Health education interaction system and method based on virtual reality and deep learning

Abstract

The invention relates to the technical field of education, in particular to a health education interaction system and method based on virtual reality and deep learning. The health education interaction system creatively integrates motion sensing games, VR video teaching, personalized reading and a body measurement motion module, and all-around and immersive learning experience is created; the system motivates the participation degree of a user through a motion sensing game, interactive learning is achieved through VR video teaching, personalized reading meets the requirements of the user, and the health state is monitored through motion measurement. The intelligent analysis module is used for deeply mining data and providing personalized health assessment and suggestions; the self-learning unit continuously optimizes a knowledge base and a recommendation algorithm, and constructs a dynamic health knowledge graph; the system breaks through the limitation of traditional health education, improves the knowledge propagation efficiency and acceptance, effectively promotes the formation and maintenance of health behaviors, and brings revolutionary change to the field of health education.

Description

Technical Field

[0001] The present invention relates to the field of educational technology, and in particular to a health education interactive system and method based on virtual reality and deep learning. Background Art

[0002] With increasing awareness of health, health education has become a crucial topic in modern society. Traditional health education methods rely primarily on written materials, lectures, and simple online courses. These methods often suffer from poor interactivity, low personalization, and ineffective learning outcomes. In recent years, the application of virtual reality (VR) technology and deep learning algorithms in education has brought new opportunities for health education.

[0003] Currently, there are already some products on the market that attempt to apply VR technology to health education. These products use three-dimensional visualizations to showcase the human body's structure and physiological processes, providing learners with a more intuitive learning experience. However, most of these systems remain at the level of simple visual presentation, lacking in-depth interaction and personalization. For example, while some VR health education software allows users to tour the human body in a virtual environment, it cannot dynamically adjust content based on the user's learning progress and interests, nor can it provide targeted health advice.

[0004] Meanwhile, AI-based health education platforms are beginning to emerge. These platforms leverage machine learning algorithms to analyze users' learning behaviors and health data, providing personalized learning content and health recommendations. However, these systems are often limited to text and two-dimensional images, making it difficult for users to truly understand and experience complex health information.

[0005] The closest existing technology is a health education system that combines VR with simple AI capabilities. This system provides basic VR teaching content and can recommend relevant learning materials based on user preferences. However, this system still has the following technical issues:

[0006] 1. Insufficient interactivity: Although VR technology is used, user participation is still limited, making it difficult to convert learned knowledge into actual health behaviors.

[0007] 2. Low personalization: The system lacks deep learning support and cannot provide truly personalized content and recommendations based on the user's learning performance and health status.

[0008] 3. Weak data integration capabilities: The system cannot effectively integrate users’ learning data, physiological data, and behavioral data, making it difficult to form a comprehensive health assessment.

[0009] 4. Untimely feedback: Lack of real-time monitoring and analysis capabilities, unable to provide users with immediate health feedback and guidance.

[0010] 5. Fixed learning path: The system cannot dynamically adjust teaching content and difficulty based on the user's learning progress and mastery. Summary of the Invention

[0011] The present invention aims to solve the above technical problems and provide a health education interactive system and method based on virtual reality and deep learning, which can provide users with a highly interactive, personalized, and real-time feedback health education experience.

[0012] The present invention proposes a health education interactive system and method based on virtual reality and deep learning, including:

[0013] Somatosensory game module, used for:

[0014] Generate somatosensory game scenes;

[0015] Matching the user's body movements with virtual objects in the game;

[0016] The VR video teaching module is connected to the somatosensory game module data and is used to:

[0017] Play health education videos;

[0018] Set answer nodes in the video;

[0019] Dynamically adjust subsequent video content based on the user's answer at the question node;

[0020] The health education reading module is connected to the VR video teaching module data and is used to:

[0021] Provide personalized reading content based on user patterns;

[0022] Record users' reading habits and learning progress;

[0023] The physical exercise module is connected to the data of the health education reading module and is used to:

[0024] Perform user fitness tests;

[0025] Record user's motion data;

[0026] The intelligent analysis module is connected to the physical exercise module data and is used to:

[0027] Analyze users' learning data, vital signs data and exercise data;

[0028] Generate personalized health assessment reports and recommendations;

[0029] The self-learning unit is connected to the data of the intelligent analysis module and is used to:

[0030] Based on deep learning algorithms, continuously optimize the system's knowledge base and recommendation algorithms;

[0031] Build and update health knowledge graphs.

[0032] Preferably, the VR video teaching module includes:

[0033] Video playback unit, used to play health education video content;

[0034] A video retrieval unit, used to retrieve relevant video content based on keywords selected by the user;

[0035] Answering node setting unit, used to set interactive answering points in the video;

[0036] Among them, when the answering node setting unit detects that the user has reached the answering node, it pauses the video playback and triggers the question display, and continues to play the video after the user completes the answer.

[0037] Preferably, the health education reading module includes:

[0038] A user mode selection unit, used to allow the user to select gender and age group;

[0039] Content recommendation unit, used to recommend personalized health information, graphic tutorials, interactive topics and question-and-answer evaluations based on user patterns;

[0040] Reading habit recording unit, used to record the user's reading content and reading time;

[0041] The content recommendation unit dynamically adjusts the difficulty and theme of the recommended content based on the user's reading habits and learning progress.

[0042] Preferably, the body movement measurement module includes:

[0043] Warm-up test unit, used to warm up the user before the formal test;

[0044] Project test unit, used to perform multiple physical fitness test projects;

[0045] Score statistics unit, used to calculate and record the user's score in each test item;

[0046] The item testing unit dynamically adjusts the difficulty and sequence of subsequent test items based on the user's warm-up test results.

[0047] Preferably, the intelligent analysis module includes:

[0048] Data fusion unit, used to integrate user data from different modules;

[0049] A deep learning analysis unit, used to analyze the fused data using a deep neural network;

[0050] A health risk assessment unit, configured to generate a health risk assessment report for the user based on the analysis results;

[0051] A personalized recommendation generation unit, used to generate targeted health improvement recommendations based on health risk assessment results;

[0052] Among them, the deep learning analysis unit adopts transfer learning technology, which can quickly adapt to new health problems under limited sample conditions.

[0053] As an advantage, it also includes:

[0054] The interactive module is connected to the data of the intelligent analysis module and is used to:

[0055] Receive natural language questions from users;

[0056] understand the intent of the question and retrieve relevant information from the health knowledge base;

[0057] Generate targeted responses and output them to the user in voice form.

[0058] Preferably, the interaction module includes:

[0059] A speech recognition unit, used to convert user voice input into text;

[0060] Natural language understanding unit, used to analyze text content and extract keywords and intent;

[0061] Knowledge retrieval unit, used to find relevant information in the health knowledge base;

[0062] An answer generation unit, used to generate a complete answer based on the retrieval results;

[0063] A speech synthesis unit, used to convert text answers into speech output;

[0064] Among them, the natural language understanding unit adopts a deep learning model based on the attention mechanism, which can accurately capture key information in long texts.

[0065] Preferably, the somatosensory game module includes:

[0066] A motion capture unit, used to capture the user's body movements in real time;

[0067] A virtual scene generation unit, used to create 3D health education game scenes;

[0068] An interaction matching unit, used to match user actions with operations on virtual objects;

[0069] Game logic control unit, used to manage game rules and processes;

[0070] Among them, the interactive matching unit adopts an algorithm based on deep reinforcement learning, which can adapt to the action characteristics of different users and improve the accuracy and fluency of matching.

[0071] As an advantage, it also includes:

[0072] The sensor array is connected to the intelligent analysis module for:

[0073] Real-time collection of user's physiological data, including heart rate, blood pressure, body temperature, etc.;

[0074] Transmit the collected data to the intelligent analysis module for processing;

[0075] The intelligent analysis module generates a more comprehensive and timely health status assessment based on the real-time physiological data collected by the sensor array and combined with the user's learning and exercise data.

[0076] The interactive health education method based on virtual reality and deep learning includes the following steps:

[0077] a) User authentication:

[0078] Verify user identity through facial recognition technology;

[0079] Retrieve the user's historical learning progress and health data;

[0080] b) Teaching content display:

[0081] Select appropriate VR teaching videos, somatosensory games, or reading materials based on the user's learning progress and interests;

[0082] Present teaching content through VR headsets or other display devices;

[0083] c) Vital signs information collection:

[0084] Use sensor arrays to monitor the user's physiological indicators in real time;

[0085] Associate the collected vital sign data with the current learning content;

[0086] d) Interactive teaching:

[0087] Set up question-and-answer nodes in VR videos and adjust subsequent content based on user answers;

[0088] Provide real-time feedback and guidance based on user actions in motion-sensing games;

[0089] e) Learning feedback:

[0090] Generate personalized health assessment reports based on the user's learning performance, physical data and interaction;

[0091] Use deep learning algorithms to analyze user data and provide targeted health advice and learning plans;

[0092] f) Knowledge graph update:

[0093] Collect new questions and feedback from users during the learning process;

[0094] Utilize self-learning algorithms to continuously optimize and expand the health knowledge graph;

[0095] g) Learning progress record:

[0096] Track and record users' learning time, completed courses and test scores;

[0097] Synchronize learning progress data to user accounts and support multi-device access;

[0098] Among them, the deep learning algorithm in step e) adopts multimodal data fusion technology, which can comprehensively analyze the user's video learning behavior, somatosensory game performance, reading habits and physiological indicators, thereby generating more comprehensive and personalized health recommendations.

[0099] The interactive health education system of this invention achieves a paradigm shift in health education at a macro level through the innovative integration of virtual reality technology and deep learning algorithms. The system is no longer a simple knowledge transfer tool, but an intelligent, personalized health management platform that comprehensively improves users' health literacy and behavior.

[0100] From an overall architectural perspective, the system of the present invention achieves seamless integration and collaborative operation of multiple functional modules. Core components such as the VR video teaching module, somatosensory gaming module, health education reading module, physical exercise module, and intelligent analysis module work together to form a closed-loop health education ecosystem. This design not only improves the overall efficiency of the system but also provides users with a comprehensive and coherent learning experience.

[0101] In terms of inter-module collaboration, this invention cleverly resolves the disconnect between theory and practice in traditional health education. For example, the knowledge imparted by the VR video teaching module can be immediately put into practice and reinforced in the somatosensory game module, while the data collected by the physical exercise module can be used by the intelligent analysis module to adjust the teaching content and difficulty. This close collaboration ensures the coherence and effectiveness of the learning process.

[0102] From the perspective of complementary and superimposed effects, the various modules of the present invention form a positive synergistic effect. The immersive and intuitive nature of VR technology, combined with the personalized recommendations of deep learning algorithms, not only improves learning efficiency but also greatly enhances user engagement and satisfaction. The combination of real-time vital sign monitoring and intelligent analysis provides users with timely and accurate health feedback, which is difficult to achieve in traditional health education.

[0103] At the micro level, this invention has achieved innovative breakthroughs in multiple technical details:

[0104] 1. In terms of VR content presentation, the system adopts a highly interactive design, allowing users to not only see, but also touch and manipulate virtual human structures and physiological processes. This multi-sensory learning approach significantly enhances knowledge comprehension and retention.

[0105] 2. In terms of personalized recommendation algorithms, the system uses advanced deep learning models that can comprehensively analyze users' learning behaviors, physiological data, and health conditions to provide highly customized learning content and health recommendations.

[0106] 3. In terms of data analysis, the system has developed an innovative multimodal data fusion algorithm that can effectively integrate health-related data from different sources to form a comprehensive and accurate health portrait.

[0107] 4. In terms of real-time feedback mechanism, the system realizes real-time monitoring of the user's physiological state and learning status through smart wearable devices and VR devices, and provides immediate guidance and adjustment in combination with deep learning algorithms.

[0108] 5. In the design of learning paths, the system adopts dynamic difficulty adjustment technology, which can optimize the learning content and difficulty in real time according to the user's performance, ensuring that learning is always at the best challenge level.

[0109] In summary, the interactive health education system of this invention effectively addresses many issues existing in traditional health education through technological innovation and intelligent design. It not only improves the efficiency and acceptance of health knowledge dissemination, but also effectively promotes the formation and maintenance of healthy behaviors. This new health education model is expected to play a significant role in improving public health literacy, preventing chronic diseases, and promoting healthy lifestyles, possessing broad application prospects and enormous social value. BRIEF DESCRIPTION OF THE DRAWINGS

[0110] Figure 1 It is the overall logic block diagram of the system of the present invention;

[0111] Figure 2 This is a logic block diagram of the somatosensory game module of the present invention;

[0112] Figure 3This is a logic block diagram of the VR video teaching module of the present invention;

[0113] Figure 4 It is a logic block diagram of the health education reading module of the present invention;

[0114] Figure 5 This is a logic block diagram of the body motion measurement module of the present invention;

[0115] Figure 6 This is a logic block diagram of the intelligent analysis module of the present invention;

[0116] Figure 7 This is a logic block diagram of the interaction module of the present invention. DETAILED DESCRIPTION

[0117] Please refer to the attached Figure 1-7 The present invention provides a health education interactive system and method based on virtual reality and deep learning. This system innovatively combines virtual reality technology and deep learning algorithms to provide users with an immersive, personalized health education experience. The following describes specific implementations of the present invention in detail.

[0118] The health education interactive system of the present invention includes a somatosensory game module 1, a VR video teaching module 2, a health education reading module 3, a physical exercise module 4, an intelligent analysis module 5, and a self-learning unit 6. These modules work together through data connections to form a complete health education ecosystem.

[0119] The motion-sensing game module 1 is a crucial component of the system. Its primary function is to generate the motion-sensing game scene and match the user's body movements with virtual objects in the game. Preferably, the motion-sensing game module 1 utilizes high-precision motion capture technology, enabling real-time tracking of the user's movements. For example, the module can use a combination of a depth camera and an inertial measurement unit (IMU) to collect data on the user's skeletal joint positions at a frequency of 60Hz. This high sampling rate ensures a smooth gaming experience with no noticeable lag.

[0120] In one embodiment of the present invention, the motion sensing game module 1 also includes an action recognition algorithm based on a long short-term memory (LSTM) network. Specifically, the algorithm's input is the user's skeletal joint position sequence, and its output is the recognized action category. The core formula of the algorithm is as follows:

[0121] ,

[0122] ,

[0123] tanh ,

[0124] in, is hidden state, is the unit state, is the input sequence, is the weight matrix, is the bias term, is the activation function, In this way, the somatosensory game module 1 can accurately recognize the user's complex movements, thereby providing a richer and more interactive game experience.

[0125] VR Video Teaching Module 2 is data-connected to Somatosensory Game Module 1 and is responsible for playing health education videos, setting question-answering nodes, and dynamically adjusting subsequent video content based on user responses. This interactive video learning method significantly improves user engagement and learning outcomes. For example, in a VR video about heart health, the system might set a question-answering node after explaining the structure of the heart, asking the user, "What is the main function of the heart?" Based on the user's response, the system can decide whether to continue the explanation in depth or review the previous content.

[0126] One of the innovations of the present invention is that the VR video teaching module 2 adopts an adaptive learning algorithm. This algorithm is based on the Bayesian knowledge tracking model, and its core formula is as follows:

[0127] ,

[0128] in, Indicates time The user's knowledge state at the time Represents the user's observation behavior (such as answering questions). Through this model, the system can estimate the user's mastery of each knowledge point in real time, thereby dynamically adjusting the difficulty and depth of the video content.

[0129] Health Education Reading Module 3 is connected to the VR Video Teaching Module 2 to provide personalized reading content based on user patterns and track their reading habits and learning progress. This module is designed to provide users with comprehensive health knowledge input. For example, for a 35-year-old male user, the system might recommend articles on topics such as workplace stress management and common health issues among middle-aged people.

[0130] When recording a user's reading habits, the Health Education Reading Module 3 collects data such as reading time, reading speed, and article topic preferences. This data is processed through a multi-layer perceptron (MLP) model to generate a user's reading interest vector. The mathematical representation of the model is as follows:

[0131] ,

[0132] ,

[0133] softmax ,

[0134] in, is the input feature, and is the hidden layer, is the output interest vector, and In this way, the system can accurately grasp the user's reading preferences and provide more personalized content recommendations.

[0135] The VR video teaching module 2 includes a video playback unit 21, a video retrieval unit 22, and a question answering node setting unit 23. This modular design enables the system to flexibly manage and display teaching content.

[0136] The video playback unit 21 is responsible for playing health education video content. Preferably, this unit supports 4K resolution and 60fps video playback to ensure a high-quality visual experience in a VR environment. Furthermore, the video playback unit 21 also integrates video buffering technology, dynamically adjusting video quality based on the user's network conditions to ensure a smooth playback experience.

[0137] The video retrieval unit 22 is used to retrieve relevant video content based on user-selected keywords. The present invention uses deep learning-based video content understanding technology to extract semantic features from the visual, audio, and text information of the video. Specifically, the system uses a multimodal Transformer model, whose core formula is as follows:

[0138] Attention softmax ,

[0139] in, 、 、 denote query, key, and value matrices respectively, Through this attention mechanism, the system can effectively integrate information from different modalities and achieve more accurate video retrieval.

[0140] The answer node setting unit 23 sets interactive answer points in the video. When it detects that the user has reached an answer node, the unit pauses the video playback and triggers the question display. Preferably, the difficulty of the question is dynamically adjusted based on the user's previous performance. For example, if the user answers three simple questions correctly in a row, the system may provide a more challenging question at the next node. This adaptive difficulty adjustment mechanism helps maintain the user's learning interest and motivation.

[0141] The health education reading module 3 includes a user mode selection unit 31, a content recommendation unit 32, and a reading habit recording unit 33. This design fully considers the individual differences of users and provides each user with a tailored learning experience.

[0142] The user mode selection unit 31 allows the user to select gender and age group. This basic information is an important basis for personalized recommendations. For example, the system may recommend more content about osteoporosis prevention for female users over 50 years old, while recommending more content about physical training for male users aged 20-30 years old.

[0143] The content recommendation unit 32 recommends personalized health information, illustrated tutorials, interactive topics, and question-and-answer assessments based on user patterns. This invention employs a hybrid recommendation algorithm based on collaborative filtering and content-based recommendations. Its core concept is to combine user historical behavior data with content features to generate more accurate recommendations. The main steps of the algorithm are as follows:

[0144] 1. Calculate the user-content interaction matrix

[0145] 2. Use matrix decomposition technology to obtain latent vector representations of users and content

[0146] 3. Calculate content feature vector

[0147] 4. Combine latent vectors and feature vectors to predict user interest in content using deep neural networks

[0148] Specifically, the formula for predicting interest can be expressed as:

[0149] ReLU ,

[0150] in, and are the latent vectors of users and content respectively, is the feature vector of the content, and are network parameters, is the sigmoid function.

[0151] The reading habit recording unit 33 is responsible for recording the user's reading content and reading time. This data is not only used for personalized recommendations, but also for generating learning reports for the user. For example, the system can estimate the user's knowledge absorption efficiency based on the user's reading time and reading speed. If the user's reading speed for a certain type of content is found to be significantly slower, it may mean that the user is not familiar with the topic. The system will adjust the recommendation strategy accordingly and provide more basic knowledge content.

[0152] The detailed description above demonstrates that the interactive health education system leverages the strengths of virtual reality technology and deep learning algorithms to provide users with a comprehensive, personalized, and interactive health education platform. The system's various modules work closely together to form a complete learning loop, effectively enhancing users' health knowledge and awareness.

[0153] The interactive health education system of the present invention further includes a physical fitness module 4, which is data-connected to the health education reading module 3 and is used to perform a user's physical fitness test and record the user's exercise data. The introduction of physical fitness module 4 enables the system to comprehensively assess the user's health status, focusing not only on theoretical knowledge learning but also on the improvement of practical physical fitness.

[0154] The physical test module 4 includes a warm-up test unit 41, an item test unit 42, and a score statistics unit 43. This modular design makes the physical test process more scientific and reasonable, and can comprehensively evaluate the user's physical fitness.

[0155] The warm-up test unit 41 warms up the user before the formal test. Preferably, the warm-up test includes a series of simple stretching and low-intensity aerobic exercises, such as jogging in place, arm circles, etc. This process usually lasts for 5 to 10 minutes, with the purpose of increasing the user's body temperature and muscle flexibility, and reducing the risk of injury in subsequent tests. In one embodiment of the present invention, the warm-up test unit 41 also integrates a real-time heart rate monitoring function, which determines whether the warm-up is sufficient by analyzing the changes in the user's heart rate. For example, the system may set a target heart rate range, such as 50% to 60% of the maximum heart rate, and only when the user's heart rate is stable within this range will it enter the next stage of testing.

[0156] The item test unit 42 is responsible for executing multiple physical fitness test items. The system of the present invention provides a comprehensive set of test items, including but not limited to:

[0157] 1. Cardiopulmonary endurance test: such as a 12-minute treadmill test or a step test;

[0158] 2. Muscle strength test: such as push-ups, sit-ups, etc.;

[0159] 3. Flexibility test: such as sit-and-reach test;

[0160] 4. Balance ability test: such as single-leg standing test;

[0161] 5. Coordination test: such as alternating touch test;

[0162] Preferably, the item testing unit 42 dynamically adjusts the difficulty and order of subsequent test items based on the user's warm-up test results. For example, if the system detects a slow recovery of the user's heart rate during the warm-up, the cardiopulmonary endurance test may be reduced in intensity or rescheduled later in the test sequence. This dynamic adjustment mechanism effectively improves the safety and scientific nature of the test.

[0163] The score statistics unit 43 is used to calculate and record the user's score in each test item. The present invention adopts a standardized scoring system based on age and gender. Specifically, the system uses Z scores (standard scores) to evaluate the user's performance in each test item. The calculation formula of Z scores is as follows:

[0164] ,

[0165] in, is the user's raw score, is the average score for the same age group and gender. is the standard deviation. In this way, the system can fairly evaluate the performance of users of different ages and genders and provide more personalized health recommendations.

[0166] The health education interactive system of the present invention also includes an intelligent analysis module 5, which is data-connected to the physical measurement and exercise module 4. The intelligent analysis module 5 is one of the core components of the system, responsible for analyzing the user's learning data, physical sign data, and exercise data, and generating personalized health assessment reports and recommendations.

[0167] In a preferred embodiment of the present invention, the intelligent analysis module 5 includes a data fusion unit 51, a deep learning analysis unit 52, a health risk assessment unit 53, and a personalized recommendation generation unit 54. This structural design enables the system to comprehensively and deeply analyze user data and provide high-quality health guidance.

[0168] The data fusion unit 51 is responsible for integrating user data from different modules. This process involves processing multi-source heterogeneous data, which presents a complex technical challenge. This invention employs a multimodal data fusion algorithm based on tensor decomposition. The core idea of this algorithm is to represent different types of data as high-dimensional tensors and then extract common hidden features through tensor decomposition techniques. Specifically, the algorithm uses CANDECOMP / PARAFAC (CP) decomposition, whose mathematical expression is as follows:

[0169] ,

[0170] in, is the original data tensor, 、 、 is the factor vector after decomposition, is the rank of the decomposition. In this way, the system can effectively integrate data from multiple sources such as somatosensory games, VR video learning, reading behavior, and physical test results, providing a unified data representation for subsequent analysis.

[0171] The deep learning analysis unit 52 uses a deep neural network to analyze the fused data. The present invention adopts an innovative multi-task learning framework that can simultaneously predict the user's health risks, learning effects, and exercise performance. The network structure of this framework is as follows:

[0172] 1. Shared feature extraction layer: Use multi-layer convolutional neural network to extract underlying features

[0173] 2. Task-specific fully connected layers: Design specific fully connected layers for different prediction tasks

[0174] 3. Soft parameter sharing mechanism: realizing knowledge transfer between different tasks through L2 regularization

[0175] This multi-task learning framework not only improves the model's generalization ability but also effectively alleviates the data sparsity problem. For example, even if some users lack exercise data, the system can still make reasonable health assessments based on their learned behavior and physical sign data.

[0176] The health risk assessment unit 53 generates a health risk assessment report for the user based on the results of deep learning analysis. The present invention adopts a risk assessment model based on fuzzy logic, which can handle the uncertainty and ambiguity of data. The core of the model is a set of fuzzy rules, such as:

[0177] If BMI is high AND cardiorespiratory endurance is low AND sedentary time is long, then the risk of cardiovascular disease is high

[0178] Linguistic variables such as higher and lower are defined using fuzzy sets. The final risk score is derived through fuzzy reasoning using the centrifugal method. This approach not only considers the combined impact of multiple risk factors but also provides a more user-friendly interpretation of risk.

[0179] Based on the health risk assessment results, the personalized recommendation generation unit 54 generates targeted health improvement recommendations. This invention utilizes a recommendation generation algorithm based on a knowledge graph and reinforcement learning. The system first constructs a vast knowledge graph based on health domain knowledge, containing entities such as various health concepts, risk factors, and intervention measures, as well as their relationships. It then uses a reinforcement learning algorithm to perform path search on this knowledge graph, generating a personalized health improvement plan for the user.

[0180] Preferably, the personalized recommendation generating unit 54 also integrates a progress tracking and adjustment mechanism. The system regularly evaluates the user's implementation of the recommendations and dynamically adjusts the recommendations based on feedback. For example, if it is found that the user is struggling to stick to 30 minutes of aerobic exercise per day, the system may adjust the recommendation to three 10-minute short exercises per day to improve user compliance.

[0181] The health education interactive system of the present invention further includes an interactive module 7, which is data-connected to the intelligent analysis module 5. The introduction of the interactive module 7 greatly improves the user-friendliness of the system, allowing users to interact with the system through natural language and obtain health information and suggestions.

[0182] Interaction Module 7's primary functions include receiving natural language questions from users, understanding the question's intent, retrieving relevant information from the health knowledge base, and finally generating targeted responses and outputting them to the user in voice format. This interactive approach not only allows users to access health information at any time but also provides more personalized health consultation services.

[0183] In one embodiment of the present invention, interaction module 7 utilizes the latest natural language processing technology. The system uses an intent recognition model based on BERT (Bidirectional Encoder Representations from Transformers), enabling accurate understanding of the underlying semantics of user questions. Furthermore, the system integrates knowledge graph reasoning technology, enabling multi-hop reasoning within the health knowledge graph based on the user's question, providing a more comprehensive and in-depth answer.

[0184] For example, when a user asks how to prevent diabetes, the system not only provides direct preventive measures but also offers personalized recommendations based on the user's personal health data (such as BMI, exercise habits, etc.). If the user further inquires about the reasons for a particular recommendation, the system can provide a scientific explanation based on the knowledge graph, such as how controlling carbohydrate intake affects blood sugar levels.

[0185] Preferably, the interaction module 7 also includes a sentiment analysis function that can identify the emotional state of the user's voice. For example, if the system detects anxiety or frustration in the user's tone, it will adjust the tone and content of the answer accordingly to provide more psychological support and encouragement.

[0186] The detailed description above demonstrates that the interactive health education system not only provides comprehensive health assessments and personalized recommendations, but also enables users to easily access health information through advanced interactive technology. This innovative design, combining virtual reality, deep learning, and natural language processing, provides users with an intelligent, user-friendly, and efficient health education platform.

[0187] The interactive module 7 of the present invention further includes a speech recognition unit 71, a natural language understanding unit 72, a knowledge retrieval unit 73, an answer generation unit 74, and a speech synthesis unit 75. This fine module division enables the system to efficiently and accurately process the user's speech input and provide natural and fluent speech output.

[0188] The speech recognition unit 71 is responsible for converting the user's speech input into text. This invention employs an end-to-end speech recognition model based on deep learning, specifically using the encoder-decoder structure of the Transformer architecture. A significant advantage of this model is its ability to handle long speech inputs, which is particularly important when users raise complex health questions. The core attention mechanism of the model can be expressed as:

[0189] ,

[0190] Among them, Q, K, and V represent query, key, and value respectively. This attention mechanism enables the model to capture long-range dependencies in speech and improve recognition accuracy.

[0191] The natural language understanding unit 72 is responsible for analyzing the text content and extracting keywords and intents. The present invention uses an innovative multi-task learning framework to simultaneously perform intent classification and entity recognition. The network structure of this framework includes:

[0192] 1. Shared BERT encoding layer

[0193] 2. Task-specific bidirectional LSTM layer

[0194] 3. CRF (Conditional Random Field) output layer for entity recognition

[0195] 4.Softmax output layer for intent classification

[0196] This design not only improves the model's generalization capabilities but also effectively handles specialized terms and complex expressions in the health field. For example, when a user asks, "I've been feeling tired lately. Is this a vitamin D deficiency?" the system accurately identifies the symptom entity "fatigue" and the intent of asking for possible causes.

[0197] The knowledge retrieval unit 73 searches for relevant information in the health knowledge base. The present invention employs a semantic retrieval method based on a knowledge graph. The system first maps the user query to entities and relationships in the knowledge graph and then uses a graph neural network (GNN) for multi-hop reasoning. The core update formula of the GNN is as follows:

[0198] ,

[0199] in, represents the feature vector of node v, represents the neighbor set of node v, and is a learnable parameter. Through multi-layer graph convolution, the system can capture complex semantic relationships and provide more comprehensive and in-depth health information.

[0200] The answer generation unit 74 generates a complete answer based on the search results. This invention uses a generative model based on the GPT (Generative Pre-trained Transformer) and introduces an innovative knowledge fusion mechanism. Specifically, the system dynamically incorporates relevant information from the knowledge graph during the decoding process, ensuring that the generated answer is both fluent and natural, as well as professional and accurate. The model's loss function includes language model loss and knowledge fusion loss:

[0201] ,

[0202] in, is the standard language model cross entropy loss, is the knowledge fusion loss, is a hyperparameter that balances the two.

[0203] The speech synthesis unit 75 converts the text responses into speech output. This system utilizes the latest Neural Text-to-Speech (Neural TTS) technology, using a model based on Tacotron 2 and WaveNet. This approach generates highly natural speech with excellent prosody and emotional expressiveness. For example, when the system provides positive health advice, the synthesized speech has an encouraging tone; when explaining serious health risks, the speech becomes more serious and concerned.

[0204] The somatosensory game module 1 of the present invention further comprises a motion capture unit 11, a virtual scene generation unit 12, an interactive matching unit 13 and a game logic control unit 14. This modular design enables the system to provide an immersive and highly interactive health education game experience.

[0205] The motion capture unit 11 is responsible for capturing the user's body movements in real time. This invention utilizes an innovative multimodal motion capture technology that combines data from a depth camera, an inertial measurement unit (IMU), and a pressure sensor. The system uses a Kalman filter algorithm to fuse these data sources. Its state update equation is as follows:

[0206] ,

[0207] in, is the current state vector, is the state transition matrix, is the control vector, This method can provide high-precision motion capture results while ensuring real-time performance.

[0208] The virtual scene generation unit 12 is responsible for creating a 3D health education game scene. This invention utilizes programmatic generation technology to dynamically generate a suitable gaming environment based on the user's health status and learning goals. For example, for a user who needs to improve their cardiopulmonary function, the system might generate a running scene with various obstacles; for a user who needs to practice balance, the system might create a virtual environment simulating a yoga class.

[0209] The interactive matching unit 13 is responsible for matching user actions with virtual object manipulation. This invention utilizes an algorithm based on deep reinforcement learning that adapts to the motion characteristics of different users, improving matching accuracy and smoothness. The core of the algorithm is a Double Q-Network, whose update formula is as follows:

[0210] ,

[0211] in, is the action-value function, It's a state. It's action. It's a reward. is the discount factor, is the learning rate. Through this method, the system can continuously optimize the action matching strategy and provide users with a more natural and intuitive interactive experience.

[0212] The game logic control unit 14 is responsible for managing game rules and progress. This invention has designed a dynamic difficulty adjustment (DDA) system that can adjust the game difficulty in real time based on user performance. The system uses a Bayesian optimization algorithm to find the optimal difficulty parameter, with the objective function defined as:

[0213] ,

[0214] in, is the difficulty parameter vector, is the weight coefficient. By optimizing this function, the system can strike a balance between maintaining user engagement, learning effect, and health benefits.

[0215] The health education interactive system of the present invention further includes a sensor array 8, which is data-connected to the intelligent analysis module 5. The introduction of the sensor array 8 enables the system to monitor the user's physiological state in real time, providing important data support for health assessment and personalized recommendations.

[0216] Sensor array 8 includes various types of sensors, such as heart rate sensors, blood pressure monitors, and thermometers. These sensors are non-invasive and portable, enabling continuous data collection while the user performs daily activities. For example, the heart rate sensor utilizes photoplethysmography (PPG) technology, enabling 24 / 7 monitoring via wearable devices such as smartwatches.

[0217] This invention features innovative design for sensor data processing. The system uses an anomaly detection algorithm based on a long short-term memory (LSTM) network to identify abnormal patterns in physiological data. The core formula of the algorithm is as follows:

[0218] ,

[0219] ,

[0220] ,

[0221] ,

[0222] ,

[0223] in, 、 、 They are forget gate, input gate and output gate, is the cell state, is the hidden state, is the input vector, and are weight and bias parameters. Through this method, the system can promptly detect abnormal changes in the user's physiological state and trigger corresponding health warnings.

[0224] The data quality assessment algorithm is primarily used during the vital sign information collection phase to identify and filter technical anomalies that arise during the sensor collection process. These anomalies include signal loss due to loose sensors, noise caused by environmental interference, packet loss during transmission, and abnormal readings caused by device failure. The LSTM anomaly detection algorithm, applied during the intelligent analysis phase, identifies anomalies in the user's physiological state, such as a sudden increase in heart rate during normal activity, unusual blood pressure fluctuations, body temperature outside the normal range, and abnormal combinations of multiple physiological indicators. These anomalies represent health issues and require detection based on data quality assurance. The system uses an LSTM network to analyze time-series physiological data and identify patterns that could pose potential health risks. This is a two-stage anomaly detection process: first, ensuring data quality, then analyzing this reliable data for physiological anomalies.

[0225] Preferably, the sensor array 8 also includes a data compression and encryption module for protecting the user's privacy data. The system uses a data compression algorithm based on compressed sensing, the mathematical expression of which is:

[0226] ,

[0227] in, is the compressed signal, is the measurement matrix, It is the original signal. This method can significantly reduce the amount of data transmission while retaining the key information of the signal. For encryption, the system uses homomorphic encryption technology, which allows data analysis in an encrypted state, further enhancing the security of the data. It is mainly used for simple statistical analysis and basic feature calculations, rather than complete deep learning models; it is applied to the preliminary processing stage of the user's sensitive physiological data, such as the calculation of basic statistics such as heart rate average, standard deviation, and blood pressure fluctuation range; it adopts a hybrid architecture of "homomorphic encryption + secure multi-party computing (MPC)"; simple addition operations are implemented through homomorphic encryption; complex nonlinear calculations (such as the activation function in LSTM) are implemented through secure multi-party computing protocols; data remains encrypted on the server side, and the processing results are decrypted only under authorized conditions; restricted processing of approximate calculations: For nonlinear operations that must be performed in the encrypted domain, the system limits the polynomial approximation to within 3rd order, and clearly defines the upper limit of the approximate error. When the predicted error exceeds the threshold, the system will trigger a more secure but computationally more expensive MPC protocol; a trade-off mechanism between model accuracy and privacy protection is established, allowing users to set the privacy protection level in the UI interface;

[0228] This design ensures that data privacy is protected without significantly compromising analysis accuracy due to the limitations of homomorphic encryption. Rather than attempting to run a full deep neural network under fully homomorphic encryption (which is unrealistic given current technology), the system employs a layered secure computing strategy, selecting appropriate privacy-preserving techniques based on the type of computation.

[0229] In this way, the present invention achieves a practical balance between data security and model effectiveness, which is also an important innovation of the system in terms of privacy protection.

[0230] To further illustrate how the system implements deep learning capabilities while protecting privacy, the following are the specific strategies adopted by this invention:

[0231] Model segmentation: Split the deep neural network model into a linear part that can be executed in the encrypted domain and a nonlinear part that requires plaintext processing;

[0232] Secure computation protocols: For operations that cannot be performed efficiently under homomorphic encryption, use secure multi-party computation (MPC) protocols such as Yao's obfuscated circuits or the GMW protocol;

[0233] Customized simplified models: To adapt to the limitations of encryption calculations, the system uses specially designed simplified models in some application scenarios, such as linear SVM instead of complex neural networks;

[0234] Federated learning integration: Part of the calculation is performed on the user's local device, and only the intermediate results are transmitted, reducing the amount of data that needs to be encrypted;

[0235] This multi-level secure computing architecture is the key innovation of this invention in data privacy protection. It not only ensures the security of user health data but also maintains the effectiveness of the system analysis function.

[0236] The present invention also provides an interactive health education method based on virtual reality and deep learning. This method includes steps such as user identity authentication, teaching content presentation, vital sign information collection, interactive teaching, learning feedback, knowledge graph update, and learning progress recording. This method design fully reflects the systematic and personalized features of the present invention and can provide users with a comprehensive health education experience.

[0237] In the user identity authentication step, the present invention adopts a multimodal biometric technology that combines facial recognition, voiceprint recognition and behavioral feature analysis. The system uses a multimodal fusion network based on the attention mechanism, and its core formula is as follows:

[0238] ,

[0239] in, is the fused feature vector, is the attention weight of each modality, is the feature extraction function of each modality, This method not only improves the accuracy of identity authentication, but also enhances the security of the system.

[0240] During the teaching content presentation step, the system dynamically selects appropriate VR teaching videos, somatosensory games, or reading materials based on the user's learning progress and interests. This invention uses a content recommendation algorithm based on a multi-armed bandit (MBT) to strike a balance between exploring new content and leveraging known, effective content. The core of the algorithm is the upper confidence bound (UCB) strategy:

[0241] ,

[0242] in, It's content The average reward, is the total number of attempts, It's content This method can effectively personalize the learning path and improve the user's learning efficiency and interest.

[0243] During the vital sign information collection step, the system uses a sensor array to monitor the user's physiological indicators in real time. The present invention has developed an innovative data quality assessment algorithm that can automatically identify and process noisy data. The algorithm is based on a combination of wavelet transform and machine learning, and its mathematical expression is:

[0244] ,

[0245] in, is the data quality score, is the wavelet transform, is the signal-to-noise ratio, Entropy is the signal entropy, is a trained random forest model. In this way, the system can ensure the reliability of physiological data used for health assessment.

[0246] In the interactive teaching step, the system sets up question-and-answer nodes in the VR video and provides real-time feedback based on the user's actions in the somatosensory game. This invention uses an interactive strategy optimization algorithm based on reinforcement learning, which can dynamically adjust the interaction mode according to the user's response. The core of the algorithm is the policy gradient method, and its update formula is:

[0247] ,

[0248] in, is the strategy parameter, is the learning rate, is a performance measurement function. Through this method, the system can continuously optimize the teaching interaction strategy and improve the user's learning experience and effect.

[0249] In the learning feedback step, the system generates a personalized health assessment report based on the user's learning performance, physical data, and interaction. This paper develops a multi-dimensional health scoring model that comprehensively considers factors such as knowledge mastery, behavioral changes, and improvements in physiological indicators. The mathematical expression of the model is:

[0250] ,

[0251] in, The weight of each dimension is determined by expert knowledge and machine learning algorithms. This comprehensive scoring method can provide users with more objective and valuable health feedback.

[0252] The detailed description above demonstrates that the interactive health education method of this invention leverages the strengths of virtual reality and deep learning technologies to provide users with a comprehensive, personalized, and highly interactive health education experience. Each step of this method has been carefully designed, incorporating multiple innovative technologies. This method not only effectively enhances users' health knowledge and awareness, but also monitors and improves their physical condition in real time.

[0253] In the knowledge graph update step, the present invention adopts a dynamic knowledge graph construction technology. The system can continuously collect new health knowledge points and relationships from the user's learning process and integrate them into the existing knowledge graph. This process uses a knowledge fusion algorithm based on graph neural network, and its core formula is as follows:

[0254] ,

[0255] in, is the eigenvector of node v, is the set of neighbors of node v, and is a learnable parameter. In this way, the system can continuously expand and optimize the health knowledge base to provide users with the latest and most relevant health information.

[0256] In the learning progress recording step, the present invention designs a multi-dimensional learning progress tracking system. This system not only records the user's learning time and course completion status, but also constructs a detailed knowledge mastery map by analyzing the user's performance on various health topics. The system uses an ability assessment model based on Item Response Theory (IRT), whose mathematical expression is:

[0257] ,

[0258] in, is the probability of answering a question correctly, is the user's capability parameter, 、 、 These are the discrimination, difficulty, and guessing parameters of the questions. Through this method, the system can accurately assess the user's mastery of various health knowledge points, providing an important basis for subsequent personalized learning.

[0259] Preferably, the method of the present invention also includes an emotion calculation step for analyzing the user's emotional state during the learning process. The system uses a multimodal emotion recognition technology that combines facial expression analysis, speech emotion recognition, and physiological signal processing. Specifically, the system uses a deep learning model based on the attention mechanism, and its architecture is as follows:

[0260] 1. Feature extraction layer: extract features of visual, audio and physiological signals respectively;

[0261] 2. Modality fusion layer: uses the self-attention mechanism to fuse features of different modalities;

[0262] 3. Temporal modeling layer: Use bidirectional LSTM to capture the temporal dynamics of emotions;

[0263] 4. Classification layer: Use the softmax function to output the final emotion category;

[0264] This affective computing technology enables the system to promptly identify users' learning emotions and adjust teaching strategies accordingly, such as providing encouragement when users feel frustrated, or increasing the fun of teaching content when users feel bored.

[0265] The method of the present invention also places special emphasis on privacy protection and data security. The system employs end-to-end encryption technology during data transmission and storage. Specifically, the system uses a privacy-preserving machine learning method based on homomorphic encryption, allowing data analysis to be performed in an encrypted state. The core of this technology is the partially homomorphic encryption algorithm, whose encryption function can be expressed as:

[0266] ,

[0267] in, It is a plaintext message. and is a randomly chosen integer, This encryption method ensures the security of the user's health data throughout the entire processing process and effectively prevents privacy leaks.

[0268] The above detailed description demonstrates the significant innovation and advancement of the interactive health education system and method of the present invention in terms of technological implementation. By cleverly combining cutting-edge technologies such as virtual reality, deep learning algorithms, knowledge graphs, and multimodal interaction, the system creates a comprehensive, intelligent, and personalized health education platform. This innovation not only enhances the effectiveness and appeal of health education but also provides users with real-time health monitoring and personalized recommendations, potentially playing a significant role in improving public health literacy and preventing chronic diseases.

[0269] The system and method of this invention also boast excellent scalability and adaptability. Through a modular design and advanced machine learning algorithms, the system is able to continuously learn and optimize, adapting to the needs of diverse user groups and the evolving health knowledge landscape. This design lays a solid technical foundation for future applications in telemedicine, chronic disease management, mental health, and other fields.

[0270] In summary, the virtual reality and deep learning-based interactive health education system and method provided by this invention represents a significant technological innovation in the field of health education. It not only effectively enhances users' health knowledge and awareness, but also truly encourages them to develop healthy lifestyles through immersive experiences and personalized guidance. This innovative health education model is expected to have a profound and positive impact on public health.

[0271] To validate the superiority of the virtual reality and deep learning-based health education interactive system and method, we designed a set of comparative experiments. The experiments simulated an eight-week health education course for 100 adults aged 25-55 who were seeking to improve their health.

[0272] Example 1 adopts the complete system of the present invention, including all functional modules such as VR video teaching, somatosensory games, personalized reading, real-time vital sign monitoring and intelligent analysis.

[0273] Comparative Example 1 adopts the traditional online health education method, mainly learning through web videos and text materials, without VR and somatosensory interaction functions.

[0274] Comparative Example 2 uses a simplified VR health education system with basic VR video functions, but lacks personalized recommendations and intelligent analysis functions supported by deep learning.

[0275] We set the following key indicators to evaluate the effectiveness of the system:

[0276] 1. Knowledge mastery: Assess learners’ understanding and retention of health knowledge through standardized tests.

[0277] 2. Degree of behavior change: Assess the extent to which learners adopt healthy behaviors in their daily lives.

[0278] 3. Learning engagement: measures learners’ enthusiasm and persistence.

[0279] 4. Improvement of physiological indicators: Measure changes in learners’ key health indicators (such as BMI, blood pressure, etc.).

[0280] 5. User satisfaction: Assess learners’ satisfaction with the system through questionnaire survey.

[0281] The detection method is as follows:

[0282] 1. Knowledge Mastery: Standardized tests are administered before, during, and at the end of the course, covering all health topics covered in the course.

[0283] 2. Degree of Behavior Change: Participants were asked to complete a weekly health behavior diary, recording their diet, exercise, sleep, etc. Researchers conducted a quantitative analysis of the diary.

[0284] 3. Learning participation: The system automatically records the user's learning time, login frequency, number of completed tasks and other data.

[0285] 4. Improvement of physiological indicators: Before and after the experiment, professional medical staff will conduct a comprehensive physical examination and measure various physiological indicators.

[0286] 5. User satisfaction: An anonymous questionnaire survey was conducted at the end of the course and rated using a 5-point Likert scale.

[0287] Table 1. Comparison of experimental results of Example 1, Comparative Example 1 and Comparative Example 2

[0288] index Example 1 Comparative Example 1 Comparative Example 2 Improved knowledge 85 60 70 Degree of behavior change 75 45 55 Average weekly study time 5.5 3.2 4.1 Average decrease in BMI 1.8 0.6 1 User satisfaction 4.7 / 5 3.5 / 5 4.0 / 5

[0289] The experimental results in Table 1 are analyzed and discussed as follows:

[0290] 1. Knowledge Mastery: Example 1 significantly outperformed the other two methods in terms of knowledge mastery. This is likely due to the immersive learning experience provided by VR technology and the personalized learning path supported by deep learning algorithms. The VR environment made complex health knowledge more intuitive and accessible, while personalized recommendations ensured that the learning content remained relevant to the learner's zone of proximal development.

[0291] 2. Behavior Change: Example 1 performed best in promoting behavior change. This demonstrates that the system not only imparts knowledge but also effectively translates it into action. The somatosensory gaming module likely played a significant role in this regard, making healthy behaviors more engaging and habitual through gamification.

[0292] 3. Learning Engagement: In terms of average weekly learning time, participants in Example 1 invested the most time. This demonstrates the strong appeal and stickiness of the system. The novelty and immersive nature of VR technology, combined with the personalized content provided by the intelligent recommendation system, boosted learner engagement.

[0293] 4. Improved Physiological Indicators: Example 1 demonstrated the most significant improvement, particularly in terms of BMI reduction. This result is likely due to the system's comprehensive health intervention. Real-time vital sign monitoring provides timely health feedback, while the intelligent analysis module provides precise improvement recommendations based on this data.

[0294] 5. User Satisfaction: Example 1 received the highest satisfaction score. This demonstrates that the system not only excels in objective effectiveness but also wins user approval in terms of subjective experience. High levels of personalization and interactivity are likely key factors in improving user satisfaction.

[0295] According to the experimental results, Example 1 is significantly better than the other two methods and can be considered the best embodiment. In this embodiment, all modules of the system work together to give full play to the advantages of virtual reality technology and deep learning algorithms.

[0296] Specifically, the VR video teaching module provides users with an immersive learning environment, making abstract health knowledge concrete and easy to understand. For example, when explaining the cardiovascular system, users can observe the process of blood flowing through blood vessels in an immersive way, which greatly enhances the comprehensibility and retention of knowledge.

[0297] The somatosensory game module promotes the transformation of knowledge into behavior by integrating health knowledge with physical activity. For example, in a game simulating sports rehabilitation, users are required to complete a series of movements using correct posture. This not only deepens their understanding of proper exercise methods but also directly cultivates healthy exercise habits.

[0298] The role of intelligent analysis modules and personalized recommendation systems is also crucial. By analyzing users' learning, vital, and behavioral data in real time, the system accurately assesses their health and learning needs, providing the most appropriate learning content and health recommendations. This highly personalized approach maximizes learning efficiency and increases user engagement and satisfaction.

[0299] The real-time vital sign monitoring function provides users with intuitive health feedback, which is difficult to achieve through traditional health education. For example, when the system detects that the user's heart rate has reached the ideal range during aerobic exercise, it will immediately provide positive feedback. This timely encouragement can effectively enhance the user's healthy behavior.

[0300] Overall, these experimental results fully demonstrate the superiority of the proposed system. Not only does it significantly outperform traditional methods in terms of knowledge transfer and behavioral change, it also delivers a superior learning experience and more pronounced health improvements. This approach, which organically combines advanced technologies such as virtual reality, deep learning, and real-time monitoring, opens a new path in the field of health education and holds broad application prospects.

Claims

1. A health education interactive system based on virtual reality and deep learning, characterized by: include: Somatosensory game module, used for: Generate somatosensory game scenes; Matching the user's body movements with virtual objects in the game; The VR video teaching module is connected to the somatosensory game module data and is used to: Play health education videos; Set answer nodes in the video; Dynamically adjust subsequent video content based on the user's answer at the question node; The health education reading module is connected to the VR video teaching module data and is used to: Provide personalized reading content based on user patterns; Record users' reading habits and learning progress; The physical exercise module is connected to the data of the health education reading module and is used to: Perform user fitness tests; Record user's motion data; The intelligent analysis module is connected to the physical exercise module data and is used to: Analyze users' learning data, vital signs data and exercise data; Generate personalized health assessment reports and recommendations; The self-learning unit is connected to the data of the intelligent analysis module and is used to: Based on deep learning algorithms, continuously optimize the system's knowledge base and recommendation algorithms; Build and update health knowledge graphs.

2. The system according to claim 1, wherein: The VR video teaching module includes: Video playback unit, used to play health education video content; A video retrieval unit, used to retrieve relevant video content based on keywords selected by the user; Answering node setting unit, used to set interactive answering points in the video; Among them, when the answering node setting unit detects that the user has reached the answering node, it pauses the video playback and triggers the question display, and continues to play the video after the user completes the answer.

3. The system according to claim 1, wherein: The health education reading module includes: A user mode selection unit, used to allow the user to select gender and age group; Content recommendation unit, used to recommend personalized health information, graphic tutorials, interactive topics and question-and-answer evaluations based on user patterns; Reading habit recording unit, used to record the user's reading content and reading time; The content recommendation unit dynamically adjusts the difficulty and theme of the recommended content based on the user's reading habits and learning progress.

4. The system according to claim 1, wherein: The body motion measurement module includes: Warm-up test unit, used to warm up the user before the formal test; Project test unit, used to perform multiple physical fitness test projects; Score statistics unit, used to calculate and record the user's score in each test item; The item testing unit dynamically adjusts the difficulty and sequence of subsequent test items based on the user's warm-up test results.

5. The system according to claim 1, wherein: The intelligent analysis module includes: Data fusion unit, used to integrate user data from different modules; A deep learning analysis unit, used to analyze the fused data using a deep neural network; A health risk assessment unit, configured to generate a health risk assessment report for the user based on the analysis results; A personalized recommendation generation unit, used to generate targeted health improvement recommendations based on health risk assessment results; Among them, the deep learning analysis unit adopts transfer learning technology, which can quickly adapt to new health problems under limited sample conditions.

6. The system according to claim 1, wherein: Also includes: The interactive module is connected to the data of the intelligent analysis module and is used to: Receive natural language questions from users; understand the intent of the question and retrieve relevant information from the health knowledge base; Generate targeted responses and output them to the user in voice form.

7. The system according to claim 6, characterized in that The interaction module includes: A speech recognition unit, used to convert user voice input into text; Natural language understanding unit, used to analyze text content and extract keywords and intent; Knowledge retrieval unit, used to find relevant information in the health knowledge base; An answer generation unit, used to generate a complete answer based on the retrieval results; A speech synthesis unit, used to convert text answers into speech output; Among them, the natural language understanding unit adopts a deep learning model based on the attention mechanism, which can accurately capture key information in long texts.

8. The system according to claim 1, wherein: The somatosensory game module includes: A motion capture unit, used to capture the user's body movements in real time; A virtual scene generation unit, used to create 3D health education game scenes; An interaction matching unit, used to match user actions with operations on virtual objects; Game logic control unit, used to manage game rules and processes; Among them, the interactive matching unit adopts an algorithm based on deep reinforcement learning, which can adapt to the action characteristics of different users and improve the accuracy and fluency of matching.

9. The system according to claim 1, wherein: Also includes: The sensor array is connected to the intelligent analysis module for: Real-time collection of user's physiological data, including heart rate, blood pressure, body temperature, etc.; Transmit the collected data to the intelligent analysis module for processing; The intelligent analysis module generates a more comprehensive and timely health status assessment based on the real-time physiological data collected by the sensor array and combined with the user's learning and exercise data.

10. A health education interactive method based on virtual reality and deep learning, using the system according to any one of claims 1 to 9, characterized in that: The following steps are involved: a) User Authentication: Verify user identity through facial recognition technology; Retrieve the user's historical learning progress and health data; b) Teaching content display: Select appropriate VR teaching videos, somatosensory games, or reading materials based on the user's learning progress and interests; Present teaching content through VR headsets or other display devices; c) Vital signs information collection: Use sensor arrays to monitor the user's physiological indicators in real time; Associate the collected vital sign data with the current learning content; d) Interactive teaching: Set up question-and-answer nodes in VR videos and adjust subsequent content based on user answers; Provide real-time feedback and guidance based on user actions in motion-sensing games; e) Learning feedback: Generate personalized health assessment reports based on the user's learning performance, physical data and interaction; Use deep learning algorithms to analyze user data and provide targeted health advice and learning plans; f) Knowledge graph update: Collect new questions and feedback from users during the learning process; Utilize self-learning algorithms to continuously optimize and expand the health knowledge graph; g) Learning progress record: Track and record users' learning time, completed courses and test scores; Synchronize learning progress data to user accounts and support multi-device access; Among them, the deep learning algorithm in step e) adopts multimodal data fusion technology, which can comprehensively analyze the user's video learning behavior, somatosensory game performance, reading habits and physiological indicators, thereby generating more comprehensive and personalized health recommendations.

Images