College course teaching archive electronic auxiliary arrangement system and method

Abstract

The invention provides a college course teaching archive electronic auxiliary arrangement system and method, and the system comprises a multi-source data collection terminal which is used for collecting teaching data and environment data; the behavior fusion model module is used for generating multi-dimensional behavior characteristics through the teaching data, constructing a space-time behavior graph and generating a behavior dependency relationship of teacher-student interaction in a classroom through the space-time behavior graph; the dynamic cloud storage module is used for synchronizing the teaching data and the environment data in real time and allocating storage resources according to presetting; the automatic teaching evaluation module is used for generating behavior diversity evaluation, comprehensive learning trend prediction and teaching abnormity feedback through the spatio-temporal behavior graphs of different time nodes, multi-source data acquisition, behavior analysis, dynamic storage and teaching evaluation are integrated, college teaching management is promoted to be upgraded from informatization to intellectualization, and the teaching efficiency is improved. The method has important academic value and application value.

Description

Technical Field

[0001] The present invention relates to the fields of educational informatization and artificial intelligence technology, and particularly relates to an electronic auxiliary sorting system and method for teaching archives of college courses.

Background Art

[0002] The teaching archives of college courses are an indispensable part of college teaching management. In the prior art, the electronic management methods of college teaching archives mainly involve:

[0003] Attendance and behavior recording: In traditional technology, a single video monitoring device is used for classroom attendance and behavior collection, and the accuracy is limited by video clarity and noise interference, especially showing poor performance in scenarios with a large number of students.

[0004] Data storage and management: Currently, most adopt local storage methods, lacking intelligent partitioning and priority management, resulting in low storage efficiency and high storage expansion costs;

[0005] Teaching evaluation and feedback: The prior art only supports evaluation in a single dimension (such as exam scores), and fails to conduct comprehensive analysis by combining classroom behavior data and students' long-term performance trends.

[0006] For example, some existing systems generate course summary reports based on fixed templates, but cannot provide personalized suggestions for different courses and students. Generally speaking, these technical means lack the ability of multi-dimensional collaborative processing and are difficult to meet the needs of refined teaching management in colleges and universities.

[0007] It can be seen from the above cases in the prior art that the following problems exist in the prior art:

[0008] Low accuracy of behavior analysis: Existing systems are difficult to identify behavior patterns in complex classroom scenarios, such as interactions between students and attention distribution;

[0009] Insufficient real-time performance and adaptability: In a dynamic classroom environment, the ability to process large-scale data in real time is weak and is easily affected by noise and network latency;

[0010] Single evaluation index: New indexes such as behavior diversity and participation contribution rate cannot be quantified, and the evaluation results have limited support for optimizing teaching strategies;

[0011] High storage cost and insufficient scalability: Traditional local storage systems cannot be efficiently expanded, cloud storage is not fully utilized, and there is a lack of dynamic partitioning optimization strategies.

[0012] Therefore, it is necessary to study an electronic auxiliary sorting system and method for teaching archives of college courses to address the deficiencies of the prior art and solve or alleviate one or more of the above problems.

Summary of the Invention

[0013] In view of this, the present invention provides an electronic auxiliary sorting system and method for teaching files in colleges and universities, which integrates multi-source data collection, behavior analysis, dynamic storage and teaching evaluation, promotes the upgrading of college teaching management from "informationization" to "intelligence", and has important academic value and application value.

[0014] On the one hand, the present invention provides an electronic auxiliary sorting system for teaching files in colleges and universities, and the electronic auxiliary sorting system for teaching files in colleges and universities includes:

[0015] A multi-source data collection terminal for collecting teaching data and environmental data;

[0016] A behavior fusion model module for generating multi-dimensional behavior features through a CRNN network model and teaching data and constructing a spatio-temporal behavior graph, and generating a behavior dependence relationship of teacher-student interaction in the classroom through the spatio-temporal behavior graph;

[0017] A dynamic cloud storage module for synchronizing teaching data and environmental data in real time and allocating storage resources according to a preset;

[0018] An automated teaching evaluation module for generating behavior diversity evaluation, comprehensive learning trend prediction and teaching anomaly feedback through spatio-temporal behavior graphs at different time nodes;

[0019] One end of the behavior fusion model module is connected to the multi-source data collection terminal, and the other end is connected to the automated teaching evaluation module through the dynamic cloud storage module.

[0020] In the above aspect and any possible implementation manner, a further implementation manner is provided. The multi-source data collection terminal includes a multi-spectral video collection device, an audio collection device and an environmental monitoring device, and the multi-spectral video collection device, the audio collection device and the environmental monitoring device are all arranged in the teaching area.

[0021] In the above aspect and any possible implementation manner, a further implementation manner is provided. The teaching data includes classroom voice and student action posture data, and the environmental data is environmental parameters in the teaching area, and the environmental parameters include noise level, temperature and humidity.

[0022] In the above aspect and any possible implementation manner, a further implementation manner is provided. The classroom voice is obtained through the audio collection device, the student action posture data is obtained through the combination of visible light and infrared light by the multi-spectral video collection device, and the environmental parameters in the teaching area are obtained through the environmental monitoring device.

[0023] For the aspects and any possible implementation manners described above, a further implementation manner is provided, in which the multi-dimensional behavior feature generation process is obtained by fusing the dynamic behavior data of students with static features through a hybrid convolutional-recursive neural network model.

[0024] For the aspects and any possible implementation manners described above, a further implementation manner is provided, in which after the behavior fusion model module generates the behavior dependency relationships of the teacher-student interaction in the classroom, the accuracy and robustness of the behavior feature recognition are enhanced through an optimization algorithm of contrastive learning.

[0025] For the aspects and any possible implementation manners described above, a further implementation manner is provided, in which the real-time synchronization of the classroom-side environment data and behavior data in the dynamic cloud storage module has a latency of less than 30 milliseconds, and the storage resource allocation is based on the behavior frequency and importance.

[0026] For the aspects and any possible implementation manners described above, a further implementation manner is provided, in which the behavior diversity evaluation, comprehensive learning trend prediction, and teaching anomaly feedback are specifically as follows:

[0027] Behavior diversity evaluation: Generate a score based on the range of student behavior changes and interaction participation.

[0028] Comprehensive learning trend prediction: Generate a personalized learning assessment report for students by combining their long-term classroom performance.

[0029] Teaching anomaly feedback: Generate suggestions for optimizing teaching strategies by detecting behavior anomalies in real time.

[0030] For the aspects and any possible implementation manners described above, a further implementation manner is provided, in which the electronic auxiliary sorting system for college course teaching archives further includes a cloud collaboration server, and the cloud collaboration server includes a teaching plan module, a behavior analysis module, a dynamic evaluation module, and an anomaly handling module, and the teaching plan module, the behavior analysis module, the dynamic evaluation module, and the anomaly handling module are all connected to the dynamic cloud storage module.

[0031] For the aspects and any possible implementation manners described above, a further implementation manner is provided for an electronic auxiliary sorting method for college course teaching archives, which is completed through the electronic auxiliary sorting system for college course teaching archives described above. The electronic auxiliary sorting method for college course teaching archives includes the following steps:

[0032] S1: Collect teaching data and environmental data;

[0033] S2: Generate multi-dimensional behavior features through the teaching data and construct a spatio-temporal behavior graph, and generate the behavior dependency relationships of the teacher-student interaction in the classroom through the spatio-temporal behavior graph.

[0034] S2: Synchronize teaching data and environmental data in real time and allocate storage resources according to preset settings;

[0035] S4: Generate behavior diversity evaluation, comprehensive learning trend prediction, and teaching anomaly feedback through spatio-temporal behavior graphs at different time nodes.

[0036] Compared with the prior art, the present invention can achieve the following technical effects:

[0037] 1. Excellent multi-modal data fusion performance: Through CRNN and spatio-temporal graph modeling, the accuracy of behavior recognition is increased to over 95%;

[0038] 2. Significantly improved storage efficiency: By adopting a dynamic partitioning strategy, the storage cost is reduced by about 30%, and the data access latency is shortened to within 20 milliseconds;

[0039] 3. Scientific and comprehensive evaluation indicators: New indicators such as behavior diversity index and participation contribution rate fill the gaps in the teaching evaluation system and provide more valuable evaluation results.

[0040] Of course, it is not necessary for any product implementing the present invention to achieve all of the above-mentioned technical effects simultaneously.

BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0042] Figure 1 It is a structural diagram of an electronic auxiliary sorting system for college course teaching archives provided by an embodiment of the present invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] In order to better understand the technical solutions of the present invention, the following will describe the embodiments of the present invention in detail with reference to the drawings.

[0044] It should be clear that the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0045] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "the", and "said" used in the embodiments of the present invention and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise.

[0046] As Figure 1 shown, the present invention provides an electronic auxiliary sorting system for college course teaching archives, and the electronic auxiliary sorting system for college course teaching archives includes:

[0047] A multi-source data acquisition terminal for acquiring teaching data and environmental data;

[0048] A behavior fusion model module for generating multi-dimensional behavior features through a CRNN network model and teaching data, constructing a spatio-temporal behavior graph, and generating a behavior dependence relationship of teacher-student interaction in the classroom through the spatio-temporal behavior graph;

[0049] A dynamic cloud storage module for synchronizing teaching data and environmental data in real time and allocating storage resources according to preset settings;

[0050] An automated teaching evaluation module for generating behavior diversity evaluation, comprehensive learning trend prediction, and teaching anomaly feedback through spatio-temporal behavior graphs at different time nodes;

[0051] One end of the behavior fusion model module is connected to the multi-source data acquisition terminal, and the other end is connected to the automated teaching evaluation module through the dynamic cloud storage module.

[0052] The multi-source data acquisition terminal includes a multi-spectral video acquisition device, an audio acquisition device, and an environmental monitoring device, and the multi-spectral video acquisition device, the audio acquisition device, and the environmental monitoring device are all arranged in the teaching area.

[0053] The teaching data includes classroom voice and student action posture data, and the environmental data is environmental parameters in the teaching area, and the environmental parameters include noise level, temperature, and humidity.

[0054] The classroom voice is acquired through the audio acquisition device, the student action posture data is acquired through the combination of visible light and infrared light by the multi-spectral video acquisition device, and the environmental parameters in the teaching area are acquired through the environmental monitoring device.

[0055] The process of generating the multi-dimensional behavior features is obtained by fusing the dynamic behavior data and static features of students through a hybrid convolutional-recursive neural network model.

[0056] After generating the behavior dependence relationship of teacher-student interaction in the classroom, the behavior fusion model module also enhances the accuracy and robustness of behavior feature recognition through an optimization algorithm of contrastive learning.

[0057] The real-time synchronization of classroom end environmental data and behavior data in the dynamic cloud storage module has a delay of less than 30 milliseconds, and the storage resource allocation is based on behavior frequency and importance.

[0058] The behavior diversity evaluation, comprehensive learning trend prediction, and teaching anomaly feedback are specifically as follows:

[0059] Behavior diversity evaluation: Generate a score based on the range of students' behavior changes and their interaction participation;

[0060] Comprehensive learning trend prediction: Generate a personalized learning assessment report for students by combining their long-term classroom performance;

[0061] Teaching anomaly feedback: Generate suggestions for optimizing teaching strategies by detecting behavior anomalies in real time.

[0062] The electronic auxiliary sorting system for college course teaching archives further includes a cloud collaboration server, which includes a teaching plan module, a behavior analysis module, a dynamic evaluation module, and an anomaly handling module. The teaching plan module, behavior analysis module, dynamic evaluation module, and anomaly handling module are all connected to the dynamic cloud storage module.

[0063] The present invention also provides a method for electronically assisting in sorting college course teaching archives, which is completed through the above-mentioned electronic auxiliary sorting system for college course teaching archives. The method for electronically assisting in sorting college course teaching archives includes the following steps:

[0064] S1: Collect teaching data and environmental data;

[0065] S2: Generate multi-dimensional behavior features from the teaching data and construct a spatio-temporal behavior graph, and generate the behavior dependence relationship of teacher-student interaction in the classroom through the spatio-temporal behavior graph;

[0066] S2: Synchronize the teaching data and environmental data in real time and allocate storage resources according to the preset;

[0067] S4: Generate behavior diversity evaluation, comprehensive learning trend prediction, and teaching anomaly feedback through the spatio-temporal behavior graphs at different time nodes.

[0068] In a specific embodiment, an electronic auxiliary sorting system for college course teaching archives includes:

[0069] A cloud collaboration server for running software for electronically assisting in sorting college course teaching archives. The software includes a teaching plan module, a behavior analysis module, a dynamic evaluation module, and an anomaly handling module;

[0070] A multi-source data acquisition terminal, including:

[0071] A multi-spectral video acquisition device that combines visible light and infrared light to collect students' action and posture data;

[0072] An audio acquisition device for recording classroom voices;

[0073] An environmental monitoring device for collecting classroom environmental parameters (including noise level, temperature, and humidity);

[0074] A behavior fusion model module implemented based on the following algorithm:

[0075] Adopt a hybrid convolutional-recursive neural network (CRNN) model to fuse students' dynamic behavior data with static features to generate multi-dimensional behavior feature vectors;

[0076] Construct a specific spatio-temporal behavior graph to capture the behavior dependence relationship of teacher-student interaction in the classroom;

[0077] Use an optimization algorithm based on contrast learning to enhance the accuracy and robustness of behavior feature recognition;

[0078] A dynamic cloud storage module that supports:

[0079] Real-time synchronization of classroom-side environmental data and behavior data with a latency of less than 30 milliseconds;

[0080] A dynamic storage partition that preferentially allocates high-performance storage resources according to behavior frequency and importance;

[0081] An automated teaching evaluation module, including:

[0082] Behavior diversity scoring, generating a score based on the range of students' behavior changes and interaction participation;

[0083] Comprehensive learning trend prediction, generating a personalized learning assessment report for students by combining long-term classroom performance;

[0084] A teaching anomaly feedback module that generates suggestions for optimizing teaching strategies by detecting behavior anomalies in real time.

[0085] In a specific embodiment, the behavior fusion model module processes data through the following steps:

[0086] Convert the behavior trajectory collected by the video into a spatio-temporal graph structure, define nodes to represent students' actions, and edges to represent the dependence relationship between behaviors;

[0087] Adopt a dynamic weighting mechanism to classify students' behaviors, with the weight range from 0.1 to 1.0, and dynamically adjust according to behavior frequency and influence weight;

[0088] Optimize the model robustness through the contrast learning algorithm, and have the ability to adapt to low-resolution videos and data noise.

[0089] In a specific embodiment, the automated teaching evaluation module generates the following quantitative indicators based on the multi-dimensional behavior feature vectors:

[0090] Behavior diversity index (range from 0 to 10), measuring the richness of students' classroom behaviors;

[0091] The classroom participation contribution rate, calculated based on the behavior frequency and quality of each student, ranges from 0 to 100%;

[0092] The long-term learning trend chart predicts the future learning performance change trend based on a time series model.

[0093] In a specific embodiment, classroom behavior data and environmental data are collected through a multi-source data collection terminal. The data includes:

[0094] The student movement trajectory (frame rate is 30 to 60 frames per second);

[0095] The classroom noise level (decibel range is 30 to 90);

[0096] The dynamic adjustment range of the video resolution is from 720P to 1080P;

[0097] The collected data is transmitted to the cloud collaboration server, and the behavior fusion model module processes the data to generate a spatio-temporal behavior map;

[0098] The automated teaching evaluation module generates a student performance score by combining the behavior diversity index and the classroom participation contribution rate;

[0099] The dynamic cloud storage module allocates storage resources according to the data importance and triggers policy optimization suggestions during behavior anomaly detection;

[0100] Finally, a course summary report and a student personalized learning assessment report are generated and archived to the cloud.

[0101] The overall system architecture of the present invention is as follows:

[0102] Cloud collaboration server: Runs the software for electronically assisting in organizing college course teaching archives, supporting data processing, dynamic storage, and teaching evaluation;

[0103] Multi-source data collection terminal: Integrates multi-spectral video, audio, and environmental monitoring devices to collect behavior data and environmental data;

[0104] Behavior fusion model module: Combines convolutional neural network (CNN), recurrent neural network (RNN), and spatio-temporal behavior map to analyze behavior characteristics in real time;

[0105] Dynamic cloud storage module: Adopts an intelligent partitioning and priority storage mechanism to optimize data transmission and access efficiency;

[0106] Automated teaching evaluation module: Generates a teaching evaluation report based on dimensions such as behavior diversity index and participation contribution rate;

[0107] Teaching anomaly feedback module: Real-time detects classroom abnormal behaviors and provides suggestions for optimizing teaching strategies.

[0108] The main inventive points of the present invention include:

[0109] Behavior fusion model design:

[0110] Use spatio-temporal behavior graphs to capture the interaction relationships among students;

[0111] Enhance the robustness to low-quality data through contrastive learning;

[0112] Use hierarchical feature extraction technology to combine dynamic behaviors and static features.

[0113] Dynamic storage strategy:

[0114] The data partitioning mechanism classifies and stores data according to behavior frequency and importance;

[0115] Hot data is preferentially stored on high-performance nodes, and historical data is transferred to low-cost storage.

[0116] Evaluation index design:

[0117] Behavior diversity index: Quantify the richness of students' classroom behaviors based on the types and frequencies of behaviors;

[0118] Participation contribution rate: Calculate students' classroom performance by integrating behavior frequency and interaction quality;

[0119] Long-term trend analysis: Predict students' future performance through time series models.

[0120] Example 1: Collection and behavior analysis based on single classroom data

[0121] Step 1: Data collection

[0122] Hardware preparation: Install a multi-spectral video acquisition device at the center of the classroom ceiling, and adjust the angle to cover all seat areas; install an audio acquisition device (with a sensitivity higher than -38 dB) near the podium to capture clear classroom voices; install an environmental monitoring device in the corner of the classroom to monitor the noise level, temperature (20°C to 30°C), and humidity (40% to 60%) in real time.

[0123] Data collection settings: Set the frame rate to 30 frames per second, the video resolution to 1080P, the audio sampling rate to 16 kHz through the acquisition terminal, and the environmental monitoring device records data every 5 seconds.

[0124] Data collection task: Record the complete data of a 45-minute course, including students' actions, classroom voices, and environmental parameters.

[0125] Step 2: Data processing and analysis

[0126] Data upload: The acquisition terminal uploads the acquired data to the cloud collaboration server in real time through the WiFi6 transmission module, with the delay controlled within 30 milliseconds.

[0127] Behavior trajectory modeling: The behavior fusion model module receives video data and converts the student action trajectory into a spatio-temporal behavior graph.

[0128] Node definition: Each node represents a specific action of a student, such as raising a hand or lowering the head to write.

[0129] Edge definition: Adjacent nodes represent the time-dependent relationship of consecutive actions.

[0130] Weight assignment: The behavior frequency weight is set between 0.1 and 1.0, and higher weights are assigned to important behaviors (such as interacting with the teacher).

[0131] Optimization algorithm processing: The data robustness is enhanced through the contrast learning algorithm, especially the ability to analyze behaviors in low-light and high-noise scenarios.

[0132] Step 3: Automatically generate a behavior analysis report

[0133] The generated report includes:

[0134] The classroom behavior feature vector of each student;

[0135] Behavior diversity index (5.8 / 10, relatively high);

[0136] Classroom participation contribution rate (78%).

[0137] Example 2: Intelligent partitioning and transmission optimization of dynamic cloud storage

[0138] Step 1: Set the data storage strategy

[0139] Intelligent partitioning rule: Mark the classroom behavior data (such as the frequency of students raising their hands and the number of classroom discussions) as "high-frequency updated data" and store it preferentially on high-performance nodes with a storage period of 180 days; mark the environmental data (such as the noise level) as "low-frequency updated data" and store it in the low-cost cloud storage area with a storage period of 30 days.

[0140] Bandwidth optimization strategy: Dynamically allocate bandwidth according to the real-time requirements of the classroom to ensure that the transmission delay of classroom behavior data is controlled within 20 milliseconds.

[0141] Step 2: Data transmission and storage

[0142] The multi-spectral video data (10GB / hour) is uploaded to the cloud storage through the edge cache node, and the cache strategy preferentially processes the behavior data;

[0143] Environmental monitoring data (5KB / minute) is directly transmitted to the low-priority storage area.

[0144] Step 3: Cloud data access and feedback

[0145] The teacher accesses the statistical results of the behavior data through the management interface, and the loading time is less than 2 seconds;

[0146] Environmental data is automatically archived for subsequent generation of long-term trend charts.

[0147] Example 3: Comprehensive teaching evaluation based on multi-dimensional data

[0148] Step 1: Behavior diversity scoring

[0149] Upload the behavior trajectory data collected in the classroom to the cloud collaboration server;

[0150] The system analyzes the behavior types and change ranges of each student through the spatio-temporal behavior map, for example:

[0151] Student A: The behavior types include raising hands, lowering the head to write, and classroom discussion, a total of 3 types;

[0152] Student B: The behavior types include raising the head and lowering the head to write, a total of 2 types;

[0153] Student C: There is no obvious behavior change, only listening.

[0154] The system generates behavior diversity scores (7.5 / 10, 5.0 / 10, 2.0 / 10) for students A, B, and C respectively.

[0155] Step 2: Classroom participation contribution rate

[0156] Based on the classroom interaction data, the system calculates the interaction times and quality of each student:

[0157] Student A: Interacts 5 times, of which 3 times are high-quality interactions (asking questions and participating in discussions);

[0158] Student B: Interacts 2 times, both of which are low-quality interactions;

[0159] Student C: There is no interaction record.

[0160] The system generates a contribution rate based on the weights (high-quality interaction weight 0.8, low-quality interaction weight 0.3):

[0161] Student A: Contribution rate 85%;

[0162] Student B: Contribution rate 40%;

[0163] Student C: Contribution rate 0%.

[0164] Step 3: Learning trend prediction

[0165] Generate a trend chart of classroom performance for the next 3 months based on the time series analysis of students' behavior data. Example 4: Abnormal behavior detection and strategy optimization

[0166] Step 1: Abnormal behavior detection

[0167] The system monitors students' classroom behavior in real time and detects abnormal events:

[0168] Student A: Frequently leaves the seat (5 times per class);

[0169] Student B: No movement for a long time (lasting 20 minutes).

[0170] The abnormal detection model automatically marks the time period when abnormal behavior occurs and generates a warning.

[0171] Step 2: Suggestions for strategy optimization

[0172] For Student A, it is recommended to adjust the seat arrangement or provide activities with higher participation;

[0173] For Student B, it is recommended to add interactive links in the course to enhance the sense of classroom participation.

[0174] Comparative example and effect summary

[0175]

[0176] Through the above embodiments, it can be seen that the performance of the present invention in behavior recognition, data management and teaching evaluation is significantly better than that of traditional systems, and at the same time provides a novel teaching strategy optimization ability.

[0177] Part of the working principle:

[0178] The electronic auxiliary sorting system for college course teaching files of the present invention realizes multi-dimensional data collection, processing, storage and analysis of the classroom teaching process through the close cooperation of a series of hardware and software components, and then provides an automated teaching evaluation and feedback mechanism. The main components of the system include: cloud collaboration server, multi-source data collection terminal, behavior fusion model module, dynamic cloud storage module, automated teaching evaluation module and teaching anomaly feedback module. Next, the working principle of the present invention and the technical means and parameters of each step will be described in detail.

[0179] System structure and component functions

[0180] 1. Cloud collaboration server

[0181] The cloud collaboration server is the core control center of the present invention, responsible for executing all data processing, behavior analysis, storage management, and teaching evaluation tasks. The server is connected to all other modules via the Internet and a high-speed transmission network. The main functions of the server include:

[0182] Data processing: Receive and process video, audio, and environmental data transmitted from the acquisition terminals.

[0183] Behavior analysis and fusion: Analyze the behavior of students in the classroom through the behavior fusion model module to generate behavior feature vectors.

[0184] Dynamic storage management: Use the dynamic cloud storage module to store and manage teaching data to ensure real-time updates during the teaching process.

[0185] Teaching evaluation and feedback: The automated teaching evaluation module generates a teaching evaluation report based on student behavior data and generates a personalized learning assessment report for students in combination with behavior diversity, classroom participation, etc.

[0186] Teaching anomaly feedback: When the system detects abnormal classroom behaviors (such as frequent leaving of seats, long periods of no interaction, etc.), it will generate policy optimization suggestions through the teaching anomaly feedback module.

[0187] 2. Multi-source data acquisition terminals

[0188] The multi-source data acquisition terminals consist of the following three parts:

[0189] Multi-spectral video acquisition device: Combine visible light and infrared light to collect the action and posture data of students, ensuring accurate identification of student behavior even in low-light environments. The parameters of the video acquisition device are set as follows: the resolution range is from 720P to 1080P, and the frame rate is set to 30 to 60 frames per second to ensure clear images and the ability to capture the subtle movements of students.

[0190] Audio acquisition device: Used to record the voice content in the classroom, with a sampling rate of 16 kHz and a sensitivity of -38 dB, to capture the communication content between teachers and students and assist in behavior recognition.

[0191] Environmental monitoring device: Collect the temperature (20°C to 30°C), humidity (40% to 60%), and noise level (30 to 90 decibels) of the classroom environment. These data are used to evaluate the impact of the classroom environment on student behavior and learning and provide a basis for subsequent teaching strategy adjustments.

[0192] All the collected raw data is uploaded to the cloud collaboration server through the transmission protocol to ensure real-time and stable synchronization of the data.

[0193] 3. Behavior fusion model module

[0194] The behavior fusion model module is the core algorithm part of the present invention and processes data based on the following steps:

[0195] Convert the behavior trajectory into a spatio-temporal graph: Convert the student behavior trajectory collected by the video into a spatio-temporal graph structure. Each node represents a specific behavior of the student (such as raising a hand, answering a question, lowering the head to write, etc.), and the edges between the nodes represent the temporal dependence relationship between behaviors. This step adopts the combination of a convolutional neural network (CNN) and a recurrent neural network (RNN) (i.e., CRNN) to effectively extract the temporal features and spatial features of the behavior.

[0196] In the present invention, CRNN is an end-to-end training model. CNN extracts a feature sequence from the given picture, and RNN (bi-LSTM) makes predictions based on the feature sequence generated by the convolutional layer. CTC outputs the predicted character sequence of the RNN layer into labels. CNN+RNN+CTC is integrated into a complete network, and the entire network architecture is called CRNN (Convolution Recurrent Neural Network), which can be trained under a loss function. The specific feature extraction process is as follows: In CRNN, the convolutional component is composed of convolutional and max-pooling layers, that is, the fully connected layer is removed from the standard convolutional network. The role of the convolutional layer is to extract the feature sequence from the input image. Each feature vector in the feature sequence is generated column by column from left to right in the feature maps. That is to say, the i-th feature vector is composed of the feature vectors generated by the i-th column of all feature maps, and the width of each column is set to one pixel. Because of the translational invariance of the convolutional layer, max-pooling, and element-wise activation functions, each column of the feature maps corresponds to a rectangular region in the original image, and this rectangular region is called the receptive field. Moreover, this rectangular region is in the same order as the corresponding column from left to right in the feature maps. Then it can be considered that each feature vector in the feature sequence also corresponds to each receptive field in the original image.

[0197] Sequence labeling is specifically as follows: A bidirectional RNN is connected after the convolutional layer. The RNN predicts each feature sequence generated in the convolutional layer as a character sequence. The reasons for choosing the RNN are as follows: First, the RNN has a strong ability to capture the context information of a sequence. In the above feature extraction, it can be seen that a wide character may be described by several consecutive receptive fields. Using context for sequence recognition based on images is more effective than treating individual characters separately, and for some ambiguous characters, they can be well distinguished after observing their context information. Second, the RNN can also perform weight updates through back-propagation, enabling the connection of the CNN and the RNN into a complete network. Third, the RNN can process sequences of any length. In the above, the height of the input image is fixed, which is to fix the size of each receptive field (the width of each receptive field is one pixel, which is also fixed), and images of any width can be processed.

[0198] Network training is specifically as follows: Assume that the training data is X = {I i , l i}, i , where I i is the training image, l i is the true label sequence, y i is the sequence output after the input image I i passes through the CRNN network. The objective equation is to minimize the negative log-likelihood of the conditional probability.

[0199]

[0200] The CNN network draws on the VGG architecture. The height value of the input image is fixed at 32. SGD is used during network training, and Adadelta is used for optimization.

[0201] Applied to the present invention, behavior classification and weighting mechanism: Dynamically adjust the parameters of the classification system according to the frequency and influence weight of students' behaviors. The weight value range is from 0.1 to 1.0. The weights of behavior categories (such as raising hands, writing, talking, etc.) are dynamically adjusted. Frequent behaviors (such as asking questions and interacting with teachers) are assigned higher weights, while low-frequency behaviors (such as occasional walking) are assigned lower weights.

[0202] Contrastive learning optimization algorithm: Based on the contrastive learning algorithm, optimize low-resolution videos and noisy environmental data to improve the model's adaptability to complex data. Optimize feature recognition through the backpropagation algorithm, enabling the model to better recognize the diverse behaviors of students in the classroom.

[0203] 4. Dynamic Cloud Storage Module

[0204] The dynamic cloud storage module provides data storage management functions to ensure the efficiency and scalability of the system. The main functions include:

[0205] Intelligent Partitioning Mechanism: Dynamically store data in partitions according to the priority of the data. Classroom behavior data (such as the number of student interactions and the frequency of raising hands) belongs to high-frequency updated data and is preferentially stored in high-performance nodes; environmental data (such as noise level and temperature) belongs to low-frequency updated data and is stored in low-cost cloud storage areas. The storage period for high-frequency data is 180 days, and for low-frequency data is 30 days.

[0206] Bandwidth Optimization Strategy: For real-time requirements, the system dynamically adjusts the cloud transmission bandwidth according to the network conditions and data update frequency to ensure that the synchronous transmission delay of high-frequency updated data is less than 50 milliseconds.

[0207] 5. Automated Teaching Evaluation Module

[0208] The automated teaching evaluation module generates teaching evaluations for students based on multi-dimensional behavior feature vectors. The main steps include:

[0209] Behavior Diversity Scoring: Generate a behavior diversity score by analyzing the different types and ranges of behaviors demonstrated by students in the classroom. The scoring range is from 0 to 10. The more types of behaviors and the greater the variation, the higher the score.

[0210] Classroom Participation Contribution Rate: Calculate the participation contribution rate based on the frequency and quality of students' behaviors. The scoring range is from 0 to 100%. Behaviors with high frequency and high quality (such as frequent interactions and answering questions) will be given a higher contribution rate.

[0211] Long-term Learning Trend Prediction: Based on students' behavior data and classroom participation, the system uses a time series model to predict students' future learning trends, including their participation, learning effects, and subject development directions.

[0212] 6. Teaching Abnormality Feedback Module

[0213] The role of the teaching abnormality feedback module is to monitor classroom abnormal behaviors in real time and generate optimization strategies. For example: Abnormality Detection: When it is detected that a student frequently leaves their seat (more than 3 times per class) or has no interaction for a long time (more than 20 minutes), the system will mark the student as "abnormal".

[0214] Strategy Optimization Suggestions: For abnormal behaviors, the system will generate optimization suggestions through the teaching feedback module, such as adjusting the seat arrangement and providing more interaction opportunities.

[0215] Explanation of the function and specific steps of each technical means in the present invention

[0216] Multi-source data collection and transmission:

[0217] By collecting multi-spectral video, audio, and environmental data, comprehensive classroom data is provided. These data are comprehensively analyzed on the cloud collaboration server to ensure multi-dimensional teaching evaluation.

[0218] The collected data not only includes students' behaviors but also the impact of the classroom environment on learning, such as noise level and temperature, which may directly affect students' attention and learning effects.

[0219] Key role of the behavior fusion model module:

[0220] Using the hybrid convolutional-recursive neural network (CRNN) model, it can effectively fuse static and dynamic features, capture the temporal relationship of students' behaviors, and further improve the accuracy of behavior recognition.

[0221] The construction of the spatio-temporal graph and the contrast learning optimization algorithm solve the problems of low-resolution video and noise influence,

[0222] ensuring that the system can still maintain high accuracy even in complex environments.

[0223] Dynamic cloud storage module:

[0224] The dynamic storage partition strategy effectively reduces storage costs, preferentially stores frequently updated data, and reduces the storage burden on the system.

[0225] The bandwidth optimization strategy ensures the real-time nature of data. Especially when a large amount of data needs to be processed quickly during the teaching process, it can ensure seamless connection.

[0226] Innovation of the automated teaching evaluation module:

[0227] This module not only provides a behavior diversity score but also comprehensively considers the frequency and quality of students' behaviors, providing a more accurate classroom evaluation report for teachers.

[0228] The long-term learning trend prediction function combined with time series analysis can provide teachers with the future learning development direction of students, facilitating the implementation of personalized education.

[0229] Real-time feedback role of the teaching anomaly feedback module:

[0230] Through real-time monitoring, the system can timely detect students' abnormal behaviors, provide corresponding optimization suggestions, help teachers adjust teaching strategies, and ensure students' continuous participation and good learning effects.

[0231] As can be seen from the above embodiments, the present invention designs a new system integrating multi-source data collection, behavior analysis, dynamic storage, and teaching evaluation to solve the following problems: how to improve the accuracy and adaptability of classroom behavior analysis and enhance the robustness to dynamic scenarios; how to reduce storage costs and improve data access efficiency; how to design a multi-dimensional evaluation system to generate personalized suggestions by integrating behavior, homework, and exam data; how to assist classroom improvement through a teaching feedback mechanism and enhance teaching effects. The difficulties in solving the above problems are as follows: Behavior recognition needs to combine multi-modal data (video, audio, environmental data), which poses high requirements for model design, algorithm optimization, and hardware adaptation. In addition, achieving the real-time performance and economy of cloud storage under limited bandwidth and latency conditions is also a major challenge. The present invention solves the above problems and promotes the upgrade of college teaching management from "informatization" to "intelligence", having important academic value and application value.

[0232] The above has introduced in detail a system and method for electronically assisting the collation of college course teaching archives provided by the embodiments of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation of the present application.

[0233] As used in the specification and claims, certain terms are used to refer to specific components. Those skilled in the art should understand that hardware manufacturers may use different terms to refer to the same component. The specification and claims do not use the difference in names as a way to distinguish components, but use the difference in functions of components as the criterion for distinction. As mentioned throughout the specification and claims, "comprising" and "including" are open-ended terms, so they should be interpreted as "including / including but not limited to". "Substantially" means within an acceptable error range. Those skilled in the art can solve the technical problem within a certain error range and basically achieve the technical effect. The subsequent description in the specification is for the purpose of describing the preferred embodiments of the present application, but the description is for the purpose of explaining the general principles of the present application and is not intended to limit the scope of the present application. The protection scope of the present application shall be determined by the scope defined by the appended claims.

[0234] It should also be noted that the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, so that a commodity or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such a commodity or system. Without further limitation, the element defined by the statement "including one..." does not exclude the existence of another identical element in the commodity or system including the said element.

[0235] It should be understood that the term "and / or" used herein is merely a description of the associated relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: the existence of A alone, the simultaneous existence of A and B, and the existence of B alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0236] The above description shows and describes several preferred embodiments of the present application. However, as mentioned above, it should be understood that the present application is not limited to the form disclosed herein, should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be within the scope of the application concept described herein, through the above teachings or the technology or knowledge in the relevant field for modification. And any changes and modifications made by those skilled in the art that do not depart from the spirit and scope of the present application shall fall within the protection scope of the appended claims of the present application.

Claims

1. A system for electronic auxiliary arrangement of university course teaching archives, characterized in that: The electronic auxiliary arrangement system for university course teaching archives includes: Multi-source data acquisition terminal, used to collect teaching data and environmental data; The behavior fusion model module is used to generate multi-dimensional behavior features and construct spatiotemporal behavior graphs through the CRNN network model and teaching data, and to generate the behavior dependency relationship of teacher-student interaction in the classroom through the spatiotemporal behavior graphs; Dynamic cloud storage module, used to synchronize teaching data and environmental data in real time and allocate storage resources according to presets; Automated teaching evaluation module, which is used to generate behavior diversity evaluation, comprehensive learning trend prediction and teaching abnormality feedback through spatiotemporal behavior graphs at different time nodes; One end of the behavior fusion model module is connected to a multi-source data acquisition terminal, and the other end is connected to an automated teaching evaluation module through a dynamic cloud storage module.

2. The electronic auxiliary arrangement system for university course teaching archives according to claim 1 is characterized in that: The multi-source data acquisition terminal comprises a multi-spectral video acquisition device, an audio acquisition device and an environmental monitoring device, and the multi-spectral video acquisition device, the audio acquisition device and the environmental monitoring device are all arranged in the teaching area.

3. The electronic auxiliary arrangement system for university course teaching archives according to claim 2 is characterized in that: The teaching data includes classroom voice and student movement and posture data, and the environmental data is environmental parameters in the teaching area, and the environmental parameters include noise level, temperature and humidity.

4. The electronic auxiliary arrangement system for university course teaching archives according to claim 3 is characterized in that: The classroom speech is acquired through an audio acquisition device, the student's movement and posture data is acquired through a multi-spectral video acquisition device combined with visible light and infrared light acquisition, and the environmental parameters in the teaching area are acquired through an environmental monitoring device.

5. The electronic auxiliary arrangement system for university course teaching archives according to claim 1 is characterized in that: The multi-dimensional behavior feature generation process obtains the students' dynamic behavior data and static features by fusing them through a hybrid convolutional-recurrent neural network model.

6. The electronic auxiliary arrangement system for university course teaching archives according to claim 1 is characterized in that: After generating the behavioral dependency relationship of the teacher-student interaction in the classroom, the behavior fusion model module also enhances the accuracy and robustness of the behavior feature recognition through the optimization algorithm of contrastive learning.

7. The electronic auxiliary arrangement system for university course teaching archives according to claim 1 is characterized in that: The dynamic cloud storage module synchronizes classroom environment data and behavior data in real time with a delay of less than 30 milliseconds, and storage resources are allocated according to behavior frequency and importance.

8. The electronic auxiliary arrangement system for university course teaching archives according to claim 1 is characterized in that: The behavior diversity evaluation, comprehensive learning trend prediction and teaching abnormality feedback are specifically as follows: Behavioral diversity assessment: Generate scores based on the range of student behavior changes and interactive participation; Comprehensive learning trend prediction: Generate personalized learning assessment reports for students based on long-term classroom performance; Feedback on teaching anomalies: Generate suggestions for optimizing teaching strategies by detecting behavioral anomalies in real time.

9. The electronic auxiliary arrangement system for university course teaching archives according to claim 1 is characterized in that: The electronic auxiliary organization system for university course teaching archives also includes a cloud collaborative server, which includes a teaching plan module, a behavior analysis module, a dynamic evaluation module and an exception handling module. The teaching plan module, behavior analysis module, dynamic evaluation module and exception handling module are all connected to the dynamic cloud storage module.

10. A method for electronically assisting the arrangement of university course teaching archives, which is accomplished by the electronically assisting the arrangement of university course teaching archives as described in any one of claims 1 to 9, characterized in that: The electronic auxiliary arrangement method of the university course teaching archives comprises the following steps: S1: Collect teaching data and environmental data; S2: Generate multi-dimensional behavior characteristics through teaching data and construct spatiotemporal behavior graphs, and generate behavioral dependencies between teachers and students in the classroom through spatiotemporal behavior graphs; S2: synchronize teaching data and environmental data in real time and allocate storage resources according to preset settings; S4: Generate behavior diversity evaluation, comprehensive learning trend prediction, and teaching anomaly feedback through spatiotemporal behavior graphs at different time nodes.

Images