Virtual reality interactive immersive digital education system

Abstract

The invention relates to the field of artificial intelligence education, and discloses a virtual reality interaction immersive digital education system, which comprises a data acquisition unit for acquiring user behavior data through a sensor array to obtain a user interaction original data set; performing multi-dimensional feature extraction on the user interaction original data set to obtain a user cognitive state feature vector; the dynamic attention distribution model is constructed based on the user cognitive state feature vector, and the user attention thermodynamic diagram data is generated. According to the user cognitive state feature and the attention thermodynamic diagram data, personalized learning path suggestions are automatically generated and mapped to the virtual scene, targeted learning content is provided for the user, and the user experience is improved. The learning process is optimized, and the system can adjust the information presentation density and the interaction complexity of the virtual scene by analyzing the cognitive load index of the user in real time, so as to avoid cognitive overload, help the user maintain the optimal learning state and improve the learning efficiency.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence education technology, specifically to a virtual reality interactive immersive digital education system. Background Technology

[0002] With the rapid development of information technology and artificial intelligence, virtual reality (VR), as an emerging interactive technology, has been widely applied in various fields, especially in education, where its immersive learning environment has brought about tremendous changes to traditional education models. Virtual reality technology can generate three-dimensional virtual environments through computers, allowing learners to participate in educational activities as if they were actually there. However, most existing virtual reality education systems focus on visual and auditory presentation, while neglecting the dynamic monitoring and feedback of the user's cognitive state.

[0003] Traditional virtual reality education systems typically lack real-time assessment of users' cognitive load and cannot adjust the density of information presentation and the complexity of interactions according to users' cognitive states, thus affecting learning outcomes. To address this issue, researchers have begun to focus on how to achieve personalized and intelligent educational experiences based on user behavior data, cognitive state analysis, and dynamic adaptation to virtual scenes.

[0004] Existing systems typically rely on static learning paths and fixed interaction methods, failing to adapt to factors such as user attention distribution and cognitive load. Therefore, providing a more intelligent and personalized learning experience in virtual reality environments has become a pressing issue in the field of educational technology. Summary of the Invention

[0005] The purpose of this invention is to provide a virtual reality interactive immersive digital education system to solve the problem that existing systems typically rely on static learning paths and fixed interaction methods, and cannot adaptively adjust according to factors such as the user's attention distribution and cognitive load.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a virtual reality interactive immersive digital education system, comprising: The data acquisition unit collects user behavior data through a sensor array to obtain the raw user interaction dataset; it performs multi-dimensional feature extraction on the raw user interaction dataset to obtain the user cognitive state feature vector; and it constructs a dynamic attention distribution model based on the user cognitive state feature vector to generate user attention heatmap data. The learning path generation unit locates the user's focus area in a 3D virtual teaching scene based on user attention heatmap data; combines the content attributes of the focus area with the user's cognitive state feature vector to generate learning path suggestions through a knowledge graph reasoning engine; and maps the learning path suggestions to virtual scene elements to obtain a dynamic scene adaptation parameter set. The cognitive assessment unit uses a dynamic scene adaptation parameter set to drive the virtual scene rendering engine and generate an immersive learning environment; it collects user interaction feedback data in the immersive environment through an eye-tracking device; and it assesses cognitive load based on the interaction feedback data and the user's cognitive state feature vector to obtain a cognitive load index. The reinforcement learning unit adjusts the information presentation density and interaction complexity in the virtual scene based on the cognitive load index to generate teaching strategy parameters; the teaching strategy parameters are input into the reinforcement learning model for dynamic optimization to obtain the teaching intervention plan; and the personalized teaching intervention plan is synchronized to the cloud education platform through the communication module.

[0007] Furthermore, user behavior data is collected through a sensor array to obtain a user cognitive state feature vector, including: The original user interaction dataset is preprocessed with spatiotemporal alignment, and a dynamic time warping algorithm is used to eliminate differences in sensor sampling frequencies. A feature fusion model is constructed based on an attention mechanism to extract user behavior features and generate a primary feature fusion matrix. The primary feature fusion matrix is ​​input into a bidirectional long short-term memory network for temporal pattern mining to identify cognitive transition features in user behavior sequences; a support vector machine classifier is then used to label the cognitive transition features with state labels to generate a cognitive state label sequence. A cognitive state association graph is constructed using graph neural networks, with sensor nodes as vertices and feature correlations as edges, to mine implicit association rules between cross-modal features; and a user cognitive state feature vector is generated based on a dynamically weighted primary feature fusion matrix of association rules.

[0008] Furthermore, a dynamic attention distribution model is constructed based on the user's cognitive state feature vector to generate user attention heatmap data, including: The user's cognitive state feature vector is input into the spatiotemporal attention encoder, which consists of a self-attention mechanism and a convolutional neural network module. The attention mechanism captures the dynamic correlation weights between different cognitive dimensions in the feature vector, and the convolutional module extracts the local cognitive state temporal pattern to generate the attention base encoding matrix. A 3D virtual scene coordinate system is constructed, and the basic attention encoding matrix is ​​mapped to the spatial position of interactive objects in the scene; a Gaussian mixture model is used to perform spatial clustering of the cognitive weights in the encoding matrix to generate an attention probability density cloud; the probability density cloud is dynamically smoothed by combining the user's historical gaze trajectory data to form an initial attention distribution field; A reinforcement learning-driven attention calibration mechanism is introduced, using real-time user feedback as a reward signal to adjust the weight parameters of the initial attention distribution field through the network; differentiated reward functions are designed for different cognitive states to generate a dynamic attention weight map; The dynamic attention weight map is fused with the virtual scene depth map at the pixel level. The discrete weight values ​​are converted into a continuous heat map distribution using a bilinear interpolation algorithm. The weight values ​​are mapped to gradient levels through a color mapping table and superimposed into the scene rendering pipeline to generate user attention heat map data.

[0009] Furthermore, based on user attention heatmap data, a dynamic scene adaptation parameter set is obtained, including: Spatial clustering analysis was performed on user attention heatmap data, and algorithms were used to identify high-density attention regions. Combined with the virtual scene coordinate system, the geometric center coordinates of the focus of attention were generated. Core attention regions with a continuous focus duration of more than 3 seconds were extracted by threshold segmentation and marked as first-level focus. The remaining regions were divided into second- to fourth-level focus according to the attention weight in descending order, forming a focus region topology map. A knowledge graph of scene content attributes is constructed, in which content resources in the virtual scene are labeled as knowledge nodes and the logical relationships between nodes are defined. The attribute labels of knowledge nodes in the focal area are extracted by natural language processing technology, and combined with the confusion index in the user's cognitive state feature vector, the optimal association path is matched in the knowledge graph to generate learning path suggestions. The learning path suggestions are input into the scene element mapping engine; the learning path suggestions are parsed according to the rule base to generate dynamic scene operation instructions; The dynamic scene operation instructions are synchronized with the temporal changes of user attention heatmap data by using a spatiotemporal alignment algorithm, and the parameter change process is smoothed by using Bézier curves to obtain a dynamic scene adaptation parameter set.

[0010] Furthermore, by leveraging a dynamic scene adaptation parameter set to drive the virtual scene rendering engine, an immersive learning environment is generated, including: The dynamic scene adaptation parameter set is parsed into spatial layout instructions, content rendering instructions, and interaction control instructions; spatial layout instructions are executed through a layered rendering pipeline; content rendering instructions are input into a multimodal content fusion engine; and an event-driven interactive system is built based on interaction control instructions. The system simulates realistic collision feedback between elements using a physics engine, optimizes interactive response speed by combining operation feedback delay thresholds in a parameter set, dynamically generates interactive interfaces using user permission configuration data, and synchronously outputs visual, auditory, and tactile signals through multi-threaded rendering technology to create an immersive learning environment.

[0011] Furthermore, by collecting user interaction feedback data in the immersive environment using eye-tracking devices, a cognitive load index is obtained, including: Raw eye-tracking data of users in an immersive environment is collected using eye-tracking devices; timestamps of state changes of dynamic elements in the virtual scene are recorded to obtain a spatiotemporally aligned dataset of eye-tracking data and scene events. The spatiotemporal aligned dataset is deconstructed into multiple dimensions, and the sliding window algorithm is used to extract eye features: the eye features include pupil dynamic features, fixation behavior features, saccade behavior features, and blinking behavior features, generating an eye movement interaction feature vector; A dual-channel cognitive load assessment model is constructed. The first channel is a physiological signal fusion network, which concatenates the parameters in the eye-tracking interaction feature vector and the user's cognitive state feature vector, and extracts cross-modal correlation features through 1D-CNN. The second channel is a scene context encoder, which uses a Transformer structure to perform temporal encoding of the time stamps of scene element state changes. The outputs of the two channels are dynamically weighted and fused through an attention mechanism to generate the original cognitive load score. The original score is dynamically calibrated, and the user's historical cognitive load baseline data is introduced as a reference value. An adaptive threshold adjustment algorithm is used to eliminate the influence of individual differences. When the current score exceeds 1.5 times the standard deviation of the baseline value, a secondary verification mechanism is triggered, and the cognitive load index is generated by combining the scene complexity index.

[0012] Furthermore, based on the cognitive load index, the information presentation density and interaction complexity in the virtual scene are adjusted to generate teaching strategy parameters, including: A cognitive load scenario parameter mapping matrix is ​​constructed, and five threshold intervals of the cognitive load index are defined. The five thresholds include extremely low, low, medium, high and extremely high. An initial adjustment rule set of information presentation density and interaction complexity is preset for each interval. The initial rule set is dynamically optimized using a reinforcement learning framework. The rate of change of the user's cognitive load index is used as the reward signal, and the magnitude of the scene parameter adjustment is used as the action space to generate a scene parameter adjustment strategy. A multimodal buffering mechanism is developed to activate auxiliary cognitive channels for cognitive load states: the cognitive channels include visual channels, auditory channels and tactile channels. This mechanism simultaneously reduces the information presentation density of the main scene, forming a cognitive load diversion effect. The scenario parameter adjustment strategy is integrated with the multimodal buffer mechanism parameters to generate a teaching strategy parameter set.

[0013] Furthermore, the sensor array includes a wearable motion capture sensor, an eye tracker, a skin conductance sensor, and an EEG acquisition headband. The user behavior characteristics include limb movement frequency, eye gaze hotspot distribution, skin conductance transient peak value, and EEG alpha wave to beta wave power ratio.

[0014] Furthermore, the spatial layout instructions include element coordinates, hierarchical relationships, and viewpoint parameters; the content rendering instructions include model materials, lighting effects, and multimodal content switching rules; and the interactive control instructions include collision detection thresholds, operation feedback delays, and user permission configurations.

[0015] Furthermore, it also includes: building a cross-user behavior analysis matrix on the cloud education platform and integrating user teaching intervention plan data; mining group learning patterns based on the analysis matrix through federated learning algorithms to generate a group cognitive evolution trend prediction model; and feeding the group cognitive evolution trend prediction model back to the local virtual teaching system to form a closed-loop optimization mechanism.

[0016] Compared with the prior art, the beneficial effects of the present invention are: (1) Based on the user's cognitive state characteristics and attention heatmap data, the present invention automatically generates personalized learning path suggestions and maps them into a virtual scene to provide users with targeted learning content and optimize the learning process.

[0017] (2) By analyzing the user’s cognitive load index in real time, the system can adjust the information presentation density and interaction complexity of the virtual scene to avoid cognitive overload, help users maintain the best learning state, and improve learning efficiency.

[0018] (3) This invention collects user behavior data through multiple sensors and combines it with various physiological signals such as eye tracking, skin conductance response, and electroencephalogram to achieve a comprehensive assessment of the user's cognitive state and provide data support for teaching intervention.

[0019] (4) Through virtual reality technology and dynamic scene adaptation, the system can provide users with an immersive learning environment, enabling users to concentrate their attention during the learning process and enhance their sense of immersion and participation.

[0020] (5) This invention analyzes cross-user behavioral data through federated learning algorithms, mines group learning patterns, and thus predicts the evolutionary trend of group cognition, providing decision support for education administrators. Attached Figure Description

[0021] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1 This invention provides a technical solution: a virtual reality interactive immersive digital education system, comprising: The data acquisition unit collects user behavior data through a sensor array to obtain the raw user interaction dataset; it performs multi-dimensional feature extraction on the raw user interaction dataset to obtain the user cognitive state feature vector; and it constructs a dynamic attention distribution model based on the user cognitive state feature vector to generate user attention heatmap data. The system comprises a sensor array, a system composed of multiple sensors of different types arranged and combined according to certain rules; the raw user interaction dataset is a collection of unprocessed data generated by the sensor array through user interaction with the surrounding environment or system; multi-dimensional feature extraction analyzes and mines data from multiple different angles, aspects, or feature dimensions in the raw user interaction dataset to extract representative and key features; the user cognitive state feature vector is obtained through multi-dimensional feature extraction and is used to describe the user's cognitive state; the dynamic attention distribution model is a mathematical model used to describe the distribution of user attention at different times, in different regions, or on different objects; and the user attention heatmap data is generated based on the dynamic attention distribution model and is used to graphically display the distribution of user attention. The learning path generation unit locates the user's focus area in a 3D virtual teaching scene based on user attention heatmap data; combines the content attributes of the focus area with the user's cognitive state feature vector to generate learning path suggestions through a knowledge graph reasoning engine; and maps the learning path suggestions to virtual scene elements to obtain a dynamic scene adaptation parameter set. The three-dimensional virtual teaching scene utilizes computer graphics, virtual reality, and other technologies to create a teaching environment with a three-dimensional spatial feel; the user attention focus area is determined based on user attention heatmap data within the three-dimensional virtual teaching scene, identifying the area where user attention is most concentrated; the focus area content attributes describe various characteristics of the teaching content contained within the focus area, such as the type of knowledge point (concept, principle, skill, etc.), difficulty level, and relevance to preceding and following knowledge points; the knowledge graph reasoning engine is a reasoning system built upon a knowledge graph; the learning path suggestion is generated based on the results of the knowledge graph reasoning engine, providing users with suggestions regarding learning order, learning content selection, etc.; and the dynamic scene adaptation parameter set is a set of parameters obtained by mapping the learning path suggestion to virtual scene elements. The cognitive assessment unit uses a dynamic scene adaptation parameter set to drive the virtual scene rendering engine and generate an immersive learning environment; it collects user interaction feedback data in the immersive environment through an eye-tracking device; and it assesses cognitive load based on the interaction feedback data and the user's cognitive state feature vector to obtain a cognitive load index. Among them, the virtual scene rendering engine is a core software component in computer graphics, responsible for calculating and processing information such as models, materials, and lighting of virtual scenes to ultimately generate images or videos that can be displayed on display devices; immersive learning environments are created through technologies such as virtual reality and augmented reality, allowing users to fully immerse themselves in the environment as if they were in a real learning scenario; eye-tracking devices are devices that can monitor and record users' eye movement trajectories, fixation points, blink frequencies, and other eye behaviors in real time; interactive feedback data refers to various data generated when users interact with virtual scenes or systems in immersive learning environments; cognitive load assessment is the process of evaluating and analyzing the degree of cognitive burden experienced by users during the learning process; and the cognitive load index is an indicator used to quantify the degree of cognitive load experienced by users. The reinforcement learning unit adjusts the information presentation density and interaction complexity in the virtual scene according to the cognitive load index to generate teaching strategy parameters; the teaching strategy parameters are input into the reinforcement learning model for dynamic optimization to obtain the teaching intervention plan; the personalized teaching intervention plan is synchronized to the cloud education platform through the communication module. Among them, information presentation density refers to the quantity and frequency of information displayed to users in a virtual scene; interaction complexity is the ease or difficulty for users to interact with the system or virtual objects in a virtual scene; teaching strategy parameters are a set of parameters obtained by adjusting information presentation density and interaction complexity according to the cognitive load index, and these parameters are used to guide the implementation of teaching strategies in virtual scenes; reinforcement learning model is a machine learning model that learns the optimal strategy by interacting with the environment; teaching intervention plan is generated based on the dynamic optimization results of the reinforcement learning model, and is a specific intervention measure and strategy for the user's learning process; the communication module is a component responsible for data transmission and communication between different devices or systems.

[0024] It should be noted that during operation, the data acquisition unit extracts features from multiple dimensions to build models and generate heatmaps, which can accurately grasp user cognition and attention, providing a basis for subsequent work; the learning path generation unit locates the focus based on the heatmap and generates suggestions adapted to the virtual scene by combining knowledge graphs, realizing personalized learning path planning; the cognitive assessment unit creates an immersive environment by using the rendering engine, collects feedback to assess cognitive load, and understands the learning status in a timely manner; the reinforcement learning unit adjusts and optimizes teaching parameters according to cognitive load, generates intervention plans and synchronizes them to the cloud, which can dynamically adapt to different users, realize personalized and intelligent teaching, improve learning effectiveness and experience, and make full use of technological advantages to meet diverse educational needs.

[0025] In one embodiment, user behavior data is collected through a sensor array to obtain a user cognitive state feature vector, including: The original user interaction dataset is preprocessed with spatiotemporal alignment, and a dynamic time warping algorithm is used to eliminate differences in sensor sampling frequencies. A feature fusion model is constructed based on an attention mechanism to extract user behavior features and generate a primary feature fusion matrix. The primary feature fusion matrix is ​​input into a bidirectional long short-term memory network for temporal pattern mining to identify cognitive transition features in user behavior sequences; a support vector machine classifier is then used to label the cognitive transition features with state labels to generate a cognitive state label sequence. A cognitive state association graph is constructed using graph neural networks, with sensor nodes as vertices and feature correlations as edges, to mine implicit association rules between cross-modal features; and a user cognitive state feature vector is generated based on a dynamically weighted primary feature fusion matrix of association rules.

[0026] This design collects data through a sensor array and generates a user cognitive state feature vector through steps such as spatiotemporal alignment, feature fusion, temporal mining, state labeling, and association map construction. It comprehensively integrates multi-source sensor data, eliminates sampling differences, mines user behavior temporal patterns and cross-modal correlations, accurately portrays user cognitive states, and provides a reliable foundation for subsequent attention analysis, scene adaptation, etc. It helps to deeply understand the cognitive changes of users in the learning process and provides a strong basis for personalized teaching.

[0027] In one embodiment, a dynamic attention distribution model is constructed based on the user's cognitive state feature vector to generate user attention heatmap data, including: The user's cognitive state feature vector is input into the spatiotemporal attention encoder, which consists of a self-attention mechanism and a convolutional neural network module. The attention mechanism captures the dynamic correlation weights between different cognitive dimensions in the feature vector, and the convolutional module extracts the local cognitive state temporal pattern to generate the attention base encoding matrix. A 3D virtual scene coordinate system is constructed, and the basic attention encoding matrix is ​​mapped to the spatial position of interactive objects in the scene; a Gaussian mixture model is used to perform spatial clustering of the cognitive weights in the encoding matrix to generate an attention probability density cloud; the probability density cloud is dynamically smoothed by combining the user's historical gaze trajectory data to form an initial attention distribution field; A reinforcement learning-driven attention calibration mechanism is introduced, using real-time user feedback as a reward signal to adjust the weight parameters of the initial attention distribution field through the network; differentiated reward functions are designed for different cognitive states to generate a dynamic attention weight map; The dynamic attention weight map is fused with the virtual scene depth map at the pixel level. The discrete weight values ​​are converted into a continuous heat map distribution using a bilinear interpolation algorithm. The weight values ​​are mapped to gradient levels through a color mapping table and superimposed into the scene rendering pipeline to generate user attention heat map data.

[0028] This design inputs the cognitive state feature vector into the spatiotemporal attention encoder, which generates an attention heatmap through coordinate mapping, clustering, smoothing, and calibration. This heatmap can intuitively present the distribution of user attention. By combining multiple algorithms and models, it captures dynamic attention changes and takes into account real-time operation feedback for calibration, making the heatmap more accurately reflect the user's attention focus. This provides key information for scene adaptation and helps optimize the presentation of teaching content.

[0029] In one embodiment, a dynamic scene adaptation parameter set is obtained based on user attention heatmap data, including: Spatial clustering analysis was performed on user attention heatmap data, and algorithms were used to identify high-density attention regions. Combined with the virtual scene coordinate system, the geometric center coordinates of the focus of attention were generated. Core attention regions with a continuous focus duration of more than 3 seconds were extracted by threshold segmentation and marked as first-level focus. The remaining regions were divided into second- to fourth-level focus according to the attention weight in descending order, forming a focus region topology map. Construct a knowledge graph of scene content attributes, label content resources in virtual scenes as knowledge nodes, and define the logical relationships between nodes; extract attribute labels of knowledge nodes in the focus area through natural language processing technology, combine them with the confusion index in the user's cognitive state feature vector, match the optimal association path in the knowledge graph, and generate learning path suggestions; The learning path suggestions are input into the scene element mapping engine; the learning path suggestions are parsed according to the rule base to generate dynamic scene operation instructions; The dynamic scene operation instructions are synchronized with the temporal changes of user attention heatmap data by using a spatiotemporal alignment algorithm, and the parameter change process is smoothed by using Bézier curves to obtain a dynamic scene adaptation parameter set.

[0030] This design analyzes attention heatmaps and combines them with scene knowledge graphs to generate learning path suggestions. Through mapping and synchronization, a dynamic scene adaptation parameter set is obtained. Based on the user's attention focus, combined with scene content and cognitive state, the learning path is intelligently planned, and scene parameters are dynamically adjusted to make the virtual scene more in line with user needs, thereby improving the learning experience and effectiveness.

[0031] In one embodiment, a virtual scene rendering engine is driven by a dynamic scene adaptation parameter set to generate an immersive learning environment, including: The dynamic scene adaptation parameter set is parsed into spatial layout instructions, content rendering instructions, and interaction control instructions; spatial layout instructions are executed through a layered rendering pipeline; content rendering instructions are input into a multimodal content fusion engine; and an event-driven interactive system is built based on interaction control instructions. The system simulates realistic collision feedback between elements using a physics engine, optimizes interactive response speed by combining operation feedback delay thresholds in a parameter set, dynamically generates interactive interfaces using user permission configuration data, and synchronously outputs visual, auditory, and tactile signals through multi-threaded rendering technology to create an immersive learning environment.

[0032] This design parses the parameter set into multiple instructions and generates an immersive learning environment through layered rendering and multimodal fusion. It can fully leverage the advantages of virtual scenes, optimize all aspects from spatial layout and content presentation to interactive control, simulate real feedback, dynamically generate interfaces, and synchronously output multi-sensory signals, allowing users to fully immerse themselves in learning and enhancing the immersion and interactivity of learning.

[0033] In one embodiment, a cognitive load index is obtained by collecting user interaction feedback data in an immersive environment using an eye-tracking device, including: Raw eye-tracking data of users in an immersive environment is collected using eye-tracking devices; timestamps of state changes of dynamic elements in the virtual scene are recorded to obtain a spatiotemporally aligned dataset of eye-tracking data and scene events. Multi-dimensional feature deconstruction is performed on the spatiotemporal aligned dataset, and eye features are extracted using the sliding window algorithm: eye features include pupil dynamic features, fixation behavior features, saccade behavior features, and blinking behavior features, generating an eye movement interaction feature vector; A dual-channel cognitive load assessment model is constructed. The first channel is a physiological signal fusion network, which concatenates the parameters in the eye-tracking interaction feature vector and the user's cognitive state feature vector, and extracts cross-modal correlation features through 1D-CNN. The second channel is a scene context encoder, which uses a Transformer structure to perform temporal encoding of the time stamps of scene element state changes. The outputs of the two channels are dynamically weighted and fused through an attention mechanism to generate the original cognitive load score. The original score is dynamically calibrated, and the user's historical cognitive load baseline data is introduced as a reference value. An adaptive threshold adjustment algorithm is used to eliminate the influence of individual differences. When the current score exceeds 1.5 times the standard deviation of the baseline value, a secondary verification mechanism is triggered, and the cognitive load index is generated by combining the scene complexity index.

[0034] This design uses eye-tracking devices to collect data, and through feature deconstruction, model evaluation, and calibration, a cognitive load index is obtained. This index can assess the user's cognitive load from multiple dimensions, and by combining physiological signals and contextual information, it eliminates individual differences and accurately reflects the user's cognitive stress during learning, providing quantitative indicators for adjusting teaching strategies.

[0035] In one embodiment, the information presentation density and interaction complexity in the virtual scene are adjusted according to the cognitive load index to generate teaching strategy parameters, including: Construct a parameter mapping matrix for cognitive load scenarios, define five threshold ranges for the cognitive load index, including extremely low, low, medium, high and extremely high thresholds, and preset an initial set of adjustment rules for information presentation density and interaction complexity for each range; The initial rule set is dynamically optimized using a reinforcement learning framework. The rate of change of the user's cognitive load index is used as the reward signal, and the magnitude of the scene parameter adjustment is used as the action space to generate a scene parameter adjustment strategy. Develop a multimodal buffering mechanism to activate auxiliary cognitive channels for cognitive load states: cognitive channels include visual, auditory, and tactile channels. This mechanism simultaneously reduces the information presentation density of the main scene, forming a cognitive load diversion effect. The scenario parameter adjustment strategy is integrated with the multimodal buffer mechanism parameters to generate a teaching strategy parameter set.

[0036] This design adjusts scenario parameters based on the cognitive load index, optimizes rules through reinforcement learning, develops a buffer mechanism, and generates teaching strategy parameters. It can dynamically adapt to different cognitive load states, optimize information presentation and interaction complexity, reduce cognitive pressure through multimodal diversion, achieve personalized and intelligent teaching strategy adjustments, and improve learning efficiency.

[0037] In one embodiment, the sensor array includes a wearable motion capture sensor, an eye tracker, a skin conductance sensor, and an EEG acquisition headband. User behavior characteristics include limb movement frequency, eye gaze hotspot distribution, skin conductance transient peak value, and EEG alpha wave to beta wave power ratio.

[0038] This design allows the sensor array to encompass a variety of devices, and user behavior characteristics to include multiple indicators. It comprehensively collects multi-dimensional user behavior data, from physical movements to physiological signals, providing rich material for cognitive state analysis. This makes the assessment of user state more accurate and comprehensive, and helps to gain a deeper understanding of the user's physical and mental responses during the learning process.

[0039] In one embodiment, the spatial layout instructions include element coordinates, hierarchical relationships, and viewpoint parameters; the content rendering instructions include model materials, lighting effects, and multimodal content switching rules; and the interactive control instructions include collision detection thresholds, operation feedback delays, and user permission configurations.

[0040] This design incorporates multiple parameters in spatial layout, content rendering, and interactive control commands, allowing for precise control over all aspects of the virtual scene, from element positions to interactive details. This ensures that the scene is accurately presented and operates according to teaching needs, providing users with a smooth and personalized learning environment and enhancing teaching effectiveness.

[0041] In one embodiment, the method further includes: constructing a cross-user behavior analysis matrix on a cloud-based education platform and integrating user teaching intervention plan data; mining group learning patterns based on the analysis matrix using a federated learning algorithm to generate a group cognitive evolution trend prediction model; and feeding the group cognitive evolution trend prediction model back to the local virtual teaching system to form a closed-loop optimization mechanism.

[0042] This design constructs an analysis matrix in the cloud, uncovers group patterns, and feeds them back to the local system to form a closed loop. It leverages the powerful computing capabilities of the cloud to process large amounts of user data, discovers group learning patterns, provides optimization directions for the local teaching system, achieves continuous improvement of the teaching system, and enhances the overall teaching quality.

[0043] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A virtual reality interactive immersive digital education system, characterized in that, include: The data acquisition unit collects user behavior data through a sensor array to obtain the raw user interaction dataset; it performs multi-dimensional feature extraction on the raw user interaction dataset to obtain the user cognitive state feature vector; and it constructs a dynamic attention distribution model based on the user cognitive state feature vector to generate user attention heatmap data. The learning path generation unit locates the user's focus area in a 3D virtual teaching scene based on user attention heatmap data; and generates learning path suggestions by combining the content attributes of the focus area with the user's cognitive state feature vector through a knowledge graph reasoning engine. The learning path suggestions are mapped to virtual scene elements to obtain a dynamic scene adaptation parameter set; The cognitive assessment unit uses a dynamic scene adaptation parameter set to drive the virtual scene rendering engine and generate an immersive learning environment; it collects user interaction feedback data in the immersive environment through an eye-tracking device; and it assesses cognitive load based on the interaction feedback data and the user's cognitive state feature vector to obtain a cognitive load index. The reinforcement learning unit adjusts the information presentation density and interaction complexity in the virtual scene based on the cognitive load index to generate teaching strategy parameters; the teaching strategy parameters are input into the reinforcement learning model for dynamic optimization to obtain the teaching intervention plan; and the personalized teaching intervention plan is synchronized to the cloud education platform through the communication module.

2. The virtual reality interactive immersive digital education system according to claim 1, characterized in that, User behavior data is collected through a sensor array to obtain a user cognitive state feature vector, including: The original user interaction dataset is preprocessed with spatiotemporal alignment, and a dynamic time warping algorithm is used to eliminate differences in sensor sampling frequencies. A feature fusion model is constructed based on an attention mechanism to extract user behavior features and generate a primary feature fusion matrix. The primary feature fusion matrix is ​​input into a bidirectional long short-term memory network for temporal pattern mining to identify cognitive transition features in user behavior sequences; a support vector machine classifier is then used to label the cognitive transition features with state labels to generate a cognitive state label sequence. A cognitive state association graph is constructed using graph neural networks, with sensor nodes as vertices and feature correlations as edges, to mine implicit association rules between cross-modal features; and a user cognitive state feature vector is generated based on a dynamically weighted primary feature fusion matrix of association rules.

3. The virtual reality interactive immersive digital education system according to claim 2, characterized in that, A dynamic attention distribution model is constructed based on user cognitive state feature vectors to generate user attention heatmap data, including: The user's cognitive state feature vector is input into the spatiotemporal attention encoder, which consists of a self-attention mechanism and a convolutional neural network module. The attention mechanism captures the dynamic correlation weights between different cognitive dimensions in the feature vector, and the convolutional module extracts the local cognitive state temporal pattern to generate the attention base encoding matrix. A 3D virtual scene coordinate system is constructed, and the basic attention encoding matrix is ​​mapped to the spatial position of interactive objects in the scene; a Gaussian mixture model is used to perform spatial clustering of the cognitive weights in the encoding matrix to generate an attention probability density cloud; the probability density cloud is dynamically smoothed by combining the user's historical gaze trajectory data to form an initial attention distribution field; A reinforcement learning-driven attention calibration mechanism is introduced, using real-time user feedback as a reward signal to adjust the weight parameters of the initial attention distribution field through the network; differentiated reward functions are designed for different cognitive states to generate a dynamic attention weight map; The dynamic attention weight map is fused with the virtual scene depth map at the pixel level. The discrete weight values ​​are converted into a continuous heat map distribution using a bilinear interpolation algorithm. The weight values ​​are mapped to gradient levels through a color mapping table and superimposed into the scene rendering pipeline to generate user attention heat map data.

4. The virtual reality interactive immersive digital education system according to claim 3, characterized in that, Based on user attention heatmap data, a set of dynamic scene adaptation parameters is obtained, including: Spatial clustering analysis was performed on user attention heatmap data, and algorithms were used to identify high-density attention regions. Combined with the virtual scene coordinate system, the geometric center coordinates of the focus of attention were generated. Core attention regions with a continuous focus duration of more than 3 seconds were extracted by threshold segmentation and marked as first-level focus. The remaining regions were divided into second- to fourth-level focus according to the attention weight in descending order, forming a focus region topology map. A knowledge graph of scene content attributes is constructed, in which content resources in the virtual scene are labeled as knowledge nodes and the logical relationships between nodes are defined. The attribute labels of knowledge nodes in the focal area are extracted by natural language processing technology, and combined with the confusion index in the user's cognitive state feature vector, the optimal association path is matched in the knowledge graph to generate learning path suggestions. The learning path suggestions are input into the scene element mapping engine; the learning path suggestions are parsed according to the rule base to generate dynamic scene operation instructions; The dynamic scene operation instructions are synchronized with the temporal changes of user attention heatmap data by using a spatiotemporal alignment algorithm, and the parameter change process is smoothed by using Bézier curves to obtain a dynamic scene adaptation parameter set.

5. The virtual reality interactive immersive digital education system according to claim 4, characterized in that, By leveraging a dynamic scene adaptation parameter set to drive a virtual scene rendering engine, an immersive learning environment is generated, including: The dynamic scene adaptation parameter set is parsed into spatial layout instructions, content rendering instructions, and interaction control instructions; spatial layout instructions are executed through a layered rendering pipeline; content rendering instructions are input into a multimodal content fusion engine; and an event-driven interactive system is built based on interaction control instructions. The system simulates realistic collision feedback between elements using a physics engine, optimizes interactive response speed by combining operation feedback delay thresholds in a parameter set, dynamically generates interactive interfaces using user permission configuration data, and synchronously outputs visual, auditory, and tactile signals through multi-threaded rendering technology to create an immersive learning environment.

6. The virtual reality interactive immersive digital education system according to claim 5, characterized in that, By collecting user interaction feedback data in an immersive environment using eye-tracking devices, a cognitive load index is obtained, including: Raw eye-tracking data of users in an immersive environment is collected using eye-tracking devices; timestamps of state changes of dynamic elements in the virtual scene are recorded to obtain a spatiotemporally aligned dataset of eye-tracking data and scene events. The spatiotemporal aligned dataset is deconstructed into multiple dimensions, and the sliding window algorithm is used to extract eye features: the eye features include pupil dynamic features, fixation behavior features, saccade behavior features, and blinking behavior features, generating an eye movement interaction feature vector; A dual-channel cognitive load assessment model is constructed. The first channel is a physiological signal fusion network, which concatenates the parameters in the eye-tracking interaction feature vector and the user's cognitive state feature vector, and extracts cross-modal correlation features through 1D-CNN. The second channel is a scene context encoder, which uses a Transformer structure to perform temporal encoding of the time stamps of scene element state changes. The outputs of the two channels are dynamically weighted and fused through an attention mechanism to generate the original cognitive load score. The original score is dynamically calibrated, and the user's historical cognitive load baseline data is introduced as a reference value. An adaptive threshold adjustment algorithm is used to eliminate the influence of individual differences. When the current score exceeds 1.5 times the standard deviation of the baseline value, a secondary verification mechanism is triggered, and the cognitive load index is generated by combining the scene complexity index.

7. The virtual reality interactive immersive digital education system according to claim 6, characterized in that, Based on the cognitive load index, the information presentation density and interaction complexity in the virtual scene are adjusted to generate teaching strategy parameters, including: A cognitive load scenario parameter mapping matrix is ​​constructed, and five threshold intervals of the cognitive load index are defined. The five thresholds include extremely low, low, medium, high and extremely high. An initial adjustment rule set of information presentation density and interaction complexity is preset for each interval. The initial rule set is dynamically optimized using a reinforcement learning framework. The rate of change of the user's cognitive load index is used as the reward signal, and the magnitude of the scene parameter adjustment is used as the action space to generate a scene parameter adjustment strategy. A multimodal buffering mechanism is developed to activate auxiliary cognitive channels for cognitive load states: the cognitive channels include visual channels, auditory channels and tactile channels. This mechanism simultaneously reduces the information presentation density of the main scene, forming a cognitive load diversion effect. The scenario parameter adjustment strategy is integrated with the multimodal buffer mechanism parameters to generate a teaching strategy parameter set.

8. The virtual reality interactive immersive digital education system according to claim 7, characterized in that, The sensor array includes a wearable motion capture sensor, an eye tracker, a skin conductance sensor, and an EEG acquisition headband. The user behavior characteristics include limb movement frequency, eye gaze hotspot distribution, skin conductance transient peak value, and EEG alpha wave to beta wave power ratio.

9. The virtual reality interactive immersive digital education system according to claim 8, characterized in that: The spatial layout instructions include element coordinates, hierarchical relationships, and viewpoint parameters; the content rendering instructions include model materials, lighting effects, and multimodal content switching rules; and the interactive control instructions include collision detection thresholds, operation feedback delays, and user permission configurations.

10. The virtual reality interactive immersive digital education system according to claim 9, characterized in that, Also includes: Build a cross-user behavior analysis matrix on the cloud-based education platform and integrate user teaching intervention program data; Based on the analysis matrix, a federated learning algorithm is used to mine group learning patterns and generate a group cognitive evolution trend prediction model. The group cognitive evolution trend prediction model is then fed back to the local virtual teaching system to form a closed-loop optimization mechanism.

Images