A method and device for detecting an abnormal driving state

Abstract

The application provides a driving state anomaly detection method and device, wherein the method comprises: obtaining navigation planning data, driving trajectory, front road environment data, and operation signal generated by the driver operating the vehicle of the vehicle; calculating the characteristic value of the navigation compliance feature according to the navigation planning data and the driving trajectory; calculating the response delay amount feature and the response delay stability feature of the driver's operation response to the road sign, the speed stability feature and the direction stability feature of the driver operating the vehicle, and the perception response integrity feature according to the front road environment data and the operation signal; and determining whether the driving state of the driver is abnormal according to the navigation compliance feature, the response delay amount feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature calculated in the plurality of continuous sliding time windows. The method ensures the accuracy and timeliness of the detection.

Description

Technical Field

[0001] This application relates to the field of safe driving technology, and in particular to a method and device for detecting abnormal driving conditions. Background Technology

[0002] With the continuous growth of car ownership, traffic accidents caused by driver abnormalities (such as drunk driving or fatigued driving) occur frequently. Therefore, real-time and accurate detection of driver abnormalities during driving is of great significance for ensuring road safety.

[0003] Currently, the main methods for detecting abnormal driving conditions include: detection based on driver facial images and detection based on driving behavior.

[0004] Driver fatigue detection methods based on driver facial images involve capturing images of the driver's face using an in-vehicle camera and analyzing facial features such as eyelid closure and yawning to determine the level of fatigue. However, this method is highly susceptible to changes in lighting conditions and facial obstructions such as the driver wearing glasses or sunglasses, making it difficult to guarantee accuracy under different driving environments.

[0005] Driving behavior-based detection methods refer to judging driving status by monitoring the vehicle's dynamic performance. For example, by monitoring operational indicators such as whether the vehicle frequently deviates from its lane or whether there are abnormal fluctuations in speed, an alarm is triggered when these indicators exceed preset thresholds. This method focuses on abnormal behavior at the vehicle's operational level. However, when the driver exhibits obvious abnormalities at the operational level, the driving condition is often already in a stage of continuous deterioration. By the time an alarm is issued, the time left for the driver or vehicle system to take countermeasures is very limited. Summary of the Invention

[0006] In view of this, the purpose of this application is to provide a method and device for detecting abnormal driving conditions, so as to ensure the accuracy and timeliness of detection.

[0007] In a first aspect, embodiments of this application provide a method for detecting abnormal driving conditions, including: It can acquire real-time navigation planning data, driving trajectory, forward road environment data, and operation signals generated by the driver during the driving process. Based on the navigation planning data and the driving trajectory within the current sliding time window, calculate the feature value of the navigation compliance feature; Based on the forward-looking road environment data and the operation signal within the current sliding time window, the characteristic values ​​of the response delay and response delay stability of the driver's operation response to road signs, the characteristic values ​​of the speed stability and directional stability of the driver's vehicle control, and the characteristic value of the perception response integrity are calculated; wherein, the perception response integrity is used to characterize the coverage ratio of the driver's actual operation response to the road signs that should be responded to. Based on the feature values ​​of the navigation compliance feature, the response delay feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature calculated within multiple consecutive sliding time windows, it is determined whether the driver's driving state is abnormal.

[0008] In conjunction with the first aspect, this application provides a first possible implementation of the first aspect, wherein calculating the feature value of the navigation compliance feature based on the navigation planning data and the driving trajectory within the current sliding time window includes: Extract navigation action points from the navigation planning data within the current sliding time window; the navigation action points are the positioning points in the navigation planning data that indicate that the driver needs to perform steering or fork operations; Spatial matching is performed between the driving trajectory within the current sliding time window and the navigation action points, and the number of navigation action points that the driving trajectory did not pass through correctly is counted. Calculate the ratio of the number of navigation action points that were not traversed correctly to the total number of navigation action points, and use this ratio as the feature value of the navigation compliance feature within the current sliding time window.

[0009] In conjunction with the first aspect, this application provides a second possible implementation of the first aspect, wherein calculating the characteristic values ​​of the response delay amount and the response delay stability characteristics of the driver's operational response to the road sign based on the forward-looking road environment data and the operation signal within the current sliding time window includes: From the forward-looking road environment data within the current sliding time window, identify road signs that require driver action, and record the first time each road sign is identified as the anchor point time for each road sign; For each road sign identified within the current sliding time window, starting from the anchor point time of the road sign, within a detection window of a preset duration, it is detected whether there is an operation response action corresponding to the road sign in the operation signal. If there is, the time when the operation response action first appears is taken as the response time of the road sign. For each road sign that has an operational response action, calculate the time difference between the response time of the road sign and the anchor point time, and use it as the driver's response time to the road sign; The road signs within the current sliding time window are divided into types. For each type of road sign, the average response time of all road signs under that type is calculated. Then, the ratio of this average to the driver's baseline response time for that type of road sign is calculated. This ratio is used as the driver's response delay ratio for that type of road sign. The mean of all types of response delay ratios is calculated as a feature value of the response delay amount characteristic of the driver's operation response to road signs within the current sliding time window; wherein, the response delay amount characteristic characterizes the driver's response delay level to road signs; Calculate the standard deviation of the response delay ratio of all road signs within the current sliding time window, and use this standard deviation as the feature value of the response delay stability characteristic of the driver's operation response to all road signs within the current sliding time window; wherein, the response delay stability characteristic characterizes the degree of dispersion of the driver's response delay to multiple road signs.

[0010] In conjunction with the first aspect, this application provides a third possible implementation of the first aspect, wherein the operation signal includes a vehicle speed sequence; the step of calculating the characteristic values ​​of the speed stability feature and the characteristic values ​​of the directional stability feature of the driver's vehicle operation based on the forward-looking road environment data within the current sliding time window and the operation signal includes: Calculate the ratio between the standard deviation and the mean of the vehicle speed sequence within the current sliding time window, and use this ratio as a feature value of the speed stability characteristic of the driver's vehicle control. The vehicle lateral offset and lane width at each sampling moment within the current sliding time window are extracted from the forward-looking road environment data; wherein, the vehicle lateral offset is the lateral distance of the vehicle center from the current lane center; For each sampling time, the ratio of the vehicle's lateral offset to the lane width is calculated as the normalized lateral offset. Calculate the standard deviation of all the normalized lateral offsets within the current sliding time window, and use it as the lane deviation. Then, use the lane deviation as a feature value of the directional stability characteristic of the driver's vehicle control.

[0011] In conjunction with the second possible implementation of the first aspect, this application provides a fourth possible implementation of the first aspect, wherein calculating the feature value of the perception response integrity feature based on the forward-looking road environment data and the operation signal within the current sliding time window includes: For each road sign identified that requires driver action, determine whether the driver has performed an action corresponding to the road sign within a preset detection window. The number of road signs that are determined not to have taken any action within the current sliding time window is counted as the number of unresponsive signs. Calculate the ratio of the number of unresponsive signs to the total number of road signs identified within the current sliding time window, and use this ratio as the feature value of the perceived response integrity feature.

[0012] In conjunction with the first aspect, this application provides a fifth possible implementation of the first aspect, wherein determining whether the driver's driving state is abnormal based on the feature values ​​of the navigation compliance feature, the response delay feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature calculated within multiple consecutive sliding time windows includes: For each of the navigation compliance feature, the response delay feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature, a temporal feature value sequence for each feature is determined within multiple consecutive sliding time windows, including the current sliding time window; the temporal feature value sequence includes the feature value of the feature in each sliding time window within the multiple consecutive sliding time windows. For each of the aforementioned features, the ratio of the feature value of each sliding time window in the time-series feature value sequence of that feature to the individual baseline value corresponding to that feature is calculated, and this ratio is used as the deviation factor for each sliding time window corresponding to that feature. The number of windows in each sliding time window corresponding to this feature whose deviation multiple exceeds a preset deviation threshold is counted as the number of abnormal windows for this feature. The total number of abnormal windows for all features is calculated to obtain the total number of abnormal windows. Based on the relationship between the total number of abnormal windows and a preset abnormal threshold, it is determined whether there are abnormal signs in the driving behavior. If no abnormal signs are found, the driver's driving status is determined to be normal. If any abnormal signs are found, the system determines whether the driver's driving state is abnormal and the cause of the abnormal driving state based on the first preset deviation multiple range of each feature under the abnormal bodily function caused by the substance and the second preset deviation multiple range under fatigue, as well as the deviation multiple of each feature in the current sliding time window.

[0013] In conjunction with the fifth possible implementation of the first aspect, this application provides a sixth possible implementation of the first aspect, wherein determining whether the driver's driving state is abnormal and the cause of the abnormal driving state includes: For each feature, if the deviation multiple of the feature in the current sliding time window falls within the first preset deviation multiple range corresponding to the feature, then the feature is taken as the first feature, and the sum of the product results of each first feature is taken as the body abnormality score according to the product of the preset weight of each first feature and the specified base; and if the deviation multiple of the feature in the current sliding time window falls within the second preset deviation multiple range corresponding to the feature, then the feature is taken as the second feature, and the sum of the product results of each second feature is taken as the fatigue score according to the product of the preset weight of each second feature and the specified base. If the abnormal score of the body is greater than or equal to the first preset threshold, it is determined that the driver's driving state is abnormal, and the abnormal driving state is caused by a physical dysfunction caused by a substance. If the abnormal score of the body is less than the first preset threshold and the fatigue score is greater than or equal to the second preset threshold, then it is determined that the driver's driving state is abnormal and the abnormal driving state is caused by fatigue; the first preset threshold is greater than the second preset threshold. If the abnormal score of the body is less than the first preset threshold and the fatigue score is less than the second preset threshold, then for each feature, the variance of the feature value of the feature in each sliding time window is calculated to obtain the inter-window variance of the feature, and the mean of all the inter-window variances is calculated. If the average value is greater than the third preset threshold, it is determined that the driver's driving state is abnormal and that the abnormal driving state is caused by distraction. If the average value is greater than or equal to the fourth preset threshold and less than or equal to the third preset threshold, it is determined that the driver's driving state is abnormal and that the abnormal driving state is caused by fatigue. If the average value is less than the fourth preset threshold, it is determined that the driver's driving status is not abnormal and is considered a normal fluctuation.

[0014] In conjunction with the sixth possible implementation of the first aspect, this application provides a seventh possible implementation of the first aspect, wherein the method further includes: In the initial stage of establishing personal baseline values ​​for each feature, feature values ​​of each feature during driving are collected for the navigation compliance feature, the speed stability feature, the direction stability feature, and the perception response integrity feature, until the cumulative driving time reaches a first preset time and the cumulative number of recognitions of each type of road sign involved in the response delay feature reaches a preset number. The average value of the feature values ​​of each feature that has been collected is then calculated as the initial personal baseline value for each feature. For the response delay quantity characteristic and the response delay stability characteristic, preset empirical values ​​are used as the initial personal baseline values ​​for each characteristic; After establishing the individual baseline values ​​for each characteristic for the first time, when it is determined that the driver's driving behavior is not abnormal, the individual baseline values ​​for each characteristic are updated in the following manner:

[0015] in, For the updated individual baseline value, The individual baseline value before the update. The first preset learning rate, The feature value is the value of the feature calculated within the current sliding time window.

[0016] In conjunction with the second possible implementation of the first aspect, this application provides an eighth possible implementation of the first aspect, wherein the method further includes updating the response baseline duration of various road signs in the following manner: When it is determined that the driver's driving behavior is not abnormal, for each type of road sign, the average response time of each road sign in that type of road sign is calculated within the current sliding time window. Update the response baseline duration for this type of road sign using the following formula:

[0017] in, The updated response baseline duration for this type of road sign. The baseline response time for this type of road sign before the update. To preset the learning rate, The mean of the response duration is denoted as .

[0018] Secondly, embodiments of this application also provide a detection device for abnormal driving conditions, comprising: The acquisition module is used to acquire in real time the vehicle's navigation planning data, driving trajectory, forward road environment data, and operation signals generated by the driver during the driving process; The first calculation module is used to calculate the feature value of the navigation compliance feature based on the navigation planning data and the driving trajectory within the current sliding time window; The second calculation module is used to calculate, based on the forward-looking road environment data and the operation signal within the current sliding time window, the characteristic values ​​of the response delay feature, the characteristic values ​​of the response delay stability feature, the characteristic values ​​of the speed stability feature and the directional stability feature of the driver's operation response to road signs, and the characteristic value of the perception response integrity feature; wherein, the perception response integrity feature is used to characterize the coverage ratio of the driver's actual operation response to the road signs that should be responded to. The determination module is used to determine whether the driver's driving state is abnormal based on the feature values ​​of the navigation compliance feature, the response delay feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature calculated within multiple consecutive sliding time windows.

[0019] This application provides a method and apparatus for detecting abnormal driving states. It acquires real-time navigation planning data, driving trajectory, forward-looking road environment data, and driver operation signals during driving. Within a current sliding time window, it calculates feature values ​​for navigation compliance, response delay, response delay stability, speed stability, directional stability, and perception response integrity. Based on these feature values ​​calculated over multiple consecutive sliding time windows, it determines whether the driver's driving state is abnormal. On one hand, this method does not rely on capturing and analyzing driver facial images using an onboard camera. Instead, it detects abnormalities based on the driver's actual response to navigation planning data and road signs. Therefore, it is not affected by changes in lighting or facial obstructions such as the driver wearing glasses or sunglasses. It maintains stable detection capabilities under different driving environments, thus ensuring detection accuracy. On the other hand, this method reflects the driver's execution of navigation instructions by calculating the feature values ​​of navigation compliance characteristics, and reflects the driver's perception and response ability to road signs by calculating the feature values ​​of response delay quantity characteristics and response delay stability characteristics. These features measure the driver's performance at the cognitive decision-making level, rather than whether the vehicle has experienced lane departure, abnormal speed fluctuations, or other phenomena at the operational level. Since changes in the driver's cognitive decision-making level often precede obvious abnormalities at the operational level, this method can detect abnormal signs at an earlier stage of deterioration in driving condition, giving the driver or vehicle system more time to respond, thereby ensuring the timeliness of detection.

[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart of a method for detecting abnormal driving conditions provided in an embodiment of this application is shown; Figure 2 This illustration shows a schematic diagram of a front view area provided in an embodiment of this application; Figure 3 A flowchart illustrating the calculation of a navigation compliance feature provided in an embodiment of this application is shown; Figure 4 A schematic diagram of the structure of a driving state abnormality detection device provided in an embodiment of this application is shown. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0024] Currently, the main methods for detecting abnormal driving states include: detection based on driver facial images and detection based on driving behavior.

[0025] Driver fatigue detection methods based on driver facial images involve capturing images of the driver's face using an in-vehicle camera and analyzing facial features such as eyelid closure and yawning to determine the level of fatigue. However, this method is highly susceptible to changes in lighting conditions and facial obstructions such as the driver wearing glasses or sunglasses, making it difficult to guarantee accuracy under different driving environments.

[0026] Driving behavior-based detection methods refer to judging driving status by monitoring the vehicle's dynamic performance. For example, by monitoring operational indicators such as whether the vehicle frequently deviates from its lane or whether there are abnormal fluctuations in speed, an alarm is triggered when these indicators exceed preset thresholds. This method focuses on abnormal behavior at the vehicle's operational level. However, when the driver exhibits obvious abnormalities at the operational level, the driving condition is often already in a stage of continuous deterioration. By the time an alarm is issued, the time left for the driver or vehicle system to take countermeasures is very limited.

[0027] Based on this, this application provides a method and apparatus for detecting abnormal driving conditions to ensure the accuracy and timeliness of detection, which will be described below through embodiments.

[0028] To facilitate understanding of this embodiment, a method for detecting abnormal driving conditions disclosed in this application will first be described in detail. For example... Figure 1 As shown, the process includes the following steps S101-S104: S101: Real-time acquisition of vehicle navigation planning data, driving trajectory, forward road environment data, and operation signals generated by the driver during driving.

[0029] In this embodiment, navigation planning data refers to the route guidance information generated by the vehicle navigation system based on the destination set by the driver and the current road information. It includes the route the vehicle is about to travel and the turning or branching operation instructions that the driver needs to perform at each location point along the route.

[0030] Driving trajectory refers to the sequence of actual driving trajectory points collected by a positioning system (such as GPS positioning) at certain time intervals during the actual driving process of a vehicle, reflecting the vehicle's true travel path.

[0031] Forward-looking road environment data refers to the image information of the road ahead of the vehicle collected by a camera installed in front of the vehicle. After image recognition processing, structured information such as lane line positions, lane widths, and the type and content of road signs (i.e., traffic signs) can be obtained. For example, Figure 2 As shown, forward view refers to the camera's shooting direction being towards the vehicle's forward direction (i.e., the driving direction), that is, facing the road scene in front of the vehicle to obtain information about the road environment directly related to the current driving operation.

[0032] Operating signals refer to signals generated by the driver's control of the vehicle, including vehicle speed sequence, gear signal, steering wheel angle signal, brake pressure signal, etc. These signals are transmitted through the vehicle's CAN bus and reflect the driver's real-time control actions on the vehicle.

[0033] In one possible implementation, when performing step S101 to acquire the aforementioned multi-source data, the specific steps can be as follows: S1011: Real-time acquisition of the vehicle's original navigation planning data, original driving trajectory, original forward road environment data, and original operation signals generated by the driver during the driving process.

[0034] In this embodiment, the raw navigation planning data is transmitted once every 200 milliseconds. The acquisition frequency of the raw driving trajectory is related to the positioning system. The acquisition interval of image frames in the raw forward-looking road environment data is approximately 33 milliseconds. The raw operation signals are transmitted via the vehicle's CAN bus at a fixed frequency, for example, at a transmission interval of 10 milliseconds.

[0035] S1012: Perform synchronization alignment and anomaly filtering on the original navigation planning data, original driving trajectory, original forward-looking road environment data, and original operation signals to obtain preprocessed navigation planning data, driving trajectory, forward-looking road environment data, and operation signals.

[0036] Specifically, due to differences in the acquisition frequency and transmission latency of various data sources, directly using unprocessed raw data will affect the accuracy of subsequent feature calculations. To address this issue, this step first aligns the data with different sampling rates to the same timestamp system. Specifically, the data with the lowest sampling frequency, i.e., the navigation planning data, is used as the time base. Data from other data streams that fall between two adjacent navigation planning data timestamps are aggregated or interpolated, ensuring that all data are output as synchronized data frames at 200 millisecond intervals. For data streams with significant differences in sampling rates, multi-rate Kalman filtering is used to eliminate system errors introduced by multi-sensor transmission delays, ensuring consistency of data across the time dimension.

[0037] Building upon this, this step also performs outlier filtering on the synchronized data, employing chi-square tests or residual analysis to identify anomalies in the observations of each data stream. When an observation is identified as an outlier, it is excluded from subsequent data usage to prevent it from interfering with subsequent feature calculations.

[0038] S102: Calculate the feature value of navigation compliance characteristics based on the navigation planning data and driving trajectory within the current sliding time window.

[0039] In this embodiment, a sliding time window refers to a fixed-length time interval that is captured by sliding forward a fixed step size on a continuous time series. For example, the window length is 5 minutes and the sliding step size is 1 minute. The current sliding time window refers to the sliding time window closest to the current time, and the data in the window is collected within the most recent period.

[0040] In one possible implementation, such as Figure 3 As shown, when performing step S102, the specific steps S1021-S1023 can be executed as follows: S1021: Extract navigation action points from the navigation planning data within the current sliding time window; navigation action points are the positioning points in the navigation planning data that indicate the driver needs to perform steering or fork operations.

[0041] In this step, the navigation planning data includes a series of navigation planning location points along the planned route. Most of these location points are simply used to indicate the route direction, and drivers can pass them directly while driving normally. Navigation action points, on the other hand, are location points that require active driver intervention, such as highway exits, intersection turning points, and ramp forks. This step filters out the location points with turning or fork operation commands from all navigation planning location points within the current sliding time window, and uses these as navigation action points. For example, within a 5-minute sliding time window, if the navigation planning data includes a section of ordinary straight road and an exit ramp, only the location point corresponding to the exit ramp will be extracted as a navigation action point.

[0042] S1022: Spatial match the driving trajectory within the current sliding time window with navigation action points, and count the number of navigation action points that the driving trajectory did not pass through correctly.

[0043] In this step, the sequence of actual vehicle trajectory points (i.e., the driving trajectory) collected within the current sliding time window is spatially compared with the positioning coordinates of each navigation action point. If the distance between the coordinates of a navigation action point and a point on the actual vehicle trajectory is less than a preset distance threshold (e.g., 10 meters), and the navigation system does not trigger route replanning after the vehicle passes the location of the navigation action point, then the navigation action point is determined to have been correctly passed. Conversely, if the vehicle leaves without passing the navigation action point, or if the navigation system triggers route replanning after passing it, then the navigation action point is determined to have not been correctly passed. After performing the above judgment on all navigation action points, the number of navigation action points determined to have not been correctly passed is counted.

[0044] S1023: Calculate the ratio of the number of navigation action points that were not traversed correctly to the total number of navigation action points, and use this ratio as the feature value of the navigation compliance feature within the current sliding time window.

[0045] For example, if there are 4 navigation action points within the current sliding time window, of which 3 were correctly traversed and 1 was not, then the feature value of the navigation compliance feature is 0.25. The lower the feature value of the navigation compliance feature, the more accurately the driver can execute driving operations according to the navigation planned path; the higher the feature value of the navigation compliance feature, the more the driver's ability to execute navigation commands has decreased.

[0046] S103: Based on the forward-looking road environment data and operation signals within the current sliding time window, calculate the characteristic values ​​of the response delay, response delay stability, speed stability, and directional stability of the driver's operation response to road signs, as well as the characteristic value of the perception response integrity. The perception response integrity is used to characterize the coverage ratio of the driver's actual operation response to the road signs that should be responded to.

[0047] In one possible implementation, when performing step S103, which calculates the characteristic values ​​of the response delay amount and the response delay stability characteristics of the driver's operational response to road signs based on the forward-looking road environment data and operation signals within the current sliding time window, the specific steps S10311-S10316 can be performed as follows: S10311: Identify road signs that require driver action from the forward-looking road environment data within the current sliding time window, and record the first time each road sign is identified as the anchor point time for each road sign.

[0048] In this step, the forward-looking road environment data includes information on the type and content of road signs obtained through image recognition processing. Types of road signs include, for example, speed limit signs, prohibitory signs, and directional signs.

[0049] This step iterates through all road sign recognition results within the current sliding time window, filtering out those that require driver action. Road signs requiring driver action are those whose traffic rules dictate corresponding driving maneuvers, such as speed limit signs requiring drivers to adjust their speed, prohibitory signs prohibiting drivers from entering or overtaking, and directional signs requiring drivers to travel in a designated direction. Warning signs (such as those indicating sharp bends ahead) are not included in the scope of road signs requiring driver action because they typically do not require immediate action from the driver.

[0050] For each road sign identified as requiring action, the moment it first appears in the forward-looking road environment data within the current time window is recorded. That is, the moment corresponding to the frame in which the forward-looking camera first identifies the road sign within the current time window is recorded as the anchor point moment for that road sign, denoted as t0. If the same road sign is identified consecutively in multiple frames, only the moment of the first identification is used as its anchor point moment.

[0051] S10312: For each road sign identified within the current sliding time window, starting from the anchor point time of the road sign, within a detection window of a preset duration, detect whether there is an operation response action corresponding to the road sign in the operation signal. If there is, the time when the operation response action first appears is taken as the response time of the road sign.

[0052] In this step, for each road sign identified within the current sliding time window, a detection window of a preset duration, such as 5 seconds, is opened starting from its anchor point time t0. Within this detection window, the vehicle speed sequence, gear signal, steering wheel angle signal, and brake pressure signal in the operation signals are scanned frame by frame to determine whether the driver has made an operation response action corresponding to the road sign within the detection window.

[0053] For example, for a speed limit sign, the judgment condition for its corresponding operation response action can be: the value of the braking pressure signal exceeds a preset braking threshold (e.g., 1 bar) and the duration exceeds a preset duration (e.g., 200 milliseconds). If an operation response action that meets the judgment condition is detected within the detection window, the time corresponding to the frame that first meets the judgment condition is recorded as the response time of the road sign, denoted as t1.

[0054] In this embodiment, if no operation response action corresponding to the road sign is detected within the detection window, the road sign is considered an unresponsive event.

[0055] It should be noted that before detecting the operational response, the driving scenario at the anchor point can be filtered to remove invalid events unsuitable for inclusion in the calculation. For example, if the vehicle speed at anchor point t0 is lower than a preset speed threshold (e.g., 5 km / h), it indicates that the vehicle is in a congested or stopped state. In this case, the driver's behavior is not a normal response to the road sign, and the road sign is excluded. Similarly, if the driver is already performing braking at anchor point t0 (i.e., the brake pressure signal value has exceeded a preset braking threshold), it is impossible to distinguish whether the subsequent braking action is a response to the road sign or a continuation of the previous action, and the road sign is also excluded.

[0056] S10313: For each road sign with an operational response action, calculate the time difference between the response time of the road sign and the anchor point time, and use it as the driver's response time to the road sign.

[0057] In this step, for each road sign recorded as having an operational response action, its response time Δt = t1 is calculated. t0 is the time difference between the response time of the road sign and the anchor point time, measured in seconds. This response time reflects the time it takes for the driver to take action from seeing the road sign to initiating the necessary maneuver.

[0058] S10314: Divide the road signs within the current sliding time window into categories. For each category of road signs, calculate the average response time of all road signs under that category. Then calculate the ratio of this average to the driver's baseline response time for that category of road signs. Use this ratio as the driver's response delay ratio for that category of road signs.

[0059] In this step, all road signs recorded as having operational response actions within the current sliding time window are first divided into different type groups according to their type, such as speed limit signs, prohibitory signs, and instruction signs. For each type of road sign, the response time of all road signs under that type is summarized, and the arithmetic mean of these response times is calculated to obtain the average response time of that type of road sign within the current sliding time window.

[0060] The average response time is then divided by the driver's baseline response time to this type of road sign. The resulting ratio is the response delay ratio for that type of road sign. The baseline response time is the driver's average response time to this type of road sign under normal driving conditions. This can be determined through the collection and statistical analysis of historical driving data, or it can be set as a preset empirical value when the system is first used and then gradually updated through self-learning. The response delay ratio is a dimensionless value. A value of 1 indicates that the driver's average response speed to this type of road sign is comparable to their baseline level; a value greater than 1 indicates that the average response speed is slower than the baseline; and a value less than 1 indicates that the average response speed is faster than the baseline.

[0061] For example, the speed limit sign group has three road signs with response times of 1.2 seconds, 1.5 seconds and 1.8 seconds respectively, with an average response time of 1.5 seconds. However, the baseline response time for drivers to speed limit signs is 1.2 seconds. Therefore, the response delay ratio of the speed limit sign group is 1.25, which means that drivers' average response speed when they see speed limit signs is 25% slower than their normal level.

[0062] S10315: Calculate the mean of all types of response delay ratios as a feature value of the response delay quantity characteristic of the driver's operation response to road signs within the current sliding time window; wherein, the response delay quantity characteristic characterizes the driver's response delay level to road signs.

[0063] In this step, the response delay ratios of various road signs are summarized, and the arithmetic mean of these response delay ratios is calculated. This mean is the characteristic value of the response delay quantity feature. The response delay quantity feature integrates the response delay of different types of road signs into a single overall index, reflecting the driver's overall response delay level to road signs within the current sliding time window. The closer this characteristic value is to 1, the more the driver's overall response speed is comparable to their normal level; the larger the characteristic value, the slower the driver's overall response to road signs, indicating a significant decrease in information processing speed and operational execution speed. For example, if the current sliding time window involves three types of road signs—speed limit signs, prohibitory signs, and instruction signs—with response delay ratios of 1.25, 1.10, and 1.05 respectively, then the characteristic value of the response delay quantity feature is the average of these three, 1.13, indicating that the driver's overall response to road signs is approximately 13% slower than their normal level.

[0064] S10316: Calculate the standard deviation of the response delay ratio of all road signs within the current sliding time window, and use this standard deviation as the characteristic value of the response delay stability feature of the driver's operation response to all road signs within the current sliding time window; wherein, the response delay stability feature characterizes the degree of dispersion of the driver's response delay to multiple road signs.

[0065] In this step, the response delay ratios of all road signs recorded as having operational responses within the current sliding time window are aggregated to form a delay ratio dataset. The standard deviation of this delay ratio dataset is calculated, representing the degree to which each response delay ratio deviates from its mean. This standard deviation is used as the characteristic value of the response delay stability feature. This characteristic value reflects the consistency of the driver's response speed across multiple road signs. A small standard deviation (e.g., 0.1 to 0.2) indicates that the driver's response delay to each road sign is relatively stable and consistent, indicating normal driving behavior. A large standard deviation (e.g., exceeding 0.5) indicates significant fluctuations in the driver's response (e.g., some road signs respond close to normal levels while others are severely delayed). This instability usually points to fluctuations in the driver's cognitive executive function, possibly due to physical abnormalities caused by substances (such as alcohol or drugs) or severe fatigue.

[0066] In one possible implementation, the operation signal includes a vehicle speed sequence; when performing step S103, which calculates the characteristic values ​​of the speed stability feature and the characteristic values ​​of the directional stability feature of the driver's vehicle control based on the forward-looking road environment data and the operation signal within the current sliding time window, the specific steps S10321-S10324 can be performed as follows: S10321: Calculate the ratio between the standard deviation and the mean of the vehicle speed sequence within the current sliding time window, and use this ratio as the feature value of the speed stability characteristic of the driver's vehicle control.

[0067] In this step, the vehicle speed sequence refers to the data sequence of vehicle speed changes over time contained in the operating signal. Within the current sliding time window, the vehicle speed values ​​at all sampling moments are extracted to form the vehicle speed sequence. The standard deviation of this vehicle speed sequence is then calculated. and mean Then calculate the ratio of the two (i.e., the coefficient of variation): , The coefficient of variation is the eigenvalue of the speed stability characteristic. It is a dimensionless value that eliminates the influence of average vehicle speed on the measurement of fluctuation, allowing for a fair comparison of speed stability under different speed conditions. When the vehicle is stationary or at extremely low speeds (e.g., average speed below 5 km / h), the eigenvalue of the speed stability characteristic is set to zero to avoid abnormally amplified ratios due to extremely small mean values. A larger eigenvalue indicates less smooth driver control of the vehicle speed, with unstable operations such as rapid acceleration and deceleration; a smaller eigenvalue indicates smoother speed changes and more stable driving operation.

[0068] S10322: Extract the vehicle lateral offset and lane width at each sampling moment within the current sliding time window from the forward-looking road environment data; where the vehicle lateral offset is the lateral distance of the vehicle center from the current lane center.

[0069] In this step, the forward-looking road environment data includes vehicle lateral offset and lane width, which can be obtained through image processing. Vehicle lateral offset refers to the lateral distance of the vehicle's center from the current lane centerline, denoted by L, in meters. A positive value indicates the vehicle's center is offset to the left of the lane centerline, and a negative value indicates offset to the right. Lane width refers to the lateral distance between the left and right lane lines of the current lane, denoted by W, in meters, typically ranging from 3.0 meters to 3.75 meters.

[0070] S10323: For each sampling time, calculate the ratio of the vehicle's lateral offset to the lane width as the normalized lateral offset.

[0071] In this step, for each extracted sampling time point, the vehicle's lateral offset L and lane width W are used to calculate their ratio, lnorm = L / W. After normalization, the value of lnorm typically ranges from 0 at the lane center to approximately ±0.5 near the lane edge, and its absolute value exceeds 0.5 when outside the lane lines. This normalization process allows for a fair comparison of the degree of vehicle lateral offset on roads with different lane widths. For example, the normalized offset values ​​corresponding to the same 0.3-meter lateral offset differ on a narrow road with a lane width of 3.0 meters and a wide road with a lane width of 3.75 meters, thus more reasonably reflecting the degree of relative position deviation of the vehicle within the lane.

[0072] S10324: Calculate the standard deviation of all normalized lateral offsets within the current sliding time window, and use it as the lane deviation. The lane deviation is then used as a feature value of the directional stability characteristic of the driver's vehicle control.

[0073] In this step, the normalized lateral offsets of all sampling times within the current sliding time window are summarized to form a normalized offset sequence. The standard deviation of this normalized offset sequence is then calculated. The standard deviation is the degree to which each normalized lateral offset deviates from its mean. This refers to the lane departure rate within the current sliding time window. As a characteristic value of directional stability, lane departure rate reflects the driver's ability to maintain vehicle stability within the lane. A smaller lane departure rate (i.e., a characteristic value of directional stability) (e.g., 0.00 to 0.05) indicates that the vehicle hardly deviates from the center of the lane, and the driver's control of the steering wheel is extremely stable. A larger lane departure rate (e.g., exceeding 0.18) indicates that the vehicle exhibits significant lateral swaying, and the driver's lateral control ability has significantly decreased, possibly due to impaired cognitive function.

[0074] In one possible implementation, when performing step S103 to calculate the feature value of the perception response integrity feature based on the forward-looking road environment data and operation signal within the current sliding time window, the specific steps S10331-S10333 can be performed as follows: S10331: For each road sign that requires the driver to take action, determine whether the driver has made an operation response action corresponding to the road sign within a preset detection window.

[0075] In this step, the road signs that require driver action are the same as those identified in step S10311. For each road sign, the same detection window and operation response action determination conditions as in step S10312 are used. That is, starting from the anchor point time t0 of the road sign, within a detection window of a preset duration (e.g., 5 seconds), it is detected whether there is an operation response action corresponding to the road sign in the operation signal.

[0076] S10332: Count the number of road signs that are determined not to have taken any action within the current sliding time window, and use this as the number of unresponsive signs.

[0077] In this step, all road signs identified as requiring driver action within the current sliding time window are iterated through. Each sign is checked to see if it has been determined that no action has been taken, and the number of road signs determined to be non-responsive is counted. This number is recorded as the non-responsive sign count. For example, if there are 8 road signs requiring driver action within the current sliding time window, and 2 of them did not detect a corresponding action within the detection window, then the non-responsive sign count is 2.

[0078] S10333: Calculate the ratio of the number of unresponsive signs to the total number of road signs identified within the current sliding time window, and use this ratio as the feature value of the perception response integrity feature.

[0079] In this step, the total number of road signs identified within the current sliding time window refers to the total number of road signs requiring driver action. The ratio of the number of unresponsive signs to this total number is calculated; this ratio is the feature value of the perception-response integrity feature, ranging from 0 to 1. A higher feature value indicates that the driver missed responding to more road signs, resulting in less complete perception of road environment information; a lower feature value indicates that the driver responded to most of the road signs that should have been responded to, resulting in more complete perception coverage. The perception-response integrity feature, along with the response delay quantity feature and response delay stability feature, complements each other, jointly characterizing the driver's perception and response ability to road signs from two dimensions: response delay level and response coverage ratio.

[0080] S104: Based on the feature values ​​of navigation compliance, response delay, response delay stability, speed stability, direction stability, and perception response integrity calculated within multiple consecutive sliding time windows, determine whether the driver's driving state is abnormal.

[0081] In one possible implementation, when performing step S104, the following steps S1041-S1046 can be specifically performed: S1041: For each of the following features: navigation compliance feature, response delay amount feature, response delay stability feature, speed stability feature, direction stability feature, and perception response integrity feature, determine the temporal feature value sequence of each feature in multiple consecutive sliding time windows, including the current sliding time window; the temporal feature value sequence includes the feature value of the feature in each sliding time window in the multiple consecutive sliding time windows.

[0082] In this step, for each of the six features mentioned above, feature values ​​from the most recent N consecutive sliding time windows, including the current sliding time window, are collected. N is a preset number of windows, for example, N is 10. The feature values ​​from these N sliding time windows are arranged in chronological order to form the temporal feature value sequence of that feature.

[0083] S1042: For each feature, calculate the ratio of the feature value of each sliding time window in the time series feature value sequence of that feature to the personal baseline value corresponding to that feature, and use it as the deviation multiple of each sliding time window corresponding to that feature.

[0084] In this step, for each feature, the feature value of each sliding time window in the time-series feature value sequence is divided by the corresponding personal baseline value to obtain the deviation factor *dev* for each sliding time window. The personal baseline value is the average level of the feature under normal driving conditions, reflecting the driver's personalized driving behavior benchmark. The deviation factor measures the degree of deviation of the feature value in the current sliding time window from the personal baseline value. When the deviation factor is close to 1, it indicates that the value of the feature in the current window is comparable to the driver's normal level; when the deviation factor is greater than 1, it indicates that the feature value is higher than the normal level, and the larger the deviation factor, the more severe the deviation.

[0085] For example, taking navigation compliance features as an example, Table 1 shows the time-series feature value sequence of navigation compliance features and the calculation process of deviation multiple, as shown in Table 1 below:

[0086] Table 1 S1043: Count the number of windows in each sliding time window corresponding to this feature whose deviation multiple exceeds the preset deviation threshold, and use this number as the abnormal window count of this feature.

[0087] In this step, for each feature, all deviation factors of that feature within N sliding time windows are iterated through, and it is determined whether the deviation factor of each sliding time window exceeds a preset deviation threshold. If it exceeds the threshold, the sliding time window is marked as an abnormal window for that feature; otherwise, it is marked as a normal window for that feature. The number of times the feature is marked as an abnormal window within the N sliding time windows is counted, which is the number of abnormal windows for that feature.

[0088] S1044: Calculate the sum of the number of abnormal windows for all features to obtain the total number of abnormal windows, and determine whether there are abnormal signs in the driving behavior based on the relationship between the total number of abnormal windows and the preset abnormal threshold.

[0089] In this step, the number of abnormal windows for the six features is summed to obtain the total number of abnormal windows. The total number of abnormal windows is then compared with a preset abnormality threshold. When the total number of abnormal windows exceeds the preset abnormality threshold, it indicates that there are significant deviations in multiple sliding time windows for multiple features, confirming abnormal signs in driving behavior; when the total number of abnormal windows does not exceed the preset abnormality threshold, it is determined that there are no abnormal signs in driving behavior.

[0090] S1045: If there are no abnormal signs, it is determined that the driver's driving status is not abnormal.

[0091] In this step, if no abnormal signs are observed, it indicates that the deviations of the current features are within the normal fluctuation range, and the driver's overall driving state is at a normal level. Therefore, it is directly determined that the driving state is not abnormal. At this time, only the feature values ​​of each feature within the current sliding time window are recorded for subsequent updates to the individual baseline values.

[0092] S1046: If there are abnormal signs, based on the first preset deviation multiple range of each feature under the abnormal bodily function caused by the substance and the second preset deviation multiple range under fatigue, as well as the deviation multiple of each feature in the current sliding time window, determine whether the driver's driving state is abnormal and the cause of the abnormal driving state.

[0093] In this step, when there are abnormal signs, it indicates that multiple features have shown a persistent and significant deviation across multiple sliding time windows, and further analysis is needed to determine the specific reasons for these abnormal signs.

[0094] This step pre-defines the deviation multiple ranges for two main cause categories: The first preset deviation multiple range corresponds to bodily dysfunction caused by substances (such as drinking alcohol, taking drugs, etc.). Under this cause category, the deviation multiple of each feature usually shows a serious deviation, such as serious deviation of navigation compliance feature, drastic deterioration of speed stability feature and direction stability feature, and significant abnormality of perception response integrity feature.

[0095] The second preset deviation multiple range corresponds to fatigue. Under this category, the deviation multiple of each feature usually shows a moderate degree of deviation, and the response delay feature shows a global and uniform increase.

[0096] By matching the deviation factor of each feature within the current sliding time window with the two preset ranges mentioned above, it is possible to determine whether the driving state is indeed abnormal and the specific cause of the abnormal state.

[0097] In one possible implementation, when performing step S1046, the specific steps S10461-S10467 can be performed as follows: S10461: For each feature, if the deviation multiple of the feature in the current sliding time window falls within the first preset deviation multiple range corresponding to the feature, then the feature is taken as the first feature, and the product of the preset weight of each first feature and the specified base is calculated. The sum of the product results of each first feature is taken as the body abnormality score; and if the deviation multiple of the feature in the current sliding time window falls within the second preset deviation multiple range corresponding to the feature, then the feature is taken as the second feature, and the product of the preset weight of each second feature and the specified base is calculated. The sum of the product results of each second feature is taken as the fatigue score.

[0098] In this step, for each feature within the current sliding time window, the deviation factor of the feature is compared with the first preset deviation factor range (abnormal bodily function caused by substances) and the second preset deviation factor range (fatigue) corresponding to the feature.

[0099] For example, Table 2 shows the first preset deviation multiple range and the second preset deviation multiple range for each feature.

[0100]

[0101] Table 2 For each feature, if the deviation multiple of the feature falls within the first preset deviation multiple range, it indicates that the performance of the feature conforms to the pattern of abnormal bodily function caused by substances. Then, the feature is marked as the first feature, the product of the preset weight of the feature and the specified base (such as 1) is calculated, and the sum of the product results of each first feature is calculated to obtain the body abnormality score.

[0102] If the deviation multiple of the feature falls within the second preset deviation multiple range, it indicates that the performance of the feature conforms to the fatigue pattern. Then, the feature is marked as the second feature, the product of the preset weight of the feature and the specified base (such as 1) is calculated, and the sum of the product results of each second feature is calculated to obtain the fatigue score.

[0103] The preset weights of each feature can be set according to the degree of contribution of the feature to distinguishing the cause of the abnormality. For example, the weights of navigation compliance feature and perception response integrity feature are set to twice that of other features, because these two features have a strong ability to distinguish between physical dysfunction and fatigue caused by substances.

[0104] S10462: If the abnormal score of the body is greater than or equal to the first preset threshold, it is determined that the driver's driving state is abnormal, and the abnormal driving state is caused by a physical dysfunction caused by a substance.

[0105] In this step, the calculated bodily abnormality score is compared with a first preset threshold (e.g., 8 points). If the bodily abnormality score reaches or exceeds the first preset threshold, it indicates that the deviation pattern of each feature is highly consistent with the typical manifestation of bodily dysfunction caused by a substance. Therefore, it is determined that the driving state is abnormal, and the cause of the abnormality is attributed to bodily dysfunction caused by a substance. At this time, corresponding countermeasures can be triggered, such as issuing an alarm, limiting the vehicle speed, or calling for rescue.

[0106] S10463: If the abnormal score of the body is less than the first preset threshold and the fatigue score is greater than or equal to the second preset threshold, it is determined that the driver's driving state is abnormal and the abnormal driving state is caused by fatigue; the first preset threshold is greater than the second preset threshold.

[0107] In this step, when the abnormality score does not reach the first preset threshold, but the fatigue score reaches or exceeds the second preset threshold (e.g., 6 points), it indicates that although the deviation pattern of each characteristic does not completely match the typical manifestation of bodily dysfunction caused by substances, it is quite consistent with the typical manifestation of fatigue. Considering that bodily dysfunction caused by substances has a more serious and urgent impact on driving safety, a higher judgment threshold is set (the first preset threshold is greater than the second preset threshold), while the judgment threshold for fatigue is relatively low. In this case, it is determined that the driving state is abnormal, and the cause of the abnormality is fatigue.

[0108] S10464: If the abnormal score of the body is less than the first preset threshold and the fatigue score is less than the second preset threshold, then for each feature, calculate the variance of the feature value in each sliding time window, obtain the inter-window variance of the feature, and calculate the mean of all inter-window variances.

[0109] In this step, when neither of the two scores reaches its corresponding judgment threshold, it indicates that matching based solely on deviations from the multiple range is insufficient to clearly determine the abnormal state and its cause, and further analysis using the stability of each feature over time is required.

[0110] For each feature, using the time-series feature value sequence obtained in step S1041, the variance of the feature value over N consecutive sliding time windows is calculated. The inter-window variance reflects the degree of fluctuation of the feature value between different sliding time windows. Then, the arithmetic mean of the inter-window variances of all features is calculated to obtain a comprehensive variance index.

[0111] S10465: If the average value is greater than the third preset threshold, it is determined that the driver's driving state is abnormal and that the abnormal driving state is caused by distraction.

[0112] In this step, the mean of the variances between windows for all features is compared with a third preset threshold (e.g., 0.4). If the mean exceeds the third preset threshold, it indicates that the fluctuations of each feature between different windows are very drastic, with some windows showing abnormally high feature values ​​while others return to normal levels, exhibiting an unstable pattern. This dramatic fluctuation in feature values ​​between windows is consistent with the typical characteristics of inattention (abnormalities occur intermittently when inattention is present), therefore, it is determined that the driving state is abnormal, and the cause of the abnormality is inattention.

[0113] S10466: If the average value is greater than or equal to the fourth preset threshold and less than or equal to the third preset threshold, it is determined that the driver's driving state is abnormal and the abnormal driving state is caused by fatigue.

[0114] In this step, when the mean of the variance between windows falls between the fourth preset threshold (e.g., 0.15) and the third preset threshold (e.g., 0.4), it indicates that each feature exhibits a moderate degree of fluctuation between windows. This is not a dramatic spike or drop, but rather a certain degree of persistence and trend. This moderate fluctuation pattern is consistent with the characteristics of fatigue; that is, a driver's cognitive function fluctuates but maintains a certain degree of persistence under fatigue, unlike the drastic jumps that occur when attention is diverted. Therefore, it is determined that the driving state is abnormal, and the cause of the abnormality is fatigue.

[0115] S10467: If the average value is less than the fourth preset threshold, it is determined that the driver's driving status is not abnormal and is considered normal fluctuation.

[0116] In this step, when the mean variance between windows is less than the fourth preset threshold (e.g., 0.15), it indicates that the fluctuation range of each feature between windows is very small, and the feature values ​​are stable without significant deviation. Although the total number of abnormal windows exceeded the preset abnormal threshold in step S1044 (i.e., there are signs of abnormality), the subsequent deviation multiple range matching failed to clearly determine the abnormality (both scores were insufficient), and the variance between windows was also at a low level. In summary, the deviation of individual windows that appeared earlier may have been caused by special road conditions or temporary environmental factors, rather than a problem with the driver's driving state itself. Therefore, it is determined that the driving state is not abnormal, and the aforementioned abnormal signs are normal fluctuations.

[0117] In one possible implementation, the following steps S201-S203 may also be performed: S201: In the initial stage of establishing personal baseline values ​​for each feature, feature values ​​of each feature during driving are collected for navigation compliance feature, speed stability feature, direction stability feature and perception response integrity feature, until the cumulative driving time reaches the first preset time and the cumulative number of recognitions of each type of road sign involved in the response delay feature reaches the preset number. The average value of the feature values ​​of each feature that has been collected is calculated as the initial personal baseline value of each feature.

[0118] In this step, the initial establishment of personal baseline values ​​for each feature refers to the stage when the driver first uses this detection method and has not yet established personal driving behavior benchmark data. During this stage, feature values ​​for each feature calculated in steps S102 to S103 are continuously collected during each driving session for navigation compliance, speed stability, directional stability, and perception response integrity features, and driving time is accumulated. When the accumulated driving time reaches a first preset duration (e.g., 7 days) and the accumulated recognition count for each type of road sign (e.g., speed limit signs, prohibitory signs, instruction signs, etc.) involved in the response delay feature reaches a preset number (e.g., 30 times), it indicates that the collected data sample size is sufficient to support reliable baseline statistics. At this point, the arithmetic mean of all feature values ​​collected for each feature is calculated, and this mean is used as the initial personal baseline value for the corresponding feature.

[0119] S202: For response delay quantity characteristics and response delay stability characteristics, preset empirical values ​​are used as the initial personal baseline values ​​for each characteristic.

[0120] In this step, the calculation of response delay quantity characteristics and response delay stability characteristics depends on the response baseline duration. However, in the initial baseline establishment phase, sufficient road sign response data has not yet been accumulated to determine a reliable response baseline duration. Therefore, for the response delay quantity characteristic, its initial personal baseline value is set to a preset empirical value (e.g., 1), indicating that the driver's initial response delay level is assumed to be comparable to their normal level. Similarly, for the response delay stability characteristic, its initial personal baseline value is also set to a preset empirical value (e.g., 0.2), which is determined based on statistical analysis of a large amount of response stability data from drivers under normal driving conditions. After the initial baseline establishment is completed, these preset empirical values ​​will be gradually adjusted through incremental updates as driving data accumulates, gradually approaching the driver's true personal baseline level.

[0121] S203: After the initial establishment of individual baseline values ​​for each characteristic, when it is determined that the driver's driving behavior is not abnormal, the individual baseline values ​​for each characteristic shall be updated in the following manner:

[0122] in, For the updated individual baseline value, The individual baseline value before the update. The first preset learning rate, The feature value is the value of the feature calculated within the current sliding time window.

[0123] In this step, after establishing the initial individual baseline values ​​for each feature, a daily incremental update phase begins. Whenever step S104 determines that the driver's driving state is not abnormal (including step S1045 determining no abnormality and step S10467 determining it is within normal fluctuations), an update to the individual baseline values ​​is triggered. The update uses an exponential moving average method, as shown in the formula. The first preset learning rate (e.g., 0.02) controls the degree of influence of new data on the baseline value. The smaller the learning rate, the slower the baseline value changes, and the better it can resist the interference of short-term road condition fluctuations on the baseline. This refers to the feature value of the feature calculated in step S102 or S103 within the current sliding time window. Taking navigation compliance feature as an example, if the personal baseline value before the update is 0.92, the feature value of the current window is 0.88, and the first preset learning rate is 0.02, then the updated personal baseline value is (1 0.02)×0.92+0.02×0.88=0.9192. In this way, the individual baseline value can be gradually adjusted as the driver's driving habits change, without being drastically changed by a single occasional fluctuation, thus ensuring the stability of the baseline and the ability to continuously track the driver's personalized characteristics.

[0124] In one possible implementation, the response baseline duration of various road signs can also be updated according to the following steps S301-S302: S301: When it is determined that the driver's driving behavior is not abnormal, for each type of road sign, calculate the average response time of each road sign in that type of road sign within the current sliding time window.

[0125] In this step, the response baseline duration is the fundamental parameter for calculating the response delay ratio, and it needs to be continuously updated as driving data accumulates to maintain its accuracy. The condition for triggering the update is the same as in step S203, that is, it is executed when it is determined in step S104 that the driving state is not abnormal. For each type of road sign (such as speed limit signs, prohibitory signs, and instruction signs), the response duration of each road sign calculated in step S10313 is obtained from all road signs corresponding to that type of road sign within the current sliding time window, and the arithmetic mean of these response durations is calculated. This mean reflects the driver's average response speed to that type of road sign within the current window.

[0126] S302: Update the response baseline duration for this type of road sign according to the following formula:

[0127] in, The updated response baseline duration for this type of road sign. The baseline response time for this type of road sign before the update. To preset the learning rate, This represents the average response time.

[0128] In this step, the response baseline duration is updated using a second preset learning rate. (For example, 0.01), this learning rate is different from the first preset learning rate. Smaller. This is because the response baseline duration reflects the driver's basic reaction ability to road signs, which is affected by long-term factors such as the driver's age and driving experience, and changes extremely slowly. Therefore, a smaller learning rate is used to maintain a high degree of stability in the baseline. For example, if the response baseline duration before updating a certain type of road sign is 1.2 seconds, the average response duration of the current window is 1.35 seconds, and the preset learning rate is 0.01, then the updated response baseline duration will be (1... 0.01)×1.2+0.01×1.35=1.2015 seconds. Through this slow update method, the response baseline duration can gradually adapt to the driver's actual reaction level during long-term driving, without significant deviation due to abnormal fluctuations in individual windows.

[0129] Through steps S201 to S203 and S301 to S302 described above, this detection method achieves the initial establishment and continuous updating of individual baseline values ​​for various features and response baseline durations for various road signs. This self-learning mechanism ensures that the deviation factor calculation relied upon for anomaly detection always uses the driver's normal driving level as a reference benchmark, effectively avoiding misjudgments caused by differences in driving styles among different drivers and guaranteeing the accuracy of detection. Furthermore, updates are only performed when the driving state is determined to be normal, preventing driving data from abnormal states from being mixed into the baseline and ensuring the reliability of the baseline values.

[0130] Based on the same technical concept, embodiments of this application also provide a detection device for abnormal driving conditions, such as... Figure 4 As shown, it includes: The acquisition module 401 is used to acquire in real time the vehicle's navigation planning data, driving trajectory, forward road environment data, and operation signals generated by the driver controlling the vehicle during the driving process. The first calculation module 402 is used to calculate the feature value of the navigation compliance feature based on the navigation planning data and the driving trajectory within the current sliding time window; The second calculation module 403 is used to calculate, based on the forward-looking road environment data and the operation signal within the current sliding time window, the characteristic values ​​of the response delay feature, the characteristic values ​​of the response delay stability feature, the characteristic values ​​of the speed stability feature and the direction stability feature of the driver's vehicle operation, and the characteristic value of the perception response integrity feature; wherein, the perception response integrity feature is used to characterize the coverage ratio of the driver's actual operation response to the road sign that should be responded to. The determination module 404 is used to determine whether the driver's driving state is abnormal based on the feature values ​​of the navigation compliance feature, the response delay feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature calculated within multiple consecutive sliding time windows.

[0131] Optionally, when the first calculation module 402 calculates the feature value of the navigation compliance feature based on the navigation planning data and the driving trajectory within the current sliding time window, it is specifically used for: Extract navigation action points from the navigation planning data within the current sliding time window; the navigation action points are the positioning points in the navigation planning data that indicate that the driver needs to perform steering or fork operations; Spatial matching is performed between the driving trajectory within the current sliding time window and the navigation action points, and the number of navigation action points that the driving trajectory did not pass through correctly is counted. Calculate the ratio of the number of navigation action points that were not traversed correctly to the total number of navigation action points, and use this ratio as the feature value of the navigation compliance feature within the current sliding time window.

[0132] Optionally, when the second calculation module 403 calculates the characteristic values ​​of the response delay amount and the response delay stability characteristics of the driver's operation response to the road sign based on the forward-looking road environment data and the operation signal within the current sliding time window, it is specifically used for: From the forward-looking road environment data within the current sliding time window, identify road signs that require driver action, and record the first time each road sign is identified as the anchor point time for each road sign; For each road sign identified within the current sliding time window, starting from the anchor point time of the road sign, within a detection window of a preset duration, it is detected whether there is an operation response action corresponding to the road sign in the operation signal. If there is, the time when the operation response action first appears is taken as the response time of the road sign. For each road sign that has an operational response action, calculate the time difference between the response time of the road sign and the anchor point time, and use it as the driver's response time to the road sign; The road signs within the current sliding time window are divided into types. For each type of road sign, the average response time of all road signs under that type is calculated. Then, the ratio of this average to the driver's baseline response time for that type of road sign is calculated. This ratio is used as the driver's response delay ratio for that type of road sign. The mean of all types of response delay ratios is calculated as a feature value of the response delay amount characteristic of the driver's operation response to road signs within the current sliding time window; wherein, the response delay amount characteristic characterizes the driver's response delay level to road signs; Calculate the standard deviation of the response delay ratio of all road signs within the current sliding time window, and use this standard deviation as the feature value of the response delay stability characteristic of the driver's operation response to all road signs within the current sliding time window; wherein, the response delay stability characteristic characterizes the degree of dispersion of the driver's response delay to multiple road signs.

[0133] Optionally, the operation signal includes a vehicle speed sequence; when the second calculation module 403 calculates the characteristic values ​​of the speed stability characteristics and the directional stability characteristics of the driver's vehicle control based on the forward-looking road environment data and the operation signal within the current sliding time window, it is specifically used for: Calculate the ratio between the standard deviation and the mean of the vehicle speed sequence within the current sliding time window, and use this ratio as the feature value of the speed stability characteristic of the driver's vehicle control. Extract the vehicle lateral offset and lane width at each sampling moment within the current sliding time window from the forward-looking road environment data; where the vehicle lateral offset is the lateral distance of the vehicle center from the current lane center. For each sampling time, the ratio of the vehicle's lateral offset to the lane width is calculated as the normalized lateral offset. Calculate the standard deviation of all normalized lateral offsets within the current sliding time window, and use it as the lane deviation. The lane deviation is then used as a feature value of the driver's directional stability characteristics.

[0134] Optionally, when the second calculation module 403 calculates the feature value of the perception response integrity feature based on the forward-looking road environment data and the operation signal within the current sliding time window, it is specifically used for: For each road sign identified that requires driver action, determine whether the driver has performed an action corresponding to the road sign within a preset detection window. The number of road signs that are determined not to have taken any action within the current sliding time window is counted as the number of unresponsive signs. Calculate the ratio of the number of unresponsive signs to the total number of road signs identified within the current sliding time window, and use this ratio as the feature value of the perceived response integrity feature.

[0135] Optionally, when the determination module 404 determines whether the driver's driving state is abnormal based on the feature values ​​of the navigation compliance feature, the response delay feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature calculated within multiple consecutive sliding time windows, it is specifically used for: For each of the navigation compliance feature, the response delay feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature, a temporal feature value sequence for each feature is determined within multiple consecutive sliding time windows, including the current sliding time window; the temporal feature value sequence includes the feature value of the feature in each sliding time window within the multiple consecutive sliding time windows. For each of the aforementioned features, the ratio of the feature value of each sliding time window in the time-series feature value sequence of that feature to the individual baseline value corresponding to that feature is calculated, and this ratio is used as the deviation factor for each sliding time window corresponding to that feature. The number of windows in each sliding time window corresponding to this feature whose deviation multiple exceeds a preset deviation threshold is counted as the number of abnormal windows for this feature. The total number of abnormal windows for all features is calculated to obtain the total number of abnormal windows. Based on the relationship between the total number of abnormal windows and a preset abnormal threshold, it is determined whether there are abnormal signs in the driving behavior. If no abnormal signs are found, the driver's driving status is determined to be normal. If any abnormal signs are found, the system determines whether the driver's driving state is abnormal and the cause of the abnormal driving state based on the first preset deviation multiple range of each feature under the abnormal bodily function caused by the substance and the second preset deviation multiple range under fatigue, as well as the deviation multiple of each feature in the current sliding time window.

[0136] Optionally, the determination module 404, in determining whether the driver's driving state is abnormal and the cause of the abnormal driving state, is specifically used for: For each feature, if the deviation multiple of the feature in the current sliding time window falls within the first preset deviation multiple range corresponding to the feature, then the feature is taken as the first feature, and the product of the preset weight of each first feature and the specified base is calculated. The sum of the product results of each first feature is taken as the body abnormality score. And if the deviation multiple of the feature in the current sliding time window falls within the second preset deviation multiple range corresponding to the feature, then the feature is taken as the second feature, and the product of the preset weight of each second feature and the specified base is calculated. The sum of the product results of each second feature is taken as the fatigue score. If the abnormal score of the body is greater than or equal to the first preset threshold, it is determined that the driver's driving state is abnormal, and the abnormal driving state is caused by a physical dysfunction caused by a substance. If the abnormal score of the body is less than the first preset threshold and the fatigue score is greater than or equal to the second preset threshold, then it is determined that the driver's driving state is abnormal and the abnormal driving state is caused by fatigue; the first preset threshold is greater than the second preset threshold. If the abnormal score of the body is less than the first preset threshold and the fatigue score is less than the second preset threshold, then for each feature, the variance of the feature value of the feature in each sliding time window is calculated to obtain the inter-window variance of the feature, and the mean of all the inter-window variances is calculated. If the average value is greater than the third preset threshold, it is determined that the driver's driving state is abnormal and that the abnormal driving state is caused by distraction. If the average value is greater than or equal to the fourth preset threshold and less than or equal to the third preset threshold, it is determined that the driver's driving state is abnormal and that the abnormal driving state is caused by fatigue. If the average value is less than the fourth preset threshold, it is determined that the driver's driving status is not abnormal and is considered to be a normal fluctuation.

[0137] Optionally, the device further includes: The third calculation module is used to collect feature values ​​of each feature during the driving process for the navigation compliance feature, speed stability feature, direction stability feature and perception response integrity feature during the initial establishment of personal baseline values ​​for each feature, until the cumulative driving time reaches a first preset time and the cumulative number of recognitions of each type of road sign involved in the response delay feature reaches a preset number, and calculates the average value of the feature values ​​of each feature collected, as the initial personal baseline value of each feature; The determination module is used to use preset experience values ​​as the initial personal baseline values ​​for the response delay quantity feature and the response delay stability feature. The first update module is used to update the individual baseline values ​​of each feature after the initial establishment of the individual baseline values ​​for each feature, and when it is determined that the driver's driving behavior is not abnormal, in the following manner:

[0138] in, For the updated individual baseline value, The individual baseline value before the update. The first preset learning rate, The feature value is the value of the feature calculated within the current sliding time window.

[0139] Optionally, the device further includes: The fourth calculation module is used to calculate the average response time of each road sign in the current sliding time window for each type of road sign when it is determined that the driver's driving behavior is not abnormal. The second update module is used to update the response baseline duration of this type of road sign according to the following formula:

[0140] in, The updated response baseline duration for this type of road sign. The baseline response time for this type of road sign before the update. To preset the learning rate, The mean of the response duration is denoted as .

[0141] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0142] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection between devices or modules may be electrical, mechanical, or other forms.

[0143] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0144] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0145] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0146] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the claims.

Claims

1. A method for detecting abnormal driving conditions, characterized in that, include: It can acquire real-time navigation planning data, driving trajectory, forward road environment data, and operation signals generated by the driver during the driving process. Based on the navigation planning data and the driving trajectory within the current sliding time window, calculate the feature value of the navigation compliance feature; Based on the forward-looking road environment data and the operation signal within the current sliding time window, the characteristic values ​​of the response delay and response delay stability of the driver's operation response to road signs, the characteristic values ​​of the speed stability and directional stability of the driver's vehicle control, and the characteristic value of the perception response integrity are calculated; wherein, the perception response integrity is used to characterize the coverage ratio of the driver's actual operation response to the road signs that should be responded to. Based on the feature values ​​of the navigation compliance feature, the response delay feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature calculated within multiple consecutive sliding time windows, it is determined whether the driver's driving state is abnormal.

2. The method according to claim 1, characterized in that, The step of calculating the feature values ​​of navigation compliance characteristics based on the navigation planning data and the driving trajectory within the current sliding time window includes: Extract navigation action points from the navigation planning data within the current sliding time window; the navigation action points are the positioning points in the navigation planning data that indicate that the driver needs to perform steering or fork operations; Spatial matching is performed between the driving trajectory within the current sliding time window and the navigation action points, and the number of navigation action points that the driving trajectory did not pass through correctly is counted. Calculate the ratio of the number of navigation action points that were not traversed correctly to the total number of navigation action points, and use this ratio as the feature value of the navigation compliance feature within the current sliding time window.

3. The method according to claim 1, characterized in that, The step of calculating the characteristic values ​​of the response delay amount and the characteristic values ​​of the response delay stability of the driver's operation response to road signs based on the forward-looking road environment data and the operation signal within the current sliding time window includes: From the forward-looking road environment data within the current sliding time window, identify road signs that require driver action, and record the first time each road sign is identified as the anchor point time for each road sign; For each road sign identified within the current sliding time window, starting from the anchor point time of the road sign, within a detection window of a preset duration, it is detected whether there is an operation response action corresponding to the road sign in the operation signal. If there is, the time when the operation response action first appears is taken as the response time of the road sign. For each road sign that has an operational response action, calculate the time difference between the response time of the road sign and the anchor point time, and use it as the driver's response time to the road sign; The road signs within the current sliding time window are divided into types. For each type of road sign, the average response time of all road signs under that type is calculated. Then, the ratio of this average to the driver's baseline response time for that type of road sign is calculated. This ratio is used as the driver's response delay ratio for that type of road sign. The mean of all types of response delay ratios is calculated as a feature value of the response delay amount characteristic of the driver's operation response to road signs within the current sliding time window; wherein, the response delay amount characteristic characterizes the driver's response delay level to road signs; Calculate the standard deviation of the response delay ratio of all road signs within the current sliding time window, and use this standard deviation as the feature value of the response delay stability characteristic of the driver's operation response to all road signs within the current sliding time window; wherein, the response delay stability characteristic characterizes the degree of dispersion of the driver's response delay to multiple road signs.

4. The method according to claim 1, characterized in that, The operation signal includes a vehicle speed sequence; the step of calculating the characteristic values ​​of the speed stability feature and the characteristic values ​​of the directional stability feature of the driver's vehicle control based on the forward-looking road environment data within the current sliding time window and the operation signal includes: Calculate the ratio between the standard deviation and the mean of the vehicle speed sequence within the current sliding time window, and use this ratio as a feature value of the speed stability characteristic of the driver's vehicle control. The vehicle lateral offset and lane width at each sampling moment within the current sliding time window are extracted from the forward-looking road environment data; wherein, the vehicle lateral offset is the lateral distance of the vehicle center from the current lane center; For each sampling time, the ratio of the vehicle's lateral offset to the lane width is calculated as the normalized lateral offset. Calculate the standard deviation of all the normalized lateral offsets within the current sliding time window, and use it as the lane deviation. Then, use the lane deviation as a feature value of the directional stability characteristic of the driver's vehicle control.

5. The method according to claim 3, characterized in that, The step of calculating the feature value of the perception response integrity feature based on the forward-looking road environment data and the operation signal within the current sliding time window includes: For each road sign identified that requires driver action, determine whether the driver has performed an action corresponding to the road sign within a preset detection window. The number of road signs that are determined not to have taken any action within the current sliding time window is counted as the number of unresponsive signs. Calculate the ratio of the number of unresponsive signs to the total number of road signs identified within the current sliding time window, and use this ratio as the feature value of the perceived response integrity feature.

6. The method according to claim 1, characterized in that, The method of determining whether the driver's driving state is abnormal based on the feature values ​​of the navigation compliance feature, the response delay feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature calculated within multiple consecutive sliding time windows includes: For each of the navigation compliance feature, the response delay feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature, a temporal feature value sequence for each feature is determined within multiple consecutive sliding time windows, including the current sliding time window; the temporal feature value sequence includes the feature value of the feature in each sliding time window within the multiple consecutive sliding time windows. For each of the aforementioned features, the ratio of the feature value of each sliding time window in the time-series feature value sequence of that feature to the individual baseline value corresponding to that feature is calculated, and this ratio is used as the deviation factor for each sliding time window corresponding to that feature. The number of windows in each sliding time window corresponding to this feature whose deviation multiple exceeds a preset deviation threshold is counted as the number of abnormal windows for this feature. The total number of abnormal windows for all features is calculated to obtain the total number of abnormal windows. Based on the relationship between the total number of abnormal windows and a preset abnormal threshold, it is determined whether there are abnormal signs in the driving behavior. If no abnormal signs are found, the driver's driving status is determined to be normal. If any abnormal signs are found, the system determines whether the driver's driving state is abnormal and the cause of the abnormal driving state based on the first preset deviation multiple range of each feature under the abnormal bodily function caused by the substance and the second preset deviation multiple range under fatigue, as well as the deviation multiple of each feature in the current sliding time window.

7. The method according to claim 6, characterized in that, The determination of whether the driver's driving state is abnormal and the reasons for the abnormal driving state include: For each feature, if the deviation multiple of the feature in the current sliding time window falls within the first preset deviation multiple range corresponding to the feature, then the feature is taken as the first feature, and the product of the preset weight of each first feature and the specified base is calculated. The sum of the product results of each first feature is taken as the body abnormality score. And if the deviation multiple of the feature in the current sliding time window falls within the second preset deviation multiple range corresponding to the feature, then the feature is taken as the second feature, and the product of the preset weight of each second feature and the specified base is calculated. The sum of the product results of each second feature is taken as the fatigue score. If the abnormal score of the body is greater than or equal to the first preset threshold, it is determined that the driver's driving state is abnormal, and the abnormal driving state is caused by a physical dysfunction caused by a substance. If the abnormal score of the body is less than the first preset threshold and the fatigue score is greater than or equal to the second preset threshold, then it is determined that the driver's driving state is abnormal and the abnormal driving state is caused by fatigue; the first preset threshold is greater than the second preset threshold. If the abnormal score of the body is less than the first preset threshold and the fatigue score is less than the second preset threshold, then for each feature, the variance of the feature value of the feature in each sliding time window is calculated to obtain the inter-window variance of the feature, and the mean of all the inter-window variances is calculated. If the average value is greater than the third preset threshold, it is determined that the driver's driving state is abnormal and that the abnormal driving state is caused by distraction. If the average value is greater than or equal to the fourth preset threshold and less than or equal to the third preset threshold, it is determined that the driver's driving state is abnormal and that the abnormal driving state is caused by fatigue. If the average value is less than the fourth preset threshold, it is determined that the driver's driving status is not abnormal and is considered to be a normal fluctuation.

8. The method according to claim 7, characterized in that, The method further includes: In the initial stage of establishing personal baseline values ​​for each feature, feature values ​​of each feature during driving are collected for the navigation compliance feature, the speed stability feature, the direction stability feature, and the perception response integrity feature, until the cumulative driving time reaches a first preset time and the cumulative number of recognitions of each type of road sign involved in the response delay feature reaches a preset number. The average value of the feature values ​​of each feature that has been collected is then calculated as the initial personal baseline value for each feature. For the response delay quantity characteristic and the response delay stability characteristic, preset empirical values ​​are used as the initial personal baseline values ​​for each characteristic; After establishing the individual baseline values ​​for each characteristic for the first time, when it is determined that the driver's driving behavior is not abnormal, the individual baseline values ​​for each characteristic are updated in the following manner: in, For the updated individual baseline value, The individual baseline value before the update. The first preset learning rate, The feature value is the value of the feature calculated within the current sliding time window.

9. The method according to claim 3, characterized in that, The method also includes updating the response baseline duration of various road signs in the following ways: When it is determined that the driver's driving behavior is not abnormal, for each type of road sign, the average response time of each road sign in that type of road sign is calculated within the current sliding time window. Update the response baseline duration for this type of road sign using the following formula: in, The updated response baseline duration for this type of road sign. The baseline response time for this type of road sign before the update. To preset the learning rate, The mean of the response duration is denoted as .

10. A detection device for abnormal driving conditions, characterized in that, include: The acquisition module is used to acquire in real time the vehicle's navigation planning data, driving trajectory, forward road environment data, and operation signals generated by the driver during the driving process; The first calculation module is used to calculate the feature value of the navigation compliance feature based on the navigation planning data and the driving trajectory within the current sliding time window; The second calculation module is used to calculate, based on the forward-looking road environment data and the operation signal within the current sliding time window, the characteristic values ​​of the response delay feature, the characteristic values ​​of the response delay stability feature, the characteristic values ​​of the speed stability feature and the directional stability feature of the driver's operation response to road signs, and the characteristic value of the perception response integrity feature; wherein, the perception response integrity feature is used to characterize the coverage ratio of the driver's actual operation response to the road signs that should be responded to. The determination module is used to determine whether the driver's driving state is abnormal based on the feature values ​​of the navigation compliance feature, the response delay feature, the response delay stability feature, the speed stability feature, the direction stability feature, and the perception response integrity feature calculated within multiple consecutive sliding time windows.

Images