A tablet computer use time alert method and device

Abstract

The application relates to the technical field of data processing, in particular to a tablet computer use time warning method and device, which comprises the following steps: monitoring a touch event sequence of a target user on a tablet computer; determining a touch control point sequence according to the time interval between adjacent touch events in the touch event sequence; performing trajectory fitting on the touch control point sequence according to a Bezier curve to obtain a touch trajectory; determining an operation intensity label of the target user according to the touch event sequence and the touch trajectory; monitoring a head posture video of the target user; and obtaining a video frame sequence of the head posture video. The application can accurately evaluate the operation intensity of the user by monitoring the touch event sequence of the user on the tablet computer and fitting the touch trajectory according to the time interval between events and the Bezier curve, and the accurate evaluation is helpful for judging whether the user has the risk of over operation.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and device for alerting tablet computer usage time. Background Technology

[0002] Traditional methods typically rely on simple touch event statistics or fixed time interval assessments, failing to accurately capture the user's actual operational intensity and fatigue level. This approach struggles to provide personalized assessments based on each user's behavioral differences and actual needs. Furthermore, traditional methods are often based on static rules or settings, failing to dynamically monitor changes in user behavior in real time. User operational intensity and fatigue levels fluctuate with time and usage, and traditional methods struggle to adapt to these changes. Moreover, traditional methods often employ a uniform alert mechanism, unable to flexibly adjust based on different users' habits, physical conditions, or fatigue levels. Regardless of the user's actual situation, alerts are usually standardized, easily leading to excessive warnings or unnecessary reminders, impacting user experience. Additionally, traditional methods often provide a coarse assessment of fatigue, struggling to accurately predict fatigue levels using actual physiological data (such as head posture and eye movements). Therefore, they easily miss early warnings of fatigue, potentially causing users to experience excessive fatigue without realizing it. Finally, traditional methods typically focus only on monitoring operational behavior, paying less attention to health risk prevention. Even with prolonged device use, traditional methods may not promptly alert users to eye or neck fatigue, potentially leading to the accumulation of health problems. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a method and device for alerting the usage time of a tablet computer.

[0004] The technical solution adopted to solve the above-mentioned technical problem is: a method for alerting the usage time of a tablet computer, comprising:

[0005] The system monitors the target user's touch event sequence on the tablet computer, determines the touch control point sequence based on the time interval between adjacent touch events in the touch event sequence, and performs trajectory fitting on the touch control point sequence based on a Bezier curve to obtain the touch trajectory.

[0006] The target user's operation intensity label is determined based on the touch event sequence and the touch trajectory;

[0007] Monitor the head posture video of the target user, obtain the video frame sequence of the head posture video, and perform key point detection on the video frame sequence according to the pre-trained key point detection model to obtain the key point set of each video frame in the video frame sequence.

[0008] The target user's action pattern feature sequence is determined by the key point set of each video frame in the video frame sequence according to a preset sliding window. Fatigue prediction is then performed by the action pattern feature sequence according to a pre-trained fatigue prediction model to obtain the fatigue intensity label of the target user.

[0009] The operation intensity label and the fatigue intensity label are matched with a preset label mapping table to obtain the usage intensity level of the target user, and usage warnings are issued based on the usage intensity level.

[0010] Preferably, the touch event sequence includes at least one touch event, and the touch control trajectory includes at least two touch events, wherein the touch event includes touch point coordinates and duration.

[0011] Preferably, determining the touch control point sequence based on the time interval between adjacent touch events in the touch event sequence includes:

[0012] Determine the time interval between adjacent touch events in the touch event sequence;

[0013] The average time interval of the touch event sequence is determined based on the time interval between adjacent touch events in the touch event sequence, and the average time interval is used as the dynamic threshold of the interval of the touch event sequence.

[0014] If the time interval between adjacent touch events in the touch event sequence is less than the dynamic threshold of the interval, then the adjacent touch events are regarded as a pair of adjacent touch points;

[0015] The Euclidean distance between the adjacent touch point pairs is determined based on the touch point coordinates of the adjacent touch point pairs;

[0016] If the Euclidean distance between the adjacent touch point pairs is less than a preset distance threshold, the adjacent touch point pairs are used as touch control points. After processing all touch events in the touch event sequence, a touch control point sequence is obtained.

[0017] Preferably, the touch control point sequence is fitted with a Bézier curve to obtain the touch trajectory, including:

[0018] The first touch control point in the touch control point sequence is taken as the starting point of the Bézier curve, and the last touch control point in the touch control point sequence is taken as the ending point of the Bézier curve.

[0019] Determine the chord length between adjacent touch control points in the touch control point sequence, and determine the cumulative chord length corresponding to the touch control point sequence based on the chord length between adjacent touch control points in the touch control point sequence;

[0020] The chord lengths between adjacent touch control points in the touch control point sequence are normalized based on the cumulative chord length to obtain the control parameters of the Bézier curve.

[0021] The value of the Bezier basis function is determined based on the control parameters of the Bezier curve. The target function is obtained by minimizing the value of the Bezier basis function with the coordinates of the corresponding touch control points in the touch control point sequence.

[0022] The objective function is solved using the least squares method to obtain multiple intermediate control points of the Bézier curve. The touch trajectory is then determined based on the starting point, multiple intermediate control points, and the ending point of the Bézier curve.

[0023] Preferably, determining the target user's operation intensity label based on the touch event sequence and the touch trajectory includes:

[0024] Remove the touch events corresponding to the touch control point sequence from the touch event sequence to obtain the press touch event sequence;

[0025] Based on the duration of each touch event in the press touch event sequence, the press touch event sequence is divided into a single-point click touch event sequence and a long-press touch event sequence;

[0026] The number of single-point clicks in the single-point click event sequence and the number of long-press operations in the long-press event sequence are counted to determine the trajectory length of the touch trajectory.

[0027] The number of single-point clicks, the number of long-press operations, and the trajectory length are weighted according to preset weighting coefficients to obtain the operation intensity coefficient of the target user;

[0028] The target user's operation intensity label is determined based on the target user's operation intensity coefficient.

[0029] Preferably, the set of key points includes key points of the nose, left eye, right eye, left ear, right ear, center of the neck, left shoulder, and right shoulder.

[0030] Preferably, determining the target user's action pattern feature sequence through the key point set of each video frame in the video frame sequence using a preset sliding window includes:

[0031] The preset sliding window is moved in the video frame sequence, and the fatigue feature set of the target user is determined based on the key point set of all video frames within the sliding window;

[0032] Each fatigue feature in the fatigue feature set is weighted according to a preset fatigue weight to obtain the comprehensive fatigue feature of the target user;

[0033] The feature dispersion of the comprehensive fatigue feature is determined based on the standard deviation and mean of the comprehensive fatigue feature. The feature dispersion is the motion pattern feature of the sliding window. When the sliding window moves to the end of the video frame sequence, the motion pattern feature sequence corresponding to the target user is obtained.

[0034] Preferably, the fatigue feature set includes posture stability features, limb movement discrete features, and joint angle fluctuation features. The posture stability features represent the head sway amplitude of the target user. The posture stability features are calculated by summing the squared differences between the coordinates of the nose key point at each moment and the mean coordinates of the nose key point within the sliding window, and then averaging the sum of the squared differences over all moments. The limb movement discrete features represent the regularity of the target user's arm movements. The limb movement discrete features are calculated by calculating the Euclidean distance between the left and right shoulder key points at each moment, and then calculating the standard deviation of the Euclidean distance over all moments. The joint angle fluctuation features represent the fluctuation of the target user's neck angle. The joint angle fluctuation features are calculated by calculating the angle between the nose key point and the center neck key point at each moment using an inverse trigonometric function, and then calculating the variance of the angle over all moments.

[0035] Preferably, the fatigue prediction model includes a bidirectional LSTM layer, an attention layer, and a classification layer. The bidirectional LSTM layer is used to capture the sequential dependence features of the action pattern feature sequence. The attention layer is used to focus on key fatigue accumulation periods in the action pattern feature sequence to obtain the attention weights corresponding to the sequential dependence features of each period. The sequential dependence features of each period are weighted according to the attention weights to obtain comprehensive temporal dependence features. The classification layer is used to classify the comprehensive temporal dependence features for fatigue according to the fully connected layer to obtain fatigue probability, and to perform label mapping according to the fatigue probability to obtain fatigue intensity labels.

[0036] The technical solution adopted to solve the above-mentioned technical problem is: a tablet computer usage time warning device, which is applicable to the aforementioned tablet computer usage time warning method, including:

[0037] A touch monitoring unit is used to monitor the touch event sequence of a target user on a tablet computer, determine the touch control point sequence based on the time interval between adjacent touch events in the touch event sequence, and perform trajectory fitting on the touch control point sequence based on a Bezier curve to obtain a touch trajectory.

[0038] An operation classification unit is configured to determine the operation intensity label of the target user based on the touch event sequence and the touch trajectory.

[0039] A key point detection unit is used to monitor the head posture video of the target user, obtain the video frame sequence of the head posture video, and perform key point detection on the video frame sequence according to a pre-trained key point detection model to obtain the key point set of each video frame in the video frame sequence.

[0040] The fatigue prediction unit is used to determine the action pattern feature sequence of the target user through the key point set of each video frame in the video frame sequence according to a preset sliding window, and to perform fatigue prediction through the action pattern feature sequence according to the pre-trained fatigue prediction model to obtain the fatigue intensity label of the target user.

[0041] The warning unit is used to match the operation intensity label and the fatigue intensity label with a preset label mapping table to obtain the usage intensity level of the target user, and to issue a usage warning based on the usage intensity level.

[0042] The beneficial effects of the present invention are as follows: (1) The present invention can accurately assess the user's operation intensity by monitoring the user's touch event sequence on the tablet computer and fitting the touch trajectory according to the time interval between events and the Bezier curve. This accurate assessment helps to determine whether the user is at risk of over-operation. Furthermore, by performing key point detection on the target user's head posture video, extracting the action pattern feature sequence, and analyzing it through a pre-trained fatigue prediction model, the user's fatigue intensity can be predicted in real time. This not only monitors the degree of fatigue but also provides early warnings to avoid fatigue or discomfort caused by prolonged overuse of the device. (2) The present invention combines operation intensity labels and fatigue intensity labels and generates personalized usage intensity levels through a preset label mapping table. It can provide more accurate warnings based on the user's actual state. This personalized assessment system avoids the "one-size-fits-all" warning mode and provides more humane reminders. Furthermore, by monitoring and analyzing the user's operation intensity and fatigue intensity in real time, the system can issue warnings when the user may be over-fatigued or over-operated, which helps to reduce the health risks caused by prolonged use of the tablet computer, such as eye fatigue and neck fatigue, thereby improving the user experience. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the overall method steps in one embodiment of the present invention;

[0044] Figure 2 This is a schematic diagram of the overall device architecture in one embodiment of the present invention.

[0045] Reference numerals: 1. Touch monitoring unit; 2. Operation classification unit; 3. Key point detection unit; 4. Fatigue prediction unit; 5. Usage warning unit. Detailed Implementation

[0046] Example 1, as Figure 1 As shown, the present invention proposes a method for alerting users about the usage time of a tablet computer, comprising:

[0047] S1. Monitor the target user's touch event sequence on the tablet computer, determine the touch control point sequence based on the time interval between adjacent touch events in the touch event sequence, and perform trajectory fitting on the touch control point sequence based on the Bézier curve to obtain the touch trajectory;

[0048] S2. Determine the target user's operation intensity label based on the touch event sequence and touch trajectory;

[0049] S3. Monitor the head posture video of the target user, obtain the video frame sequence of the head posture video, and perform key point detection on the video frame sequence according to the pre-trained key point detection model to obtain the key point set of each video frame in the video frame sequence.

[0050] S4. Based on the preset sliding window, determine the action pattern feature sequence of the target user through the key point set of each video frame in the video frame sequence. Based on the pre-trained fatigue prediction model, perform fatigue prediction through the action pattern feature sequence to obtain the fatigue intensity label of the target user.

[0051] S5. Match the operation intensity label and fatigue intensity label with the preset label mapping table to obtain the target user's usage intensity level, and issue usage warnings based on the usage intensity level.

[0052] In this invention, the touch event sequence refers to the time sequence of all events in which a user interacts with the screen while using a tablet computer; the time interval refers to the time difference between adjacent touch events, i.e., the time interval between each touch operation by the user; the touch control point sequence refers to a set of data points formed by the touch position and time captured in the touch event sequence, and the control points usually reflect the user's movement trajectory on the touch screen; the Bézier curve is a mathematical curve often used to describe smooth, fluid trajectories, and it can generate smooth curves through control points, and is widely used in computer graphics and animation; the touch trajectory refers to the path formed by the user's touch points on the screen, and is obtained by fitting the control point sequence with a Bézier curve; the operation intensity label refers to classifying these behaviors into different intensity levels, such as mild, moderate, and severe, by analyzing the user's touch event sequence and touch trajectory; the head posture video refers to the video recording of changes in the user's head posture captured by a camera, which can help determine the user's attention, fatigue state, etc.; and the video frame sequence refers to each frame of the head posture video arranged in chronological order. Each frame represents a specific moment in the video; the keypoint detection model is a machine learning model used to detect and locate the coordinates of key parts of a person in an image or video. These keypoints help determine the user's posture and actions; the keypoint set refers to the collection of all detected keypoints in the video frame, representing the coordinate positions of various parts of the person's body; the sliding window is a time series analysis method that uses a fixed-length window to gradually slide and extract features from different time periods. Here, the sliding window is used to extract features from a sequence of video frames; the action pattern feature sequence is a feature extracted from the keypoint set in the video frame sequence, reflecting the user's action patterns over different time periods; the label mapping table is a table used to map operation intensity labels and fatigue intensity labels to usage intensity levels, typically containing predefined rules, such as different warning levels corresponding to different operations and fatigue intensities; usage warnings refer to warnings issued by the system based on the user's operation intensity and fatigue intensity labels, reminding the user whether they need to rest or reduce device use to prevent health risks due to overuse.

[0053] Example 2: The present invention proposes a method for alerting the usage time of a tablet computer. Compared with Example 1, this example further includes: a touch event sequence including at least one touch event, and a touch control trajectory including at least two touch events, wherein the touch event includes touch point coordinates and duration.

[0054] In an optional embodiment, determining the touch control point sequence based on the time interval between adjacent touch events in the touch event sequence includes:

[0055] A1. Determine the time interval between adjacent touch events in the touch event sequence;

[0056] A2. Determine the average time interval of the touch event sequence based on the time interval between adjacent touch events in the touch event sequence, and use the average time interval as the dynamic threshold of the interval of the touch event sequence.

[0057] A3. If the time interval between adjacent touch events in the touch event sequence is less than the dynamic threshold, then the adjacent touch events are regarded as a pair of nearby touch points.

[0058] A4. Determine the Euclidean distance between adjacent touch point pairs based on the touch point coordinates of adjacent touch point pairs;

[0059] A5. If the Euclidean distance between adjacent touch point pairs is less than the preset distance threshold, then the adjacent touch point pairs are used as touch control points. After processing all touch events in the touch event sequence, the touch control point sequence is obtained.

[0060] It should be noted that in a touch event sequence, the average time interval between all adjacent touch events is calculated. This average can represent a temporal pattern in the touch event sequence and is also used as a dynamic threshold to further analyze the correlation of events. The threshold calculated based on the average time interval of the touch event sequence is used to determine whether the time between adjacent touch events is close enough. If the time interval between adjacent events is less than this dynamic threshold, these touch events are considered to be part of the same operation. A pair of adjacent touch points refers to a pair of touch points formed by two touch events that are close enough in time. If the time interval between these two events is less than the dynamic threshold, they are considered to be a pair of adjacent touch points.

[0061] In an optional embodiment, the touch control point sequence is fitted with a Bézier curve to obtain the touch trajectory, including:

[0062] B1. Take the first touch control point in the touch control point sequence as the starting point of the Bézier curve, and the last touch control point in the touch control point sequence as the ending point of the Bézier curve.

[0063] B2. Determine the chord length between adjacent touch control points in the touch control point sequence, and determine the cumulative chord length corresponding to the touch control point sequence based on the chord length between adjacent touch control points in the touch control point sequence;

[0064] B3. Normalize the chord length between adjacent touch control points in the touch control point sequence based on the cumulative chord length to obtain the control parameters of the Bézier curve;

[0065] B4. Determine the value of the Bezier basis function based on the control parameters of the Bezier curve, and minimize the target by using the value of the Bezier basis function and the corresponding touch point coordinates in the touch control point sequence to obtain the target function;

[0066] B5. Solve the objective function using the least squares method to obtain multiple intermediate control points of the Bézier curve. Determine the touch trajectory based on the starting point, multiple intermediate control points, and the ending point corresponding to the Bézier curve.

[0067] It's important to note that a Bézier curve is a type of curve commonly used in graphic design and computer graphics. It's typically defined by a start point, an end point, and several control points. Bézier curves are used to smoothly fit and describe paths or trajectories. Depending on the number of control points, Bézier curves have different orders, commonly including linear, quadratic, and cubic Bézier curves. Chord length refers to the straight-line distance between two points. For a sequence of touch control points, the chord length is the distance between adjacent touch control points, reflecting the spatial relationship between them. Cumulative chord length provides a metric for measuring the progress of touch points along a path, often used for path normalization to ensure uniform scale across different curve segments. Control parameters are used to define the shape of the Bézier curve; they are typically the positions of control points and the proportional relationships between adjacent points. Adjusting these parameters can change the curvature of the curve; Bézier basis functions are the mathematical representation of Bézier curves, describing how control points affect the shape of the curve. Each basis function is associated with a corresponding control point, and the weight of the basis function determines the contribution of that control point to the curve shape; minimizing the objective function refers to adjusting the control points of the curve during the optimization process to minimize a defined objective function (e.g., an error function). The objective function usually represents the error of fitting the Bézier curve to the sequence of touch control points. Minimizing the objective function means minimizing the error between the fitted touch trajectory and the actual trajectory; the least squares method is a commonly used optimization method used to minimize the sum of squared errors between a set of data points and the fitted curve. In the process of fitting a Bézier curve, the least squares method is used to calculate the optimal position of the control points to minimize the fitting error.

[0068] In an optional embodiment, determining the target user's operation intensity label based on the touch event sequence and touch trajectory includes:

[0069] C1. Remove the touch events corresponding to the touch control point sequence from the touch event sequence to obtain the press touch event sequence;

[0070] C2. Based on the duration of each touch event in the press touch event sequence, the press touch event sequence is divided into single-point click touch event sequence and long press touch event sequence;

[0071] C3. Count the number of single-point clicks in the single-point click touch event sequence, count the number of long-press operations in the long-press touch event sequence, and determine the trajectory length of the touch trajectory;

[0072] C4. Weight the number of single clicks, the number of long presses, and the trajectory length according to the preset weighting coefficients to obtain the target user's operation intensity coefficient;

[0073] C5. Determine the target user's operation intensity label based on the target user's operation intensity coefficient.

[0074] It should be noted that in the weighted calculation, the weight coefficient is a value used to determine the proportion of each indicator in the final operation intensity calculation. For example, the number of single clicks, the number of long press operations, and the trajectory length can have different weight coefficients to reflect their different impacts on the operation intensity.

[0075] In an optional embodiment, the keypoint set includes a nose keypoint, a left eye keypoint, a right eye keypoint, a left ear keypoint, a right ear keypoint, a center neck keypoint, a left shoulder keypoint, and a right shoulder keypoint.

[0076] In an optional embodiment, the action pattern feature sequence of the target user is determined by using a preset sliding window through the key point set of each video frame in the video frame sequence, including:

[0077] D1. Move the preset sliding window through the video frame sequence, and determine the fatigue feature set of the target user based on the key point set of all video frames within the sliding window;

[0078] D2. Weight each fatigue feature in the fatigue feature set according to the preset fatigue weight to obtain the comprehensive fatigue features of the target user;

[0079] D3. Determine the feature dispersion of the comprehensive fatigue features based on the standard deviation and mean of the comprehensive fatigue features. The feature dispersion is the motion pattern feature of the sliding window. When the sliding window moves to the end of the video frame sequence, the motion pattern feature sequence corresponding to the target user is obtained.

[0080] It should be noted that fatigue weight refers to the importance of each fatigue feature when analyzing fatigue features. Different fatigue features may have different impacts on judging the user's fatigue level. Therefore, it is necessary to assign a weight value to each feature, usually based on experience or actual data. Feature dispersion measures the volatility of a feature through standard deviation. In this scenario, feature dispersion refers to the range of variation of the overall fatigue features, used to describe the regularity and stability of the target user's actions. High dispersion indicates that the actions change more and the fatigue state is unstable, while low dispersion indicates that the user's state is relatively stable.

[0081] In an optional embodiment, the fatigue feature set includes posture stability features, limb movement discrete features, and joint angle fluctuation features. Posture stability features represent the head sway amplitude of the target user. The posture stability features are calculated by summing the squared differences between the coordinates of the nose keypoint at each moment and the mean coordinates of the nose keypoint within the sliding window, and then averaging the sum of the squared differences over all moments. Limb movement discrete features represent the regularity of the target user's arm movements. The limb movement discrete features are calculated by calculating the Euclidean distance between the left and right shoulder keypoints at each moment, and then calculating the standard deviation of the Euclidean distance over all moments. Joint angle fluctuation features represent the fluctuation of the target user's neck angle. The joint angle fluctuation features are calculated by calculating the angle between the nose keypoint and the center neck keypoint at each moment using an inverse trigonometric function, and then calculating the variance of the angle over all moments.

[0082] It should be noted that inverse trigonometric functions are the inverse operations of trigonometric functions, such as arcsine, arccosine, and arctangent. Here, inverse trigonometric functions are used to calculate the angle based on the distance between the center of the nose and the center of the neck, thereby measuring the movement state of the neck.

[0083] In an optional embodiment, the fatigue prediction model includes a bidirectional LSTM layer, an attention layer, and a classification layer. The bidirectional LSTM layer is used to capture the sequential dependence features of the action pattern feature sequence. The attention layer is used to focus on the key fatigue accumulation period in the action pattern feature sequence to obtain the attention weights corresponding to the sequential dependence features of each period. The sequential dependence features of each period are weighted according to the attention weights to obtain the comprehensive temporal dependence features. The classification layer is used to classify the comprehensive temporal dependence features for fatigue according to the fully connected layer to obtain the fatigue probability. The fatigue probability is then used for label mapping to obtain the fatigue intensity label.

[0084] It should be noted that the bidirectional LSTM layer is a type of recurrent neural network (RNN) capable of capturing long-term dependencies. By simultaneously considering the forward and backward sequence information of the input data, the bidirectional LSTM enhances the model's ability to capture temporal dependencies, making it particularly suitable for processing sequential data. Fatigue accumulation periods refer to the time periods in a time series where fatigue gradually accumulates and affects the body or behavior. The characteristics of these periods are crucial for fatigue prediction, and the attention mechanism improves the accuracy of model prediction by focusing on these periods. Temporal dependency features refer to the temporal relationships in the data, that is, the features at a certain moment are affected by the data at the preceding and following moments. Through this temporal dependency, LSTM can capture long-term temporal dependencies in the data.

[0085] Example 3, as Figure 2As shown, the present invention proposes a tablet computer usage time warning device, which is applicable to the aforementioned tablet computer usage time warning method, comprising:

[0086] Touch monitoring unit 1 is used to monitor the touch event sequence of the target user on the tablet computer, determine the touch control point sequence based on the time interval between adjacent touch events in the touch event sequence, and perform trajectory fitting on the touch control point sequence based on the Bezier curve to obtain the touch trajectory;

[0087] Operation classification unit 2 is used to determine the operation intensity label of the target user based on the touch event sequence and touch trajectory;

[0088] Key point detection unit 3 is used to monitor the head posture video of the target user, obtain the video frame sequence of the head posture video, and perform key point detection on the video frame sequence according to the pre-trained key point detection model to obtain the key point set of each video frame in the video frame sequence.

[0089] The fatigue prediction unit 4 is used to determine the action pattern feature sequence of the target user through the key point set of each video frame in the video frame sequence according to the preset sliding window, and to perform fatigue prediction through the action pattern feature sequence according to the pre-trained fatigue prediction model to obtain the fatigue intensity label of the target user.

[0090] The warning unit 5 is used to match the operation intensity label and fatigue intensity label with a preset label mapping table to obtain the usage intensity level of the target user, and issue usage warnings based on the usage intensity level.

[0091] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for alerting users about the usage time of a tablet computer, characterized in that, include: The system monitors the target user's touch event sequence on the tablet computer, determines the touch control point sequence based on the time interval between adjacent touch events in the touch event sequence, and performs trajectory fitting on the touch control point sequence based on a Bezier curve to obtain the touch trajectory. The target user's operation intensity label is determined based on the touch event sequence and the touch trajectory; Monitor the head posture video of the target user, obtain the video frame sequence of the head posture video, and perform key point detection on the video frame sequence according to the pre-trained key point detection model to obtain the key point set of each video frame in the video frame sequence. The target user's action pattern feature sequence is determined by the key point set of each video frame in the video frame sequence according to a preset sliding window. Fatigue prediction is then performed by the action pattern feature sequence according to a pre-trained fatigue prediction model to obtain the fatigue intensity label of the target user. The operation intensity label and the fatigue intensity label are matched with a preset label mapping table to obtain the usage intensity level of the target user, and usage warnings are issued based on the usage intensity level.

2. The method for alerting tablet computer usage time according to claim 1, characterized in that, The touch event sequence includes at least one touch event, and the touch control trajectory includes at least two touch events, wherein the touch event includes touch point coordinates and duration.

3. The method for alerting a tablet computer regarding usage time according to claim 2, characterized in that, Determining the touch control point sequence based on the time interval between adjacent touch events in the touch event sequence includes: Determine the time interval between adjacent touch events in the touch event sequence; The average time interval of the touch event sequence is determined based on the time interval between adjacent touch events in the touch event sequence, and the average time interval is used as the dynamic threshold of the interval of the touch event sequence. If the time interval between adjacent touch events in the touch event sequence is less than the dynamic threshold of the interval, then the adjacent touch events are regarded as a pair of adjacent touch points; The Euclidean distance between the adjacent touch point pairs is determined based on the touch point coordinates of the adjacent touch point pairs; If the Euclidean distance between the adjacent touch point pairs is less than a preset distance threshold, the adjacent touch point pairs are used as touch control points. After processing all touch events in the touch event sequence, a touch control point sequence is obtained.

4. The method for alerting tablet computer usage time according to claim 3, characterized in that, The touch control point sequence is fitted with a Bézier curve to obtain the touch trajectory, including: The first touch control point in the touch control point sequence is taken as the starting point of the Bézier curve, and the last touch control point in the touch control point sequence is taken as the ending point of the Bézier curve. Determine the chord length between adjacent touch control points in the touch control point sequence, and determine the cumulative chord length corresponding to the touch control point sequence based on the chord length between adjacent touch control points in the touch control point sequence; The chord lengths between adjacent touch control points in the touch control point sequence are normalized based on the cumulative chord length to obtain the control parameters of the Bézier curve. The value of the Bezier basis function is determined based on the control parameters of the Bezier curve. The target function is obtained by minimizing the value of the Bezier basis function with the coordinates of the corresponding touch control points in the touch control point sequence. The objective function is solved using the least squares method to obtain multiple intermediate control points of the Bézier curve. The touch trajectory is then determined based on the starting point, multiple intermediate control points, and the ending point of the Bézier curve.

5. The method for alerting tablet computer usage time according to claim 4, characterized in that, Determining the target user's operation intensity label based on the touch event sequence and the touch trajectory includes: Remove the touch events corresponding to the touch control point sequence from the touch event sequence to obtain the press touch event sequence; Based on the duration of each touch event in the press touch event sequence, the press touch event sequence is divided into a single-point click touch event sequence and a long-press touch event sequence; The number of single-point clicks in the single-point click event sequence and the number of long-press operations in the long-press event sequence are counted to determine the trajectory length of the touch trajectory. The number of single-point clicks, the number of long-press operations, and the trajectory length are weighted according to preset weighting coefficients to obtain the operation intensity coefficient of the target user; The target user's operation intensity label is determined based on the target user's operation intensity coefficient.

6. The method for alerting a tablet computer regarding usage time according to claim 5, characterized in that, The set of key points includes the key points of the nose, left eye, right eye, left ear, right ear, center of the neck, left shoulder, and right shoulder.

7. The method for alerting a tablet computer regarding usage time according to claim 6, characterized in that, Based on a preset sliding window, the action pattern feature sequence of the target user is determined through the key point set of each video frame in the video frame sequence, including: The preset sliding window is moved in the video frame sequence, and the fatigue feature set of the target user is determined based on the key point set of all video frames within the sliding window; Each fatigue feature in the fatigue feature set is weighted according to a preset fatigue weight to obtain the comprehensive fatigue feature of the target user; The feature dispersion of the comprehensive fatigue feature is determined based on the standard deviation and mean of the comprehensive fatigue feature. The feature dispersion is the motion pattern feature of the sliding window. When the sliding window moves to the end of the video frame sequence, the motion pattern feature sequence corresponding to the target user is obtained.

8. A method for alerting tablet computer usage time according to claim 7, characterized in that, The fatigue feature set includes posture stability features, limb movement discrete features, and joint angle fluctuation features. The posture stability features represent the head sway amplitude of the target user. The posture stability features are calculated by summing the squared differences between the coordinates of the nose key point at each moment and the mean of the nose key point coordinates within the sliding window, and then averaging the sum of the squared differences over all moments. The limb movement discrete features represent the regularity of the target user's arm movements. The limb movement discrete features are calculated by calculating the Euclidean distance between the left and right shoulder key points at each moment, and then calculating the standard deviation of the Euclidean distance over all moments. The joint angle fluctuation features represent the fluctuation of the target user's neck angle. The joint angle fluctuation features are calculated by using an inverse trigonometric function to calculate the angle between the nose key point and the center neck key point at each moment, and then calculating the variance of the angle over all moments.

9. A method for alerting the usage time of a tablet computer according to claim 8, characterized in that, The fatigue prediction model includes a bidirectional LSTM layer, an attention layer, and a classification layer. The bidirectional LSTM layer is used to capture the sequential dependence features of the action pattern feature sequence. The attention layer is used to focus on key fatigue accumulation periods in the action pattern feature sequence to obtain the attention weights corresponding to the sequential dependence features of each period. The sequential dependence features of each period are weighted according to the attention weights to obtain comprehensive temporal dependence features. The classification layer is used to classify the comprehensive temporal dependence features for fatigue based on the fully connected layer to obtain fatigue probability, and to perform label mapping based on the fatigue probability to obtain fatigue intensity labels.

10. A tablet computer usage time warning device, applicable to the tablet computer usage time warning method of claim 9, characterized in that, include: Touch monitoring unit (1), the touch monitoring unit (1) is used to monitor the touch event sequence of the target user on the tablet computer, determine the touch control point sequence according to the time interval between adjacent touch events in the touch event sequence, and perform trajectory fitting on the touch control point sequence according to the Bezier curve to obtain the touch trajectory; Operation classification unit (2), the operation classification unit (2) is used to determine the operation intensity label of the target user according to the touch event sequence and the touch trajectory; The key point detection unit (3) is used to monitor the head posture video of the target user, obtain the video frame sequence of the head posture video, and perform key point detection on the video frame sequence according to the pre-trained key point detection model to obtain the key point set of each video frame in the video frame sequence. The fatigue prediction unit (4) is used to determine the action pattern feature sequence of the target user through the key point set of each video frame in the video frame sequence according to the preset sliding window, and perform fatigue prediction through the action pattern feature sequence according to the pre-trained fatigue prediction model to obtain the fatigue intensity label of the target user. The warning unit (5) is used to match the operation intensity label and the fatigue intensity label with a preset label mapping table to obtain the usage intensity level of the target user, and to issue a usage warning based on the usage intensity level.

Images