Intelligent monitoring and early warning system for physical education classroom exercise load
Patent Information
- Application Number
- CN202610390597.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-27
- Publication Date
- 2026-06-26
AI Technical Summary
Existing technologies for monitoring exercise load in physical education classes suffer from problems such as high equipment costs, cumbersome wearing, easy skin friction damage or cross-infection, significant data susceptibility to environmental radio frequency interference, and hysteresis defects that prevent prediction of future trends in exercise load.
Using non-contact visual feature extraction and temporal load depth modeling technology, high-definition cameras are used to acquire student movement video data, extract the coordinate sequence of key points of the skeleton, calculate multi-dimensional motion feature vectors, use long short-term memory networks to predict future changes in exercise load, and match them with individual physical fitness benchmarks to generate early warnings.
It enables proactive pre-emptive risk prediction of exercise load in physical education classes, improves the scientific nature and accuracy of sports injury prevention, adapts to individual dynamic physical conditions, and reduces equipment costs and discomfort issues.
Smart Images
Figure CN122290203A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of exercise monitoring, and in particular to an intelligent monitoring and early warning system for exercise load in physical education classes. Background Technology
[0002] With the advancement of smart campus construction, digital monitoring of physical education classes has become an important means to improve teaching quality and ensure student safety during exercise. Currently, physical education class exercise load monitoring is mainly conducted through the following methods:
[0003] One approach is based on wearable sensors, such as heart rate belts or smart bracelets. While this approach can obtain direct physiological parameters, it suffers from drawbacks in group teaching scenarios, including high equipment costs, cumbersome wearing, difficulties in charging and maintenance, and the risk of skin friction damage or cross-infection. Furthermore, the data collected by the sensors is often significantly affected by environmental radio frequency interference.
[0004] Second, there are monitoring solutions based on conventional computer vision. Existing technologies primarily focus on real-time statistics of movement frequency and displacement speed. However, these solutions suffer from significant lag: the system can only issue an alert when a student has already exhibited obvious movement deformation or a substantial decrease in speed (i.e., fatigue has occurred). In actual teaching, due to individual differences in physical fitness, the same intensity of movement produces completely different physiological stresses for different students, and fatigue is a non-linear cumulative process. Existing technologies cannot predict the future trend of exercise load before significant changes occur in visual characteristics, preventing teachers from intervening in advance to adjust the teaching pace and effectively preventing sports injuries and sudden physiological risks. Summary of the Invention
[0005] This application aims to at least partially address one of the technical problems in the related art.
[0006] Therefore, one objective of this application is to propose an intelligent monitoring and early warning system for physical education class exercise load. Through non-contact visual feature extraction and time-series load depth modeling technology, the system transforms exercise load monitoring from passive post-event feedback to proactive pre-event risk prediction, thereby achieving accurate early warning of physical education class load management in the dimension of individual dynamic adaptability.
[0007] To achieve the above objectives, the first aspect of this application proposes an intelligent monitoring and early warning system for physical education class exercise load, comprising:
[0008] The video acquisition module is used to acquire raw motion video data of the students to be monitored during multiple class periods, continuously captured by high-definition cameras in the sports field.
[0009] The feature extraction module is used to extract the skeleton key point coordinate sequence of the student to be monitored at each time period based on the original motion video data; and to calculate a multi-dimensional motion feature vector based on the displacement change of the skeleton key point coordinate sequence in three-dimensional space. The multi-dimensional motion feature vector includes at least the action speed, the joint angle change amplitude and the action interval time.
[0010] The temporal reconstruction module is used to construct a temporal motion feature matrix by arranging the multidimensional motion feature vectors in chronological order.
[0011] The trend prediction module is used to calculate the real-time exercise load value at the current moment based on the multi-dimensional motion feature matrix, input the time-series motion feature matrix into the pre-constructed load trend prediction model, and output the predicted exercise load change range of the student to be monitored in a preset short period of time in the future.
[0012] The risk assessment module is used to match the predicted exercise load change range with a safe load threshold generated based on the student's individual physical fitness benchmark, and to determine whether the predicted exercise load change range has a risk trend of exceeding the safe load threshold.
[0013] The early warning execution module is used to generate trend overload early warning information for the students to be monitored when a risk trend is determined, and to transmit the trend overload early warning information to the teacher management terminal in real time.
[0014] In addition, the intelligent monitoring and early warning system for physical education class exercise load proposed in this application may also have the following additional technical features:
[0015] In one embodiment of this application, the feature extraction module includes:
[0016] The velocity calculation unit is used to calculate the instantaneous velocity of the action by the displacement of the skeleton key points between consecutive video frames, and to calculate the average value within a preset period;
[0017] The joint angle calculation unit is used to calculate the angle between adjacent bone vectors to obtain the maximum joint angle change range for each action cycle;
[0018] Intermittent recognition unit is used to identify the duration of the static state of key points in the skeleton as the action interval time;
[0019] The feature standardization unit is used to normalize the velocity, joint angle change amplitude and action interval time to form multidimensional motion features for time sequence reconstruction.
[0020] In one embodiment of this application, the training steps of the load trend prediction model are as follows:
[0021] Obtain the multidimensional motion feature vector sequence of sample students during historical motion processes and the corresponding measured motion load data;
[0022] The multidimensional motion feature vector sequence of the sample is used as input, and the measured motion load data is used as a label to iteratively train the load trend prediction model.
[0023] During training, a nonlinear mapping relationship between the motion feature decay slope and the motion load increase rate is established by adjusting the weight coefficients of the forget gate and the input gate.
[0024] The training of the load trend prediction model is completed when the residual between the predicted output of the load trend prediction model and the labeled label is less than a preset value.
[0025] In one embodiment of this application, the risk determination module includes:
[0026] The slope analysis unit is used to calculate the slope of the predicted range of motion load changes over time.
[0027] The threshold comparison unit is used to determine whether the upper limit of the prediction interval reaches the safe load threshold, thereby judging the risk of trend overload.
[0028] In one embodiment of this application, the feature extraction module further includes:
[0029] The motion speed decay rate calculation unit is used to calculate the decreasing trend of speed between continuous motion cycles;
[0030] The motion instability analysis unit is used to calculate the root mean square error of the motion trajectory deviating from the standard template.
[0031] The feature fusion unit is used to incorporate the motion speed decay rate and motion instability into the temporal motion feature vector to form an enhanced temporal motion feature vector.
[0032] In one embodiment of this application, the step of obtaining the safety load threshold is as follows:
[0033] The real-time movement frequency of the student to be monitored in the current movement state is acquired in real time;
[0034] The real-time movement frequency is compared with the tolerance benchmark in the historical archive to calculate the current physical exertion deviation. Based on the physical exertion deviation, the safe load threshold is increased or decreased in real time so that the safe load threshold matches the real-time physiological state of the student to be monitored.
[0035] In one embodiment of this application, the student to be monitored has a unique identification ID.
[0036] In one embodiment of this application, the teacher management terminal includes:
[0037] The visualization unit is used to simultaneously display the real-time exercise load value and the predicted exercise load change range in the form of a dynamic curve on the display interface.
[0038] The risk identification unit is used to distinguish different risk levels in the range of exercise load changes by color identification, and to trigger the visualization display unit to provide a prompt when the prediction curve touches the safe load threshold.
[0039] The intelligent monitoring and early warning system for physical education class exercise load according to this application can transform exercise load monitoring from passive post-event feedback to proactive pre-event risk prediction. By capturing the nonlinear mapping relationship between movement characteristics and energy metabolism through long short-term memory networks, it can achieve the prospective identification of students' potential fatigue state. Combined with individual historical records, it dynamically corrects the safe load threshold to ensure the consistency between early warning judgment and students' real-time functional status, thereby improving the scientificity and accuracy of sports injury prevention in large-class teaching environments in physical education classes.
[0040] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description
[0041] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:
[0042] Figure 1 This is a block diagram of the intelligent monitoring and early warning system for physical education class exercise load in this application;
[0043] Figure 2 This is a training block diagram for the load trend prediction model of this application; Detailed Implementation
[0044] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.
[0045] The following description, in conjunction with the accompanying drawings, describes an embodiment of the intelligent monitoring and early warning system for physical education class exercise load according to this application.
[0046] like Figure 1 As shown in the figure, the intelligent monitoring and early warning system for physical education class exercise load in this application embodiment may include:
[0047] The video acquisition module is used to acquire raw motion video data of the students to be monitored during multiple class periods, continuously captured by high-definition cameras in the sports field. It should be noted that each student to be monitored has a unique identification ID, which can ensure that subsequent data can be correlated. At this time, the system uses multi-target tracking to assign a unique identification ID to each student in the picture and outputs individual sub-video streams with identification IDs.
[0048] The feature extraction module is used to extract the skeleton key point coordinate sequence of the student to be monitored at each time period based on the original motion video data. ,in Representing the Each joint The three-dimensional spatial coordinates at that moment;
[0049] Based on the displacement changes of the skeleton key point coordinate sequence in three-dimensional space, a multidimensional motion feature vector is calculated. This multidimensional motion feature vector includes at least movement speed, joint angle variation amplitude, and movement interval time. Movement speed reflects the intensity of the movement; higher speed means greater instantaneous energy consumption. Joint angle variation amplitude reflects the range of work done during the movement. Movement interval time reflects fatigue recovery; for the same movement intensity, a shorter interval time indicates that the heart and muscles are under high pressure, while a longer interval time indicates a certain degree of physical recovery. This feature extraction module provides standardized temporal feature components for subsequent temporal reconstruction.
[0050] In one embodiment of this application, the feature extraction module includes:
[0051] The velocity calculation unit calculates the instantaneous velocity of a movement by measuring the displacement of skeleton keypoints across consecutive video frames and then calculates the average value over a preset period. The unit converts the coordinates of discrete points into continuous physical vectors by calculating the Euclidean distance between adjacent frames. Calculating the average value reflects the intensity of a student's continuous movement within a given area. Furthermore, considering potential pixel jitter or keypoint jumps due to occlusion during camera capture, using instantaneous velocity for warnings might result in false alarms due to positional shifts. Calculating the average value over a preset period acts as a low-pass filter, eliminating high-frequency noise and ensuring stable and statistically significant data output to subsequent steps.
[0052] ;
[0053] in, represent The instantaneous velocity of the target at any given moment; represent Constantly refer to the three-dimensional spatial position vector of key points; This parameter typically refers to the center of mass or core area of the human body (such as the center point of the hip). In three-dimensional space, this parameter is essentially a coordinate axis. ; Represents the previous sampling time (i.e. (Time) Spatial position vector of reference key point; This represents the time sampling interval, which depends on the frame rate of the high-definition camera (e.g., when the frame rate is 30fps). Second); This represents the straight-line distance between two points.
[0054] Next, set a preset period. (For example (Or 150 video frames), within this window, a weighted average of the instantaneous velocity is calculated to obtain the average velocity:
[0055] ,in For period The total number of sampling frames within.
[0056] The joint angle calculation unit is used to calculate the angle between adjacent bone vectors to obtain the maximum joint angle change range for each action cycle;
[0057] Specifically, the system starts from the coordinate sequence Three key points with anatomical relevance were selected.
[0058] Taking the knee joint as an example, select the hip point. Knee joint points and ankle joint In a three-dimensional coordinate system, construct two points intersecting at a key center point. Skeletal vectors:
[0059] Upper limb / thigh vector: ;
[0060] Lower limb / calf vector: ;
[0061] Using the dot product formula, calculate the real-time cosine value between two vectors to obtain the included angle. ;
[0062] ;
[0063] The system can determine an action cycle by detecting the peaks and troughs of the velocity wave and extract the data within that cycle. maximum value and minimum value The difference is calculated as the maximum joint angle variation. It can quantify whether a movement is performed correctly; for the same movement, a smaller change in angle means less work done by the muscles.
[0064] The intermittent recognition unit is used to identify the duration of the static state of key skeletal points as the action interval time; the system calculates the cumulative displacement D of the student's center of mass (or other core key points) over a certain time span, and simultaneously sets a physiological information threshold. (Used to filter the swaying caused by human respiration or to maintain balance); if If so, it is determined that the period is in a static state; if If the displacement is detected as zero, then the person is determined to be in motion during that period. Furthermore, when the person is determined to be stationary, the timer begins to accumulate, and when the displacement returns to its original position... The timer stops, and the accumulated duration is recorded as the action interval time. The degree of recovery of the phosphagen and lactate systems is determined by the interval time.
[0065] The feature standardization unit is used to normalize the velocity, joint angle change amplitude and action interval time to form multidimensional motion features for time sequence reconstruction.
[0066] Specifically, a linear transformation is performed on each feature dimension using Min-Max normalization or Z-Score normalization, transforming the measured... and Mapped to Within the interval, the influence of dimensions is eliminated, and a standardized multidimensional motion feature vector is formed. .
[0067] In one embodiment of this application, the feature extraction module further includes:
[0068] The motion speed decay rate calculation unit is used to calculate the decreasing trend of speed between continuous motion cycles;
[0069] Specifically, continuous extraction Average speed value of each action cycle The rate of change of speed over time can be calculated using the least squares method or the first-order difference method. :
[0070] Among them, if A negative value with a continuously increasing absolute value indicates that the student has experienced a significant drop in speed.
[0071] The motion instability analysis unit is used to calculate the root mean square error of the motion trajectory deviating from the standard template.
[0072] ;in, To preset the key point trajectory sequence of standard movements, This is a sequence of measured coordinates. The smaller the value, the more stable the student's movements. An abnormally high level indicates a disruption in the movement trajectory.
[0073] The feature fusion unit is used to incorporate the motion speed decay rate and motion instability into the temporal motion feature vector to help predict the trend of motion load change.
[0074] Specifically, the basic feature vector With enhanced features Dimensional concatenation is performed to form an enhanced temporal motion feature vector: .
[0075] The temporal reconstruction module is used to construct a temporal motion feature matrix by arranging the multidimensional motion feature vectors in chronological order.
[0076] The trend prediction module is used to calculate the real-time exercise load value at the current moment based on the multi-dimensional motion feature vector, input the time-series motion feature vector into the pre-constructed load trend prediction model, and output the predicted exercise load change range of the student to be monitored in a preset short period of time in the future.
[0077] In one embodiment of this application, the load trend prediction model is a training model based on a Long Short-Term Memory (LSTM) network.
[0078] The model consists of an input layer, hidden layers, fully connected layers, and an output layer. The input layer receives the temporal feature matrix generated by the temporal reconstruction module. Its dimension and temporal motion feature vector Feature dimensions and time window size Matching. The hidden layer uses... The structure updates the cell state through the synergistic action of forget gates, input gates, and output gates, extracting the nonlinear temporal evolution patterns in motion features. Fully connected layers hide the terminal states of the hidden layer outputs. Mapped to a continuous numerical space.
[0079] The output layer synchronously outputs the real-time motion load value at the current moment. And the expected value of predicted exercise load within the future preset time period. With variance parameter Among them, by and Together they constitute the predicted range of motion load changes. (like Its activation function uses a linear function or The function ensures the non-negativity of the load value.
[0080] in, ; These are the normalized average velocity, maximum joint angle variation, and movement interval, respectively. ;
[0081] ,in, Represents the current moment The system rewinds a preset duration (or preset number of frames) in the past. The dynamic sliding window formed by ) This represents the size of the time window.
[0082] like Figure 2 As shown, the training steps for this model are as follows:
[0083] Obtain the multidimensional motion feature sequence of sample students during historical motion processes and the corresponding measured motion load data;
[0084] The multidimensional motion feature sequence of the sample is used as input, and the measured motion load data is used as a label to iteratively train the load trend prediction model.
[0085] First, obtain the multidimensional motion feature vector sequence of the sample students during the historical motion process. Specifically, by collecting historical sports videos of varying intensities during classroom sessions, normalized feature components were extracted and then analyzed by time window. Stacking forms a sample temporal motion feature matrix .
[0086] Simultaneously, the measured exercise load data obtained from the physiological monitoring devices worn by the sample students were collected. In this embodiment, the measured data uses metabolic equivalences (METs) as the label, and its calculation formula is as follows:
[0087] ,in, This represents the oxygen consumption per unit body weight during exercise (ml / kg·min), with a constant of 3.5 representing the baseline oxygen consumption at rest. The system will use the time-series feature matrix... Equivalent to metabolism Accurate alignment is performed on the timeline to construct training sample pairs.
[0088] Will As input, the forget gate weights within the Long Short-Term Memory (LSTM) network are adjusted. With input gate weights Establish a mapping relationship between motor degeneration and physiological load evolution:
[0089] During training, a nonlinear mapping relationship between the motion feature decay slope and the motion load increase rate is established by adjusting the weight coefficients of the forget gate and the input gate.
[0090] Forget Gate adjustment logic: Establish action interval time The connection with physiological recovery. During training, if... Large values and accompanied by measured load The loss function drives the model to reduce the weight coefficients of the terms corresponding to the forget gate, resulting in a significant decrease in the output of the forget gate. The value approaches 0, thus physically simulating the erasure of historical load memory by physical recovery during rest periods.
[0091] Input gate adjustment logic: Establish motion speed decay rate and motion instability. Related to functional overload.
[0092] When the model recognizes Negative (deceleration) and When the value remains high or increases, the training procedure increases the negative weight gain of the corresponding input term in the input gate to improve the model's ability to capture the fatigue characteristics of deceleration and energy consumption.
[0093] when Enlargement accompanied by When the number of inputs increases dramatically, the training program increases the input gate ratio. The sensitivity weights of the terms enable the model to identify the nonlinear loss of physiological energy caused by motion deformation.
[0094] During the iteration process, the predicted motion load value and the labeled value output by the model are calculated using a loss function (such as mean squared error, MSE). The residual between the values is considered. When the residual is less than a preset threshold, the training of the load trend prediction model is complete.
[0095] The risk assessment module is used to determine the predicted range of motion load changes. With safe load thresholds generated based on individual student physical fitness benchmarks The matching process is performed to determine whether the predicted range of motion load changes has a risk trend that exceeds the safe load threshold.
[0096] In one embodiment of this application, the step of obtaining the safety load threshold is as follows:
[0097] Real-time acquisition of the real-time movement frequency of the student to be monitored in the current movement state. ;
[0098] The real-time action frequency Withstand benchmarks in the aforementioned historical archives Compare and calculate the current physical exertion deviation. The safe load threshold is adjusted in real time based on the physical exertion deviation to either increase or decrease the threshold, thereby ensuring the safe load threshold remains unchanged. It matches the real-time physiological state of the student to be monitored.
[0099] It should be noted that the aforementioned tolerance benchmark This serves as the baseline frequency for the student to perform standard movements when in a good functional state.
[0100] like This indicates that the student is performing exceptionally well or is under extremely high intensity.
[0101] like Combined with the previous If the movements become distorted and the frequency decreases, it is considered that the person is physically exhausted.
[0102] ;in, This is the sensitivity coefficient. The preset basic threshold for the archive.
[0103] If the student's movement frequency steadily increases and Extremely low, indicating excellent functional condition today, system adjustment This allows for higher-intensity training without false alarms; if a student's movement frequency drops abnormally, it indicates that their physiological function is at a low point that day, and the system will proactively adjust accordingly. .
[0104] In one embodiment of this application, the risk determination module includes:
[0105] The slope analysis unit is used to calculate the slope of the predicted exercise load variation range over time; if the load value is not high but the slope is extremely steep, it indicates that the student is engaging in explosive exercise beyond their metabolic compensation capacity. Even with the current... It is far from the safe load threshold, but if the slope Even if the risk is extremely high, the system will still determine that the risk is about to occur.
[0106] The threshold comparison unit is used to determine the upper limit of the prediction interval. Has the safe load threshold been reached? This allows us to assess the risk of trend overload.
[0107] The early warning execution module is used to generate trend overload early warning information for the students to be monitored when a risk trend is determined, and to transmit the trend overload early warning information to the teacher management terminal in real time.
[0108] The teacher management terminal includes:
[0109] The visualization unit is used to simultaneously display the real-time exercise load value and the predicted exercise load change range in the form of a dynamic curve on the display interface.
[0110] The risk identification unit is used to distinguish different risk levels in the range of exercise load changes by color identification, and to trigger the visualization display unit to provide a prompt when the prediction curve touches the safe load threshold.
[0111] Specifically, the visualization unit receives scalars output from the backend in real time. With interval sequence In terms of drawing logic, Rendered as a real-time trajectory line, of Region rendering is based on predictive rendering bands. When the edge of the predictive rendering band (i.e., Touching or exceeding the dynamic threshold line At that time, the system performs pixel-level color overflow filling to achieve a strong visual warning of the risk.
[0112] In summary, the intelligent monitoring and early warning system for physical education class exercise load in this application can transform exercise load monitoring from passive post-event feedback to proactive pre-event risk prediction. By capturing the nonlinear mapping relationship between motion characteristics and energy metabolism through long short-term memory networks, it achieves the prospective identification of students' potential fatigue state. Furthermore, by dynamically correcting the safe load threshold in conjunction with individual historical records, it ensures the consistency between early warning judgment and students' real-time functional status, thereby improving the scientific nature and accuracy of sports injury prevention in large-class teaching environments in physical education classes.
[0113] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0114] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.
[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A smart monitoring and early warning system for exercise load in physical education classes, characterized in that, include: The video acquisition module is used to acquire raw motion video data of the students to be monitored during multiple class periods, continuously captured by high-definition cameras in the sports field. The feature extraction module is used to extract the skeleton key point coordinate sequence of the student to be monitored at each time period based on the original motion video data; and to calculate a multi-dimensional motion feature vector based on the displacement change of the skeleton key point coordinate sequence in three-dimensional space. The multi-dimensional motion feature vector includes at least the action speed, the joint angle change amplitude and the action interval time. The temporal reconstruction module is used to construct a temporal motion feature matrix by arranging the multidimensional motion feature vectors in chronological order. The trend prediction module is used to calculate the real-time exercise load value at the current moment based on the multi-dimensional motion feature vector, input the time-series motion feature matrix into the pre-constructed load trend prediction model, and output the predicted exercise load change range of the student to be monitored in a preset short period of time in the future. The risk assessment module is used to match the predicted exercise load change range with a safe load threshold generated based on the student's individual physical fitness benchmark, and to determine whether the predicted exercise load change range has a risk trend of exceeding the safe load threshold. The early warning execution module is used to generate trend overload early warning information for the students to be monitored when a risk trend is determined, and to transmit the trend overload early warning information to the teacher management terminal in real time.
2. The intelligent monitoring and early warning system for physical education class exercise load according to claim 1, characterized in that, The feature extraction module includes: The velocity calculation unit is used to calculate the instantaneous velocity of the action by the displacement of the skeleton key points between consecutive video frames, and to calculate the average value within a preset period; The joint angle calculation unit is used to calculate the angle between adjacent bone vectors to obtain the maximum joint angle change range for each action cycle; Intermittent recognition unit is used to identify the duration of the static state of key points in the skeleton as the action interval time; The feature standardization unit is used to normalize the velocity, joint angle change amplitude and action interval time to form multidimensional motion features for time sequence reconstruction.
3. The intelligent monitoring and early warning system for physical education class exercise load according to claim 1, characterized in that, The training steps for the load trend prediction model are as follows: Obtain the multidimensional motion feature vector sequence of sample students during historical motion processes and the corresponding measured motion load data; The multidimensional motion feature vector sequence of the sample is used as input, and the measured motion load data is used as a label to iteratively train the load trend prediction model. During training, a nonlinear mapping relationship between the motion feature decay slope and the motion load increase rate is established by adjusting the weight coefficients of the forget gate and the input gate. The training of the load trend prediction model is completed when the residual between the predicted output of the load trend prediction model and the labeled label is less than a preset value.
4. The intelligent monitoring and early warning system for physical education class exercise load according to claim 1, characterized in that, The risk assessment module includes: The slope analysis unit is used to calculate the slope of the predicted range of motion load changes over time. The threshold comparison unit is used to determine whether the upper limit of the prediction interval reaches the safe load threshold, thereby judging the risk of trend overload.
5. The intelligent monitoring and early warning system for physical education class exercise load according to claim 1, characterized in that, The feature extraction module further includes: The motion speed decay rate calculation unit is used to calculate the decreasing trend of speed between continuous motion cycles; The motion instability analysis unit is used to calculate the root mean square error of the motion trajectory deviating from the standard template. The feature fusion unit is used to incorporate the motion speed decay rate and motion instability into the temporal motion feature vector to form an enhanced temporal motion feature vector.
6. The intelligent monitoring and early warning system for physical education class exercise load according to claim 1, characterized in that, The steps for obtaining the safe load threshold are as follows: The real-time movement frequency of the student to be monitored in the current movement state is acquired in real time; The real-time movement frequency is compared with the tolerance benchmark in the historical archive to calculate the current physical exertion deviation. Based on the physical exertion deviation, the safe load threshold is increased or decreased in real time to match the real-time physiological state of the student to be monitored.
7. The intelligent monitoring and early warning system for physical education class exercise load according to claim 1, characterized in that, Each student to be monitored has a unique identification ID.
8. The intelligent monitoring and early warning system for physical education class exercise load according to claim 1, characterized in that, The teacher management terminal includes: The visualization unit is used to simultaneously display the real-time exercise load value and the predicted exercise load change range in the form of a dynamic curve on the display interface. The risk identification unit is used to distinguish different risk levels in the range of exercise load changes by color identification, and to trigger the visualization display unit to provide a prompt when the prediction curve touches the safe load threshold.