Eye fatigue degree detection method and device, electronic equipment and readable storage medium

Abstract

This application discloses a method, device, electronic device, and readable storage medium for detecting eye fatigue, belonging to the field of terminal visual recognition technology. The eye fatigue detection method includes: constructing a three-dimensional face model; acquiring multiple consecutive frames of face images, and determining eyelid opening and closing parameters corresponding to each frame of the face image based on the three-dimensional face model; obtaining an eyelid opening and closing parameter sequence based on a sliding window of preset length; determining a closed-eye judgment threshold based on the eyelid opening and closing parameter sequence and a preset initial value for eye closure judgment; dividing the eyelid opening and closing parameters that reach the closed-eye judgment threshold into multiple closed-eye data groups according to their temporal continuity within the eyelid opening and closing parameter sequence; determining a valid closed-eye data group based on the difference between each eyelid opening and closing parameter in each closed-eye data group and the closed-eye judgment threshold; determining the blinking frequency based on the valid closed-eye data group; and determining eye fatigue based on the blinking frequency.

Description

Technical Field

[0001] This application belongs to the field of terminal visual recognition technology, specifically relating to an eye fatigue detection method, device, electronic device, and readable storage medium. Background Technology

[0002] With the rapid iteration of smart devices and the increasing screen time, various electronic terminals need to detect user eye fatigue in order to alleviate it by adjusting screen brightness, color, and other factors. Research shows that blinking behavior is highly correlated with fatigue levels; when fatigued, blinking frequency decreases significantly, while eye closure time increases. Therefore, eye fatigue can be assessed by detecting blinking frequency and eye closure time.

[0003] Currently, blink detection technologies mostly rely on two-dimensional planar images for analysis, which are easily affected by head posture shifts and changes in shooting angle, making it difficult to accurately depict the true movement of the eyelids and resulting in large errors in eye feature recognition. At the same time, these technologies often use a single fixed threshold for simple judgment without combining time-series data for refined processing, and cannot distinguish between noise interference, instantaneous eyelid tremors and real blinking actions. This leads to low blink counting accuracy and poor detection robustness, which in turn affects the accuracy of eye fatigue assessment. Summary of the Invention

[0004] The purpose of this application is to provide a method, apparatus, electronic device, and readable storage medium for detecting eye fatigue, which can improve the accuracy of blink counting and overall detection robustness, thereby improving the accuracy of eye fatigue assessment.

[0005] In a first aspect, embodiments of this application provide a method for detecting eye fatigue, including:

[0006] A 3D face model is constructed, which is used to represent the movement of the eyelids;

[0007] Acquire multiple consecutive frames of face images, and determine the eyelid opening and closing parameters corresponding to each frame of face image based on the 3D face model. The eyelid opening and closing parameters are used to characterize the opening and closing ratio relative to the preset fully open eye state.

[0008] Based on a sliding window of preset length, a sequence of eyelid opening and closing parameters is obtained;

[0009] The threshold for determining eye closure is determined based on the eyelid opening and closing parameter sequence and the preset initial value for eye closure determination.

[0010] In the eyelid opening and closing parameter sequence, the eyelid opening and closing parameters that reach the eye-closing judgment threshold are divided into multiple eye-closing data groups according to the temporal continuity relationship;

[0011] The effective closed-eye data group is determined based on the difference between each eyelid opening and closing parameter and the closed-eye judgment threshold in each closed-eye data group;

[0012] Based on the effective closed-eye data set, the blink frequency was determined;

[0013] The degree of eye fatigue is determined based on the blinking frequency.

[0014] Secondly, embodiments of this application provide an eye fatigue detection device, comprising:

[0015] The building module is used to construct a 3D face model, which is used to represent the eyelid movement state;

[0016] The first determining module is used to acquire multiple consecutive frames of face images and determine the eyelid opening and closing parameters corresponding to each frame of face image based on the three-dimensional face model. The eyelid opening and closing parameters are used to characterize the opening and closing ratio relative to the preset fully open eye state.

[0017] The first processing module is used to obtain the eyelid opening and closing parameter sequence based on a sliding window of preset length;

[0018] The second determining module is used to determine the eye-closing judgment threshold based on the eyelid opening and closing parameter sequence and the preset initial value for eye-closing judgment;

[0019] The second processing module is used to divide the eyelid opening and closing parameters that reach the eye-closing judgment threshold into multiple eye-closing data groups according to the temporal continuity relationship in the eyelid opening and closing parameter sequence.

[0020] The third determination module is used to determine the valid closed eye data group based on the difference between the eyelid opening and closing parameters and the closed eye judgment threshold in each closed eye data group.

[0021] The fourth determination module is used to determine the blink frequency based on the effective eye-closing data set;

[0022] The fifth determination module is used to determine eye fatigue based on blink frequency.

[0023] Thirdly, embodiments of this application provide an electronic device, which includes a processor and a memory. The memory stores a program or instructions that can run on the processor. When the program or instructions are executed by the processor, they implement the steps of the eye fatigue detection method provided in the first aspect.

[0024] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the eye fatigue detection method provided in the first aspect.

[0025] Fifthly, embodiments of this application provide a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled, and the processor is used to run programs or instructions to implement the steps of the eye fatigue detection method provided in the first aspect.

[0026] In a sixth aspect, embodiments of this application provide a computer program product stored in a storage medium, which is executed by at least one processor to implement the steps of the eye fatigue detection method provided in the first aspect.

[0027] In this embodiment, a three-dimensional face model is constructed to characterize the eyelid movement state. This three-dimensional spatial representation replaces traditional two-dimensional planar analysis, mitigating recognition interference caused by head posture shifts and changes in shooting angle, reducing measurement deviations in eyelid opening and closing parameters. The three-dimensional face model determines the eyelid opening and closing parameters corresponding to each frame of the face image, objectively representing the opening and closing ratio relative to a preset fully open state, further ensuring the stability of parameter acquisition. Then, a sliding window of preset length is used to extract continuous eyelid opening and closing parameters, forming a temporally coherent sequence of eyelid opening and closing parameters, ensuring the integrity and continuity of the temporal data. Subsequently, the eyelid opening and closing parameters are combined with... The parameter sequence and preset initial values ​​for eye closure determination determine the eye closure judgment threshold. The eyelid opening and closing parameters that reach the eye closure judgment threshold are divided into multiple eye closure data groups according to their temporal continuity, completing a reasonable initial screening of eye closure segments. Finally, the difference between each eyelid opening and closing parameter in each eye closure data group and the eye closure judgment threshold is calculated. Based on all the differences in each eye closure data group, the valid eye closure data group is determined, effectively distinguishing real blinks from noise, instantaneous eyelid tremors and other interference signals, reducing false and missed judgments. Based on the valid eye closure data group, the blink frequency is determined, which effectively improves the accuracy of blink frequency and the overall detection robustness, thereby ensuring the accuracy of the obtained eye fatigue level. Attached Figure Description

[0028] Figure 1 This is a schematic flowchart of an eye fatigue detection method provided in an embodiment of this application;

[0029] Figure 2 This is a schematic diagram of the eye state corresponding to different eyelid opening and closing parameters provided in the embodiments of this application;

[0030] Figure 3 This is a schematic diagram of the waveform signal of the eyelid opening and closing parameters provided in the embodiments of this application;

[0031] Figure 4 This is a schematic diagram of the eyelid opening and closing parameter estimation and blink frequency counting processing flow provided in the embodiments of this application;

[0032] Figure 5 This is a schematic diagram of the blink counting process using a sliding window provided in an embodiment of this application;

[0033] Figure 6 This is a schematic diagram of the process of applying eye fatigue detection to fatigue assessment provided in the embodiments of this application;

[0034] Figure 7 This is a structural block diagram of the eye fatigue detection device provided in the embodiments of this application;

[0035] Figure 8 This is a structural block diagram of the electronic device proposed in the embodiments of this application. Detailed Implementation

[0036] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0037] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0038] The following description, in conjunction with the accompanying drawings, details an eye fatigue detection method, apparatus, electronic device, and readable storage medium provided in this application through specific embodiments and application scenarios.

[0039] like Figure 1 As shown in the figure, this application provides a method for detecting eye fatigue, including:

[0040] S100: Construct a 3D face model, which is used to represent the movement of the eyelids;

[0041] S110: Acquire multiple consecutive frames of face images, and determine the eyelid opening and closing parameters corresponding to each frame of face image based on the three-dimensional face model. The eyelid opening and closing parameters are used to characterize the opening and closing ratio relative to the preset fully open eye state.

[0042] S120: Obtain the eyelid opening and closing parameter sequence based on a sliding window of preset length;

[0043] S130: Determine the eye-closing judgment threshold based on the eyelid opening and closing parameter sequence and the preset initial value for eye closure judgment;

[0044] S140: In the eyelid opening and closing parameter sequence, the eyelid opening and closing parameters that reach the eye-closing judgment threshold are divided into multiple eye-closing data groups according to the temporal continuity relationship;

[0045] S150: Determine the effective closed-eye data group based on the difference between each eyelid opening and closing parameter and the closed-eye judgment threshold in each closed-eye data group;

[0046] S160: Determine blink frequency based on effective closed-eye data set.

[0047] S170: Determines eye fatigue level based on blink frequency.

[0048] Among them, the 3D face model is a model built based on the 3D structure of the head to represent the movement state of the eyelids in 3D space; the eyelid opening and closing parameters are parameters defined on the 3D face model to represent the opening and closing ratio of the eyelids relative to the preset fully open state; the sliding window is a data window of preset length used to extract continuous eyelid opening and closing parameters in time sequence. The preset length can be the time length or the length according to the time sequence frame rate; the eyelid opening and closing parameter sequence is a set of continuous time sequence eyelid opening and closing parameters obtained by the sliding window; the preset initial value for eye closure judgment is the initial setting value used to calculate the adaptive eye closure judgment threshold; the eye closure judgment threshold is the judgment value used to judge whether the eyelids have reached the degree of eye closure, and it is the dividing point that distinguishes between the open state and the half-closed state. When the eyelid opening and closing parameters reach the eye-closing judgment threshold, the eyelid is determined to have entered the closed or half-closed stage. Combined with the subsequent eye-opening judgment threshold, various blinking behaviors such as normal blinking and half-blinking (not fully closed) can be fully identified. The closed eye data set is a set of data obtained by dividing the eyelid opening and closing parameter sequence that reaches the eye-closing judgment threshold according to the temporal continuity. The difference is the numerical difference between a single eyelid opening and closing parameter and the eye-closing judgment threshold within the closed eye data set. The cumulative difference in the time domain is the total value obtained by accumulating all differences in the closed eye data set in the time dimension. The effective closed eye data set is the data set representing the actual blinking action. The blinking frequency is calculated by the number of effective closed eye data sets and the corresponding time period.

[0049] In the above technical solution, a three-dimensional face model is constructed to characterize the eyelid movement state. This three-dimensional spatial representation replaces traditional two-dimensional planar analysis, mitigating recognition interference caused by head posture shifts and changes in shooting angle, and reducing measurement deviations in eyelid opening and closing parameters. The three-dimensional face model determines the eyelid opening and closing parameters corresponding to each frame of the face image, objectively representing the opening and closing ratio relative to a preset fully open eye state, further ensuring the stability of parameter acquisition. Then, a sliding window of preset length is used to extract continuous eyelid opening and closing parameters, forming a temporally coherent sequence of eyelid opening and closing parameters, ensuring the integrity and continuity of the temporal data. Subsequently, the eyelid opening and closing parameters are combined with... The parameter sequence and preset initial values ​​for eye closure determination determine the eye closure judgment threshold. The eyelid opening and closing parameters that reach the eye closure judgment threshold are divided into multiple eye closure data groups according to their temporal continuity, completing a reasonable initial screening of eye closure segments. Finally, the difference between each eyelid opening and closing parameter in each eye closure data group and the eye closure judgment threshold is calculated. Based on all the differences in each eye closure data group, the valid eye closure data group is determined, effectively distinguishing real blinks from noise, instantaneous eyelid tremors and other interference signals, reducing false and missed judgments. Based on the valid eye closure data group, the blink frequency is determined, which effectively improves the accuracy of blink frequency and the overall detection robustness, thereby ensuring the accuracy of the obtained eye fatigue level.

[0050] For example, a 3D face model can be constructed based on the head structure; face images are acquired from a continuous video with a resolution of 640×360 and a frame rate of 30Hz, and after grayscale conversion and noise reduction preprocessing, eye feature points are extracted and mapped to the 3D face model; and the eyelid opening and closing parameters corresponding to each frame of face image are determined; for the 30Hz frame rate video, a preset fixed-length sliding window is used to extract continuous eyelid opening and closing parameters to obtain an eyelid opening and closing parameter sequence. The sequence is combined with the preset initial value for eye closure judgment to determine an adaptive eye closure judgment threshold. The parameters in the sequence that reach the threshold are divided into multiple eye closure data groups according to the temporal continuity relationship. The difference between each parameter in each group and the eye closure judgment threshold is calculated and accumulated to obtain the cumulative difference in the time domain. Based on this, the effective eye closure data groups are selected, and the number of groups is counted to determine the final blinking frequency.

[0051] For example, the eyelid opening and closing parameters are defined using a three-dimensional face model, and the eyelid opening and closing parameters are shown in the following formula:

[0052] P=P basis +w×(P target -P basis ); Formula 1

[0053] Among them, P basis P represents the 3D eyelid coordinates with eyes fully open. target Here, P represents the 3D eyelid coordinates with eyes closed, w is the eyelid opening / closing parameter (ranging from 0 to 1), and P is the eyelid coordinates at the current degree of opening / closing. w=0 represents fully open eyes, and w=1 represents fully closed eyes. Figure 2As shown.

[0054] For example, a corresponding training dataset is constructed through rendering, and a neural network model is trained based on the dataset. The 3D face model takes the eye image and head pose from the 2D face image as input and outputs the corresponding eyelid opening and closing parameters. The 3D face model uses head pose information and image information as input parameters, and the fusion of head pose information is achieved through an end-to-end neural network without additional step-by-step processing. The 3D face model redefines the eyelid opening and closing measurement standard, replacing the traditional eye aspect ratio feature to improve detection robustness; and the training dataset covering multiple head poses, multiple lighting, and multiple occlusion scenarios is constructed through rendering, eliminating the need for manual sample annotation and effectively improving model training efficiency. During training, the 3D face model autonomously learns the correspondence between head pose features and eyelid opening and closing degree within the face image. In the detection and inference stage, the original face image is directly input into the trained neural network, which automatically fuses the head pose association information end-to-end and outputs accurate eyelid opening and closing parameters.

[0055] For example, during the model training phase, a fully rendered simulation dataset can be used, and various environmental conditions can be set to enable the 3D face model to converge quickly. Taking a single eye as an example during inference, the obtained waveform signal is as follows: Figure 3 As shown.

[0056] In some embodiments of this application, the eyelid opening and closing parameters corresponding to each frame of a face image are determined based on a three-dimensional face model, including:

[0057] Based on multiple facial feature points in each frame of a face image, determine the head pose information corresponding to each frame of a face image.

[0058] Based on each frame of the face image, the corresponding eye image for each frame of the face image is cropped;

[0059] The eye image and head pose information corresponding to each frame of face image are input into the 3D face model to obtain the eyelid opening and closing parameters corresponding to each frame of face image.

[0060] Among them, facial feature points are facial feature points used to calculate head posture, head posture information is the position and angle information of the head relative to the image acquisition device calculated from multiple facial feature points, and eye image is an eye region image cropped from the face image.

[0061] In the above technical solution, the corresponding head pose information is obtained by solving the facial feature points of a single frame face image. This can correct the spatial distortion caused by the imaging viewpoint deflection, compensate for the eye imaging deviation caused by the head angle shift, establish the eye position constraint relationship in three-dimensional space, and improve the detection accuracy and scene robustness of the three-dimensional face model in solving the eyelid opening and closing parameters. At the same time, the eye region image is extracted from the complete face image, and redundant pixels and invalid background features are removed. The input dimension of the model is compressed, reducing the computational overhead and storage occupation of the three-dimensional face model inference process. While ensuring stable detection accuracy, the computational power consumption is reduced and the operating load is optimized, taking into account the detection accuracy and equipment operating efficiency in complex pose scenes.

[0062] For example, such as Figure 4 As shown, S1: Input image sequence; a continuous sequence of face images is input to provide raw data support for subsequent face feature extraction and eyelid opening / closing analysis. The image sequence is then passed to S2 and S3. S2: Face feature point extraction; key face feature points are extracted from the input image sequence to provide a location basis for eye image cropping and head pose determination. The face feature points are passed to S3 and S5 respectively. S3: Eye image cropping; based on face feature points, a standardized eye region image of size 64×64×3 is cropped from the original image to eliminate interference from irrelevant regions and unify the input specifications. The standardized eye image is then passed to S6. S4: 3D face model; prior information on the 3D face is provided as a benchmark reference for solving head pose information, ensuring the accuracy of pose determination. The 3D face model information is then passed to S5. S5: Head pose information calculation; Combining the 3D face model and facial feature points, calculate the current head pose information to correct pose deviations in the eye image and improve the robustness of eyelid opening and closing parameter estimation. This head pose information is then passed to S6 and S7. S6: Eyelid opening and closing parameters; Based on standardized eye images and head pose information, estimate the eyelid opening and closing parameters for the current frame, quantify the degree of eyelid opening and closing, and pass these parameters to S7. S7: Sliding window blink frequency counting; Based on the temporal eyelid opening and closing parameter sequence, use a sliding window mechanism to count blinks and calculate blink frequency, achieving continuous and stable counting. The calculated blink frequency is then passed to S8. S8: Output blink frequency; Output the final calculated blink frequency result, completing the entire process from image input to frequency output.

[0063] In some embodiments of this application, a threshold for determining eye closure is determined based on an eyelid opening and closing parameter sequence and a preset initial value for eye closure determination, including:

[0064] From the eyelid opening and closing parameter sequence, select multiple first target eyelid opening and closing parameters that are greater than the preset initial value for eye closure determination;

[0065] The average value of multiple first target eyelid opening and closing parameters is calculated to obtain the baseline value for closed eyes;

[0066] Based on the eyelid opening and closing parameter sequence, eyelid opening and closing parameters that are smaller than the preset initial eye-opening parameter are selected as the second target eyelid opening and closing parameters;

[0067] The average value of multiple second-target eyelid opening and closing parameters is calculated to obtain the eye-opening baseline value;

[0068] The threshold for judging closed eyes is obtained based on the preset closed eye coefficient, closed eye reference value, and open eye reference value.

[0069] Among them, the first target eyelid opening and closing parameter is an eyelid opening and closing parameter in the eyelid opening and closing parameter sequence that is greater than the preset initial value for eye closure judgment and is used to calculate the eye closure benchmark value. The eye closure benchmark value is the average value calculated from multiple first target eyelid opening and closing parameters and is used to reflect the parameter level of the eye-closed state in the current sequence. The second target eyelid opening and closing parameter is an eyelid opening and closing parameter in the eyelid opening and closing parameter sequence that is less than the preset initial value for eye opening and is used to calculate the eye-opening benchmark value. The eye-opening benchmark value is the average value calculated from multiple second target eyelid opening and closing parameters and is used to reflect the parameter level of the eye-opening state in the current sequence. The preset eye closure coefficient is a weighting coefficient that is used in conjunction with the eye closure benchmark value and the eye-opening benchmark value to calculate the eye closure judgment threshold. The eye closure judgment threshold is an adaptive judgment value jointly determined by the preset eye closure coefficient, the eye closure benchmark value, and the eye-opening benchmark value.

[0070] In the above technical solution, multiple first target eyelid opening and closing parameters that are greater than the preset initial value for eye closure judgment are selected from the eyelid opening and closing parameter sequence. This can eliminate data interference that deviates significantly from the closed state. By calculating the average value of multiple first target eyelid opening and closing parameters, the eye closure reference value is obtained, which can make the eye closure reference value fit the distribution of the actual collected data. By selecting second target eyelid opening and closing parameters and calculating the eye-opening reference value, and combining the preset eye closure coefficient, the eye closure reference value, and the eye-opening reference value, the eye closure judgment threshold is obtained. This can make the eye closure judgment threshold more consistent with the actual eyelid movement law, improve the rationality of the eye closure state judgment, thereby improving the accuracy of blinking frequency and enhancing detection robustness.

[0071] For example, in the eyelid opening and closing parameter sequence, a first target eyelid opening and closing parameter with a value greater than the preset initial value for eye closure is selected. The arithmetic mean of multiple first target eyelid opening and closing parameters is calculated to obtain the eye closure baseline value. At the same time, a second target eyelid opening and closing parameter with a value less than the preset initial value for eye opening is selected, and the eye opening baseline value is calculated. Combining the preset eye closure coefficient, the eye closure baseline value, and the eye opening baseline value, the eye closure judgment threshold adapted to the current eyelid opening and closing parameter sequence is obtained.

[0072] In some embodiments of this application, a threshold for determining eye closure is obtained based on a preset eye-closing coefficient, an eye-closing reference value, and an eye-opening reference value, including:

[0073] Calculate the difference between the closed-eye reference value and the open-eye reference value to obtain the reference difference value;

[0074] Calculate the first product of the preset eye-closing coefficient and the baseline difference;

[0075] The difference between the closed-eye baseline value and the first product is calculated to obtain the closed-eye judgment threshold.

[0076] In the above technical solution, by calculating the benchmark difference between the closed-eye benchmark value and the open-eye benchmark value step by step, and the first product of the preset closed-eye coefficient and the benchmark difference, and then performing a difference operation based on the closed-eye benchmark value and the first product to obtain the closed-eye judgment threshold, the process of determining the closed-eye judgment threshold can be made to conform to the actual distribution law and benchmark value relationship of the eyelid opening and closing parameters. This ensures that the numerical calculation logic of the closed-eye judgment threshold is rigorous, the result is adaptive and stable and reliable, and thus accurately distinguishes between the half-blink state and the fully closed-eye state. This provides an accurate judgment basis for subsequent closed-eye data group screening and effective blink determination, and improves the accuracy of eyelid state recognition and the reliability of blink counting.

[0077] In some embodiments of this application, before determining the eye-closing judgment threshold based on the eyelid opening and closing parameter sequence and a preset initial value for eye-closing judgment, the method further includes:

[0078] The eyelid opening and closing parameter sequence is cropped based on a preset eye-closing threshold, and the size of the eyelid opening and closing parameters that are greater than the preset eye-closing threshold are adjusted to the size of the preset eye-closing threshold.

[0079] Among them, the preset eye-closing threshold is a critical value used to limit the amplitude of the eyelid opening and closing parameters. Signal clipping is a data smoothing method that adjusts the eyelid opening and closing parameters that are greater than the preset eye-closing threshold to the size of the preset eye-closing threshold. The eyelid opening and closing parameter sequence after signal clipping can suppress the influence of abnormal peaks on subsequent calculations.

[0080] In the above technical solution, the eyelid opening and closing parameter sequence is clipped based on a preset eye-closing threshold. The size of the eyelid opening and closing parameters that are greater than the preset eye-closing threshold is adjusted to the size of the preset eye-closing threshold. This can suppress the interference of abnormal signal peaks on subsequent benchmark and threshold calculations, ensure the stability of the eyelid opening and closing parameter sequence, and make the determination of the eye-closing and eye-opening judgment thresholds more in line with the actual eyelid movement state. This reduces misjudgments caused by abnormal signals and improves the accuracy of blink counting.

[0081] For example, before determining the eye-closing judgment threshold based on the eyelid opening and closing parameter sequence and the preset eye-closing judgment initial value, a signal trimming operation is performed on the eyelid opening and closing parameter sequence using the preset eye-closing threshold value. The eyelid opening and closing parameters in the sequence with values ​​greater than the preset eye-closing threshold value are uniformly adjusted to the preset eye-closing threshold value, thereby reducing the interference of signal mutations and abnormal peaks on the subsequent calculation process.

[0082] In some embodiments of this application, a valid eye-closing data set is determined based on the difference between each eyelid opening / closing parameter and the eye-closing judgment threshold in each eye-closing data set, including:

[0083] Based on the difference between each eyelid opening and closing parameter and the eye closure judgment threshold in each closed eye data group, the cumulative difference in the time domain is obtained;

[0084] The eye-opening judgment threshold is obtained based on the preset eye-opening coefficient, eye-opening reference value, and eye-closing reference value;

[0085] The minimum blink energy threshold is determined based on the open-eye judgment threshold and the closed-eye reference value;

[0086] Data sets with closed eyes where the cumulative difference over the time domain is greater than the minimum blink energy threshold are identified as valid closed eye data sets.

[0087] Among them, the cumulative difference in the time domain is the total value obtained by accumulating the differences between each eyelid opening and closing parameter and the eye closing judgment threshold in the closed eye data group over the time dimension; the minimum blink energy threshold is a quantitative threshold determined by the eye opening judgment threshold and the eye closing benchmark value, used to determine whether the closed eye data group is a real blink; the effective closed eye data group is the closed eye data group whose cumulative difference in the time domain is greater than the minimum blink energy threshold, used to represent a real blinking action.

[0088] In the above technical solution, the method of comparing the cumulative difference in the time domain with the minimum blink energy threshold can be adapted to the actual application scenario of low frame rate image acquisition. However, due to the frame interval limitation of low frame rate acquisition, it is impossible to guarantee that the deepest state of eyelid closure can be accurately captured every time. For example, the physiological duration of a single blink action corresponds to a 3-frame acquisition cycle, but only 2 frames of eye-closing related data can be effectively captured. At this time, the single-frame eyelid opening and closing parameters cannot fully reflect the true characteristics of the blink action. The time-domain cumulative difference can accumulate the closed-eye features of multiple frames within a closed-eye data set, transforming discrete single-frame parameters into a continuous accumulation of motion features. Even if the deepest closed-eye state is not captured, the cumulative amount can still fully characterize the overall features of the blinking action. Simultaneously, the minimum blink energy threshold serves as the feature quantification benchmark for a true blink, accurately distinguishing the effective feature accumulation of a true blink from the weak feature fluctuations of noise, instantaneous eyelid tremors, and half-blinks. The feature accumulation of interference signals falls far short of the minimum blink energy threshold, while the time-domain cumulative difference for a true blink, even with missing sampling frames, still exceeds the threshold, thus achieving accurate differentiation between true blinks and various interference signals. This setup not only solves the judgment error problem caused by incomplete sampling in low-frame-rate scenarios and balances the feature acquisition deviation caused by missing frames, but also fundamentally reduces misjudgments and omissions in blink counting, significantly improving the accuracy of blink counting at different sampling frame rates, further enhancing the robustness and practical applicability of the entire detection scheme.

[0089] For example, the minimum blink energy threshold is calculated based on the open-eye judgment threshold and the closed-eye reference value. The cumulative difference in the time domain corresponding to each closed-eye data group is compared with the minimum blink energy threshold. The closed-eye data group whose cumulative difference in the time domain is greater than the minimum blink energy threshold is determined as the valid closed-eye data group.

[0090] In some embodiments of this application, an eye-opening judgment threshold is obtained based on a preset eye-opening coefficient, an eye-opening reference value, and a eye-closing reference value, including:

[0091] Calculate the second product of the preset eye-opening coefficient and the baseline difference;

[0092] The sum of the eye-opening baseline value and the second product is calculated to obtain the eye-opening judgment threshold.

[0093] In the above technical solution, the second product is calculated based on the benchmark difference and the preset eye-opening coefficient. The second product is then summed with the eye-opening benchmark value to obtain the eye-opening judgment threshold. This allows the eye-opening judgment threshold to conform to the numerical distribution relationship of eyelid opening and closing parameters and the actual movement law. This ensures that the calculation logic of the eye-opening judgment threshold is rigorous, the numerical values ​​are adaptive and reasonable, and the judgment boundary between the half-open eye state and the fully open eye state is accurately defined. This provides a stable benchmark for the subsequent calculation of the minimum blink energy threshold and the identification of effective closed eye data groups, further improving the accuracy of eyelid state judgment and the robustness of blink counting.

[0094] In some embodiments of this application, the sum of the preset closed-eye coefficient and the preset open-eye coefficient is 1.

[0095] In the above technical solution, setting the preset closed eye coefficient and the preset open eye coefficient to a sum of 1 enables the open eye judgment threshold and the closed eye judgment threshold to form a matched and coordinated dual threshold system, ensuring that the threshold calculation logic is unified and the values ​​are reasonable, and further improving the stability and consistency of blink judgment.

[0096] Understandably, the two baseline values ​​correspond to the fully open and fully closed eye states, respectively; the open eye judgment threshold and the closed eye judgment threshold correspond to the half-open eye state and the half-blinking eye state, respectively. By setting the preset closed eye coefficient and the preset open eye coefficient to sum to 1, the two thresholds can form a continuous and complete state judgment logic within the overall variation range of the eyelid opening and closing parameters, thereby achieving accurate differentiation and reliable judgment of the eyelid opening and closing state, ensuring the accuracy of the judgment of the effective closed eye data set, and thus improving the reliability and robustness of blink counting.

[0097] For example, the preset eye-closing coefficient is set to 0.6 and the preset eye-opening coefficient is set to 0.4; or the preset eye-closing coefficient is set to 0.7 and the preset eye-opening coefficient is set to 0.3.

[0098] In some embodiments of this application, the minimum blink energy threshold is determined based on the eye-opening judgment threshold and the eye-closing reference value, including:

[0099] Calculate the difference between the closed-eye baseline value and the open-eye judgment threshold to obtain the baseline difference value;

[0100] The minimum blink energy threshold is obtained by multiplying the baseline difference by the preset calibration coefficient.

[0101] Among them, the reference difference is the numerical difference between the closed-eye reference value and the open-eye judgment threshold, which is used to reflect the parameter difference between the open-eye state and the closed-eye state. The preset calibration coefficient is the calibration parameter that is used in conjunction with the reference difference to calculate the minimum blink energy threshold. The minimum blink energy threshold is a quantitative judgment threshold obtained by multiplying the reference difference by the preset calibration coefficient.

[0102] In the above technical solution, the difference between the closed-eye reference value and the open-eye judgment threshold is calculated to obtain the reference difference value, which can reflect the parameter difference between the open-eye state and the closed-eye state. The minimum blink energy threshold is obtained by multiplying the reference difference value and the preset calibration coefficient. This allows the minimum blink energy threshold to fit the actual range of eyelid opening and closing parameters, ensuring the rationality of the judgment of effective closed-eye data sets, further reducing misjudgment and missed judgment, improving the blink counting accuracy, and enhancing the detection robustness.

[0103] For example, the difference between the closed-eye baseline value and the open-eye judgment threshold is calculated to obtain the baseline difference value. The baseline difference value is multiplied by a preset calibration coefficient to obtain the minimum blink energy threshold adapted to the current sequence. After sliding window processing, the cross-window state parameters are updated, and data such as signal extreme values ​​and time-domain cumulative differences are recorded. A repeat count variable is set to avoid the problem of repeat counting at the window edge, ensuring the continuity and integrity of blink count in time. At the same time, only the core state parameters are transmitted, and there is no need to store complete eye data, which effectively saves terminal storage resources.

[0104] For example, the minimum blink energy threshold is calculated as shown in the following formula:

[0105] ;Formula 2

[0106] in, The minimum blink energy threshold, For preset calibration coefficients, The baseline value is the value with eyes closed. The threshold for determining when eyes are open.

[0107] For example, such as Figure 5As shown, the process begins: S11: Load the eyelid opening and closing parameter sequence and cross-window state parameters; load the eyelid opening and closing parameter data in the current window, and simultaneously load the cross-window state parameters from the previous window, providing a data foundation and state support for processing within the window. The eyelid opening and closing parameter sequence and cross-window state parameters are then passed to S12. S12: Update the open-eye and closed-eye baseline values ​​using the data within the window; based on the eyelid opening and closing parameter sequence in the current window, statistically obtain the open-eye and closed-eye baseline values, providing core reference data for subsequent threshold calculations. The updated open-eye and closed-eye baseline values ​​are then passed to S13. S13: Perform signal trimming on the eyelid opening and closing parameter sequence based on a preset closed-eye threshold; adjust abnormally high-amplitude signals in the sequence that exceed the preset closed-eye threshold to the threshold value, suppressing the interference of abnormal peak values ​​on the baseline and threshold values, improving data reliability. The trimmed eyelid opening and closing parameter sequence is then passed to S14 and S15. S14: Calculate and filter the difference signal 1 between the clipped signal and the eye-closed judgment threshold; calculate the difference between the clipped signal and the eye-closed judgment threshold to obtain the difference signal 1, and filter out high-frequency noise to weaken the interference of noise on the detection. Transmit the processed difference signal 1 to S16, S17, or S18. S15: Calculate and filter the difference signal 2 between the clipped signal and the eye-open judgment threshold; calculate the difference between the clipped signal and the eye-open judgment threshold to obtain the difference signal 2, and filter out high-frequency noise to ensure the accuracy of eye-opening state detection. Transmit the processed difference signal 2 to S16, S17, or S18. S16: Eye-closed judgment threshold detection: Based on the cumulative difference of difference signal 1 and the minimum blink energy threshold, identify the effective segment of closed / half-closed eyes; accurately identify the effective signal segment of closed and half-closed eyes by combining the cumulative difference of difference signal 1 with the minimum blink energy threshold, avoiding invalid noise interference, and transmit the identified effective segment of closed / half-closed eyes to S19. S17: Eye-opening threshold detection: Based on the cumulative difference of differential signal 2 and the minimum blink energy threshold, identify the effective segments of open / half-open eyes; by combining the cumulative difference of differential signal 2 with the minimum blink energy threshold, accurately identify the effective signal segments of open and half-open eyes, improve the full-segment recognition of blinking behavior, and pass the identified effective segments of open / half-open eyes to S19. S18: Supplementary detection of rapid blinks: Supplement the recognition of short-duration blinks to avoid missed detections; supplement the recognition of special blink types such as rapid blinks to further improve the completeness of blink counting, and pass the supplementary recognition results to S19. Rapid blinks refer to blinking actions where the duration of the eyelid opening / closing parameter rising from below 0.2 to above 0.8 is less than 0.2 seconds, and the cumulative difference in the time domain is greater than 50% of the minimum blink energy threshold. This supplementary detection logic addresses the problem of short-duration blinks being easily missed in low frame rate scenarios, complementing the judgment method of cumulative difference in the time domain, further improving blink recognition in low frame rate scenarios and enhancing the completeness of counting.S19: Window Post-processing: Count the number of valid closed-eye data sets to determine the blink count of the current window; count the number of valid closed-eye data sets after filtering by multi-detection logic to obtain the blink count within the current window, completing the core counting task within the window, and pass the blink count of the current window to S20. S20: Cross-Window Parameter Update: Update state parameters such as the cumulative difference in the time domain and the repeat count variable; update cross-window state parameters such as the cumulative difference in the time domain and the repeat count variable to ensure that edge signals can be correctly accumulated when the window slides, avoid repeat counting, and provide accurate state for the next window processing.

[0108] For example, such as Figure 6 As shown, the process begins with S31: Initialize the parameter window buffer; initialize the sliding window's related parameters and data buffer to provide the initial data environment and storage support for subsequent continuous frame processing and counting. After initialization, proceed to S32. S32: Wait for new frames; wait to acquire new eye image frames to provide a data source for the continuous updating and counting of the sliding window. After acquiring a new frame, proceed to S33. S33: Window full?; determine if the current sliding window is full of data. If not, return to S32 to continue waiting for new frames; if full, transfer the current window data to S34. S34: Acquire head pose; acquire the current head pose information to support eye data reliability assessment and pose correction, and transfer the head pose information to S35 and S36. S35: Count the blink frequency of the left and right eyes separately; based on the eyelid opening and closing parameter sequence, count the effective blinks of the left and right eyes independently to provide a basis for subsequent binocular data fusion, and transfer the count results of each eye to S37. S36: Calculate the reliability of left and right eye data; combine head posture information to evaluate the reliability of left and right eye blink data, providing a basis for subsequent weighted fusion. For example, if the head posture offset angle is <15°, the data reliability is 1.0; if 15°≤offset angle≤30°, the reliability linearly decreases to 0.5; if the offset angle is >30°, the reliability is 0.2, and the reliability evaluation result is passed to S37. S37: Output the blink count of the current window; based on the left and right eye counts and reliability weights, fuse the total blink count within the current window, and pass the current window blink count to S38 and S39. S38: Slide the window and update parameters; perform a window sliding operation, update cross-window state parameters to prepare for the next round of window processing, ensuring the continuity of counting, and return to S32 after completion. S39: Calculate the current blink frequency; based on the current blink count and window duration, calculate the current blink frequency, and pass the blink frequency to S40. S40: Assess and determine eye fatigue; based on the current blinking frequency, assess the user's eye fatigue, output the current blinking frequency and eye fatigue level, and complete a full processing flow.

[0109] For example, in addition to eye fatigue detection, the number and frequency of blinks can be used to assess user engagement with content, assist in analyzing emotional state, and expand application scenarios.

[0110] The eye fatigue detection method provided in this application can be executed by an eye fatigue detection device. This application uses an eye fatigue detection device to execute the eye fatigue detection method as an example to illustrate the eye fatigue detection device provided in this application.

[0111] like Figure 7 As shown, this application embodiment provides an eye fatigue detection device 300, including:

[0112] Module 301 is used to construct a 3D face model, which is used to represent the eyelid movement state.

[0113] The first determining module 302 is used to acquire multiple consecutive frames of face images and determine the eyelid opening and closing parameters corresponding to each frame of face image based on the three-dimensional face model. The eyelid opening and closing parameters are used to characterize the opening and closing ratio relative to the preset fully open eye state.

[0114] The first processing module 303 is used to obtain a sequence of eyelid opening and closing parameters based on a sliding window of a preset length;

[0115] The second determining module 304 is used to determine the eye-closing judgment threshold based on the eyelid opening and closing parameter sequence and the preset eye-closing judgment initial value.

[0116] The second processing module 305 is used to divide the eyelid opening and closing parameters that reach the eye-closing judgment threshold into multiple eye-closing data groups according to the temporal continuity relationship in the eyelid opening and closing parameter sequence.

[0117] The third determining module 306 is used to determine the valid closed eye data group based on the difference between the eyelid opening and closing parameters in each closed eye data group and the closed eye judgment threshold.

[0118] The fourth determining module 307 is used to determine the blinking frequency based on the effective closed-eye data set.

[0119] The fifth determining module 308 is used to determine eye fatigue based on blink frequency.

[0120] In this embodiment, a three-dimensional face model is constructed to characterize the eyelid movement state. This three-dimensional spatial representation replaces traditional two-dimensional planar analysis, mitigating recognition interference caused by head posture shifts and changes in shooting angle, reducing measurement deviations in eyelid opening and closing parameters. The three-dimensional face model determines the eyelid opening and closing parameters corresponding to each frame of the face image, objectively representing the opening and closing ratio relative to a preset fully open state, further ensuring the stability of parameter acquisition. Then, a sliding window of preset length is used to extract continuous eyelid opening and closing parameters, forming a temporally coherent sequence of eyelid opening and closing parameters, ensuring the integrity and continuity of the temporal data. Subsequently, the eyelid opening and closing parameters are combined with... The parameter sequence and preset initial values ​​for eye closure determination determine the eye closure judgment threshold. The eyelid opening and closing parameters that reach the eye closure judgment threshold are divided into multiple eye closure data groups according to their temporal continuity, completing a reasonable initial screening of eye closure segments. Finally, the difference between each eyelid opening and closing parameter in each eye closure data group and the eye closure judgment threshold is calculated. Based on all the differences in each eye closure data group, the valid eye closure data group is determined, effectively distinguishing real blinks from noise, instantaneous eyelid tremors and other interference signals, reducing false and missed judgments. The number of blinks is determined based on the valid eye closure data group, which effectively improves the accuracy of blink counting and the overall detection robustness, thereby ensuring the accuracy of the obtained eye fatigue level.

[0121] The eye fatigue detection device in this application embodiment can be an electronic device or a component within an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal or other devices besides a terminal. For example, the electronic device can be a mobile phone, tablet computer, laptop computer, PDA, in-vehicle electronic device, mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc. It can also be a server, network attached storage (NAS), personal computer (PC), television (TV), ATM, or self-service machine, etc. This application embodiment does not specifically limit the device.

[0122] The eye fatigue detection device in this application embodiment can be a device with an operating system. The operating system can be Android, iOS, or other possible operating systems, and this application embodiment does not specifically limit it.

[0123] The eye fatigue detection device provided in this application embodiment can achieve... Figures 1 to 6 The various processes implemented in the embodiment of the eye fatigue detection method will not be described again here to avoid repetition.

[0124] Optionally, such as Figure 8 As shown, this application embodiment also provides an electronic device 600, including a processor 602 and a memory 604. The memory 604 stores a program or instructions that can run on the processor 602. When the program or instructions are executed by the processor 602, they implement the various steps of the above-described eye fatigue detection method embodiment and can achieve the same technical effect. To avoid repetition, they will not be described again here.

[0125] It should be noted that the electronic devices in the embodiments of this application include the aforementioned mobile electronic devices and non-mobile electronic devices.

[0126] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described eye fatigue detection method embodiments and achieve the same technical effects. To avoid repetition, they will not be described again here.

[0127] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0128] This application also provides a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled. The processor is used to run programs or instructions to implement the various processes of the above-described eye fatigue detection method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0129] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0130] This application provides a computer program product, which is stored in a storage medium and executed by at least one processor to implement the various processes of the above-described eye fatigue detection method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0131] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0132] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0133] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for detecting eye fatigue, characterized in that, include: A three-dimensional face model is constructed, which is used to represent the movement state of the eyelids; Acquire multiple consecutive frames of facial images, and determine the eyelid opening and closing parameters corresponding to each frame of the facial image based on the three-dimensional facial model. The eyelid opening and closing parameters are used to characterize the opening and closing ratio relative to a preset fully open eye state. Based on a sliding window of preset length, a sequence of eyelid opening and closing parameters is obtained; The eyelid opening and closing parameter sequence and the preset initial value for eye closure determination are used to determine the eye closure determination threshold. In the eyelid opening and closing parameter sequence, the eyelid opening and closing parameters that reach the eye-closing judgment threshold are divided into multiple eye-closing data groups according to the temporal continuity relationship; Based on the difference between the eyelid opening and closing parameters in each of the closed-eye data groups and the closed-eye judgment threshold, a valid closed-eye data group is determined. Based on the effective closed-eye data set, the blinking frequency was determined; The degree of eye fatigue is determined based on the blinking frequency.

2. The method for detecting eye fatigue according to claim 1, characterized in that, The step of determining the eyelid opening and closing parameters corresponding to each frame of the face image based on the three-dimensional face model includes: Based on multiple facial feature points in each frame of the face image, determine the head pose information corresponding to each frame of the face image; Based on each frame of the face image, the eye image corresponding to each frame of the face image is cropped; The eye image and head pose information corresponding to each frame of the face image are input into the three-dimensional face model to obtain the eyelid opening and closing parameters corresponding to each frame of the face image.

3. The method for detecting eye fatigue according to claim 1, characterized in that, The step of determining the eye-closing judgment threshold based on the eyelid opening and closing parameter sequence and the preset initial value for eye-closing judgment includes: From the eyelid opening and closing parameter sequence, select multiple first target eyelid opening and closing parameters that are greater than the preset initial value for eye closure determination; Calculate the average value of multiple first target eyelid opening and closing parameters to obtain the eye-closed baseline value; Based on the eyelid opening and closing parameter sequence, eyelid opening and closing parameters that are less than the preset initial eye-opening parameter are selected as the second target eyelid opening and closing parameters; The average value of multiple second target eyelid opening and closing parameters is calculated to obtain the eye-opening baseline value; The eye-closing judgment threshold is obtained based on the preset eye-closing coefficient, the eye-closing reference value, and the eye-opening reference value.

4. The method for detecting eye fatigue according to claim 3, characterized in that, Before determining the eye-closing judgment threshold based on the eyelid opening and closing parameter sequence and the preset initial value for eye-closing judgment, the method further includes: The eyelid opening and closing parameter sequence is processed based on a preset eye-closing threshold, and the size of the eyelid opening and closing parameters that are greater than the preset eye-closing threshold are adjusted to the size of the preset eye-closing threshold.

5. The method for detecting eye fatigue according to claim 3, characterized in that, The determination of valid closed-eye data sets based on the difference between the eyelid opening / closing parameters in each of the closed-eye data sets and the closed-eye judgment threshold includes: Based on the difference between the eyelid opening and closing parameters and the eye closure judgment threshold in each of the closed eye data groups, the cumulative difference in the time domain is obtained; The eye-opening judgment threshold is obtained based on the preset eye-opening coefficient, the eye-opening reference value, and the eye-closing reference value; The minimum blink energy threshold is determined based on the eye-opening judgment threshold and the eye-closing reference value; The closed-eye data groups whose cumulative difference over the time domain is greater than the minimum blink energy threshold are selected and determined as valid closed-eye data groups.

6. The method for detecting eye fatigue according to claim 5, characterized in that, The sum of the preset closed-eye coefficient and the preset open-eye coefficient is 1.

7. The method for detecting eye fatigue according to claim 6, characterized in that, The step of determining the minimum blink energy threshold based on the eye-opening judgment threshold and the eye-closing reference value includes: Calculate the difference between the closed-eye reference value and the open-eye judgment threshold to obtain the reference difference value; The minimum blink energy threshold is obtained by multiplying the benchmark difference by the preset calibration coefficient.

8. A device for detecting eye fatigue, characterized in that, include: A construction module is used to construct a three-dimensional face model, which is used to represent the movement state of the eyelids; The first determining module is used to acquire multiple consecutive frames of face images and determine the eyelid opening and closing parameters corresponding to each frame of the face image based on the three-dimensional face model. The eyelid opening and closing parameters are used to characterize the opening and closing ratio relative to a preset fully open eye state. The first processing module is used to obtain the eyelid opening and closing parameter sequence based on a sliding window of preset length; The second determining module is used to determine the eye-closing judgment threshold based on the eyelid opening and closing parameter sequence and the preset eye-closing judgment initial value; The second processing module is used to divide the eyelid opening and closing parameters that reach the eye-closing judgment threshold into multiple eye-closing data groups according to the temporal continuity relationship in the eyelid opening and closing parameter sequence. The third determining module is used to determine the effective closed eye data group based on the difference between the eyelid opening and closing parameters in each of the closed eye data groups and the closed eye judgment threshold. The fourth determining module is used to determine the blinking frequency based on the effective closed-eye data set; The fifth determining module is used to determine the degree of eye fatigue based on the blinking frequency.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions that can run on the processor, the program or instructions being executed by the processor to implement the steps of the eye fatigue detection method as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the eye fatigue detection method as described in any one of claims 1 to 7.

Images