A teaching knowledge mining method, system, device and medium based on VR intelligent education

Abstract

The present invention discloses a teaching knowledge mining method, system, device and medium based on VR intelligent education, which relates to the technical field of knowledge mining. It includes obtaining the first behavior data of learners and constructing a VR model; generating a first-level learning path according to the behavior data and obtaining the simulated learning effect data of each path; collecting the second behavior data, and combining with the simulated learning effect data to determine the second-level learning path; obtaining the third behavior data and the first health data after learning, and comparing them with the simulated effect data of the second-level learning path to form a feedback result. The teaching knowledge mining method based on VR intelligent education provided by the present invention reduces the unnecessary path exploration time through the prediction and simulation of the learning path, quickly determines the optimal learning path, and ensures that learners achieve the best learning effect in the shortest time. The present invention achieves better effects in terms of learning content regulation, learning efficiency and personalization.

Description

Technical Field

[0001] The present invention relates to the technical field of knowledge mining, and particularly to a teaching knowledge mining method, system, device, and medium based on VR intelligent education. Background Art

[0002] With the rapid development of information technology, online education and distance education have gradually become important components of modern education. Especially in recent years, the continuous progress of technologies such as virtual reality (VR), augmented reality (AR), artificial intelligence (AI), and big data has brought new changes to the education field. These emerging technologies provide more possibilities for teaching, break the limitations of traditional teaching methods, and make the learning process more interactive and immersive. However, despite the increasingly widespread application of these technologies in education, there are still some significant deficiencies in the personalization and interactivity of current teaching platforms, which affect the improvement of learning effects.

[0003] Online education platforms (such as Coursera, edX, Udemy, etc.) enable learners around the world to obtain high-quality educational resources through the Internet. These platforms provide rich course content in forms such as videos, texts, quizzes, etc., which helps students learn independently. However, although the content is rich, this mainly video- and text-based teaching method still lacks interactivity and is difficult to meet the needs of all learners.

[0004] Artificial intelligence technology is widely used in the education field to achieve automated personalized teaching. Through machine learning algorithms, the system can analyze the behavioral data of learners, infer their learning status, and thus recommend suitable learning content for each learner. However, most of the existing AI education technologies are based on the analysis of static data and are difficult to achieve real-time adjustment and multi-dimensional interaction.

[0005] With the continuous progress of VR technology, its application in education has gradually increased. VR technology can provide learners with immersive learning experiences and has been widely used in teaching in fields such as science, engineering, and medicine. For example, VR can simulate scenarios such as dangerous chemical experiments and complex surgical operations to help students practice safely in a virtual environment. However, existing VR teaching systems are often limited to the simulation of single scenarios and lack flexible knowledge point association and personalized learning path design.

[0006] Although these technologies have achieved certain results in education, there are still some deficiencies in the current teaching platforms in terms of personalization, interactivity, and intelligence. Most existing online education platforms are difficult to push personalized teaching content according to the actual situation of learners (such as learning ability, interests and hobbies, health status, etc.). Many systems only make simple content recommendations based on the historical performance of learners, without considering multi-dimensional data such as emotional state, physical health data, real-time feedback, etc., resulting in a relatively single learning path design and being difficult to meet the personalized needs of learners. Traditional online learning mainly relies on videos and texts, and this one-way teaching method lacks interactivity. Even in some VR teaching systems, although they can provide an immersive learning experience, most systems only support fixed 3D scene simulations and cannot interact deeply with learners. For example, learners cannot communicate with the system in multiple ways such as voice and gesture, nor can they customize the virtual scene according to their own learning needs. In addition, the feedback in the existing systems during the teaching process is relatively single, usually only obtaining the performance data of learners through quizzes, and unable to dynamically adjust teaching strategies. Although the introduction of artificial intelligence technology has made personalized recommendation and automatic adjustment of learning paths possible, most systems are still based on the analysis of offline data and lack the ability to process real-time data. The instant emotional changes or fluctuations in health status that occur during the learning process of learners are difficult to be quickly captured and responded to by the system, which leads to the system lagging behind the actual needs of learners when providing personalized teaching and being unable to adjust teaching content and difficulty in a timely manner according to the emotional state or physical condition of learners. Existing education systems rarely consider the physiological and mental health status of learners. Although some systems can collect some physiological data of users (such as heart rate, steps, etc.), these data are usually used in the field of health management rather than education. The current teaching platforms lack in-depth mining and analysis of health data and cannot dynamically adjust learning content according to the health status of learners, ignoring the impact of health factors on learning efficiency during the learning process.

[0007] Most existing education platforms organize content by courses or disciplines and lack the ability to associate knowledge across disciplines. Learners may need to combine knowledge from multiple disciplines to solve complex problems during the learning process, but existing systems are difficult to automatically establish these interdisciplinary associations. Knowledge graphs are applied in some platforms, but they are usually static and lack the ability of dynamic update and adaptive adjustment. Summary of the Invention

[0008] In view of the above problems, the present invention is proposed.

[0009] Therefore, the technical problem to be solved by the present invention is that the existing VR online education platforms have insufficient personalization, low interactivity, insufficient data utilization, and the problem of how to effectively integrate real-time health data and emotional feedback to improve teaching quality.

[0010] To solve the above technical problems, the present invention provides the following technical solution: A teaching knowledge mining method based on VR intelligent education, including obtaining the first behavior data of learners and constructing a VR model. According to the behavior data, generate a first-level learning path and obtain the simulated learning effect data of each path. Collect the second behavior data, and combine it with the simulated learning effect data to determine the second-level learning path. Obtain the third behavior data and the first health data after learning, and compare them with the simulated effect data of the second-level learning path to form a feedback result.

[0011] As a preferred solution of the teaching knowledge mining method based on VR intelligent education of the present invention, wherein: the obtaining of the first behavior data of learners includes collecting the behavior data of learners in the virtual learning environment through VR devices.

[0012] As a preferred solution of the teaching knowledge mining method based on VR intelligent education of the present invention, wherein: the generating of the first-level learning path according to the behavior data includes using the behavior data to generate multiple first-level learning paths with different content structures and presentation orders, and each learning path is simulated according to the personal situation of the learner.

[0013] As a preferred solution of the teaching knowledge mining method based on VR intelligent education of the present invention, wherein: the obtaining of the simulated learning effect data of each path includes predicting the learning effect of each path based on historical data and related features.

[0014] As a preferred solution of the teaching knowledge mining method based on VR intelligent education of the present invention, wherein: the collecting of the second behavior data includes collecting the current behavior data of learners as the second behavior data, and the data collection type is the same as the first behavior data.

[0015] As a preferred solution of the teaching knowledge mining method based on VR intelligent education of the present invention, wherein: the determining of the second-level learning path includes comparing the current behavior data of learners with the simulated effect data of each path, and determining the learning path with the best expected effect as the second-level learning path to adapt to the personalized needs and real-time status of learners.

[0016] As a preferred solution of the teaching knowledge mining method based on VR intelligent education according to the present invention, wherein: obtaining the third behavior data and the first health data after learning, and comparing them with the simulation effect data of the secondary learning path to form a feedback result includes evaluating the effectiveness of the learning process by comparing the new behavior data collected during the actual learning process with the simulation learning effect data of the optimal path, and integrating the knowledge mastery difference, learning efficiency difference, and physiological load difference into a comprehensive deviation score by the weighted scoring method, and performing feedback adjustment of knowledge mining based on the comprehensive score. When the comprehensive deviation score ≥ 0.7, and at least two-dimensional difference scores exceed their preset own thresholds, it is identified that the learner cannot adapt to the current learning intensity and knowledge mining intensity, and the system immediately adjusts the learning path to enhance the learner's adaptability by reducing the content difficulty and increasing the rest time.

[0017] When 0.4 ≤ comprehensive deviation score < 0.7 and the physiological load difference does not exceed the preset own threshold, and only one of the knowledge mastery difference and the learning efficiency difference exceeds the preset own threshold, the learning effect is optimized by providing additional interesting auxiliary materials for learning interest guidance.

[0018] When the comprehensive deviation score < 0.4, it indicates that the learning process is close to the expected effect, and the existing learning path is maintained.

[0019] Another object of the present invention is to provide a teaching knowledge mining system based on VR intelligent education, which can accurately understand the initial learning state of the learner by comprehensively collecting various behavior data of the learner, and construct a VR model adapted to its personalized needs according to the data analysis results, providing accurate basic information for the generation of the subsequent learning path. It solves the problem of insufficient personalization in the current VR online education platform.

[0020] As a preferred solution of the teaching knowledge mining system based on VR intelligent education according to the present invention, wherein: it includes a data collection module, a learning effect evaluation module, and a path evaluation module. The data collection module is used to obtain the first behavior data, the second behavior data, the third behavior data, and the first health data of the learner. The learning effect evaluation module is used to evaluate the learning effect of the learning path in combination with the expected indicators. The path evaluation module is used to compare with the simulation effect data of the secondary learning path to form a feedback result.

[0021] A computer device includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the steps of the teaching knowledge mining method based on VR intelligent education.

[0022] A computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, it implements the steps of a teaching knowledge mining method based on VR intelligent education.

[0023] Advantages of the present invention: The teaching knowledge mining method based on VR intelligent education provided by the present invention reduces unnecessary path exploration time through the prediction and simulation of the learning path, quickly determines the optimal learning path, and ensures that learners achieve the best learning effect in the shortest time. Dynamically adjusting the learning path ensures the continuity and adaptability of the personalized learning experience, enabling learners to always obtain the most suitable learning content and rhythm arrangement throughout the learning process. The present invention achieves better results in terms of learning content regulation, learning efficiency, and personalization. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

[0025] Figure 1 It is the overall flowchart of a teaching knowledge mining method based on VR intelligent education provided by the first embodiment of the present invention.

[0026] Figure 2 It is the module schematic diagram of a teaching knowledge mining system based on VR intelligent education provided by the third embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the specific embodiments of the present invention in detail with reference to the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0028] Example 1, referring to Figure 1 , which is an embodiment of the present invention, provides a teaching knowledge mining method based on VR intelligent education, including:

[0029] S1: Obtain the first behavior data of the learner and construct a VR model.

[0030] Furthermore, obtaining the first behavioral data of the learner includes collecting the learner's behavioral data in the virtual learning environment through VR devices. Obtaining the learner's behavioral data is achieved through VR devices. These devices include, but are not limited to, head-mounted displays (HMDs), handheld controllers, body tracking devices, etc. Through these devices, various behavioral data of the learner in the virtual learning environment can be comprehensively collected. The specific collection methods include:

[0031] Position and movement tracking: Using head-mounted displays and body tracking sensors, the system can record the learner's spatial position, movement path, and the orientation of the head and body. Based on this data analysis, the degree of attention of the learner to various elements in the virtual environment and movement habits can be determined.

[0032] Eye gaze tracking: Through the eye movement tracking technology integrated in the VR device, the direction of the learner's eye gaze and the fixation points can be accurately captured. The eye gaze data is used to analyze the distribution of the learner's attention, thereby judging their interest and depth of understanding of specific knowledge points.

[0033] Gesture and operation analysis: Using handheld controllers or hand tracking technology, the system can record the learner's gesture actions and interaction methods with virtual objects, such as grasping, pointing, operating virtual devices, etc. These data help to evaluate the learner's operation skills and practical application abilities.

[0034] Physiological response monitoring: Collecting the learner's heart rate, eye movement speed, etc., and these information are used to analyze the learner's emotional state and stress level.

[0035] Speech and communication analysis: If the VR environment supports voice interaction, the voice data can be collected to analyze the learner's inquiry content, intonation, speech rate, etc., in order to evaluate their language expression ability and understanding degree.

[0036] It should be noted that building a VR model is the focus of VR intelligent education. The construction of the model usually adopts teaching content analysis based on NLP and 3D model generation.

[0037] Using natural language processing (NLP) technology to perform text analysis on teaching materials, extracting keywords and core concepts, and constructing a hierarchical 3D model through the knowledge graph framework.

[0038] Using the TF-IDF algorithm to calculate the importance score of each word, setting a preset threshold, and taking the words higher than this threshold as the keyword candidate set.

[0039] Grouping similar keywords into themes through clustering algorithms to form different knowledge nodes. For example, the related content of "force" and "motion" in a physics course can be divided into a subset.

[0040] Knowledge Node Association: Using knowledge graph technology, relevant attributes and associations such as formulas and historical backgrounds are added to each node to form the initial hierarchical structure of the 3D model.

[0041] Hierarchical classification mainly includes three levels, namely the basic knowledge level, the intermediate knowledge level, and the advanced application level. The three levels respectively assist learners in understanding, expanding, and applying and mining knowledge.

[0042] The basic knowledge level is suitable for learners who are newly exposed to new knowledge. The content mainly focuses on concept introduction and simple demonstrations. The 3D model structure is relatively simple, showing the core concepts. Taking the study of physics as an example, when learning mechanics, a simple 3D animation of an object under force is shown, emphasizing the definition and basic properties of force.

[0043] For the knowledge points at the basic level, the complexity of the 3D model of each node is measured by the number of faces and polygons, and the upper limit is set to 500 faces to ensure rendering efficiency.

[0044] When the system detects that the correct rate of a single practice of the learner's mastery of the knowledge points at the basic level reaches more than 80%, the content of the intermediate level is automatically unlocked, providing more detailed interpretations such as formula derivations and more complex 3D demonstrations such as the interaction of forces between multiple objects.

[0045] The number of polygons of the 3D model in the intermediate level can be increased to 1000 - 2000 to display more details, such as the direction of force and the decomposition diagram.

[0046] The advanced application level is suitable for in-depth research or interdisciplinary applications. The 3D model will show complex systems, such as a multi-degree-of-freedom vibration system in physics, or associating the cell metabolism process in biology with chemical reactions. According to the learner's learning trajectory, the system monitors their learning time, the exploration path coverage rate is greater than 70%, and the correct rate of test questions reaches more than 90%. When the above conditions are met, the system automatically unlocks the 3D model of the advanced level and supports custom modification of parameters, such as increasing friction or changing the system boundary conditions.

[0047] It should also be noted that the data in the first row includes learning duration, interaction frequency, exploration path (i.e., the access order of knowledge points during the learning process), the correct rate of answering questions, learning speed, and attention data (through eye movement tracking and facial expression analysis).

[0048] S2: According to the behavior data, generate the first-level learning path and obtain the simulated learning effect data for each path.

[0049] Furthermore, based on the behavior data, generate first-level learning paths, including using the behavior data to generate multiple first-level learning paths with different content structures and presentation orders, and each learning path is simulated according to the learner's personal situation. According to different behavior data characteristics, the system generates multiple learning paths, which are different in content structure and presentation order. There are four learning paths, namely:

[0050] Fast progressive path: Suitable for learners with a fast knowledge acquisition speed and strong self-learning ability indicated by behavior data. The design of the path includes reducing the staying time of basic content, entering complex knowledge points faster, and encouraging interdisciplinary applications. This path mainly reduces repetitive learning by combining with the advanced application layer and enhances the extended application of knowledge.

[0051] In-depth reinforcement path: Suitable for learners with a slow learning speed and difficulty in understanding some knowledge points. The system will extend the explanation time of key knowledge points, increase practice sessions, and provide more relevant explanations between knowledge points. This path mainly realizes in-depth learning of knowledge by combining with the intermediate knowledge layer.

[0052] Exploratory path: For learners who prefer free exploration and jump-style learning between knowledge points. The system will provide more open options, guide learners to independently select the learning order, and push relevant supplementary materials.

[0053] Progressive path: For learners with zero foundation or just starting to contact the subject. The path starts from basic knowledge and gradually progresses to more complex content. The explanation of the content is step by step, ensuring that learners can learn deeper content after fully mastering the basic content. This path mainly realizes the learning of basic knowledge by combining with the basic knowledge layer.

[0054] It should be noted that the system simulates each generated first-level learning path according to the behavior data. For example, use the learner's historical learning records to simulate the learning effects under different paths, and predict the learning progress and knowledge mastery of each path. Adopt reinforcement learning algorithms (such as Q-learning) to optimize the order and content density of each knowledge point on each path to improve learning efficiency and the comprehensiveness of knowledge mastery.

[0055] During the simulation process, the evaluation indicators used by the system include completion time, improvement amplitude of mastery, and cognitive load (by measuring the changes in brain waves and eye movement data during the learning process). According to these evaluation indicators, the system will select the most suitable first-level learning path for the learner and dynamically adjust the path during the learning process.

[0056] It should also be noted that obtaining the simulation learning effect data for each path includes predicting the learning effect of each path based on historical data and relevant features. By using the method of causal inference, the causal relationship between different features is clarified. For example, the learner's emotional state and learning duration may directly affect the learning effect, and the structure and presentation order of the learning path will indirectly affect the learning effect by influencing the learner's mastery of knowledge points.

[0057] Based on these causal relationships, a structural equation model (SEM) is constructed, and the direct and indirect effects between various features are analyzed through hypothesized paths.

[0058] Random forest regression is selected to predict the learning effect of each path. The target variables predicted for each path can include the learner's mastery of knowledge points, completion time, average score improvement rate, etc.

[0059] An ensemble learning method is adopted to perform weighted averaging on the prediction results of multiple models to improve the prediction accuracy.

[0060] The learning effect prediction is mainly based on the following several indicators, and the system scores each path:

[0061] Knowledge point mastery (K): Predict the learner's mastery of each knowledge point after completing the learning under a certain path. It can be quantified as the answering correct rate or the test score of the knowledge point.

[0062] Learning duration (T): Predict the total time required for the learner to complete all tasks in this path.

[0063] Cognitive load (C): Quantify the cognitive burden in the learning process based on physiological data (such as brain waves, eye movement changes, etc.), and it is divided into three levels: low, medium, and high.

[0064] Knowledge point conversion rate (R): It refers to the degree of absorption and application of new knowledge points by the learner, and the influence on the subsequent knowledge point test is used as the evaluation criterion.

[0065] In order to select the optimal learning path, the system adopts a multi-objective optimization algorithm to comprehensively optimize the above four quantitative indicators (K, T, C, R), sets the objective function based on the four indicators, and determines the most suitable secondary learning path.

[0066] S3: Collect the second behavior data, combine it with the simulation learning effect data, and determine the secondary learning path.

[0067] Furthermore, after collecting the second row of data, feature extraction is performed. The sliding window technique is used to segment the learning behavior data and extract time series features, such as the learning duration trend in the last 10 minutes and the change in the answering accuracy rate. Time domain and frequency domain analyses are performed on the physiological and emotional data to extract key indicators such as the average value and standard deviation of the heart rate, and the spectral density of the eye movement frequency.

[0068] Statistical analysis is performed on the interaction data to calculate the total number of gesture and voice interactions, the interaction frequency, and the distribution.

[0069] The features of multi-source data, such as learning behavior, emotion, and physiological signals, are fused into a high-dimensional feature vector, and the main feature dimensions are extracted through dimensionality reduction techniques such as principal component analysis (PCA) or linear discriminant analysis (LDA).

[0070] The physiological and emotional data of learners, such as the fatigue state, may directly affect the learning behavior data, such as the learning duration and the answering accuracy rate, and thus indirectly affect the learning effect. A Bayesian network is used to quantify this causal relationship.

[0071] The relationship between the interaction data and the learning behavior can reflect the learning strategy of the learner. For example, frequent gesture interactions and high-frequency knowledge point jumps may indicate that the learner tends to be an "exploratory" learner, while fewer interactions and longer stays on a single knowledge point may indicate that the learner is a "deep learning" type.

[0072] Based on the learning behavior and physiological data in the previous 10 minutes, predict the learning effect in the next 10 minutes, such as the answering accuracy rate and the degree of knowledge point mastery. Use a long short-term memory network to predict the time series data.

[0073] It should be noted that determining the secondary learning path includes comparing the current behavior data of the learner with the simulated effect data of each path, determining the learning path with the best expected effect as the secondary learning path to adapt to the personalized needs and real-time status of the learner. When the difference between the current behavior data of the learner and the simulated effect data of each path is small or the best expected effect is not set, in order to select the optimal learning path, the system uses a multi-objective optimization algorithm to comprehensively optimize the above four quantitative indicators (K, T, C, R), sets the objective function based on the four indicators, and determines the most suitable secondary learning path.

[0074] S4: Obtain the third row of data after learning and the first health data, and compare them with the simulated effect data of the secondary learning path to form a feedback result.

[0075] Furthermore, the third behavior data and the first health data after learning are obtained and compared with the simulated effect data of the secondary learning path to form feedback results, including evaluating the effectiveness of the learning process by comparing the new behavior data collected in the actual learning process with the simulated learning effect data of the optimal path, integrating the differences in knowledge mastery, learning efficiency and physiological load into a comprehensive deviation score through a weighted scoring method, and performing feedback adjustment of knowledge mining based on the comprehensive score. When the comprehensive deviation score is ≥0.7 and the difference scores of at least two dimensions exceed the preset self-thresholds, it is identified that the learner cannot adapt to the current learning intensity and knowledge mining intensity, and the system immediately adjusts the learning path to enhance the learner's adaptability by reducing the difficulty of the content and increasing the rest time.

[0076] The first health data may be heart rate, blinking frequency, gaze time, etc.

[0077] When 0.4≤comprehensive deviation score<0.7 is identified and the physiological load difference does not exceed the preset self-threshold, and only one of the knowledge mastery difference and learning efficiency difference exceeds the preset self-threshold, the learning effect is optimized by providing additional interesting auxiliary materials to guide learning interest.

[0078] When the comprehensive deviation score is <0.4, it means that the learning process is close to the expected effect and the existing learning path is maintained.

[0079] It should be noted that three behavioral data collections were conducted. The first behavioral data collection was conducted at the beginning of the learner's entry into the virtual learning environment, with the purpose of providing basic information for the learner's personalized learning path and VR model construction. The second behavioral data collection was conducted after the learner started the first-level learning path, and was used to monitor the learner's learning process and status in real time. The data type was the same as the first behavioral data type, and was collected every 10 minutes after the learner started learning. The purpose was to dynamically adjust the learning path based on these data, generate a second-level learning path, and make the learning experience more in line with the learner's real-time needs. The third behavioral data was used to evaluate the learning effect and feedback adjustment to ensure that the goal of the second-level learning path was achieved. The data type of the third behavioral data was different from the first and second behavioral data, including but not limited to the accuracy of answering questions, the dwell time of a single test question, the number of touches, the number of answer changes, etc. By comparing the third behavioral data with the simulated learning effect data, the system can judge the applicability of the learning path and make further optimizations.

[0080] Example 2, an embodiment of the present invention, provides a teaching knowledge mining method based on VR smart education. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0081] First of all, the purpose of this experiment is to verify the effect of collecting learners' behavioral data through the VR smart education system to generate and optimize personalized learning paths. The experimental subjects are 10 students from a high school (5 of whom are science students and 5 are liberal arts students), who participated in a learning experiment on the "mechanics" knowledge of the high school physics course. During the experiment, high-end VR equipment, including head-mounted displays (HMDs), handheld controllers, and body tracking devices, were used to collect learners' first behavioral data in the virtual learning environment. The data types include position and movement tracking, eye tracking, gesture and operation analysis, physiological response monitoring, and voice interaction analysis. The experiment lasted for 5 days, with 2 hours of learning per day.

[0082] Before the experiment began, a VR teaching model based on natural language processing (NLP) technology was built for learners. The specific process is as follows:

[0083] We conduct text analysis on mechanics textbooks and extract core concepts such as "force", "motion", and "energy". We use the TF-IDF algorithm to filter out words with high importance, and use the clustering algorithm to group similar keywords to form multiple knowledge nodes.

[0084] Construct a hierarchical 3D model, including the basic knowledge layer, intermediate knowledge layer and advanced application layer. The basic knowledge layer uses simplified animation to show the basic definition of force and the 3D effect of the force on the object. The intermediate layer adds a detailed interpretation of the force direction. The advanced application layer introduces the demonstration of a multi-degree-of-freedom vibration system.

[0085] The system records learners’ behavior data in real time. The first behavior data includes learning time, exploration path, correct answer rate, attention data (time of gaze retention and change of gaze point), gesture operation frequency, physiological response (heart rate, eye movement speed), etc. These data are used to preliminarily generate personalized first-level learning paths.

[0086] In the second phase of the experiment, four primary learning paths were generated based on the first behavior data: fast progressive path, deep reinforcement path, exploratory path, and progressive path. By simulating learning effect data, the system selects the most suitable path for each learner and monitors the learner's second behavior data in real time during the learning process to dynamically adjust the learning path. The collected second behavior data is feature extracted through sliding window technology to obtain real-time learning time trends, changes in answer accuracy, and changes in physiological load.

[0087] Finally, feedback adjustment is performed by comparing the third behavior data after learning with the simulation effect data. The system adjusts the learning path according to the comprehensive deviation score and adopts different adjustment strategies according to different score results.

[0088] Table 1 Experimental data comparison table

[0089] ,

[0090] Through the analysis of the tabular data, it can be seen that using the present invention to generate and optimize personalized learning paths has significant advantages. The data shows that different path types have different impacts on the learning effects and cognitive loads of learners. For example, students adopting the rapid progression path (students A, E, I) performed well in terms of the answering accuracy rate (85%, 88%, 86%) and the exploration path coverage rate (80%, 85%, 83%), and the average heart rate was relatively low (70 - 75 beats per minute), indicating that the rapid progression path can effectively improve learning efficiency and reduce cognitive load. This path is suitable for learners who master knowledge quickly and have strong self-learning abilities.

[0091] On the other hand, the students who chose the in-depth reinforcement path (students B, F, J) had a relatively low answering accuracy rate (62% - 65%) and a relatively high cognitive load (the average heart rate was 82 - 85 beats per minute), indicating that this type of path is more challenging for learners with relatively weak knowledge understanding. However, this path provides more detailed interpretations and repeated practice opportunities, which helps to improve the knowledge mastery level of learners with weak foundations. The use of the in-depth reinforcement path can target students with slower learning speeds and ensure in-depth knowledge mastery by extending the learning time (90 - 100 minutes).

[0092] The students on the exploratory path (students C, G) performed well in terms of the exploration path coverage rate (90%, 88%) and the answering accuracy rate (78%, 79%), and the cognitive load was medium. This indicates that the learning method of free exploration helps to stimulate the interest of learners and promote active learning. However, for students with relatively slower learning speeds or weaker knowledge bases, appropriate guidance may be needed to improve learning efficiency.

[0093] Although the students on the progressive path (students D, H) had a relatively long learning time (110 - 120 minutes), their answering accuracy rate remained at a relatively high level (77% - 80%), and the cognitive load was low (the heart rate was 70 - 72 beats per minute). This path is suitable for learners with no prior knowledge. By gradually progressing from basic knowledge to complex content, it ensures the step-by-step mastery of knowledge.

[0094] In summary, the experiments show that the present invention can not only accurately construct a personalized learning model through the collection of the first behavior data, but also dynamically adjust the learning path during the learning process to meet the actual needs of the learners. Compared with the traditional fixed teaching mode, the present invention significantly improves the learning efficiency and the level of knowledge mastery, while reducing the cognitive load of the learners, and has significant advantages especially in dealing with the personalized needs of different learners. By integrating a variety of data collection and analysis methods, the present invention effectively solves the problem in the prior art that it is difficult to optimize the dynamic learning path for individuals, and has high innovation and practicality.

[0095] Example 3. Refer to Figure 2 , which is an embodiment of the present invention, provides a teaching knowledge mining system based on VR intelligent education, including a data collection module, a learning effect evaluation module, and a path evaluation module.

[0096] Among them, the data collection module is used to obtain the first behavior data, the second behavior data, the third behavior data and the first health data of the learner. The learning effect evaluation module is used to evaluate the learning effect of the learning path in combination with the expected indicators. The path evaluation module is used to compare with the simulation effect data of the secondary learning path to form a feedback result.

[0097] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that makes a contribution to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in the various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0098] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definable sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in conjunction with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0099] More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, a computer-readable medium can even be paper or other suitable media on which a program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other suitable processing as necessary, and then stored in a computer memory.

[0100] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc. It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

[0101] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A teaching knowledge mining method based on VR smart education, characterized in that: include: Obtain learners’ first behavior data and build a VR model; Generate a first-level learning path based on the behavior data, and obtain simulated learning effect data of each path; Collect the second behavior data and combine it with the simulated learning effect data to determine the secondary learning path; Obtaining the third behavior data and the first health data after learning, and comparing them with the simulation effect data of the secondary learning path to form a feedback result; The acquiring of the first behavior data of the learner includes collecting the behavior data of the learner in the virtual learning environment through a VR device; Generating the primary learning path according to the behavior data includes using the behavior data to generate a plurality of primary learning paths having different content structures and presentation orders, each learning path being simulated according to the individual circumstances of the learner; The obtaining of the simulated learning effect data of each path includes predicting the learning effect of each path based on historical data and related features; The collecting of the second behavior data includes collecting the learner's current behavior data as the second behavior data, and the collected data type is the same as the first behavior data; Determining the secondary learning path includes comparing the learner's current behavior data with the simulation effect data of each path, determining the learning path with the best expected effect as the secondary learning path, and adapting to the learner's personalized needs and real-time status; The obtaining of the third behavior data and the first health data after learning and comparing them with the simulated effect data of the secondary learning path to form a feedback result includes evaluating the effectiveness of the learning process by comparing the new behavior data collected in the actual learning process with the simulated learning effect data of the optimal path, integrating the differences in knowledge mastery, learning efficiency and physiological load into a comprehensive deviation score through a weighted scoring method, and performing feedback adjustment of knowledge mining based on the comprehensive score. When the comprehensive deviation score is ≥0.7, and the difference scores of at least two dimensions exceed the preset self-threshold value, it is identified that the learner cannot adapt to the current learning intensity and knowledge mining intensity, and the system immediately adjusts the learning path to enhance the learner's adaptability by reducing the difficulty of the content and increasing the rest time; When 0.4≤comprehensive deviation score<0.7 is identified and the physiological load difference does not exceed the preset self-threshold, and only one of the knowledge mastery difference and learning efficiency difference exceeds the preset self-threshold, the learning effect is optimized by providing additional interesting auxiliary materials to guide learning interest; When the comprehensive deviation score is <0.4, it means that the learning process is close to the expected effect and the existing learning path is maintained; The first behavior data is the data collected when the learner enters the virtual learning environment for the first time, including learning time, interaction frequency, order of accessing knowledge points during the learning process, accuracy of answering questions, learning speed, and attention data; The second behavior data is the behavior data collected at fixed time intervals after the learner enters the first-level learning path learning stage, and the data type is the same as the first behavior data; The third behavior data is the accuracy of answering questions, the time spent on a single test question, the number of touches, and the number of answer changes collected after the learner completes the learning task of the secondary learning path; First health data: heart rate, blink frequency and gaze time; Generating a first-level learning path based on the first behavior data includes using the learner's historical learning records to simulate the learning effects under different paths, predicting the learning progress and knowledge mastery of each path, using a reinforcement learning algorithm to optimize the order and content density of each knowledge point of each path, and obtaining simulated learning effect data of each path, including predicting the learning effect of each path based on historical data and related features, and using causal inference methods to clarify the causal relationship between different features; Based on these causal relationships, a structural equation model was constructed to analyze the direct and indirect effects between various characteristics through hypothesized paths; Random forest regression is used to predict the learning effect of each path. The target variables for each path prediction include the learner's knowledge mastery, completion time, and average score improvement rate. The ensemble learning method is used to perform weighted averaging of the prediction results of multiple models; Generating a secondary learning path based on the second behavior data includes segmenting the learning behavior data using a sliding window technique, extracting time series features to perform statistical analysis on the interaction data, and calculating the total number of gesture and voice interactions, interaction frequency, and distribution; The learning behavior, emotion and physiological signals of multi-source data are fused into high-dimensional feature vectors, and the main feature dimensions are extracted through principal component analysis.

2. A system using the teaching knowledge mining method based on VR smart education as claimed in claim 1, characterized in that: Including data collection module, learning effect evaluation module, and path evaluation module; The data collection module is used to obtain the learner's first behavior data, second behavior data, third behavior data and first health data; The learning effect evaluation module is used to evaluate the learning effect of the learning path in combination with the expected indicators; The path evaluation module is used to compare with the simulation effect data of the secondary learning path to form a feedback result.

3. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the teaching knowledge mining method based on VR smart education described in claim 1 are implemented.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the teaching knowledge mining method based on VR smart education described in claim 1 are implemented.

Images