Classroom learning dynamic evaluation and management system

Abstract

The invention relates to the technical field of evaluation and management, in particular to a classroom learning dynamic evaluation and management system, which comprises a multi-source heterogeneous data acquisition module, an acoustic sensor array, an optical motion capture unit, an intelligent writing track acquisition device and a brain wave monitoring terminal, the data preprocessing and feature fusion module is used for eliminating environment interference signals and generating preprocessed data; the dynamic learning state evaluation module is used for extracting three-dimensional features and outputting a dynamic evaluation matrix; the group learning portrait generation module is used for generating a multi-dimensional learning portrait in combination with a preset knowledge graph topological structure; the self-adaptive management strategy generation module is used for dynamically generating a management strategy set; and the intelligent feedback execution module is used for driving the self-adaptive illumination system, the wearable vibration reminding device and the AR enhancement projection equipment to execute physical environment regulation and control. According to the invention, personalized strategy dynamic generation is realized, and adaptive regulation and control of the classroom environment and the teaching rhythm are realized in combination with the intelligent Internet of Things equipment, so that an intelligent collaborative feedback closed loop is formed.

Description

Technical Field

[0001] The present invention relates to the technical field of evaluation management, and in particular to a classroom learning dynamic evaluation and management system. Background Art

[0002] With the development of educational informatization, classroom teaching has gradually evolved from the traditional "one-way teaching mode by teachers" to diversified teaching modes such as "teacher-student interaction, student-student collaboration, and personalized guidance". Real-time learning status perception and dynamic teaching strategy adjustment in the teaching process have gradually become an important direction for improving the quality of classroom teaching. At this stage, some schools have tried to introduce equipment such as camera monitoring and answer feedback to record classroom behavior data or student answers. However, these single-mode data collection methods have limited information dimensions and most of them rely on teachers' manual analysis after class, which cannot meet the needs of real-time, dynamic, and all-round classroom status perception and immediate intervention.

[0003] The existing technology still has the following prominent problems: First, the synchronous collection and fusion processing capabilities of multimodal data are insufficient, especially for multi-source heterogeneous data such as sound, movement, writing, and EEG in classroom scenarios, which are affected by noise interference, sensor clock drift and other problems, making it difficult to achieve high-quality timing alignment and feature fusion; Second, the granularity of classroom status assessment is coarse, relying only on surface behavioral characteristics or static questionnaire data, and lacks deep cognitive modeling of learning status; Third, there is a lack of real-time intervention and environmental linkage mechanisms. Existing classroom management is more focused on post-class summary and experience feedback, and has failed to form a complete closed loop of real-time classroom evaluation-dynamic strategy generation-instant environmental linkage. In the face of the application needs of new smart classrooms, there is an urgent need for a classroom learning dynamic evaluation and management system driven by multi-source heterogeneous data to achieve intelligent upgrades throughout the process from multi-dimensional data perception, deep status assessment, personalized strategy generation, and intelligent environment linkage. Summary of the invention

[0004] Based on the above objectives, the present invention provides a classroom learning dynamic evaluation and management system.

[0005] A classroom learning dynamic evaluation and management system, including the following modules: Multi-source heterogeneous data acquisition module: equipped with distributed acoustic sensor array, optical motion capture unit, intelligent writing trajectory collector and brain wave monitoring terminal; Data preprocessing and feature fusion module: receiving the original data stream output by the multi-source heterogeneous data acquisition module, eliminating environmental interference signals through an adaptive noise reduction algorithm, and generating time-series synchronized preprocessed data using a multimodal feature alignment technology; Dynamic learning status evaluation module: receiving the preprocessed data, extracting the three-dimensional features of classroom concentration, knowledge mastery, and collaborative participation through a deep temporal convolutional network, and outputting a dynamic evaluation matrix; Group learning portrait generation module: receiving the dynamic evaluation matrix, combining the preset knowledge graph topology structure, and generating a multi-dimensional learning portrait including individual cognitive trajectory, group collaboration heat map, and knowledge gap distribution cloud map; Adaptive management strategy generation module: receiving the multi-dimensional learning portrait, dynamically generating a management strategy set including teaching rhythm control strategy, personalized intervention path, and group reorganization suggestion through reinforcement learning algorithm; Intelligent feedback execution module: receives the management strategy set, drives the adaptive lighting system, wearable vibration reminder device, and AR enhanced projection equipment to perform physical environment control.

[0006] Furthermore, the multi-source heterogeneous data acquisition module includes: Distributed acoustic sensor array: A distributed acoustic sensor array is composed of M×N micro-MEMS microphone units, which are arranged in a grid according to the classroom space coordinates. Each microphone unit records the sound pressure signal and calculates the sound arrival direction through the delay-sum beamforming algorithm; Optical motion capture unit: It uses an infrared stereo camera to form a spatial stereo capture array, and obtains the coordinates of key points of students' limbs in the classroom area according to the principle of binocular stereo vision. The trajectory of limb movements is tracked in real time through Kalman filtering; Intelligent writing trajectory collector: Based on the combination of smart pen and electronic writing board, it records the real-time writing position and writing pressure of each pen, and calculates the writing rhythm characteristics by combining curvature and speed; Brain wave monitoring terminal: record multi-channel EEG signals through wearable EEG acquisition equipment, and perform short-time Fourier transform on EEG signals to obtain time-frequency energy distribution; The mutual information matrix is ​​further calculated for each channel feature signal.

[0007] Furthermore, the data preprocessing and feature fusion module includes: A data stream receiving unit: receiving the sound pressure signal, body movement trajectory, writing trajectory signal, and EEG signal output by the multi-source heterogeneous data acquisition module respectively; Adaptive noise reduction unit: a multi-channel collaborative adaptive filtering algorithm is used for each channel data to eliminate environmental interference signals. The noise reduction process includes: constructing a collaborative filter group for each signal channel; Multimodal feature alignment unit: Based on the dynamic time warping algorithm, each modal signal is time-synchronized and aligned, and the distance matrix between each two modal signals is defined; Time-series synchronized data output unit: outputs time-series synchronized multi-modal pre-processed data sets.

[0008] Furthermore, the dynamic learning state evaluation module includes: Feature encoding unit: divides the acoustic signal, action signal, writing signal and EEG signal into time windows respectively, and extracts multimodal feature sequences; Temporal Convolution Analysis Unit: Based on a one-dimensional deep temporal convolutional network, it performs multi-scale time-aware analysis on multimodal feature sequences; Feature decoding unit: maps multimodal temporal features into moment-by-moment evaluation values ​​of classroom concentration, knowledge mastery, and collaborative participation; Dynamic evaluation matrix generation unit: concatenates the evaluation results of T time points to generate a dynamic evaluation matrix.

[0009] Furthermore, the group learning portrait generation module includes: Individual cognitive trajectory construction: Receive the dynamic evaluation matrix output by the dynamic learning state evaluation module, calculate the time-to-time change of the learning state of each individual in chronological order, and construct an individual cognitive trajectory that describes the individual cognitive evolution trend; Group collaboration heat map generation: Based on the collaboration participation of each student recorded in the dynamic evaluation matrix, combined with the preset student grouping information or seat arrangement information, the group learning portrait generation module constructs a group collaboration relationship matrix for each time frame, and generates a group collaboration heat map that reflects the distribution of individual collaboration intimacy throughout the class; Generation of knowledge gap distribution cloud map: Combined with the preset knowledge graph topology structure, the knowledge mastery of each student recorded in the dynamic evaluation matrix is ​​mapped one by one with the specific knowledge points in the knowledge graph to generate a knowledge gap distribution cloud map.

[0010] Furthermore, the individual cognitive trajectory construction includes: Calculate the change of three-dimensional learning state: Based on the dynamic evaluation matrix, calculate the change rate of three-dimensional learning state for each learning individual in chronological order; Construct individual cognitive trajectory vectors: based on the obtained three-dimensional learning state change rate.

[0011] Furthermore, the group collaboration heat map generation includes: Definition of group collaboration relationship matrix: Based on the dynamic interactive relationship between individual collaboration participation and the collaboration participation of all other members, define the group collaboration relationship matrix; Group collaboration heat map generation: Accumulate the collaboration intensity at all times to generate a group collaboration heat map.

[0012] Furthermore, the generation of the knowledge gap distribution cloud map includes: Average mastery of all students: Based on the knowledge mastery in the dynamic evaluation matrix and the preset knowledge graph topology, the average mastery of each knowledge point among all students is calculated; Assessment of mastery differences among the whole class: Calculate the standard deviation of the mastery of each knowledge point to assess the mastery differences among the whole class.

[0013] Furthermore, the adaptive management strategy generation module includes: Multi-dimensional learning portrait analysis: Receive the multi-dimensional learning portrait output by the group learning portrait generation module, and define the classroom state description vector based on the time series characteristics and spatial distribution characteristics of the multi-dimensional learning portrait; Reinforcement learning strategy training: The classroom state description vector is input into the strategy generation network based on deep reinforcement learning, and the teacher's executable management actions are output. The strategy generation network uses a joint optimization mode of temporal difference learning and policy gradient update to define the reward function of the strategy network; Management strategy set output: After reinforcement learning training converges, a dynamic management strategy set is generated in real time based on the current classroom status.

[0014] Furthermore, the adaptive management strategy generation module includes: Management policy receiving and parsing unit: receives the management policy set output by the adaptive management policy generation module, parses each policy action, and generates corresponding physical device control instructions; Physical environment control algorithm: Based on the received control instruction set, it performs dynamic control of the physical environment for each type of equipment; Multi-device linkage feedback mechanism: Perform closed-loop difference evaluation on the actual feedback parameters and target parameters of each device during the execution process.

[0015] Beneficial effects of the present invention: The classroom learning dynamic assessment and management system provided by the present invention breaks through the limitations of traditional classroom teaching management that relies on single behavior monitoring or after-class static assessment, and constructs a classroom teaching closed-loop management system from multimodal real-time perception, multi-dimensional dynamic assessment to strategy generation and environmental linkage feedback. The system can synchronously acquire multi-source heterogeneous data such as acoustic signals, motion trajectories, writing behaviors, and EEG signals, and through adaptive noise reduction and multimodal feature alignment processing, form a high-quality, time-synchronized classroom behavior and cognitive feature sequence; combined with deep temporal convolutional networks and knowledge graph analysis, it realizes individual cognitive trajectory reconstruction, group collaboration pattern recognition, and dynamic tracking of knowledge gaps, providing classroom teaching with full-process data-driven deep cognitive assessment capabilities, greatly improving the accuracy of state perception and the precision of assessment of classroom teaching.

[0016] The present invention realizes dynamic generation of personalized strategies through reinforcement learning algorithms, and combines intelligent IoT devices to achieve adaptive control of classroom environment and teaching rhythm, forming an intelligent collaborative feedback closed loop of teaching activities, student status, and physical environment. The system has the characteristics of strong real-time performance, short feedback path, and high autonomous adjustment ability. It not only helps teachers to grasp the differences in learning status of individuals and groups in a timely manner, but also can dynamically adjust teaching strategies according to real-time evaluation results, promote the transformation of classroom management from experience-driven to data-driven, and provide technical support for building a flexible, efficient, and individualized smart classroom. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 is a system module diagram of an embodiment of the present invention; Figure 2 This is a state assessment diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0020] like Figure 1-Figure 2 As shown, a classroom learning dynamic evaluation and management system includes the following modules: Multi-source heterogeneous data acquisition module: equipped with distributed acoustic sensor array, optical motion capture unit, intelligent writing trajectory collector and brain wave monitoring terminal; Data preprocessing and feature fusion module: receives the original data stream output by the multi-source heterogeneous data acquisition module, eliminates environmental interference signals through an adaptive noise reduction algorithm, and uses multimodal feature alignment technology to generate time-series synchronized preprocessed data; Dynamic learning status evaluation module: receives preprocessed data, extracts three-dimensional features of classroom concentration, knowledge mastery, and collaborative participation through a deep temporal convolutional network, and outputs a dynamic evaluation matrix; Group learning portrait generation module: receives the dynamic evaluation matrix, combines it with the preset knowledge graph topology structure, and generates a multi-dimensional learning portrait including individual cognitive trajectory, group collaboration heat map, and knowledge gap distribution cloud map; Adaptive management strategy generation module: receives multi-dimensional learning portraits and dynamically generates a management strategy set including teaching rhythm control strategies, personalized intervention paths, and group reorganization suggestions through reinforcement learning algorithms; Intelligent feedback execution module: receives the management strategy set and drives the adaptive lighting system, wearable vibration reminder device, and AR enhanced projection equipment to perform physical environment control.

[0021] The multi-source heterogeneous data acquisition module includes: Distributed acoustic sensor array: A distributed acoustic sensor array consisting of M×N micro-MEMS microphone units is arranged in a grid according to the classroom space coordinates. Each microphone unit records the sound pressure signal separately. , where i and j represent the transverse and longitudinal numbers of the sensors, respectively. The direction of arrival (DOA) of the sound is calculated by the delay-sum beamforming algorithm, which is expressed as: ; in, , represents the spatial distance between the sensor and the reference point, c is the speed of sound, is the horizontal angle, is the pitch angle, is the output signal after beamforming, is the sound pressure signal recorded by the (i, j)th microphone, M is the number of microphones in the horizontal direction, and N is the number of microphones in the vertical direction. is the propagation delay of the sound wave from the sound source to the (i, j)th microphone; Optical motion capture unit: uses an infrared stereo camera to form a spatial stereo capture array, and obtains the coordinates of key points of students' limbs in the classroom area according to the principle of binocular stereo vision. , where k is the key point number. The limb motion trajectory is tracked in real time through Kalman filtering, and its state prediction and update are expressed as: ; ; in, is the state vector, including position and velocity components, is the state vector at time t-1, For external control input, , are process noise and measurement noise respectively, A, B, C are state transfer matrix, control matrix and observation matrix respectively, is the observed position vector at time t; Intelligent writing track collector: Based on the combination of smart pen and electronic writing board, it records the real-time writing position of each pen and writing pressure , the writing rhythm characteristics are calculated by combining curvature and speed, expressed as: ; ; in, Write the instantaneous curvature of the trajectory at time t, is the modulus of writing speed at time t, is the coordinate of the pen tip in the X direction of the writing plane at time t, is the coordinate of the pen tip in the Y direction of the writing plane at time t, is the velocity in the X direction at time t, is the velocity in the Y direction at time t, is the acceleration in the X direction at time t, is the acceleration in the Y direction at time t; Brainwave monitoring terminal: record multi-channel EEG signals through wearable EEG acquisition equipment , where i is the electrode channel number, the EEG signal is subjected to short-time Fourier transform (STFT) to obtain the time-frequency energy distribution, which is expressed as: ; in, is a short-time window function, f is frequency, t is time, is the original EEG signal of the i-th channel, is the time-frequency energy distribution of the EEG signal of the i-th channel, is an imaginary unit, is the time-integrated variable; The mutual information matrix of each channel characteristic signal is further calculated, which is expressed as: ; in, is the joint probability distribution of channel i and j signals; , They are the single channel edge distribution, is the mutual information value of channels i and j, and k is the quantization level number (the number after the EEG amplitude is discretized).

[0022] The data preprocessing and feature fusion modules include: Data stream receiving unit: receives the sound pressure signals output by multi-source heterogeneous data acquisition modules respectively , body movement trajectory , writing track signal , EEG signals ; Adaptive noise reduction unit: Multi-channel collaborative adaptive filtering algorithm is used to eliminate environmental interference signals for each channel data. The noise reduction process includes: A collaborative filter bank consisting of K reference signal channels is constructed, and the target signal is expressed as: ; Among them, the filter coefficients Real-time update through the least mean square error (LMS) algorithm, , , is the signal value of the kth reference signal channel at time t, is the kth filter weight at time t, is the LMS algorithm step size factor, is the estimation error of the target signal at time t, is the sound pressure signal collected by the (i, j)th microphone at time t, is the three-dimensional coordinate of the kth limb key point at time t, is the coordinate of the writing trajectory position at time t , is the signal of the i-th EEG electrode channel at time t; Multimodal feature alignment unit: Based on the dynamic time warping (DTW) algorithm, each modal signal is time-synchronized and aligned, and the distance matrix between each two modal signals is defined, which is expressed as: ; in, , the final alignment path satisfy , is the mode X in The feature vector of a time slice, is the mode Y in the The feature vector of a time slice, Is the modal X Frame and Mode Y The local distance of the frame, Is the modal X Frame and Mode Y The local distance of the frame; Time-series synchronized data output unit: Outputs time-series synchronized multi-modal pre-processed data set, expressed as: ; in, is the total number of sample frames.

[0023] The dynamic learning status assessment module includes: Feature encoding unit: divide the acoustic signal, action signal, writing signal and EEG signal into time windows, and the length of each window is , and extract the multimodal feature sequence, expressed as: ; ; ; ; Where N is the number of windows, , , , They represent the acoustic features, action features, writing features and EEG features in the nth time window respectively. Each feature is extracted through standard statistics and frequency domain analysis and is expressed as: ; ; ; ; Among them, E represents the mean, Z represents the kurtosis, represents standard deviation, H represents spectral entropy; Temporal convolution analysis unit: Based on a one-dimensional deep temporal convolutional network (TCN), it performs multi-scale time-aware analysis on multimodal feature sequences. The core calculation of TCN includes: For each layer l, it is expressed as: ; in, ; represents the feature sequence output by the lth layer, represents the feature sequence of the previous layer, is the convolution kernel of the lth layer, is the bias, K is the length of the convolution kernel, and d is the expansion coefficient, which controls the receptive field; Feature decoding unit: Mapping multimodal temporal features into classroom concentration , Knowledge Mastery , Collaborative participation The moment-by-moment evaluation value of is expressed as: ; in, represents the feature vector of the last layer of TCN at time t, V is the linear mapping matrix, b is the bias, is the normalization function, is the classroom concentration at time t, is the knowledge mastery at time t, is the degree of collaboration participation at time t; Dynamic evaluation matrix generation unit: The evaluation results of T time points are concatenated to generate a dynamic evaluation matrix, which is expressed as: ; M is the dynamic evaluation matrix of classroom learning, which is passed to the group learning portrait generation module, and T is the total number of frames of evaluation time.

[0024] The group learning portrait generation module includes: Individual cognitive trajectory construction: Receive the dynamic evaluation matrix output by the dynamic learning state evaluation module, calculate the time-to-time change of the learning state of each individual in chronological order, and construct an individual cognitive trajectory that describes the individual cognitive evolution trend; Group collaboration heat map generation: Based on the collaboration participation of each student recorded in the dynamic evaluation matrix, combined with the preset student grouping information or seat arrangement information, the group learning portrait generation module constructs a group collaboration relationship matrix for each time frame, and generates a group collaboration heat map that reflects the distribution of individual collaboration intimacy throughout the class; Generation of knowledge gap distribution cloud map: Combined with the preset knowledge graph topology structure, the knowledge mastery of each student recorded in the dynamic evaluation matrix is ​​mapped one by one with the specific knowledge points in the knowledge graph to generate a knowledge gap distribution cloud map.

[0025] The construction of individual cognitive trajectory includes: Calculate the change of three-dimensional learning state: Based on the dynamic evaluation matrix, the change rate of the three-dimensional learning state is calculated for each learning individual in chronological order, which is expressed as: ; ; ; in, is the change in classroom concentration at time t, is the change in knowledge mastery at time t, is the change in collaborative participation at time t; Construct individual cognitive trajectory vector: Based on the obtained three-dimensional learning state change rate, it is expressed as: ; in, is the cognitive trajectory sequence of the sth learning individual.

[0026] Group collaboration heat map generation includes: Group collaboration relationship matrix definition: based on individual collaboration participation The dynamic interactive relationship with all other members’ collaborative participation defines the group collaborative relationship matrix, which is expressed as: ; in, , is the group collaboration relationship matrix at time t, represents the ratio of the collaboration intensity between the sth student and the dth student at time t, N is the total number of students in the current class, is the degree of individual collaborative participation, is the degree of collaborative participation of others, is the sum of the squares of the collaboration intensity of the whole class; Group collaboration heat map generation: Accumulate the collaboration intensity at all times to generate a group collaboration heat map, expressed as: ; Among them, H is an N×N collaboration heat map, which reflects the distribution of collaboration intimacy between individuals during class, and T is the total number of evaluation time frames.

[0027] The generation of knowledge gap distribution cloud map includes: Average mastery of all students: based on knowledge mastery in the dynamic assessment matrix , combined with the preset knowledge graph topology structure G=(V,E), calculate the average mastery of each knowledge point among all students, expressed as: ; in, is a collection of knowledge points, is the set of knowledge point association relationships, G=(V,E) is the knowledge graph, and N is the total number of students in the current class; Assessment of mastery differences among the whole class: Calculate the standard deviation of the mastery of each knowledge point and assess the mastery differences among the whole class, expressed as: ; in, Indicates the sth student's understanding of the knowledge point The mastery of is the mastery standard deviation, It is a knowledge point Average mastery of Generating a knowledge gap distribution cloud map includes: Average mastery is the color of the cloud, with standard deviation To increase the transparency of the cloud graph, the node connections in the cloud graph are determined by the knowledge graph topology structure E.

[0028] The adaptive management strategy generation module includes: Multi-dimensional learning portrait analysis: Receive the multi-dimensional learning portrait output by the group learning portrait generation module, and define the classroom state description vector based on the time series characteristics and spatial distribution characteristics of the multi-dimensional learning portrait, which is expressed as: ; in, is the classroom status at time t, is the three-dimensional cognitive trajectory point of the s-th student, is the collaboration intensity between students s and d, and are the average mastery and standard deviation of the kth knowledge point respectively; Reinforcement learning strategy training: Describe the classroom state vector Input is a deep reinforcement learning based policy generation network, with teacher executable management actions For output, Adjust the speed of explanation, insert group discussions, add real-time questions, call on individual students, and adjust group members , the policy generation network uses the joint optimization mode of temporal difference learning and policy gradient update to define the reward function of the policy network, which is expressed as: ; Where f is the weighted evaluation function, , is the classroom management reward value at time t, and the policy network parameter The update rule is: ; in, , For the strategy network, is the learning rate, is the discount factor, is the state value function, is the value function of the classroom state at time t, It is the classroom management action generated by the system at time t; Management strategy set output: After reinforcement learning training converges, according to the current classroom status , a dynamic management policy set is generated in real time, expressed as: ; in, The management strategy combination recommended at time t includes teaching rhythm control strategy, personalized intervention path, and group reorganization suggestion. is the nth specific strategy action.

[0029] The adaptive management strategy generation module includes: Management strategy receiving and parsing unit: receives the management strategy set output by the adaptive management strategy generation module , parse each policy action and generate corresponding physical device control instructions , where the format of each control instruction is expressed as: ; Among them, target indicates the target device type, including lighting system, vibration reminder device, AR enhanced projection device, type indicates the control category, including brightness adjustment, frequency adjustment, image content update, and value indicates the specific control parameter value; Physical environment control algorithm: based on the received control instruction set , respectively, to perform dynamic control of the physical environment for various types of equipment, including: (1) Adaptive lighting system: based on the overall concentration of the class Dynamically adjust the illumination L, expressed as: ; in, , is the initial base illumination, is the illumination adjustment gain coefficient, is the historical average concentration; (2) Wearable vibration reminder device: For individuals whose concentration has significantly decreased, a vibration reminder signal is triggered. The frequency F of the reminder signal is proportional to the decrease in concentration, expressed as: ; in, , As the basic reminder frequency, is the frequency adjustment gain coefficient, Trigger threshold for concentration reminder; (3) AR enhanced projection equipment: According to the group knowledge gap distribution cloud map, the current key knowledge points and weak points hot spots are dynamically projected. The generation rule of the projection content P is expressed as: ; That is, the knowledge point with the largest standard deviation of the current mastery distribution is projected , is the standard deviation of the mastery of the kth knowledge point; Multi-device linkage feedback mechanism: actual feedback parameters of each device during execution With target parameters Perform closed-loop difference evaluation, expressed as: ; when Exceeding the device dynamic stability threshold When the adaptive compensation is triggered Achieve full closed-loop adaptive control. is the control error of the sth device, is the feedback error compensation coefficient.

[0030] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present invention is limited to these examples. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity.

Claims

1. A classroom learning dynamic evaluation and management system, characterized in that: Includes the following modules: Multi-source heterogeneous data acquisition module: equipped with distributed acoustic sensor array, optical motion capture unit, intelligent writing trajectory collector and brain wave monitoring terminal; Data preprocessing and feature fusion module: receiving the original data stream output by the multi-source heterogeneous data acquisition module, eliminating environmental interference signals through an adaptive noise reduction algorithm, and generating time-series synchronized preprocessed data using a multimodal feature alignment technology; Dynamic learning status evaluation module: receiving the preprocessed data, extracting the three-dimensional features of classroom concentration, knowledge mastery, and collaborative participation through a deep temporal convolutional network, and outputting a dynamic evaluation matrix; Group learning portrait generation module: receiving the dynamic evaluation matrix, combining the preset knowledge graph topology structure, and generating a multi-dimensional learning portrait including individual cognitive trajectory, group collaboration heat map, and knowledge gap distribution cloud map; Adaptive management strategy generation module: receiving the multi-dimensional learning portrait, and dynamically generating a management strategy set including teaching rhythm control strategy, personalized intervention path, and group reorganization suggestion through reinforcement learning algorithm; Intelligent feedback execution module: receives the management strategy set, drives the adaptive lighting system, wearable vibration reminder device, and AR enhanced projection equipment to perform physical environment control.

2. A classroom learning dynamic evaluation and management system according to claim 1, characterized in that: The multi-source heterogeneous data acquisition module includes: Distributed acoustic sensor array: A distributed acoustic sensor array is composed of M×N micro-MEMS microphone units, which are arranged in a grid according to the classroom space coordinates. Each microphone unit records the sound pressure signal and calculates the sound arrival direction through the delay-sum beamforming algorithm; Optical motion capture unit: It uses an infrared stereo camera to form a spatial stereo capture array, and obtains the coordinates of key points of students' limbs in the classroom area according to the principle of binocular stereo vision. The trajectory of limb movements is tracked in real time through Kalman filtering; Intelligent writing trajectory collector: Based on the combination of smart pen and electronic writing board, it records the real-time writing position and writing pressure of each pen, and calculates the writing rhythm characteristics by combining curvature and speed; Brain wave monitoring terminal: multi-channel EEG signals are recorded through wearable EEG acquisition equipment, and the time-frequency energy distribution is obtained by short-time Fourier transform of EEG signals. The mutual information matrix of each channel characteristic signal is further calculated.

3. A classroom learning dynamic evaluation and management system according to claim 2, characterized in that: The data preprocessing and feature fusion module includes: A data stream receiving unit: receiving the sound pressure signal, body movement trajectory, writing trajectory signal, and EEG signal output by the multi-source heterogeneous data acquisition module respectively; Adaptive noise reduction unit: a multi-channel collaborative adaptive filtering algorithm is used for each channel data to eliminate environmental interference signals. The noise reduction process includes: constructing a collaborative filter group for each signal channel; Multimodal feature alignment unit: Based on the dynamic time warping algorithm, each modal signal is time-synchronized and aligned, and the distance matrix between each two modal signals is defined; Time-series synchronized data output unit: outputs time-series synchronized multi-modal pre-processed data sets.

4. A classroom learning dynamic evaluation and management system according to claim 3, characterized in that: The dynamic learning state evaluation module includes: Feature encoding unit: divides the acoustic signal, action signal, writing signal and EEG signal into time windows respectively, and extracts multimodal feature sequences; Temporal Convolution Analysis Unit: Based on a one-dimensional deep temporal convolutional network, it performs multi-scale time-aware analysis on multimodal feature sequences; Feature decoding unit: maps multimodal temporal features into moment-by-moment evaluation values ​​of classroom concentration, knowledge mastery, and collaborative participation; Dynamic evaluation matrix generation unit: concatenates the evaluation results of T time points to generate a dynamic evaluation matrix.

5. A classroom learning dynamic evaluation and management system according to claim 4, characterized in that: The group learning portrait generation module includes: Individual cognitive trajectory construction: Receive the dynamic evaluation matrix output by the dynamic learning state evaluation module, calculate the time-to-time change of the learning state of each individual in chronological order, and construct an individual cognitive trajectory that describes the individual cognitive evolution trend; Group collaboration heat map generation: Based on the collaboration participation of each student recorded in the dynamic evaluation matrix, combined with the preset student grouping information or seat arrangement information, the group learning portrait generation module constructs a group collaboration relationship matrix for each time frame, and generates a group collaboration heat map that reflects the distribution of individual collaboration intimacy throughout the class; Generation of knowledge gap distribution cloud map: Combined with the preset knowledge graph topology structure, the knowledge mastery of each student recorded in the dynamic evaluation matrix is ​​mapped one by one with the specific knowledge points in the knowledge graph to generate a knowledge gap distribution cloud map.

6. A classroom learning dynamic evaluation and management system according to claim 5, characterized in that: The individual cognitive trajectory construction includes: Calculate the change of three-dimensional learning state: Based on the dynamic evaluation matrix, calculate the change rate of three-dimensional learning state for each learning individual in chronological order; Construct individual cognitive trajectory vectors: based on the obtained three-dimensional learning state change rate.

7. A classroom learning dynamic evaluation and management system according to claim 5, characterized in that: The group collaboration heat map generation includes: Definition of group collaboration relationship matrix: Based on the dynamic interactive relationship between individual collaboration participation and the collaboration participation of all other members, define the group collaboration relationship matrix; Group collaboration heat map generation: Accumulate the collaboration intensity at all times to generate a group collaboration heat map.

8. A classroom learning dynamic evaluation and management system according to claim 5, characterized in that: The knowledge gap distribution cloud map generation includes: Average mastery of all students: Based on the knowledge mastery in the dynamic evaluation matrix and the preset knowledge graph topology, the average mastery of each knowledge point among all students is calculated; Assessment of mastery differences among the whole class: Calculate the standard deviation of the mastery of each knowledge point to assess the mastery differences among the whole class.

9. A classroom learning dynamic evaluation and management system according to claim 8, characterized in that: The adaptive management strategy generation module includes: Multi-dimensional learning portrait analysis: Receive the multi-dimensional learning portrait output by the group learning portrait generation module, and define the classroom state description vector based on the time series characteristics and spatial distribution characteristics of the multi-dimensional learning portrait; Reinforcement learning strategy training: The classroom state description vector is input into the strategy generation network based on deep reinforcement learning, and the teacher's executable management actions are output. The strategy generation network uses a joint optimization mode of temporal difference learning and policy gradient update to define the reward function of the strategy network; Management strategy set output: After reinforcement learning training converges, a dynamic management strategy set is generated in real time based on the current classroom status.

10. A classroom learning dynamic evaluation and management system according to claim 9, characterized in that: The adaptive management strategy generation module includes: Management policy receiving and parsing unit: receives the management policy set output by the adaptive management policy generation module, parses each policy action, and generates corresponding physical device control instructions; Physical environment control algorithm: Based on the received control instruction set, it performs dynamic control of the physical environment for each type of equipment; Multi-device linkage feedback mechanism: Perform closed-loop difference evaluation on the actual feedback parameters and target parameters of each device during the execution process.

Images