Electronic class list information broadcasting method based on artificial intelligence

Abstract

This application provides an AI-based method for playing information on electronic class signs. Based on the course type, it adaptively searches for the corresponding anomaly detection distance threshold from a pre-set mapping table. It receives the pose features of multiple students wearing electronic school badges associated with the target class, transmitted within the previous first time period. The received pose features are clustered to obtain cluster center features. If the distance between any student's pose feature and the cluster center feature is greater than the anomaly detection distance threshold, a short video image containing the student is obtained. The course type and the student's short video image are input into a pre-trained model for determining abnormal classroom behavior. If abnormal behavior is detected, the electronic class sign in the classroom area displays the student's personal information and anomaly alerts.

Description

Technical Field

[0001] This application relates to the field of smart campus management technology, and in particular to an artificial intelligence-based method for displaying information on electronic class signs. Background Technology

[0002] Electronic class signs are intelligent interactive displays placed at classroom doors or in public areas of the campus. As an important component of the smart campus environment, they replace traditional paper class signs. Their core functions include dynamically displaying real-time information such as class name, class schedule, instructors, attendance status, campus notices, and weather, and they typically support touch operation. By connecting to the school's campus network and management system, electronic class signs not only provide students with convenient functions such as facial recognition attendance and personal class schedule inquiries, but also serve as a window for communication between home and school, allowing parents to track their children's arrival and departure from school via their mobile phones.

[0003] However, current electronic class boards cannot monitor student behavior during class (such as whispering or other fidgeting in lecture-based courses). Teachers still need to visually observe students who violate classroom rules to maintain class records. However, with a large number of students in a class, teachers have limited management energy. Constantly monitoring every student's behavior is unreliable for student management and distracts teachers from their teaching focus. Summary of the Invention

[0004] This application provides an AI-based electronic class sign information playback method to address the problems of existing electronic class signs being unable to monitor abnormal student classroom behavior (such as students whispering or making other small movements during theoretical classes), resulting in low reliability of student management and distraction of teachers' teaching focus.

[0005] Firstly, this application provides an artificial intelligence-based method for playing information on electronic class signs, applied to an edge server corresponding to a target class. The edge server is communicatively connected to multiple electronic class signs set in different classroom areas associated with the target class. The electronic class signs include a display module and a video capture module. The method provided in this application includes: During the preset school hours, every preset first hour, the target class's current teaching area and course type are determined according to the preset timetable of the target class. Different course types correspond to different levels of student activity. Based on the course type, the corresponding anomaly detection distance threshold is adaptively searched from a preset mapping table; Receive the positional characteristics of students from the electronic school badges worn by multiple students associated with the target class; Cluster the pose features of multiple students received to obtain the cluster center features of the pose features of multiple students. Determine the distance between each student's pose features and the cluster center features; If the distance between any student's pose feature and the cluster center feature is greater than the anomaly detection distance threshold, determine whether the student's position is within the class area based on the student's pose feature. If the student is located in the classroom area, the electronic class sign in the classroom area is controlled to face the student based on the student's posture characteristics to shoot a preset second duration, resulting in a short video image containing the student. The course type and short video images of students are input into a pre-trained model for identifying abnormal classroom behavior of students to determine whether there is any abnormality in the students' classroom behavior. The model for identifying abnormal classroom behavior of students is trained by inputting multiple first training samples into the network to be trained. Each first training sample includes the historical course type, the short video image of the historical student and its category label. The category label is used to characterize whether the historical student's classroom behavior is abnormal or not. If a student's classroom behavior is found to be abnormal, the electronic class display in the classroom area will be controlled to show the student's personal information and abnormal notification messages.

[0006] In some implementations, after determining whether a student's classroom behavior is abnormal, the method provided in this application further includes: If a student’s classroom behavior is abnormal, perform a correct tracking count for the abnormal classroom behavior of the student and obtain the previously recorded incorrect tracking count for the abnormal classroom behavior of the student under the course type. If the student's classroom behavior is not abnormal, then perform an error tracking count for the student's abnormal classroom behavior and obtain the previously recorded correct tracking count for the student's abnormal classroom behavior under the course type. If the sum of the correct tracking count and the incorrect tracking count reaches a preset threshold, the percentage of the correct tracking count is determined based on the correct tracking count and the preset threshold. If the percentage of correct tracking counts is lower than the set percentage threshold, the anomaly detection distance threshold of the course type in the mapping table is updated using the reinforcement learning model based on the percentage of correct tracking counts. Here, the anomaly detection distance threshold of the course type is the state of the reinforcement learning model, the anomaly detection distance threshold of the course type in the mapping table is the action of the reinforcement learning model, and the percentage of correct tracking counts is the reward of the reinforcement learning model. The correct tracking count and incorrect tracking count are reset to zero, and the process returns to the preset school time period. Every preset first hour, the steps are performed to determine the class area and course type of the target class at the current moment according to the preset timetable of the target class, until the percentage of correct tracking count is greater than or equal to the set percentage threshold.

[0007] In some implementations, the reinforcement learning model is a DQN network.

[0008] In some implementations, the edge server is communicatively connected to X storage modules, where X is an integer greater than 1. The edge server includes a first trusted execution environment, and each storage module includes a second trusted execution environment. The system controls the electronic class display in the classroom area to show students' personal information, including: Retrieve Y encrypted slices of data associated with the student from the second trusted execution environments of X storage modules, where Y is an integer greater than 1; In the first trusted execution environment, the Y encrypted fragment data associated with the student are decrypted to obtain the decrypted Y fragment data. Based on preset data recovery rules, the decrypted Y pieces of data are merged to recover the corresponding multiple blocks of data; By stitching together the restored data blocks, the students' personal information can be obtained. The electronic class display in the classroom area controls the display of students' personal information.

[0009] In some implementations, after a preset school start time, the method provided in this application further includes: The number of times each student's behavior was identified as abnormal classroom behavior during the pre-set school hours was counted. The personal information of students whose behavior is identified as abnormal in class more than a set threshold during the preset school hours will be pushed to the electronic class board in the teacher's office corresponding to the student's homeroom teacher.

[0010] Secondly, this application also provides an artificial intelligence-based electronic class sign information playback system, configured on an edge server corresponding to the target class. The edge server is connected to multiple electronic class signs and electronic class signs set up in different classroom areas associated with the target class. The system provided by this application includes: The class information determination unit is used to determine the class area and course type of the target class at the current time according to the preset target class's timetable during the preset school time period, every preset first hour. Different course types correspond to different student activity levels. The threshold adaptive determination unit is used to adaptively look up the corresponding anomaly detection distance threshold from a preset mapping table according to the course type; The data receiving unit is used to receive the positional characteristics of students transmitted from the electronic school badges worn by multiple students associated with the target class; The feature clustering unit is used to cluster the pose features of multiple students received, and obtain the cluster center features of the pose features of multiple students. Distance determination unit, used to determine the distance between each student's pose features and cluster center features; The student location determination unit is used to determine whether a student is in the class area based on their pose characteristics when the distance between the pose characteristics of any student and the cluster center characteristics is greater than the anomaly detection distance threshold. The video shooting control unit is used to control the electronic class sign in the classroom area to shoot a preset second duration of video images containing the student, based on the student's posture characteristics, if the student's position is in the classroom area. An abnormal classroom behavior identification unit is used to input course type and short video images of students into a pre-trained model for determining abnormal classroom behavior of students to determine whether students’ classroom behavior is abnormal. The model for determining abnormal classroom behavior of students is trained by inputting multiple first training samples into the network to be trained. Each first training sample includes the historical course type, the short video image of the historical student and its category label. The category label is used to characterize whether the historical student’s classroom behavior is abnormal or not. The playback control unit is used to control the electronic class board in the classroom area to play the student's personal information and abnormal prompts if the student's classroom behavior is abnormal.

[0011] In some embodiments, the system provided in this application further includes: The first counting unit is used to perform a correct tracking count of the abnormal classroom behavior of a student if there is an abnormality in the student's classroom behavior, and to obtain the previously recorded incorrect tracking count of the abnormal classroom behavior of the student under the course type. The second counting unit is used to perform an error tracking count for abnormal classroom behavior of students if there is no abnormality in the students' classroom behavior, and to obtain the previously recorded correct tracking count for abnormal classroom behavior of students under the course type. The counting percentage determination unit is used to determine the percentage of the correct tracking count based on the correct tracking count and the preset number threshold if the sum of the correct tracking count and the incorrect tracking count reaches a preset number threshold. The threshold update unit is used to update the anomaly detection distance threshold of the course type in the mapping table according to the percentage of correct tracking counts when the percentage of correct tracking counts is lower than a set percentage threshold. Here, the anomaly detection distance threshold of the course type is the state of the reinforcement learning model, updating the anomaly detection distance threshold of the course type in the mapping table is the action of the reinforcement learning model, and the percentage of correct tracking counts is the reward of the reinforcement learning model. The unit then resets the correct tracking count and the incorrect tracking count to zero and returns to the preset school time period. Every preset first time interval, the unit determines the class area and course type of the target class at the current time according to the preset timetable of the target class, until the percentage of correct tracking counts is greater than or equal to the set percentage threshold.

[0012] In some implementations, the reinforcement learning model is a DQN network.

[0013] In some implementations, the edge server is communicatively connected to X storage modules, where X is an integer greater than 1. The edge server includes a first trusted execution environment, each storage module includes a second trusted execution environment, and the playback control unit includes: The data acquisition module is used to acquire Y encrypted slices of data associated with students from the second trusted execution environments of X storage modules, where Y is an integer greater than 1. The data decryption module is used to decrypt the Y encrypted fragment data associated with the student in the first trusted execution environment to obtain the decrypted Y fragment data. The data recovery module is used to fuse the decrypted Y pieces of data based on preset data recovery rules and recover them into multiple corresponding blocks of data. The data stitching module is used to stitch together the recovered data blocks to obtain the student's personal information. The playback control module is used to control the electronic class display in the classroom area and display students' personal information.

[0014] In some embodiments, the system provided in this application further includes: The data statistics unit is used to count the number of times each student's behavior was identified as abnormal classroom behavior during the preset school hours; The information push unit is used to push the personal information of students whose behavior is identified as abnormal in class more than a set threshold during the preset school hours to the electronic class board in the teacher's office corresponding to the student's homeroom teacher.

[0015] This application provides an AI-based electronic class sign information display method and system. During a preset school time period, every preset first interval, based on the preset class schedule, the system determines the target class's current learning area and course type. Different course types correspond to different levels of student activity. For example, course types with student activity levels ranging from low to high are: theoretical lecture courses, self-study courses, experimental operation courses, and physical education courses.

[0016] Based on the course type, the corresponding anomaly detection distance threshold is adaptively searched from the preset mapping table. Understandably, in the classroom, the higher the student activity level corresponding to the course type, the higher the dispersion of the students' pose characteristics, and thus the higher the corresponding anomaly detection distance threshold is set.

[0017] The system receives the pose features of multiple students in a target class, transmitted via electronic school badges, within the previous time period. It then clusters these pose features to obtain cluster center features. Typically, within the same course type, the pose features of students during normal class sessions exhibit varying degrees of clustering. For example, in lecture-based courses, students sit upright and focus on the blackboard; in self-study courses, students sit in their classrooms doing homework or reading; and in experimental courses, students operate instruments, move around, take notes, or engage in discussions in the laboratory.

[0018] Furthermore, the distance between each student's pose feature and the cluster center feature is determined. If the distance between any student's pose feature and the cluster center feature is greater than the anomaly detection distance threshold, it indicates that the student's pose feature is significantly different from that of other students, and there may be abnormal classroom behavior. It is worth noting that since the anomaly detection distance threshold here was previously determined adaptively based on the course type, the reliability of determining that students may have abnormal classroom behavior is higher.

[0019] Based on the student's posture characteristics, determine whether the student is in the teaching area; specifically, based on the position information in the posture characteristics, determine whether the student is in the corresponding teaching area.

[0020] If the student is located within the classroom area, the electronic class sign in the classroom area is further controlled to face the student based on the student's posture characteristics to capture a pre-set second duration (e.g., 10 seconds), resulting in a short video image containing the student. The course type and the student's short video image are then input into a pre-trained model for identifying abnormal classroom behavior to determine if any abnormalities exist. Understandably, different course types have different criteria for judging abnormal classroom behavior. Therefore, the basis for determining student classroom behavior includes not only the student's short video image but also the course type, resulting in high reliability in identifying abnormal classroom behavior.

[0021] If a student's classroom behavior is deemed abnormal, the electronic class display in the classroom area will show the student's personal information and a notification of the abnormality. This allows teachers to view the displayed information and notifications during breaks in class and address the student's unusual behavior appropriately. This eliminates the need for teachers to constantly monitor every student, allowing them to focus their teaching efforts, and the real-time monitoring improves the reliability of student management. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart illustrating the AI-based electronic class sign information playback method provided in this application embodiment; Figure 2 A functional unit block diagram of an AI-based electronic class sign information playback system provided in an embodiment of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments made by those skilled in the art under the guidance of these embodiments are within the scope of protection of this application.

[0025] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] Please see Figure 1 This application provides an AI-based method for displaying information on electronic class signs, applied to an edge server corresponding to a target class. The edge server is communicatively connected to multiple electronic class signs located in different classroom areas associated with the target class. These different classroom areas may include, but are not limited to, the target class's classroom (for Chinese and math classes, etc.), study room, laboratory, and playground (for physical education or activity classes). The electronic class signs include a display module and a video capture module. Multiple electronic class signs can be installed at the entrances of classrooms, study rooms, and laboratories, as well as at the entrance to the playground. Specifically, the method provided in this application includes: S101: During the preset school hours, every preset first hour, the target class's current learning area and course type are determined according to the preset timetable of the target class. Different course types correspond to different levels of student activity.

[0027] The preset school hours can be 8:00 AM to 5:00 PM or 8:00 AM to 8:30 PM. The preset first duration can be 1 minute. The current teaching area can be the classroom corresponding to the target class. The course type can be lecture-based, self-study, experimental, or physical education. The corresponding course types with student activity levels from low to high are lecture-based, self-study, experimental, and physical education, respectively.

[0028] S102: Based on the course type, adaptively search for the corresponding anomaly detection distance threshold from the preset mapping table.

[0029] Understandably, the higher the student activity level corresponding to a course type, the higher the dispersion of the students' pose characteristics, and thus the higher the corresponding anomaly detection distance threshold.

[0030] Understandably, for each course type, experts have multiple groups of students take the course and cluster the pose feature sets of each group of students to obtain the cluster radius. Then, by averaging the cluster radii corresponding to multiple groups of students, the anomaly detection distance threshold for that course type can be obtained. Furthermore, a correspondence is established between the anomaly detection thresholds for each course type, thus obtaining a pre-defined mapping table.

[0031] S103: Receives the pose characteristics of students transmitted from the electronic school badges worn by multiple students associated with the target class.

[0032] For example, a target class may have multiple students with electronic school badges, and each student's electronic school badge can establish a short-range communication connection (such as a Wi-Fi connection) with an edge server.

[0033] S104: Cluster the pose features of multiple students received to obtain the cluster center features of the pose features of multiple students.

[0034] Typically, within the same course type, the posture characteristics of students in a normal class exhibit varying degrees of clustering. For example, in lecture-based courses, students sit upright in their seats, focusing on the blackboard; in self-study courses, students sit in their seats in the classroom doing homework or flipping through books; in experimental courses, students operate instruments in the laboratory, walk around, take notes, or exchange ideas; and in physical education courses, students exercise on the playground.

[0035] S105: Determine the distance between the pose features of each student and the cluster center features.

[0036] S106: If the distance between the pose feature of any student and the cluster center feature is greater than the anomaly detection distance threshold, determine whether the student's position is in the class area based on the student's pose feature.

[0037] If the distance between a student's pose feature and the cluster center feature is greater than the anomaly detection distance threshold, it indicates that the student's pose feature is significantly different from that of other students, and there may be abnormal classroom behavior (such as a student throwing a note to the next table in math class, or turning around to talk to a student at the table behind, or a student leaving the lab or resting on the lab table in lab class). It is worth noting that since the anomaly detection distance threshold here was previously adaptively determined based on the course type, the reliability of determining that a student may have abnormal classroom behavior is higher.

[0038] S107: If the student is located in the classroom area, the electronic class sign in the classroom area is controlled to face the student according to the student's posture characteristics to shoot a preset second duration, resulting in a short video image containing the student.

[0039] Since positional features can be used to locate students who may exhibit abnormal classroom behavior, the electronic class sign in the classroom area can be controlled to face the students and take pictures for a preset second duration (e.g., 10 seconds).

[0040] S108: Input the course type and short video images of students into the pre-trained model for identifying abnormal classroom behavior of students to determine whether there are any abnormalities in students' classroom behavior.

[0041] The model for identifying abnormal classroom behavior of students is trained by inputting multiple first training samples into the network to be trained. Each first training sample includes the type of history course, a short video image of a history student and its category label. The category label is used to characterize whether the history student's classroom behavior is abnormal or not.

[0042] The network to be trained can be, but is not limited to, a convolutional neural network. A convolutional neural network may include an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a fully connected layer, and an output layer. The first convolutional layer includes 64 convolutional kernels, each with a size of 3×3, and uses the ReLU activation function. The first pooling layer is a 1D convolutional layer with a pooling window size of 2. The second convolutional layer includes 128 convolutional kernels, each with a size of 3×3, and uses the ReLU activation function. The second pooling layer is also a 1D convolutional layer with a pooling window size of 2. The fully connected layer is a Dense layer containing 50 neurons and uses the ReLU activation function. The output layer uses a linear activation function.

[0043] Understandably, different course types have different criteria for judging abnormal classroom behavior (e.g., whispering among students in a Chinese class is abnormal classroom behavior, but whispering among students in a lab class is not). The basis for determining students' classroom behavior includes not only short video images of students but also the course type, making the determination of abnormal classroom behavior highly reliable.

[0044] S109: If a student's classroom behavior is abnormal, control the electronic class sign in the classroom area to display the student's personal information and abnormal prompts.

[0045] In summary, the present application provides an AI-based electronic class sign information playback method. During a preset school time period, at preset intervals, the method determines the current class area and course type of the target class based on the preset timetable of the target class. Different course types correspond to different levels of student activity. For example, the course types with student activity levels ranging from low to high are theoretical lecture courses, self-study courses, experimental operation courses, and sports courses.

[0046] Based on the course type, the corresponding anomaly detection distance threshold is adaptively searched from the preset mapping table. Understandably, the higher the student activity level corresponding to the course type, the higher the dispersion of the pose features among students, and thus the higher the corresponding anomaly detection distance threshold is set.

[0047] The system receives the pose features of multiple students in a target class, transmitted via electronic school badges, within the previous time period. It then clusters these pose features to obtain cluster center features. Typically, within the same course type, the pose features of students during normal class sessions exhibit varying degrees of clustering. For example, in lecture-based courses, students sit upright and focus on the blackboard; in self-study courses, students sit in their classrooms doing homework or reading; and in experimental courses, students operate instruments, move around, take notes, or engage in discussions in the laboratory.

[0048] Furthermore, the distance between each student's pose feature and the cluster center feature is determined. If the distance between any student's pose feature and the cluster center feature is greater than the anomaly detection distance threshold, it indicates that the student's pose feature is significantly different from that of other students, and there may be abnormal classroom behavior. It is worth noting that since the anomaly detection distance threshold here was previously determined adaptively based on the course type, the reliability of determining that students may have abnormal classroom behavior is higher.

[0049] Based on the student's posture characteristics, determine whether the student is in the teaching area; specifically, based on the position information in the posture characteristics, determine whether the student is in the corresponding teaching area.

[0050] If the student is located within the classroom area, the electronic class sign in the classroom area is further controlled to face the student based on the student's posture characteristics to capture a pre-set second duration (e.g., 10 seconds), resulting in a short video image containing the student. The course type and the student's short video image are then input into a pre-trained model for identifying abnormal classroom behavior to determine if any abnormalities exist. Understandably, different course types have different criteria for judging abnormal classroom behavior. Therefore, the basis for determining student classroom behavior includes not only the student's short video image but also the course type, resulting in high reliability in identifying abnormal classroom behavior.

[0051] If a student's classroom behavior is deemed abnormal, the electronic class display in the classroom area will show the student's personal information and a notification of the abnormality. This allows teachers to view the displayed information and notifications during breaks in class and address the student's unusual behavior appropriately. This eliminates the need for teachers to constantly monitor every student, allowing them to focus their teaching efforts, and the real-time monitoring improves the reliability of student management.

[0052] For example, a student's personal information may include, but is not limited to, the student's name, student ID, the time when abnormal classroom behavior occurred, parents' contact information, and home address.

[0053] In some embodiments, following S109 described above, the method provided in this application further includes: Step A1: If a student’s classroom behavior is abnormal, perform a correct tracking count for the abnormal classroom behavior of the student and obtain the previously recorded incorrect tracking count for the abnormal classroom behavior of the student under the course type.

[0054] Step A2: If there are no abnormalities in the student's classroom behavior, perform an error tracking count for the student's abnormal classroom behavior and obtain the previously recorded correct tracking count for the student's abnormal classroom behavior under the course type.

[0055] Step A3: If the sum of the correct tracking count and the incorrect tracking count reaches a preset threshold, then determine the percentage of the correct tracking count based on the correct tracking count and the preset threshold.

[0056] Step A4: Determine whether the percentage of correctly tracked counts is lower than the set percentage threshold. If so, proceed to step A5.

[0057] Step A5: Based on the proportion of correct tracking counts, update the anomaly detection distance threshold of course type in the mapping table using a reinforcement learning model.

[0058] Among them, the anomaly detection distance threshold of course type is the state of the reinforcement learning model, the anomaly detection distance threshold of course type in the updated mapping table is the action of the reinforcement learning model, and the percentage of correct tracking counts is the reward of the reinforcement learning model.

[0059] It should be noted that the reinforcement learning model here can be a DQN (Deep Q-Network). Unlike Q-learning, DQN does not require building a complete Q matrix; instead, it uses a neural network to estimate the value of the Q function. The neural network responsible for estimating the Q function is called the main network. This represents the parameter set of the main network, while the target network outputs target values, which are used to update the parameters of the main network. The estimated values ​​of the main network and the target values ​​of the target network can form a loss function, and the main network parameters can be updated using stochastic gradient descent with the Adam optimization algorithm. The parameters of the main network can be continuously updated based on the gradient of the loss function, allowing the loss function value to decrease and thus making the estimated values ​​of the main network more accurate. Both the main and target networks contain input layers, convolutional layers, fully connected layers, and output layers. The neurons in the input layer are responsible for feeding data to the neurons in the convolutional layers, which are responsible for extracting local features from the input data. The fully connected layers integrate these local features into global features, and the output layer is responsible for outputting the estimated value of the Q-function for each action in the current state.

[0060] Step A6: Reset the correct tracking count and the incorrect tracking count to zero, and return to the execution of step S101 above until the percentage of correct tracking count is greater than or equal to the set percentage threshold.

[0061] Based on steps A1-A6 above, the anomaly detection distance thresholds corresponding to different course types initially configured in the preset mapping relationship table can be continuously optimized, further improving the accuracy of the anomaly detection distance thresholds.

[0062] Additionally, the edge server communicates with X storage modules, where X is an integer greater than 1. The edge server includes a first trusted execution environment, and each storage module includes a second trusted execution environment. The above S109 can be specifically implemented as follows: Step B1: Obtain Y encrypted slice data associated with the student from the second trusted execution environments of X storage modules, where Y is an integer greater than 1.

[0063] For example, if student A's personal information includes 8 encrypted data segments, with 2 encrypted data segments located on storage device A, 2 on storage device B, and 4 on storage device C, then it is possible to retrieve 2 encrypted data segments from storage device A, 2 from storage device B, and 4 from storage device C. Step B2: In the first trusted execution environment, decrypt the Y encrypted fragment data associated with the student to obtain the decrypted Y fragment data.

[0064] Step B3: Based on the preset data recovery rules, merge the decrypted Y fragments of data to recover the corresponding multiple blocks of data.

[0065] For example, restore fragment data 1, fragment data 2 and fragment data 3 into 1 block of data, restore fragment data 4 and fragment data 5 into 1 block of data, and restore fragment data 6, fragment data 7 and fragment data 8 into 1 block of data.

[0066] Step B4: Concatenate the recovered data blocks to obtain the student's personal information.

[0067] For example, given three data blocks—block 1, block 2, and block 3—combining these blocks yields six different combinations. The hash value of each combination is calculated, and the combination whose hash value matches the stored hash value is identified as the student's personal information. This allows for the efficient and accurate recovery of student personal information.

[0068] Step B5: Control the electronic class display in the classroom area to display students' personal information.

[0069] Based on the steps B1-B2 above, students' personal information can be securely protected to prevent it from being stolen.

[0070] In addition, after the preset school start time, the method provided in this application embodiment further includes: counting the number of times each student's behavior is identified as abnormal classroom behavior during the preset school start time; and pushing the personal information of students whose behavior is identified as abnormal classroom behavior more than a set threshold during the preset school start time to the electronic class board in the teacher's office corresponding to the student's homeroom teacher.

[0071] In addition, this application embodiment also provides an artificial intelligence-based electronic class sign information playback system, configured on an edge server corresponding to the target class. The edge server is communicatively connected to multiple electronic class signs located in different classroom areas associated with the target class. It should be noted that the basic principle and technical effects of the artificial intelligence-based electronic class sign information playback system provided in this application embodiment are the same as those in the above embodiments. For the sake of brevity, any parts not mentioned in this application embodiment can be referred to the corresponding content in the above embodiments. Figure 2 As shown, the system provided in this application embodiment includes a class information determination unit, a threshold adaptive determination unit, a data receiving unit, a feature clustering unit, a distance determination unit, a student location determination unit, a video capture control unit, and an abnormal classroom behavior unit, wherein... The class information determination unit is used to determine the class area and course type of the target class at the current time according to the preset target class's timetable during the preset school time period, every preset first hour. Different course types correspond to different student activity levels. The threshold adaptive determination unit is used to adaptively look up the corresponding anomaly detection distance threshold from a preset mapping table according to the course type; The data receiving unit is used to receive the positional characteristics of students transmitted from the electronic school badges worn by multiple students associated with the target class; The feature clustering unit is used to cluster the pose features of multiple students received, and obtain the cluster center features of the pose features of multiple students. Distance determination unit, used to determine the distance between each student's pose features and cluster center features; The student location determination unit is used to determine whether a student is in the class area based on their pose characteristics when the distance between the pose characteristics of any student and the cluster center characteristics is greater than the anomaly detection distance threshold. The video shooting control unit is used to control the electronic class sign in the classroom area to shoot a preset second duration of video images containing the student, based on the student's posture characteristics, if the student's position is in the classroom area. An abnormal classroom behavior identification unit is used to input course type and short video images of students into a pre-trained model for determining abnormal classroom behavior of students to determine whether students’ classroom behavior is abnormal. The model for determining abnormal classroom behavior of students is trained by inputting multiple first training samples into the network to be trained. Each first training sample includes the historical course type, the short video image of the historical student and its category label. The category label is used to characterize whether the historical student’s classroom behavior is abnormal or not. The playback control unit is used to control the electronic class board in the classroom area to play the student's personal information and abnormal prompts if the student's classroom behavior is abnormal.

[0072] In some embodiments, the system provided in this application further includes: The first counting unit is used to perform a correct tracking count of the abnormal classroom behavior of a student if there is an abnormality in the student's classroom behavior, and to obtain the previously recorded incorrect tracking count of the abnormal classroom behavior of the student under the course type. The second counting unit is used to perform an error tracking count for abnormal classroom behavior of students if there is no abnormality in the students' classroom behavior, and to obtain the previously recorded correct tracking count for abnormal classroom behavior of students under the course type. The counting percentage determination unit is used to determine the percentage of the correct tracking count based on the correct tracking count and the preset number threshold if the sum of the correct tracking count and the incorrect tracking count reaches a preset number threshold. The threshold update unit is used to update the anomaly detection distance threshold of the course type in the mapping table according to the percentage of correct tracking counts when the percentage of correct tracking counts is lower than a set percentage threshold. Here, the anomaly detection distance threshold of the course type is the state of the reinforcement learning model, updating the anomaly detection distance threshold of the course type in the mapping table is the action of the reinforcement learning model, and the percentage of correct tracking counts is the reward of the reinforcement learning model. The unit then resets the correct tracking count and the incorrect tracking count to zero and returns to the preset school time period. Every preset first time interval, the unit determines the class area and course type of the target class at the current time according to the preset timetable of the target class, until the percentage of correct tracking counts is greater than or equal to the set percentage threshold.

[0073] In some implementations, the reinforcement learning model is a DQN network.

[0074] In some implementations, the edge server is communicatively connected to X storage modules, where X is an integer greater than 1. The edge server includes a first trusted execution environment, each storage module includes a second trusted execution environment, and the playback control unit includes: The data acquisition module is used to acquire Y encrypted slices of data associated with students from the second trusted execution environments of X storage modules, where Y is an integer greater than 1. The data decryption module is used to decrypt the Y encrypted fragment data associated with the student in the first trusted execution environment to obtain the decrypted Y fragment data. The data recovery module is used to fuse the decrypted Y pieces of data based on preset data recovery rules and recover them into multiple corresponding blocks of data. The data stitching module is used to stitch together the recovered data blocks to obtain the student's personal information. The playback control module is used to control the electronic class display in the classroom area and display students' personal information.

[0075] In some embodiments, the system provided in this application further includes: The data statistics unit is used to count the number of times each student's behavior was identified as abnormal classroom behavior during the preset school hours; The information push unit is used to push the personal information of students whose behavior is identified as abnormal in class more than a set threshold during the preset school hours to the electronic class board in the teacher's office corresponding to the student's homeroom teacher.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for playing electronic class information based on artificial intelligence, characterized in that, An edge server corresponding to a target class is applied, the edge server being communicatively connected to multiple electronic class signs set in different classroom areas associated with the target class, the electronic class signs including a display module and a video acquisition module, the method comprising: During the pre-set school hours, at each pre-set first hour interval, the class area and course type of the target class at the current moment are determined according to the pre-set timetable of the target class. Different course types correspond to different levels of student activity. Based on the course type, the corresponding anomaly detection distance threshold is adaptively searched from a preset mapping table; Receive the positional characteristics of students within the previous first time period transmitted by the electronic school badges worn by multiple students associated with the target class; Cluster the pose features of multiple students received to obtain cluster center features of the pose features of multiple students; Determine the distance between each student's pose features and the cluster center features; If the distance between any student's pose feature and the cluster center feature is greater than the anomaly detection distance threshold, determine whether the student's position is within the class area based on the student's pose feature. If the student is located in the classroom area, the electronic class sign in the classroom area is controlled to face the student and shoot for a preset second duration based on the student's posture characteristics, so as to obtain a short video image containing the student. The course type and the student's short video image are input into a pre-trained model for determining abnormal classroom behavior of students to determine whether the student's classroom behavior is abnormal. The model for determining abnormal classroom behavior of students is trained by inputting multiple first training samples into the network to be trained. Each first training sample includes a historical course type, a short video image of a historical student and its category label. The category label is used to characterize whether the historical student's classroom behavior is abnormal or not. If a student's classroom behavior is abnormal, the electronic class sign in the classroom area will be controlled to display the student's personal information and abnormal notification information.

2. The method according to claim 1, characterized in that, After determining whether the student's classroom behavior is abnormal, the method further includes: If the student’s classroom behavior is abnormal, a correct tracking count of the abnormal classroom behavior of the student is performed, and the previously recorded incorrect tracking count of the abnormal classroom behavior of the student under the course type is obtained. If the student's classroom behavior is not abnormal, then perform an error tracking count for the student's abnormal classroom behavior, and obtain the previously recorded correct tracking count for the student's abnormal classroom behavior under the course type. If the sum of the correct tracking count and the incorrect tracking count reaches a preset threshold, then the percentage of the correct tracking count is determined based on the correct tracking count and the preset threshold. If the percentage of correct tracking counts is lower than a set percentage threshold, the anomaly detection distance threshold of the course type in the mapping table is updated using a reinforcement learning model based on the percentage of correct tracking counts. Here, the anomaly detection distance threshold of the course type is the state of the reinforcement learning model, updating the anomaly detection distance threshold of the course type in the mapping table is the action of the reinforcement learning model, and the percentage of correct tracking counts is the reward of the reinforcement learning model. The correct tracking count and the incorrect tracking count are reset to zero, and the process returns to the step of determining the class area and course type of the target class at the current time according to the preset target class's timetable during the preset school period, every preset first time interval, until the percentage of the correct tracking count is greater than or equal to the set percentage threshold.

3. The method according to claim 2, characterized in that, The reinforcement learning model is a DQN network.

4. The method according to claim 1, characterized in that, The edge server is communicatively connected to X storage modules, where X is an integer greater than 1. The edge server includes a first trusted execution environment, and each storage module includes a second trusted execution environment. The function of controlling the electronic class sign in the classroom area to display the students' personal information includes: From the second trusted execution environments of X storage modules respectively, obtain Y encrypted slice data associated with the student, where Y is an integer greater than 1; In the first trusted execution environment, the Y encrypted fragment data associated with the student are decrypted to obtain the decrypted Y fragment data. Based on preset data recovery rules, the decrypted Y pieces of data are merged to recover the corresponding multiple blocks of data; The recovered data blocks are then stitched together to obtain the student's personal information. Control the electronic class display in the classroom area to display the students' personal information.

5. The method according to claim 1, characterized in that, After the preset school start time period, the method further includes: The number of times each student's behavior was identified as abnormal classroom behavior during the preset school hours was counted. The personal information of students whose behavior is identified as abnormal in class during the preset school hours exceeds a set threshold will be pushed to the electronic class board in the teacher's office corresponding to the student's homeroom teacher.

6. An electronic class sign information playback system based on artificial intelligence, characterized in that, An edge server configured for the target class, the edge server being communicatively connected to multiple electronic class signs and electronic class signs located in different classroom areas associated with the target class, the system comprising: The class information determination unit is used to determine the class area and course type of the target class at the current time according to the preset target class's timetable during a preset school period, at preset first intervals. Different course types correspond to different student activity levels. The threshold adaptive determination unit is used to adaptively search for the corresponding anomaly judgment distance threshold from a preset mapping table according to the course type; The data receiving unit is used to receive the position and posture characteristics of students in the previous first time period transmitted by the electronic school badges worn by multiple students associated with the target class. The feature clustering unit is used to cluster the pose features of the multiple students received, and obtain the cluster center features of the pose features of the multiple students. A distance determination unit is used to determine the distance between the pose features of each student and the cluster center features; The student location determination unit is used to determine whether the student's location is within the class area based on the student's pose features when the distance between the pose features of any student and the cluster center features is greater than the anomaly detection distance threshold. The video shooting control unit is used to control the electronic class sign in the class area to shoot a preset second duration of video for the student based on the student's posture characteristics if the student's position is in the class area, thereby obtaining a short video image containing the student. An abnormal classroom behavior identification unit is used to input the course type and the student's short video image into a pre-trained student abnormal classroom behavior determination model to determine whether the student's classroom behavior is abnormal. The student abnormal classroom behavior determination model is trained by inputting multiple first training samples into the network to be trained. Each first training sample includes a historical course type, a historical student's short video image and its category label. The category label is used to characterize whether the historical student's classroom behavior is abnormal or not. The playback control unit is used to control the electronic class sign in the classroom area to play the student's personal information and abnormal prompts if the student's classroom behavior is abnormal.

7. The system according to claim 6, characterized in that, The system also includes: The first counting unit is used to perform a correct tracking count of the abnormal classroom behavior of the student if the student's classroom behavior is abnormal, and to obtain the previously recorded incorrect tracking count of the abnormal classroom behavior of the student under the course type. The second counting unit is used to perform an error tracking count for abnormal classroom behavior of the student if there is no abnormality in the student's classroom behavior, and to obtain the previously recorded correct tracking count for abnormal classroom behavior of the student under the course type. The counting percentage determination unit is used to determine the percentage of the correct tracking count based on the correct tracking count and the preset number threshold if the sum of the correct tracking count and the incorrect tracking count reaches a preset number threshold. The threshold update unit is used to update the anomaly detection distance threshold of the course type in the mapping table according to the percentage of the correct tracking count when the percentage of the correct tracking count is lower than a set percentage threshold. The anomaly detection distance threshold of the course type represents the state of the reinforcement learning model, updating the anomaly detection distance threshold of the course type in the mapping table represents the action of the reinforcement learning model, and the percentage of the correct tracking count represents the reward of the reinforcement learning model. The unit then resets the correct tracking count and the incorrect tracking count to zero and returns to the step of determining the class area and course type of the target class at the current time according to the preset target class's timetable during a preset school period, every preset first time interval, until the percentage of the correct tracking count is greater than or equal to the set percentage threshold.

8. The system according to claim 7, characterized in that, The reinforcement learning model is a DQN network.

9. The system according to claim 6, characterized in that, The edge server is communicatively connected to X storage modules, where X is an integer greater than 1. The edge server includes a first trusted execution environment, and each storage module includes a second trusted execution environment. The playback control unit includes: The data acquisition module is used to acquire Y encrypted fragments of data associated with the student from the second trusted execution environments of X storage modules respectively, where Y is an integer greater than 1; The data decryption module is used to decrypt the Y encrypted fragment data associated with the student in the first trusted execution environment to obtain the decrypted Y fragment data. The data recovery module is used to fuse the decrypted Y pieces of data based on preset data recovery rules and recover them into multiple corresponding blocks of data. The data stitching module is used to stitch together the recovered data blocks to obtain the student's personal information; The playback control module is used to control the electronic class sign in the classroom area and play the students' personal information.

10. The system according to claim 6, characterized in that, The system also includes: The data statistics unit is used to count the number of times each student's behavior is identified as abnormal classroom behavior during the preset school hours. The information push unit is used to push the personal information of students whose behavior is identified as abnormal in class more than a set threshold during the preset school hours to the electronic class board in the teacher's office corresponding to the student's homeroom teacher.

Images