Online teaching interaction system based on cloud platform

Abstract

The invention relates to the technical field of intelligent interaction, in particular to an online teaching interaction system based on a cloud platform, and the system comprises a data acquisition module which determines the motion disturbance route complexity and motion repetition of a teaching end; the data analysis module is used for calculating action complex characterization parameters and dividing action variation tendencies; the interaction processing module is used for determining the action state of the learning end by comparing the teaching end video and the joint points of the action starting frame and the action ending frame of the learning end video according to the weak variation tendency, determining a plurality of extraction intervals according to the action with the strong variation tendency, predicting the extraction intervals to correspond to predicted joint point action graphs, and determining the action state of the learning end according to the predicted joint point action graphs. Comparing with an actual joint point action diagram in the learning end video frame in the extraction interval, and determining the action state of the learning end; the state reminding module is used for judging whether state correction reminding needs to be sent out or not. By comparing and analyzing the learning actions, the interaction efficiency and the analysis accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent interaction technology, and in particular to an online teaching interaction system based on a cloud platform. Background Art

[0002] Online interactive teaching systems leverage network, multimedia, and software technologies to break the spatial and temporal constraints of traditional teaching, enabling students and teachers to engage in remote teaching and interaction over the internet. In particular, cloud computing architecture, with its powerful computing and storage capabilities, ensures stable system operation, supports massive data storage, and flexible resource allocation. Artificial intelligence algorithms mine learning data for intelligently assisted teaching and question-answering, providing comprehensive and powerful technical support for cloud-based online interactive teaching systems. Furthermore, object detection and image classification technologies can identify specific objects in images, providing support for teaching decisions. The development of online interactive teaching systems has not only improved teaching efficiency and quality, but also provided students with a more personalized and flexible learning experience.

[0003] Chinese Patent Publication No. CN114863433A discloses an online assisted teaching system with real-time handwriting interaction, comprising: an extraction module for extracting the hand skeleton from the live broadcast image captured by the camera using a deep learning method to obtain the coordinates of the key points of the hand; a gesture recognition module for performing gesture recognition using the coordinates of the key points of the hand and switching modes based on the gesture recognition results, wherein the modes include: character input, basic graphic drawing, and sketch drawing; a character recognition module for recognizing the handwritten characters input by the user and obtaining the corresponding printed characters; a graphic drawing module for recognizing the basic graphics drawn by the user and obtaining the basic graphics after smoothing the trajectory; and an image generation module for generating and visualizing images based on the sketch drawn by the user. This invention can solve the problem of missing blackboard writing in online classes.

[0004] Chinese patent publication number: CN118656485A, discloses an online paperless language teaching interactive method, medium and equipment. By constructing a basic database for each student, the current stage learning database is obtained based on the basic database, teaching students in accordance with their aptitude is achieved, and easy-to-make mistakes vocabulary groups and easy-to-make mistakes grammar groups are constructed based on wrong questions to achieve multi-dimensional divergence of right and wrong questions. The first question group is extracted from the easy-to-make mistakes vocabulary group, and the second question group is extracted from the easy-to-make mistakes grammar group, and inserted into the question combination of the next round. While strengthening the training of wrong questions in the next round, not only the wrong questions are focused on, but the question combination of the next round can also include questions of other difficulty levels, taking into account the difficulty distribution of the question combination of each round on the student side, so that the generation of the question combination of each round is dynamic and diversified, which helps to improve learning efficiency.

[0005] However, the prior art still has the following problems:

[0006] In the existing technology, for online courses such as dance teaching and fitness teaching, most of them are in the form of one teaching end corresponding to several learning ends. When monitoring each learning end, a traversal analysis of video data is often used. When there are many learning ends, more computing power is consumed. Summary of the Invention

[0007] To this end, the present invention provides an online teaching interaction system based on a cloud platform to solve the problem in the existing technology that for online courses such as dance teaching and fitness teaching, most of them are in the form of one teaching end corresponding to several learning ends, and when monitoring each learning end, a traversal analysis of video data is often adopted. When there are many learning ends, more computing power is consumed.

[0008] To achieve the above objectives, the present invention provides an online teaching interactive system based on a cloud platform, comprising:

[0009] A data acquisition module acquires authorized teaching end videos and several learning end videos to determine the complexity of the action disturbance route and the action repetition based on the teaching end videos;

[0010] a data analysis module connected to the data acquisition module, for calculating a movement complexity representation parameter based on the movement disturbance route complexity and the movement repetition degree, and classifying movement abnormality tendencies;

[0011] The interactive processing module is connected to the data acquisition module and the data analysis module respectively, and analyzes the learning end video in response to the classification result of the action abnormality tendency, wherein:

[0012] For weak mutation tendency, compare the joint points of the corresponding action start frame and action end frame of the teaching end video and the learning end video to determine the action state of the learning end;

[0013] For strong mutation tendency, several reference intervals are determined based on the corresponding action start frame of the teaching end video, and the delay time of the reference interval relative to the time of the action start frame is determined. Several extraction intervals are determined based on the action start frame of the teaching end video and the corresponding delay time. The predicted joint point action graph corresponding to the extraction interval is compared with the actual joint point action graph in the learning end video frame within the extraction interval to determine the action state of the learning end;

[0014] The status reminder module is connected to the interaction processing module and is used to determine whether a status correction reminder needs to be issued based on the action status of the learning terminal.

[0015] Furthermore, the data acquisition module is used to determine the complexity of the action disturbance route and the action repetition, including:

[0016] Used to record the number of direction changes of each joint point, the number of action repetitions, and the duration of the teaching video;

[0017] The ratio of the number of direction changes of each joint point to the duration is used as the change frequency;

[0018] The average value of each ratio corresponding to each joint point is used to determine the complexity of the action perturbation route;

[0019] The ratio of the number of repetitions of the action and the duration of the action is determined as the action repetition degree;

[0020] The joints include the head joints and the limb joints.

[0021] Furthermore, the data analysis module is used to calculate the complex characterization parameters of the action, including:

[0022] The ratio of the complexity of the action perturbation route to the complexity of the reference action perturbation route is used to determine the first influencing factor;

[0023] The ratio of the baseline action repetition rate to the action repetition rate is used to determine the second influencing factor;

[0024] The weighted sum of the first influencing factor and the second influencing factor is used to determine the action complexity representation parameter.

[0025] Furthermore, the data analysis module is used to classify the tendency of abnormal movement, wherein:

[0026] If the action complexity representation parameter is greater than the action complexity representation parameter threshold, the action abnormality tendency is classified as a strong abnormality tendency;

[0027] If the action complexity representation parameter is less than or equal to the action complexity representation parameter threshold, the action abnormality tendency is classified as a weak abnormality tendency.

[0028] Furthermore, the interactive processing module is used to compare the joint points of the teaching end video and the learning end video corresponding to the action start frame and the action end frame, including:

[0029] for extracting joint point graphs in the action start frame and the action end frame in the teaching end video and the learning end video;

[0030] Place the joint point graphs corresponding to the starting frames of each action in the same coordinate system, scale the joint point graphs corresponding to the learning end video until they meet the stacking error standard, and record the current scaling ratio;

[0031] The joint point graphs corresponding to the end frames of each action are placed in the same coordinate system, and the joint point graphs corresponding to the learning end video are scaled based on the scaling ratio;

[0032] Used to compare the corresponding joint point graphs of each action end frame to determine the average distance difference of each joint point

[0033] Furthermore, the interactive processing module is used to compare the joint points of the corresponding action start frame and action end frame of the teaching end video and the learning end video to determine the action state of the learning end, wherein,

[0034] If the average distance difference is greater than a preset average distance difference standard threshold, determining that the learning terminal action state is a non-standard action state;

[0035] If the average distance difference is less than or equal to a preset average distance difference standard threshold, the learning terminal action state is determined to be a standard action state.

[0036] Furthermore, the interactive processing module is used to determine a number of reference intervals based on the starting frame of the corresponding action in the teaching end video, including:

[0037] To determine the moment when the corresponding action start frame of the teaching video is located;

[0038] It is used to postpone the current time, determine several reference times, determine several reference intervals with the reference times as the starting point, and record the delay time corresponding to each reference interval;

[0039] The starting time of the reference interval is the reference time, and the interval length is the predetermined duration.

[0040] Furthermore, the interactive processing module is used to determine a number of extraction intervals based on the action start frame of the teaching end video and the corresponding delay time, and predict the extraction intervals corresponding to the predicted joint point action graph, including:

[0041] Determine the time of the action start frame of the teaching end video, delay the time by a corresponding delay time, determine a number of extraction times, and determine a number of extraction intervals with the extraction time as the starting point;

[0042] Determining a correspondence between each of the extraction intervals and the reference interval based on the delay time;

[0043] Determine a reference interval corresponding to the extraction interval, extract any teaching end video frame within the reference interval, identify the joint point action graph, and obtain a predicted joint point action graph corresponding to the extraction interval;

[0044] The starting time of the extraction interval is the extraction time, and the interval length is the predetermined duration.

[0045] Furthermore, the interactive processing module is used to compare the predicted joint action graph with the actual joint action graph in the learning end video frame within the extraction interval, including:

[0046] Select any learning end video frame in each extraction interval, determine the actual joint point action graph, compare it with the corresponding predicted joint point action graph, and determine the overlap degree of the joint points corresponding to each extraction interval;

[0047] Determine whether each extraction interval meets the verification criteria;

[0048] Among them, the verification standard is that the joint point overlap corresponding to the extraction interval is greater than or equal to the preset joint point overlap threshold.

[0049] Furthermore, the interaction processing module is used to determine the action state of the learning terminal, including:

[0050] Determine the proportion of extraction intervals that meet the validation criteria;

[0051] If the ratio is greater than or equal to a predetermined ratio threshold, determining that the learning terminal action state is a standard action state;

[0052] If the ratio is less than a predetermined ratio threshold, it is determined that the action state of the learning terminal is a non-standard action state.

[0053] Compared with the prior art, the present invention sets a data acquisition module to determine the complexity of the teaching end action disturbance route and the action repetition; a data analysis module to calculate the action complexity representation parameters and classify the action mutation tendency; an interactive processing module to compare the joint points of the teaching end video and the action start frame and the action end frame of the learning end video for weak mutation tendency, determine the learning end action state, and for actions with strong mutation tendency, determine several extraction intervals, predict the predicted joint point action graph corresponding to the extraction interval, and compare it with the actual joint point action graph in the learning end video frame within the extraction interval to determine the learning end action state; a state reminder module to determine whether it is necessary to issue a state correction reminder. By performing comparative analysis on the learning action, the present invention ensures the accuracy of action analysis while saving computing power.

[0054] In particular, by calculating the complex representation parameters of the movements on the teaching end, dividing the tendency of movement mutation, and analyzing the learning end video in a targeted manner according to different mutation tendencies, in actual situations, the learning end mostly has multiple ports. If the tendency of movement mutation is not distinguished and the movements are analyzed indiscriminately, simple movements will take up too much computing power, and complex movements may not be able to get enough resource support, thereby reducing the efficiency of movement analysis. For example, the upward stretching movement can be analyzed by a simple method, but if the same analysis method as the complex movement is used, it will cause calculation overflow and waste computing power. Based on this, the present invention pre-calculates the complex representation parameters of the movements on the teaching end, and determines different analysis methods for strong mutation tendencies and weak mutation tendencies, respectively, to improve the efficiency and accuracy of movement analysis.

[0055] In particular, for actions with a weak tendency to mutate, the action state of the learning end is determined by comparing the joint points of the action start frame and the action end frame of the teaching end and the learning end. In actual situations, simple actions have a weak tendency to mutate, especially for straight up and down actions. It is only necessary to determine the starting point and the end position to determine the accuracy of the student end action through the action data of the teaching end, and then determine whether the system needs to issue a status correction reminder to the learning end to improve the efficiency and accuracy of action analysis.

[0056] In particular, for those with strong mutation tendency, the predicted joint action graph corresponding to the predicted extraction interval is compared with the actual joint action graph in the video frame of the learning end within the extraction interval to determine the action state of the learning end. In actual situations, complex actions have strong mutation tendency. When learning, the probability of non-standard action will increase. At this time, if the same analysis method as the weak mutation tendency is adopted, it may cause analysis errors, which will affect the judgment result of the action state of the learning end. Based on this, the present invention determines several extraction intervals in combination with the action video segment of the teaching end, predicts the extraction interval corresponding to the predicted joint action graph, and compares it with the actual joint action graph in the video frame of the learning end within the extraction interval to improve the efficiency and accuracy of action analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A schematic diagram of the structure of an online teaching interactive system based on a cloud platform according to an embodiment of the invention;

[0058] Figure 2 A logic block diagram for dividing the tendency of action abnormality according to an embodiment of the invention;

[0059] Figure 3 A logic diagram for comparing the joint points of the corresponding action start frame and action end frame of the teaching end video and the learning end video in an embodiment of the invention to determine the action state of the learning end;

[0060] Figure 4A logic block diagram for determining the action status of a learning terminal according to an embodiment of the invention;

[0061] Figure 5 This is a logic block diagram of an embodiment of the invention for determining whether a status correction reminder needs to be issued based on the action status of the learning terminal. DETAILED DESCRIPTION

[0062] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0063] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0064] It should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the term "connection" should be understood in a broad sense. For example, it can mean a fixed connection, a detachable connection, or an integral connection; it can mean a mechanical connection or an electrical connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0065] See also Figure 1 , Figure 1 The following is a schematic diagram of the structure of an online teaching interaction system based on a cloud platform according to an embodiment of the invention. The online teaching interaction system based on a cloud platform according to an embodiment of the invention includes:

[0066] A data acquisition module acquires authorized teaching end videos and several learning end videos to determine the complexity of the action disturbance route and the action repetition based on the teaching end videos;

[0067] a data analysis module connected to the data acquisition module, for calculating a movement complexity representation parameter based on the movement disturbance route complexity and the movement repetition degree, and classifying movement abnormality tendencies;

[0068] The interactive processing module is connected to the data acquisition module and the data analysis module, and analyzes the learning end video in response to the classification result of the action abnormality tendency, wherein:

[0069] For weak mutation tendency, compare the joint points of the corresponding action start frame and action end frame of the teaching end video and the learning end video to determine the action state of the learning end;

[0070] For strong mutation tendency, several reference intervals are determined based on the corresponding action start frame of the teaching end video, and the delay time of the reference interval relative to the time of the action start frame is determined. Several extraction intervals are determined based on the action start frame of the teaching end video and the corresponding delay time. The predicted joint point action graph corresponding to the extraction interval is compared with the actual joint point action graph in the learning end video frame within the extraction interval to determine the action state of the learning end;

[0071] The status reminder module is connected to the interaction processing module and is used to determine whether a status correction reminder needs to be issued based on the action status of the learning terminal.

[0072] Specifically, the teaching end is the terminal used by the teacher, and the learning end is the terminal used by the learner. There is no limitation on the method of obtaining the teaching end video and the learning end video. For example, it can be to obtain the video data authorized to be uploaded by the teaching end or the learning end. This will not be repeated here.

[0073] It can be understood that the complexity of the motion perturbation route represents the complexity of the joint motion route. The higher the complexity of the motion perturbation route, the greater the complexity of the action. The action repetition represents the probability of repeated actions in the whole set of actions. The lower the action repetition, the greater the complexity of the action.

[0074] Specifically, the data acquisition module is used to determine the complexity of the action perturbation route and the action repetition, including:

[0075] Used to record the number of direction changes of each joint point, the number of action repetitions, and the duration of the teaching video;

[0076] The ratio of the number of direction changes of each joint point to the duration is used as the change frequency;

[0077] The average value used to determine the corresponding change frequency of each joint point is the complexity of the action disturbance route;

[0078] The ratio of the number of repetitions of the action and the duration of the action is determined as the action repetition degree;

[0079] The joints include the head joints and the limb joints.

[0080] Specifically, the steps for determining the number of direction changes are as follows:

[0081] Calculate the position change of the joint point at each moment relative to the previous moment to obtain several direction vectors;

[0082] If it is determined that the angle between adjacent direction vectors is greater than a predetermined angle threshold, it is recorded as a direction change;

[0083] The predetermined angle threshold is selected within the interval [15°, 30°].

[0084] Specifically, the data analysis module is used to calculate the complex representation parameters of the action, including:

[0085] The ratio of the complexity of the action perturbation route to the complexity of the reference action perturbation route is used to determine the first influencing factor;

[0086] The ratio of the baseline action repetition rate to the action repetition rate is used to determine the second influencing factor;

[0087] The weighted sum of the first influencing factor and the second influencing factor is used to determine the action complexity representation parameter.

[0088] Specifically, the baseline action perturbation route complexity is calculated in advance, and several teaching end videos are obtained in advance to determine the complexity of several action perturbation routes, and the average value of the complexity of each action perturbation route is determined as the baseline action perturbation route complexity.

[0089] Specifically, the benchmark action repetition is calculated in advance, and several teaching end videos are obtained in advance to determine several action repetitions, and the average value of the action repetitions is determined as the benchmark action repetition.

[0090] Specifically, the sum of the weight coefficients of the first impact factor and the second impact factor is 1, the weight coefficient of the first impact factor is 0.65, and the weight coefficient of the second impact factor is 0.35.

[0091] See also Figure 2 , Figure 2 This is a logic block diagram of the embodiment of the invention for dividing the tendency of action abnormality. Specifically, the data analysis module is used to divide the tendency of action abnormality into modules, wherein:

[0092] If the action complexity representation parameter is greater than the action complexity representation parameter threshold, the action abnormality tendency is classified as a strong abnormality tendency;

[0093] If the action complexity representation parameter is less than or equal to the action complexity representation parameter threshold, the action abnormality tendency is classified as a weak abnormality tendency.

[0094] Specifically, the threshold value of the complex representation parameter of the action is calculated through historical data, and several teaching end videos are obtained in advance to determine several complex representation parameters of the action, and the average value of the complex representation parameters of each action is determined as the threshold value of the complex representation parameter of the action.

[0095] It is understandable that the threshold of the action complexity representation parameter is a periodically changing value and needs to be recalculated every time a change cycle comes to ensure the accuracy of the action abnormality tendency division. In this implementation, the change cycle is determined to be 15 teaching end videos.

[0096] Specifically, by calculating the complex representation parameters of the movements on the teaching end, dividing the tendency of movement mutation, and analyzing the learning end video in a targeted manner for different mutation tendencies, in actual situations, the learning end mostly has multiple ports. If the tendency of movement mutation is not distinguished and the movements are analyzed indiscriminately, simple movements will take up too much computing power, and complex movements may not get enough resource support, thereby reducing the efficiency of movement analysis. For example, the upward stretching movement can be analyzed by a simple method, but if the same analysis method as the complex movement is used, it will cause calculation overflow and waste computing power. Based on this, the present invention pre-calculates the complex representation parameters of the movements on the teaching end, and determines different analysis methods for strong mutation tendencies and weak mutation tendencies, respectively, to improve the efficiency and accuracy of movement analysis.

[0097] Specifically, the interactive processing module is used to compare the joint points of the corresponding action start frame and action end frame of the teaching end video and the learning end video, including:

[0098] for extracting joint point graphs in the action start frame and the action end frame in the teaching end video and the learning end video;

[0099] Place the joint point graphs corresponding to the starting frames of each action in the same coordinate system, scale the joint point graphs corresponding to the learning end video until they meet the stacking error standard, and record the current scaling ratio;

[0100] The joint point graphs corresponding to the end frames of each action are placed in the same coordinate system, and the joint point graphs corresponding to the learning end video are scaled based on the scaling ratio;

[0101] It is used to compare the joint point graphs corresponding to the end frames of each action and determine the average distance difference of each joint point.

[0102] It is understandable that there will be a difference in proportion between the joint point diagrams of the teaching video and the learning video. If the proportions are not adjusted, it may lead to errors in the judgment of the learning action. Therefore, it is necessary to adjust the proportions of the joint point diagrams in the teaching video and the learning video action start frame and action end frame to be consistent, and then place them in the same coordinate system and stack them for calculation.

[0103] Specifically, the stacking error standard is that the average distance difference of each joint point corresponding to the action start frame is less than the distance difference threshold. The distance difference threshold is determined based on the key point map corresponding to the learning end video, wherein the maximum distance between each joint point in the joint point map is determined, and the distance difference threshold is determined between 0.1 times and 0.15 times the maximum distance.

[0104] See also Figure 3 , Figure 3This is a logic diagram for comparing the joint points of the action start frame and the action end frame of the teaching end video and the learning end video in an embodiment of the invention to determine the action state of the learning end. Specifically, the interactive processing module is used to compare the joint points of the action start frame and the action end frame of the teaching end video and the learning end video to determine the action state of the learning end, wherein:

[0105] If the average distance difference is greater than a preset average distance difference standard threshold, determining that the learning terminal action state is a non-standard action state;

[0106] If the average distance difference is less than or equal to a preset average distance difference standard threshold, the learning terminal action state is determined to be a standard action state.

[0107] Specifically, the standard threshold of the average distance difference is determined based on the joint point map corresponding to any video frame on the teaching end, wherein the maximum distance between each joint point in the joint point map is determined, and the standard threshold of the average distance difference is set to between 0.15 times and 0.2 times the maximum distance.

[0108] Specifically, for actions with a weak tendency to mutate, the action state of the learning end is determined by comparing the joint points of the action start frame and the action end frame of the teaching end and the learning end. In actual situations, simple actions have a weak tendency to mutate, especially for straight up and down actions. It is only necessary to determine the starting point and the end position to determine the accuracy of the student end action through the action data of the teaching end, and then determine whether the system needs to issue a status correction reminder to the learning end to improve the efficiency and accuracy of action analysis.

[0109] Specifically, the interactive processing module is used to determine several reference intervals based on the starting frame of the corresponding action in the teaching end video, including:

[0110] To determine the moment when the corresponding action start frame of the teaching video is located;

[0111] It is used to postpone the current time, determine several reference times, determine several reference intervals with the reference times as the starting point, and record the delay time corresponding to each reference interval;

[0112] The starting time of the reference interval is the reference time, and the interval length is the predetermined duration.

[0113] Specifically, there is no limit on the specific value of the delay time. For example, in implementation, the delay time is determined to be 0.1 times the duration of the teaching video. Of course, those skilled in the art can also determine it according to actual conditions, which will not be repeated here.

[0114] Specifically, in implementation, the predetermined duration is selected within the interval [0.5s, 2s].

[0115] Specifically, the interactive processing module is used to determine a number of extraction intervals based on the action start frame of the teaching end video and the corresponding delay time, and predict the extraction interval corresponding to the predicted joint point action graph, including:

[0116] Determine the time of the action start frame of the teaching end video, delay the time by a corresponding delay time, determine a number of extraction times, and determine a number of extraction intervals with the extraction time as the starting point;

[0117] Determining a correspondence between each of the extraction intervals and the reference interval based on the delay time;

[0118] Determine a reference interval corresponding to the extraction interval, extract any teaching end video frame within the reference interval, identify the joint point action graph, and obtain a predicted joint point action graph corresponding to the extraction interval;

[0119] The starting time of the extraction interval is the extraction time, and the interval length is the predetermined duration.

[0120] It is understandable that due to the consistency of the delay time, there is a corresponding relationship between the extraction interval and the reference interval, that is, the extraction interval and the reference interval with the same delay time correspond to each other, which will not be elaborated here.

[0121] Specifically, the interactive processing module is used to compare the predicted joint point action graph with the actual joint point action graph in the learning end video frame within the extraction interval, including:

[0122] Select any learning end video frame in each extraction interval, determine the actual joint point action graph, compare it with the corresponding predicted joint point action graph, and determine the overlap degree of the joint points corresponding to each extraction interval;

[0123] Determine whether each extraction interval meets the verification criteria;

[0124] Among them, the verification standard is that the joint point overlap corresponding to the extraction interval is greater than or equal to the preset joint point overlap threshold.

[0125] Specifically, when solving the joint point coincidence, we can first determine the distance between each joint point of the two joint point graphs, solve the average distance, and determine the reciprocal of the average distance as the joint point coincidence;

[0126] The preset joint point overlap threshold is predetermined, wherein joint point graphs of several individuals performing single actions are obtained in advance, the average joint point overlap between the joint point graphs is calculated, and the product of the average joint point overlap and the error coefficient is determined as the preset joint point overlap threshold.

[0127] See also Figure 4 , Figure 4This is a logic block diagram for determining the action state of the learning terminal according to an embodiment of the invention. Specifically, the interaction processing module is used to determine the action state of the learning terminal, including:

[0128] Determining the proportion of the extracted intervals that meet the validation criteria, it can be understood that the proportion is obtained by calculating the ratio of the number of extracted intervals that meet the validation criteria to the total number of extracted intervals;

[0129] If the ratio is greater than or equal to a predetermined ratio threshold, determining that the learning terminal action state is a standard action state;

[0130] If the ratio is less than a predetermined ratio threshold, it is determined that the action state of the learning terminal is a non-standard action state.

[0131] Specifically, the purpose of setting the ratio threshold is to characterize the situation where most of the extracted intervals meet the verification criteria, so the ratio threshold is selected in the interval [0.5, 0.6].

[0132] Specifically, for those with strong mutation tendency, the predicted joint action graph corresponding to the predicted extraction interval is compared with the actual joint action graph in the video frame of the learning end within the extraction interval to determine the action state of the learning end. In actual situations, complex actions have a strong mutation tendency. During learning, the probability of non-standard actions will increase. At this time, if the same analysis method as the weak mutation tendency is adopted, it may cause errors in the analysis, thereby affecting the judgment result of the action state of the learning end. Based on this, the present invention determines several extraction intervals in combination with the action video segment of the teaching end, predicts the extraction interval corresponding to the predicted joint action graph, and compares it with the actual joint action graph in the video frame of the learning end within the extraction interval to improve the efficiency and accuracy of action analysis.

[0133] Please participate Figure 5 , Figure 5 This is a logic diagram of an embodiment of the invention for determining whether a state correction reminder needs to be issued based on the action state of the learning terminal. Specifically, the state reminder module determines whether a state correction reminder needs to be issued based on the action state of the learning terminal, wherein:

[0134] If the learning terminal action state is a non-standard action state, the determination system needs to issue a state correction reminder;

[0135] If the learning terminal action state is the standard action state, the determination system does not need to issue a state correction reminder.

[0136] Status correction reminders can be sent to the learning end, which will not be repeated here.

[0137] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0138] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. An online teaching interactive system based on a cloud platform, characterized in that: include: A data acquisition module acquires authorized teaching end videos and several learning end videos to determine the complexity of the action disturbance route and the action repetition based on the teaching end videos; a data analysis module connected to the data acquisition module, for calculating a movement complexity representation parameter based on the movement disturbance route complexity and the movement repetition degree, and classifying movement abnormality tendencies; The interactive processing module is connected to the data acquisition module and the data analysis module respectively, and analyzes the learning end video in response to the classification result of the action abnormality tendency, wherein: For weak mutation tendency, compare the joint points of the corresponding action start frame and action end frame of the teaching end video and the learning end video to determine the action state of the learning end; For strong mutation tendency, several reference intervals are determined based on the corresponding action start frame of the teaching end video, and the delay time of the reference interval relative to the time of the action start frame is determined. Several extraction intervals are determined based on the action start frame of the teaching end video and the corresponding delay time. The predicted joint point action graph corresponding to the extraction interval is compared with the actual joint point action graph in the learning end video frame within the extraction interval to determine the action state of the learning end; The status reminder module is connected to the interaction processing module and is used to determine whether a status correction reminder needs to be issued based on the action status of the learning terminal.

2. The online teaching interactive system based on the cloud platform according to claim 1 is characterized in that: The data acquisition module is used to determine the complexity of the action disturbance route and the action repetition, including: Used to record the number of direction changes of each joint point, the number of action repetitions, and the duration of the teaching video; The ratio of the number of direction changes of each joint point to the duration is used as the change frequency; The average value used to determine the corresponding change frequency of each joint point is the complexity of the action disturbance route; The ratio of the number of action repetitions and the duration is used to determine the action repetition degree.

3. The online teaching interactive system based on the cloud platform according to claim 1 is characterized in that: The data analysis module is used to calculate the complex characterization parameters of the action, including: The ratio of the complexity of the action perturbation route to the complexity of the reference action perturbation route is used to determine the first influencing factor; The ratio of the baseline action repetition rate to the action repetition rate is used to determine the second influencing factor; The weighted sum of the first influencing factor and the second influencing factor is used to determine the action complexity representation parameter.

4. The online teaching interactive system based on the cloud platform according to claim 1 is characterized in that: The data analysis module is used to classify the tendency of action changes, wherein: If the action complexity representation parameter is greater than the action complexity representation parameter threshold, the action abnormality tendency is classified as a strong abnormality tendency; If the action complexity representation parameter is less than or equal to the action complexity representation parameter threshold, the action abnormality tendency is classified as a weak abnormality tendency.

5. The online teaching interactive system based on the cloud platform according to claim 1 is characterized in that: The interactive processing module is used to compare the joint points of the teaching end video and the learning end video corresponding to the action start frame and the action end frame, including: for extracting joint point graphs in the action start frame and the action end frame in the teaching end video and the learning end video; Place the joint point graphs corresponding to the starting frames of each action in the same coordinate system, scale the joint point graphs corresponding to the learning end video until they meet the stacking error standard, and record the current scaling ratio; The joint point graphs corresponding to the end frames of each action are placed in the same coordinate system, and the joint point graphs corresponding to the learning end video are scaled based on the scaling ratio; It is used to compare the joint point graphs corresponding to the end frames of each action and determine the average distance difference of each joint point.

6. The online teaching interactive system based on the cloud platform according to claim 5 is characterized in that: The interactive processing module is used to compare the joint points of the teaching end video and the learning end video corresponding to the action start frame and the action end frame to determine the action state of the learning end, wherein, If the average distance difference is greater than a preset average distance difference standard threshold, determining that the learning terminal action state is a non-standard action state; If the average distance difference is less than or equal to a preset average distance difference standard threshold, the learning terminal action state is determined to be a standard action state.

7. The online teaching interactive system based on the cloud platform according to claim 1 is characterized in that: The interactive processing module is used to determine a number of reference intervals based on the starting frame of the action corresponding to the teaching end video, including: To determine the moment when the corresponding action start frame of the teaching video is located; It is used to postpone the current time, determine several reference times, determine several reference intervals with the reference times as the starting point, and record the delay time corresponding to each reference interval; The starting time of the reference interval is the reference time, and the interval length is the predetermined duration.

8. The online teaching interactive system based on the cloud platform according to claim 7 is characterized in that: The interactive processing module is used to determine a number of extraction intervals based on the action start frame of the teaching end video and the corresponding delay time, and predict the extraction interval corresponding to the predicted joint point action graph, including: Determine the time of the action start frame of the teaching end video, delay the time by a corresponding delay time, determine a number of extraction times, and determine a number of extraction intervals with the extraction time as the starting point; Determining a correspondence between each of the extraction intervals and the reference interval based on the delay time; Determine a reference interval corresponding to the extraction interval, extract any teaching end video frame within the reference interval, identify the joint point action graph, and obtain a predicted joint point action graph corresponding to the extraction interval; The starting time of the extraction interval is the extraction time, and the interval length is the predetermined duration.

9. The cloud platform-based online teaching interactive system according to claim 8, characterized in that: The interactive processing module is used to compare the predicted joint point action graph with the actual joint point action graph in the learning end video frame within the extraction interval, including: Select any learning end video frame in each extraction interval, determine the actual joint point action graph, compare it with the corresponding predicted joint point action graph, and determine the overlap degree of the joint points corresponding to each extraction interval; Determine whether each extraction interval meets the verification criteria; Among them, the verification standard is that the joint point overlap corresponding to the extraction interval is greater than or equal to the preset joint point overlap threshold.

10. The cloud platform-based online teaching interactive system according to claim 9, characterized in that: The interaction processing module is used to determine the action status of the learning terminal, including: Determine the proportion of extraction intervals that meet the validation criteria; If the ratio is greater than or equal to a predetermined ratio threshold, determining that the learning terminal action state is a standard action state; If the ratio is less than a predetermined ratio threshold, it is determined that the action state of the learning terminal is a non-standard action state.

Images