limited vigilance; transmission to an alert module of said vehicle, of the order to send a warning message to said driver; said transmission step being implemented as soon as said current vigilance level is below said detection threshold; said method being characterized in that it comprises previously in said comparison step, the following steps: - determination of the conductive profile by facial recognition; selecting the detection threshold to be adopted for said conductor, said selecting step consisting of: when said conductive profile corresponds to a profile already known and associated with a personalized detection threshold, to opt for said personalized detection threshold; - when these conditions are not met, to opt for a predefined standard detection threshold; said personalized detection threshold being calculated according to the nominal vigilance level of said driver determined at the end of a learning phase. Rather than using the same threshold of detection of the level of vigilance limit for all drivers, the method according to the invention adopts, provided that the driver of the vehicle has been recognized and has undergone a phase of prior learning, a personalized detection threshold that is deduced from the nominal alertness level of this driver. Experience shows that the physiological parameters visible on the driver's face, as well as those of control of the vehicle, corresponding to the nominal vigilance level representative of driving periods where the driver is fully awake and attentive, can vary significantly. according to the drivers. Thus, a driver may for example always have, at his nominal level of vigilance, eyelids that blink for ophthalmic or neurological reasons, In the same way, another driver may have a natural driving style associated with a frequency high trajectory correction via the steering wheel. The taking into account of this inter-individual variability in the calculation of the personalized detection threshold entails a consequent limitation of the risks of error of judgment on the part of the process according to the invention. According to preferred features of the method according to the invention, taken alone or in combination: the current and nominal vigilance levels of said conductor are determined at least partially by means of certain physiological parameters, such as the focusing time, the frequency and the eyelid closing time, the frequency and the speed of eye movements or the heart rate, said physiological parameters themselves being calculated by analyzing the face of said driver from the measurements made on the images transmitted by a first camera directed on said face; said conductive profile is determined by means of the facial analysis of its face from measurements made on the images transmitted by said first camera; the current and nominal vigilance levels of said conductor are also determined by means of certain behavioral parameters, such as the degree of oscillation of the trajectory or the frequency of correction of the trajectory via the steering wheel, said behavioral parameters being they the same ones calculated by analyzing the behavior of said vehicle from the measurements made on the images transmitted by a second camera directed towards the road, and dynamic data provided by sensors embedded in said vehicle; said on-board sensors comprise a speed sensor of said vehicle, a gearbox gear engaged sensor, a steering wheel steering angle sensor, a steering wheel rotation speed sensor and a pedal position sensor; brake and acceleration; said learning phase is performed automatically during time intervals when certain conditions relating to the time slot, the driving conditions and / or the current level of vigilance of said driver are fulfilled, said learning phase ceasing when the time spent or the number of kilometers traveled under those conditions has reached a predefined duration or number of kilometers; said learning phase is reproduced regularly after a certain driving time or a certain distance traveled by said driver, in order to ensure that said personalized detection threshold associated with said driver is always representative. said conductive profile is stored in memory when it does not correspond to any already known conducting profile; and / or - said transmitting step also comprises transmitting to said alert module of said vehicle, the order to operate certain driving assistance devices of which said vehicle is equipped.

Brief description of the drawings The description of the invention will now be continued by the detailed description of an exemplary embodiment, given below by way of illustration but without limitation, with reference to the appended drawings, in which: FIG. a block diagram of a vigilance decrease detection system adapted to implement the method according to the invention; and FIG. 2 represents a flowchart of the method for detecting the driver's vigilance decrease according to the invention. DETAILED DESCRIPTION OF THE EMBODIMENT The driver's hypovigilance detection system 1 illustrated in FIG. 1 comprises a first camera 10 directed on the driver's face, a second camera 11 facing the road, a sensor 12 vehicle speed, a sensor 13 of the engaged gear ratio, a steering wheel angle sensor 14, a steering wheel speed sensor 15, a brake pedal position sensor 16 and acceleration, a detection and control unit 20 and an alert module 30. The data collected by the two cameras 10, 11 and the sensors 12, 13, 14, 15 and 16 are conveyed in real time to the camera. detection and control unit 20. This comprises a computer 21, a storage module 22 which comprises non-volatile memory type EEPROM or FLASH and RAM, as well as input interfaces 23 and output 24. The nonvolatile memory tocke a process for detecting the vehicle driver's vigilance decrease which is implemented in this unit 20 and whose flowchart is shown in FIG. 2. According to a preferred embodiment, all the information contained in this non-volatile memory can be updated by means of communication or reading means of a data medium. The process that is implemented at the start of the rolling of the vehicle, will determine the state of alertness of the driver based on data received from the two cameras 10, 11 and sensors 12, 13, 14, 15 and 16.

If the process judges that the driver has entered a hypovigilance phase, the control unit 20 will send a signal to the alert module 30 so that the latter sends a warning message to its attention and / or actuates certain devices. driver assistance system with which the vehicle is equipped (ADAS).

We will now describe in detail and in support of the flowchart in Figure 2, the different stages of this process. In step 100, the latter determines the conductive profile by means of facial analysis of its face from measurements made on the images transmitted by the first camera 10.

The computer 21 will then compare the conductive profile determined in step 100 with all the conductive profiles stored in the non-volatile memory of the storage module 22 (step 200). If this conductive profile does not correspond to any of the known profiles, the process will record this new conductive profile in this non-volatile memory of the storage module 22 (step 300). The process then continues with step 500 during which the computer 21 compares the driver's current level of vigilance Nvc with a standard detection threshold Sm of the level of vigilance limit.

This current level of alertness Nvc of the driver is determined using the current values of certain physiological and behavioral parameters of the driver, such as the focusing time, the frequency and duration of closure of the eyelids, the frequency and the speed of the movements. eyepieces, the heart rate, the degree of oscillation of the vehicle trajectory or the frequency of correction of the trajectory via the steering wheel. These physiological and behavioral parameters are calculated by means of the simultaneous analysis: of the driver's face from measurements made on the images transmitted by the first camera 10 (position of the eyelids and pupils, orientation of the head, direction of gaze) ; and - the behavior of the vehicle based on the measurements made on the images transmitted by the second camera 11 (position of the vehicle in its lane, with respect to the obstacles of the environment and to the signaling), and dynamic data provided by the sensor 12 of the vehicle speed, the gearbox gearbox sensor 13 engaged, the steering wheel steering angle sensor 14, the steering wheel rotation speed sensor 15 and the brake pedal position sensor 16 and acceleration. The standard detection threshold Sm is determined in advance by studying a wide representative panel of the set of conductors, and stored in the non-volatile memory of the storage module 22. In order to reduce the false alerts, this one takes into account only the most risky situations (long blinks, high frequencies of long blinks, strong oscillations of trajectories). If the current Nvc level of alertness of the driver is greater than or equal to the standard detection threshold SM / the process considers that this driver is sufficiently attentive and vigilant, so that he decides not to transmit any order to the module of alert 30 (step 700). On the other hand, if the driver's current level of vigilance Nvc is below this standard detection threshold SM, the process deduces that the driver is in the hypovigilance phase, so that he decides to transmit the module to the module. alert, the order to send a warning message (visual, audible and / or haptic) to the driver (step 800). When the vehicle is equipped with certain driving aids devices, it is also possible that an activation order of these devices is also transmitted to the alert module during this step 800. We will now return to the step 200 in which the computer 21 compares the conductive profile determined in step 100 with all the known profiles, stored in the non-volatile memory of the storage module 22. If this conductive profile corresponds to one of the stored driver profiles in memory, the process will check during a step 400 if a custom detection threshold SDp of the level of vigilance limit has been associated with this driver.

As will be discussed in detail below, each personalized detection threshold SDp is calculated according to the nominal vigilance level of the driver determined at the end of a learning phase. If no personalized detection threshold SDp is still associated with the driver, the process proceeds to step 500 where the detection of driver hypovigilance is carried out as indicated above by comparing the level of current vigilance Nvc of the driver to the threshold On the other hand, if such a personalized detection threshold SDp is associated with the driver, the process proceeds to step 600 during which the computer 21 compares the driver's current level of vigilance Nvc with that driver. Personalized detection threshold SDP- If this level of current alertness Nvc of the driver is greater than or equal to the personalized detection threshold SDp, the process considers that this driver is sufficiently attentive and vigilant, so that he decides not to transmit any order to the intention of the alert module 30 (step 700). Conversely, if the current level of vigilance Nvc of the driver is lower than this custom detection threshold SDp, the process deduces that the driver is in a phase of hypovigilance, so that he decides to transmit to the module d alert, the order to send a warning message (visual, audible and / or haptic) to the driver (step 800). We will now return in more detail on the operation of the learning phase. This is done automatically in parallel with the process of detecting the driver's loss of vigilance during periods of time during which at least one of the following conditions is met: - the current schedule is within a range of associated time slot according to the circadian rhythm (biological clock) at a maximum waking state (for example between 9 and 11 o'clock or between 16 and 18 o'clock); - the driving conditions are favorable (conditions requiring attention and generating awakening such as driving in the city with traffic); and / or the current level of vigilance of the driver is greater than or equal to a second predefined detection threshold of the level of vigilance limit, different and higher than the standard detection threshold SDS used during step 500 mentioned above.

Alternatively, it can be provided that the learning is performed more selectively, only when several of these conditions are met (or even all). The learning stops when the time spent or the number of kilometers traveled under these conditions has reached a predefined duration or number of kilometers. Therefore, a personalized detection threshold SDp will be associated with the corresponding driver profile, this threshold being calculated according to the nominal vigilance level determined at the end of this learning phase.

This level of nominal driver vigilance is determined using the nominal values of certain physiological and behavioral parameters of the driver, such as the focusing time, the frequency and duration of closure of the eyelids, the frequency and the speed of the eye movements. , the heart rate, the degree of oscillation of the vehicle trajectory or the frequency of correction of the trajectory via the steering wheel. These nominal physiological and behavioral parameters are calculated by means of the simultaneous analysis: of the driver's face from the measurements made on the images transmitted by the first camera 10 (position of the eyelids and pupils, orientation of the head, direction of gaze ); and - the behavior of the vehicle based on the measurements made on the images transmitted by the second camera 11 (position of the vehicle in its lane, with respect to the obstacles of the environment and to the signaling), and dynamic data provided by the sensor 12 of the vehicle speed, the gearbox gearbox sensor 13 engaged, the steering wheel steering angle sensor 14, the steering wheel rotation speed sensor 15 and the brake pedal position sensor 16 and acceleration. After having been calculated, these nominal physiological and behavioral parameters are then saved in the non-volatile memory of the storage module 22 and associated with the conductive profile. Specifically, this level of nominal vigilance corresponds to the moment in which the values recorded for these parameters were the most favorable, provided that these are not too far removed from the average values recorded for these parameters during the learning phase ( an excessively large deviation could indeed reveal an accidental value).

In this case, a new learning phase must be performed before a personalized detection threshold is associated with the driver. In addition, in order to ensure that the personalized detection threshold SDp associated with a driver is always representative, the learning phase is reproduced regularly after a certain driving time or a certain distance traveled by this driver . According to variants not shown, the driver's current and nominal vigilance levels are determined solely by means of physiological parameters, such as the focusing time, the frequency and duration of closure of the eyelids, the frequency and the speed of the movements. eyepieces or heart rate. According to other variants not shown, the dynamic data for analyzing the behavior of the vehicle can come from other types of onboard sensors. Of course, the present invention is not limited to the embodiments described and shown, but encompasses any alternative embodiment.

Claims (9)

REVENDICATIONS1. A method for detecting the vigilance of the driver of a motor vehicle, said method comprising the following steps: comparing the current vigilance level (Nvc) of said driver at the wheel of said vehicle with a detection threshold (SDs, SDP) of the alertness level limit (500; 600); transmitting, to an alert module (30) of said vehicle, the order to send a warning message to said driver (800); said transmission step (800) being implemented as soon as said current vigilance level (Nvc) is below said detection threshold (SDs; SDP); said method being characterized in that it comprises previously in said comparison step (500; 600), the following steps: - determination of the conductive profile by facial recognition (100); selection of the detection threshold (SDs, SDP) to be adopted for said conductor (200, 400), said selection step consisting of: when said conductive profile corresponds to a profile already known and associated with a personalized detection threshold (SDP) to opt for said personalized detection threshold (SDP); - when these conditions are not met, choose a predefined standard detection threshold (SDs); said personalized detection threshold (SDP) being calculated according to the nominal vigilance level of said driver determined at the end of a learning phase. 2. A method of detecting the decline in alertness according to claim 1, characterized in that the current and nominal vigilance levels of said conductor are determined at least partially by the aid of certain physiological parameters, such as the focusing time, the frequency and the eyelid closing time, the frequency and the speed of eye movements or the heart rate, said physiological parameters themselves being calculated by analyzing the face of said driver from the measurements made on the images transmitted by a first camera (10) directed on said face. 3. A method of detecting the decline in alertness according to claim 2, characterized in that said conductive profile is determined by the facial analysis of his face from measurements made on the images transmitted by said first camera (10). 4. A method of detecting the decline in alertness according to one of claims 2 or 3, characterized in that the current and nominal vigilance levels of said driver are also determined using certain behavioral parameters, such as the degree of alertness. oscillation of the trajectory or the frequency of correction of the trajectory via the steering wheel, said behavioral parameters being themselves calculated by analyzing the behavior of said vehicle from measurements made on the images transmitted by a second camera (11) directed to the road, and dynamic data provided by sensors embedded in said vehicle (12, 13, 14, 15, 16). 5. Vigilance drop detection method according to claim 4, characterized in that said onboard sensors comprise a speed sensor (12) of said vehicle, a gearbox gearbox sensor (13) engaged, a sensor (14). ) steering wheel angle of rotation, a sensor (15) of the speed of rotation of the steering wheel and a sensor (16) of the position of the brake and acceleration pedals. 6. A method of detecting the decline in alertness according to one of claims 1 to 5, characterized in that said learning phase is performed automatically during time intervals when certain conditions relating to the time slot, the conditions. driving and / or at the current vigilance level of said driver are met, said learning phase ceasing when the time spent or the number of kilometers traveled in said conditions has reached a predefined duration or number of kilometers. 7. A method of detecting the decline in alertness according to one of claims 1 to 6, characterized in that said learning phase is reproduced regularly after a certain driving time or a certain distance traveled by said driver, to ensure that said personalized detection threshold (SDP) associated with said driver is always representative. 8. Vigilance decline detection method according to one of claims 1 to 7, characterized in that said conductive profile is stored in memory when it does not correspond to any known conductor profile. 9. Vigilance drop detection method according to one of claims 1 to 8, characterized in that said transmission step (800) also comprises transmitting to said alert module (20) of said vehicle, the order of actuate certain driving assistance devices of which said vehicle is equipped.