Dynamic examination system based on XR augmented reality and implementation method

Abstract

The invention discloses a dynamic examination system based on XR augmented reality and an implementation method, and relates to the field of XR technology and education examination fusion. Aiming at the problems of weak anti-cheating ability, insufficient foreign language scenario, limited occupational qualification examination practical operation evaluation and the like of a traditional examination, the system integrates intelligent anti-cheating, immersive scenario examination, multi-dimensional evaluation and AI digital human invigilation functions, including a complete version and a simplified version. The complete version system realizes full-process safety monitoring and scene assessment through an XR intelligent head-mounted display (supporting perspective / immersion dual-mode switching and integrating a multi-mode biological recognition module, a full-band detection module and the like), scene adaptive interaction equipment and an AI digital human invigilation engine; the simplified version is a low-cost transition scheme, a lightweight XR head-mounted display is adopted, synchronous display with a traditional computer host is supported, basic anti-cheating and scenarized functions are reserved, and the system is suitable for small and medium-sized examinations. According to the method, the examination safety and the evaluation accuracy can be remarkably improved, the deployment cost is reduced, the method is suitable for scenes such as large-scale selection examinations, foreign language evaluation and occupational qualification examinations, layered landing of the XR technology in the educational examination field is achieved, and examination evaluation is promoted to be transformed from'knowledge memory 'to'problem solution'.

Description

Technical Field

[0001] This invention relates to the field of integrating XR extended reality technology with education and examinations. Specifically, it relates to a comprehensive examination system that integrates intelligent anti-cheating, immersive scenario-based examinations, multi-dimensional assessment, and AI digital human proctoring. It is suitable for large-scale selection examinations requiring enhanced anti-cheating capabilities (such as the National College Entrance Examination, the Senior High School Entrance Examination, and civil service recruitment), foreign language proficiency assessments requiring enhanced scenario realism and interactivity (such as the College English Test Band 4 / 6, IELTS, and TOEFL), professional qualification examinations requiring simulated practical operations (such as medical and construction engineer qualification examinations), and examinations used by schools to assess student learning outcomes. The system possesses strong scalability, allowing for flexible adjustment of scenario modules, assessment dimensions, and interaction methods according to different examination types. With technological advancements, it can continuously integrate new sensing technologies and interaction modes, demonstrating its ability to evolve with technological iterations. It is a preferred solution for adapting to the evolving needs of examinations in the intelligent era. Furthermore, to meet the need for a low-cost transition, a simplified version of the system is included, replacing the traditional computer monitor solely with an XR headset, lowering the deployment threshold. Background Technology

[0002] 1. Security crisis in traditional examination rooms With the innovation of cheating technology, traditional examination rooms face multiple challenges: 1.1 Upgraded high-tech cheating methods: millimeter-wave question transmission devices (only 0.5cm³ in size), bionic skin masks (fingerprint / face forgery success rate >80%), AI real-time translation, and answer glasses (latency <100ms) and other new tools have broken through the defense range of traditional metal detectors and human proctors; 1.2 Anti-cheating systems are passive and lagging behind: existing signal jammers only cover the 10MHz-8GHz frequency band and are ineffective against terahertz waves (0.3-3THz); single biometric identification (such as facial recognition) is easily cracked, increasing the risk of proxy test-taking; 1.3 Weak data traceability: Surveillance videos in traditional examination rooms are easily tampered with, making it difficult to prove cheating, and the average annual increase in the student appeal rate is 15%; 1.4 In traditional examinations, relying on live invigilators has many shortcomings: invigilators have limited energy and it is difficult to pay attention to the subtle movements of multiple candidates at the same time; when candidates encounter equipment failures or process questions during the examination, they cannot get timely responses, which affects the progress of the examination and the candidates' state; for large-scale examinations, a large number of invigilators are required, resulting in high labor costs.

[0003] 2. The Deficiency of Contextualization in Foreign Language Tests Traditional foreign language listening and speaking tests have significant limitations: 2.1 Disconnect between listening comprehension and practical application: Relying solely on audio playback without considering real-world context (such as ambient noise from airport announcements or the facial expressions and gestures of speakers) leads to test takers being able to "understand the exam audio but not real-world conversations," resulting in an application ability assessment deviation of >30%. 2.2 Insufficient authenticity in oral interaction: Using recorded answers or a fixed question bank allows candidates to respond by memorizing templates, failing to assess genuine communication skills (such as impromptu responses and body language coordination). 2.3 Lack of multimodal assessment: Focusing only on voice content and ignoring auxiliary communication information such as body language and facial expressions, the assessment dimensions differ significantly from those in actual communication scenarios.

[0004] 3. Limitations of skills assessment in professional qualification examinations Taking the medical practitioner qualification examination as an example, traditional skills assessments have significant drawbacks: 3.1 Blurred medical image display: Chest X-rays, MRIs, and other images are displayed on paper or at low resolution on screens, resulting in the loss of details (such as tiny nodules), leading to a misjudgment rate of >25% among test takers; 3.2 Insufficient realism of the consultation scenario: Standardized patients or simulated cases are used, the scenario is fixed and the interaction is limited, which makes it impossible to simulate the complex evolution of the disease (such as the process of the symptoms of emergency patients worsening). 3.3 Single dimension of operation assessment: Skill operation (such as puncture, suturing) is scored only by the examiner’s visual observation, lacking objective data support (such as needle angle, force control), and the consistency of scores is poor (the difference in scores between different examiners is >10 points).

[0005] The application of existing XR technology in professional qualification examinations has not yet formed a systematic approach, especially lacking deep integration with anti-cheating systems. Furthermore, the practical tools lack adaptability to various scenarios, and there is room for improvement in balancing the naturalness of interaction and accuracy. In addition, there is a lack of collaborative cooperation with AI digital human proctors. All of these factors provide direction for the technological expansion of this invention. Summary of the Invention

[0006] 1. Technical problems to be solved As mentioned earlier, traditional examinations have exposed numerous problems, including security issues, due to technological advancements, while technological evolution itself provides support for addressing these challenges. Based on this reality, this invention provides a dynamic examination system based on XR (Extended Reality), achieving an organic unity of audiovisual fusion examination presentation, multimodal intelligent assessment, adaptation to scenario-based practical tools, anti-cheating functions, and AI digital human proctoring. Considering the continuous development of technology, the system needs to possess good scalability, gradually optimizing interaction methods and improving the naturalness and convenience of operation.

[0007] 2. Technical Solution 2.1 XR Anti-Cheating Core System 2.1.1 Hardware Architecture 2.1.1.1 XR Smart Headset (Core Carrier) Functions: The XR smart headset achieves scene isolation and precise monitoring through dual-mode switching, supporting "Perspective Mode" and "Immersive Mode" switching. Before the exam, the Perspective Mode allows observation of the real exam environment; after the exam starts, it switches to Immersive Mode, generating a 360° immersive virtual scene (such as a virtual library or hospital clinic), completely covering the candidate's field of vision and blocking visual contact with the real environment. Biometric verification (iris, face, fingerprint, photoelectric pulse) prevents proxy exams; eye tracking and depth cameras capture real-time eye, hand, and head movements; a full-band anti-cheating system is built using terahertz radar and distributed detection; and abnormal sound waves are identified through noise-canceling headphones. It can connect to external devices and collaborate with AI digital human proctors, forming an integrated core of "perception-processing-interaction-monitoring".

[0008] Mode switching principle: Perspective mode: Relying on the built-in binocular full-color camera (150° field of view, 8K resolution) of the head-mounted display, it captures real-time images of the real environment and eliminates distortion through image processing algorithms, and overlays virtual guidance information (such as seat signs) on the display screen to help candidates familiarize themselves with the examination room layout before the exam; Immersive Mode: After the exam starts, external environmental rendering is disabled. A seamless 360° virtual scene is output through dual silicon-based Micro-OLED screens (3840×3840 resolution per eye, 1.38-inch screen size, 3300 PPI). Combined with the head-mounted display's light-blocking structure, real-world visual input is completely blocked, retaining only system-authorized virtual content display, physically blocking traditional cheating methods. It supports dynamic focus adjustment (0.3-5 meters adaptive) and 60 PPD (pixels per degree), achieving the maximum resolution of the human eye with a Fresnel lens group. Peak brightness reaches 20,000 nits (typical value 1,500 nits), and supports HDR10+ dynamic mapping. Integrated core modules: Eye-tracking unit (accuracy ±0.5°): Real-time capture of eye movement trajectory, identification of abnormal movements (such as frequent glances at the edge of the visual field blind spot, fixed focus on suspicious areas); Multimodal biometric unit: Iris recognition (300 feature points): Static identity verification is completed by comparing the information with the information entered before the exam, with an anti-forgery capability of ≥99.9%; Dynamic facial recognition: Combines micro-expression analysis (such as concealed vocalizations and abnormal blinking frequency) to assist in identity verification, while monitoring facial movements related to cheating. It also features built-in 3D structured light scanning (constructing a three-dimensional facial model through infrared dot matrix projection to resist 2D photo / video forgery). Multi-dimensional sensing camera: 32MP dual-lens depth camera: captures hand skeletal movements (accuracy 0.1°) and subtle facial expressions, linked with a behavior analysis system to identify abnormal operations (such as hands obstructing the camera or transmitting hidden signals), and integrates AI semantic gesture recognition algorithms (supporting air gesture commands such as virtual clicks and zooming). Environmental perception function: Real-time scanning of the examination room space generates a three-dimensional boundary model, and an automatic reminder is triggered when a candidate crosses the boundary; Noise-canceling headphone module: Built-in bone conduction and active noise cancellation unit (noise reduction ≥40dB), which physically isolates external sounds while detecting abnormal sound waves (such as the vibration of the miniature speaker in the concealed earphone) through the ear canal vibration sensor to help identify cheating devices in the ear canal.

[0009] Fingerprint recognition module (500dpi resolution): Integrated on one side of the head-mounted display, it is pressed for verification before the exam and compared with the fingerprint features in the registration database, with a matching accuracy of ≥99.8%; Photoelectric pulse tracking module (sampling rate 128Hz): Using photoplethysmography (PPG), it is attached to the inner side of the head-mounted display arm (corresponding to the superficial temporal artery region). It captures subcutaneous vascular pulsation signals through a green LED and a photodetector, records the baseline pulse before the exam (60-100 beats / minute), and monitors abnormal fluctuations (≥20% deviation) in real time during the exam as a dynamic physiological verification indicator.

[0010] Full-band detection module Function: Construct a full-band electromagnetic detection network to accurately identify cheating tools such as smart wearable devices around examinees, and realize automated processing of "detection-analysis-alarm" to replace the traditional active shielding logic. Core modules: Terahertz radar (0.3-3THz): Integrated into the top of the head-mounted display, it scans in real time within a 1-3m range centered on the examinee, and can identify smart wearable devices at the 0.5cm³ level (such as ear canal miniature earphones and wrist-mounted question transmitters), with a response time of <100ms; Distributed collaborative detection unit: The head-mounted displays form a detection network through a low-power wireless protocol (latency <10ms). When a head-mounted display detects an abnormal signal, the surrounding head-mounted displays synchronously locate the signal source and accurately lock onto the specific examinee through a triangulation algorithm (positioning error ≤30cm). Signal Feature Database: Pre-stores the electromagnetic features of common cheating devices (such as frequency band, modulation method, power spectrum), compares the scanning results in real time, and distinguishes between "normal signals of the head-mounted display itself" and "external cheating signals" (identification accuracy ≥ 99.5%).

[0011] 2.1.1.2 Interactive and motion tracking devices Function: Through high-precision sensor fusion technology, it captures the candidate's operation trajectory and body movements, providing data support for the assessment of behavioral compliance and operational standardization. Core modules: Scene-adaptive interactive tools: Basic tracking unit: It adopts a six-axis IMU sensor (accelerometer + gyroscope) + infrared sensor (supporting 6DoF tracking) + UWB positioning module (1cm spatial accuracy), combined with AI motion prediction algorithm, to maintain sub-millimeter positioning accuracy during rapid operation or brief occlusion, and ensure continuous and complete trajectory recording; Written test scenario kit: Electromagnetic induction / infrared sensing dual-configuration stylus: Electromagnetic induction with 4096 levels of pressure sensitivity, supports 0-60° tilt angle recognition, and handwriting reproduction accuracy of 99.2%; Infrared sensor, supports 6DoF high-frequency tracking. Flexible digitizing tablet: integrates UWB anchor points to synchronize position and pressure data with the stylus in real time, enabling dual calibration of note coordinates and pen stroke parameters (such as accurate recording of character positions during formula derivation). Practical scenario kit: Virtual instruments (scalpels / wrenches, etc.): integrate a wide-frequency linear motor (50-500Hz) and a force feedback module to simulate operating resistance of 0.1N-10N / mm and provide real-time feedback on instrument angle (±0.1° accuracy) and force changes; Enhanced Mixed Reality Interaction: Using a depth camera to locate real-world tools (such as real tweezers), parameter guidance (such as clamping force values) is overlaid in a virtual scene to achieve fusion recording of virtual and real operation data. Head motion tracking unit (accuracy ±0.1°): monitors head turning amplitude and frequency (threshold: single time >60° or ≥3 times within 1 minute), and identifies suspicious behaviors such as transmitting signals to others.

[0012] 2.1.1.3 AI Digital Human Proctoring Engine (Cloud Collaboration + Evolutionary Adaptation) Function: As an intelligent proctoring and collaborative terminal, it relies on cloud computing power to achieve real-time monitoring and interactive response, and can be gradually transitioned to local deployment as technology develops. Core components: Digital human avatar: Based on 3D face reconstruction technology, it generates a realistic model that supports basic body language (gesture guidance) and micro-expressions (gazing / reminder nodding) to enhance the naturalness of interaction; Core features: Abnormal behavior linkage analysis: Integrating multi-dimensional data such as eye trajectory, biometrics, and device operation, and using cloud AI algorithms (recognition accuracy ≥98.5%) to mark suspicious behaviors (such as covert operations, sudden changes in pulse), triggering alerts and pushing them to the live proctor's terminal; Candidate interaction response: Supports virtual text (preferred) or noise-reduced voice (limited to necessary scenarios) calls, provides examination process guidance, equipment failure assistance, and real-time encrypted storage of interaction logs; Full-process recording and traceability: Integrating behavioral data, changes in biometrics, and interactive content to generate tamper-proof invigilation logs, supporting post-exam review and appeal verification.

[0013] 2.1.2 Anti-cheating logic flow 2.1.2.1 Pre-exam preparation phase 2.1.2.1.1 Equipment Initialization: When candidates wear XR headsets, the system automatically switches to "Perspective Mode," displays exam room instructions, and an AI digital proctor appears on the virtual interface to introduce the exam process and precautions to the candidates; 2.1.2.1.2 Multi-dimensional verification: Under the guidance of AI digital human proctor, candidates will complete the following steps in sequence: “face recognition → iris recognition → fingerprint pressing → photoelectric pulse baseline value recording”. All data will be compared with the exam registration database. Only after passing the comparison can candidates enter the exam interface. 2.1.2.1.3 Mode Switching: After verification, the AI ​​digital human proctor will prompt "The exam is about to begin" and automatically switch to "Immersive Mode" after 5 seconds, generating a virtual scene (the examinee can only see the virtual environment). At the same time, the full-band detection module is activated (the terahertz radar enters real-time scanning mode), and the scene-adaptive interactive tool automatically matches the current exam scene (such as automatically loading a virtual scalpel model for the medical exam).

[0014] 2.1.2.2 Dynamic Monitoring Stage During the Examination 2.1.2.2.1 Real-time data acquisition: A "security snapshot" is generated every 10 seconds, including: Biometrics: Iris / face matching accuracy (≥99.9%), pulse fluctuation (deviation from the baseline value ±30% is considered abnormal); Behavioral trajectories: head rotation angle (single head rotation > 45° is abnormal), eye focus area (continuous focus on the blind spot outside the field of vision > 3 seconds is abnormal), mouth movements (frequent opening and closing of the mouth when not answering questions is abnormal), hand trajectories (irregular movements of scene-adaptive interactive tools away from the operation area are abnormal). Signal environment: Terahertz radar scan results (presence or absence of abnormal signals), distributed network cooperative positioning results; Behavior-signal cross-analysis: AI digital human proctors analyze radar data in conjunction with behavioral trajectories (e.g., "detecting abnormal signals near the ear canal + the examinee frequently touching the ear canal + abnormal eye focus" is judged as high-risk cheating).

[0015] 2.1.2.2.2 Triggering of Level III Early Warning Level 1 Warning: For a single minor anomaly (such as radar detecting a weak signal but no corresponding abnormal behavior), the AI ​​digital human proctor will display a text reminder via a virtual interface: "Please stay away from electronic devices and maintain proper posture." Level 2 warning: Both signal and behavior are abnormal (such as detecting clear cheating signals + frequent head turning). The AI ​​digital human proctor locks the answer interface (only displays the warning information) and simultaneously notifies the live proctor on the proctoring terminal, waiting for manual confirmation. Level 3 warning: Severe anomaly (such as strong characteristic signal detected + biometric feature matching failure + concealed hand operation), the AI ​​digital human proctor immediately stops the current exam (freezes the exam paper interface), triggers a physical alarm (4000Hz prompt sound), the head-mounted display temporarily switches to see-through mode for proctor intervention, and records the complete abnormal trajectory (including signal characteristics, behavioral actions, and location results). 2.1.2.2.3 Student Interaction Handling: When a student calls upon the AI ​​digital proctor during the exam, the AI ​​digital proctor responds promptly and handles the issue according to its type. For general inquiries, it provides direct answers; for issues such as equipment malfunctions, it first attempts to resolve the problem remotely, and if that fails, it notifies the human proctor; for inquiries involving exam content that violate regulations, it refuses to answer and issues a warning.

[0016] 2.1.2.3 Post-exam data consolidation stage 2.1.2.3.1 After the exam, all security snapshots, biometric data, behavioral trajectories (including operation records of scene-adaptive interactive tools), signal monitoring records, AI digital human proctoring logs, etc., are automatically packaged, generate an immutable hash value, and uploaded to the National Education Examination Blockchain Platform in real time; 2.1.2.3.2 Data Retention: Blockchain-based evidence is retained for 10 years and can be traced at any time (such as retrieving the original record when a candidate appeals).

[0017] 2.2 Simplified XR Dynamic Examination System (Low-Cost Transition Solution) 2.2.1 Core Positioning and Applicable Scenarios The simplified system is designed with "low-cost transition" as its core objective, retaining the core advantages of XR technology (such as scenario-based presentation and basic anti-cheating capabilities). It replaces the traditional computer monitor simply with an XR smart headset, eliminating the need for complex interaction devices and full-band detection modules. It is suitable for: Small and medium-sized exams with limited budgets (such as routine school tests and internal corporate assessments); Transitional scenarios from traditional exams to XR exams (such as regional pilot exams); Theoretical exams with low requirements for practical assessment (such as exams for humanities courses).

[0018] 2.2.2 Simplified Hardware Architecture Design 2.2.2.1 Lightweight XR Headset (Core Device) Display features: It adopts a silicon-based LCD dual-screen (2560×1440 resolution per eye, 1200 PPI), supports 1080P full HD virtual desktop display, simulates the layout of a traditional computer screen (customizable split-screen display of exam paper and answer sheet); supports 100% immersive mode (completely covers the field of view, blocking real visual input), has traditional computer host connection interfaces (such as HDMI, USB-C), and supports synchronous display with traditional monitors.

[0019] Core modules streamlined: Basic biometric unit: integrates dynamic face recognition (100 feature points, anti-forgery capability ≥98%), eliminating iris and fingerprint recognition; Monocular camera (8 megapixels): Captures head movements (accuracy ±1°) and basic eye movements (such as prolonged deviation from the screen area); Simplified noise-canceling headphones: Only passive noise cancellation function is retained (noise reduction ≥20dB), blocking ambient noise but not detecting abnormal sound waves.

[0020] 2.2.2.2 Supporting Equipment Equipped with basic interactive input tools: an electromagnetic induction stylus; Eliminate the distributed detection network: rely on the existing signal jammers in the examination room (covering 10MHz-8GHz) to reduce equipment costs.

[0021] 2.2.3 Core Function Adjustments 2.2.3.1 Simplified Anti-Cheating Logic Pre-exam verification: Identity verification is completed solely through dynamic facial recognition, and the AI ​​digital proctoring system is simplified to a text-based interface; In-exam monitoring: Focus on capturing obvious violations such as abnormal head rotation (more than 90° in a single instance or ≥5 times within 1 minute) and eyes continuously deviating from the screen (more than 5 seconds), triggering a Level 1 warning (text reminder) or a Level 2 warning (notification to the live proctor), and canceling the physical alarm for a Level 3 warning; Data recording: Only facial recognition results, head / eye movement trajectories, and exam interface operation logs are retained, using local encrypted storage (non-blockchain evidence storage), and supporting 7-day traceability after the exam.

[0022] 2.2.3.2 Simplified Contextual Presentation Theoretical exam scenario: The virtual desktop defaults to simulating a "standard exam room" background (such as a classroom environment), which can be switched to static scenes such as "library" or "office" depending on the exam type, without dynamic environmental sound effects or interactive elements; Foreign language exam adaptation: The 3D immersive listening module is retained (simplified to a single channel), the virtual oral interaction is removed, and oral answers are still recorded (consistent with traditional exams).

[0023] 2.2.4 Cost and Efficiency Advantages Hardware costs reduced by 60%: Compared to the full version, components such as terahertz radar, distributed detection units, and high-precision interactive tools are omitted, and the cost of a single device is controlled within 40% of the full version; Reduced deployment difficulty: No complex equipment debugging is required, and candidates can complete the head-mounted display and system startup within 5 minutes, making it suitable for rapid promotion; Compatible with traditional processes: Avoids disruptive changes to the way exams are organized, reducing transitional resistance.

[0024] 2.2.5 The simplified version of the system, which is compatible with the full version, adopts a modular design and can be upgraded to the full version in the future through the following methods: Hardware-wise: Add an external infrared sensor and a terahertz radar module; Software-wise: Unlock full-band detection, multi-modal evaluation and other functional modules; At the data level: connect to the national education examination blockchain platform to upgrade data storage and evidence collection.

[0025] This simplified version allows for the gradual verification of the applicability of XR technology in examination scenarios while controlling costs, providing a transitional solution for the subsequent full-scale rollout of the complete system.

[0026] 2.3 Contextualization Upgrade of Foreign Language Examinations 2.3.1 3D Immersive Hearing Module: 2.3.1.1 Dynamic Scene Generation: 3D virtual scenes are constructed based on the exam content (such as "campus cafeteria" and "workplace meeting room" for CET-4), including environmental sound effects (5.1 channels), character expressions (100+ micro-expression libraries), and lighting changes (simulating day and night / weather). The scene details have an error of less than 5% compared to the real world. 2.3.1.2 Enhanced Listening Interaction: Candidates can interact with the scene (such as zooming in on the speaker or adjusting the ambient volume) through an XR controller (the basic form of a scene-adaptive interactive tool). The system records the interaction trajectory as an additional scoring dimension for "depth of comprehension" (weight 10%). The controller also supports motion-sensing operation (such as simulating "waving" to interrupt the conversation).

[0027] 2.3.2 Virtual Spoken Interaction Module: 2.3.2.1 3D Virtual Interlocutor: Employs real-time motion capture technology (accuracy ±0.1°), supports lip-sync (latency <50ms), and emotional feedback (such as frowning to indicate question). The language library covers 10+ languages ​​including English and Japanese, and the dialogue logic is based on an AI exam-specific model. 2.3.2.2 Multimodal assessment: Simultaneously collect speech (clarity, fluency), lip movements (pronunciation accuracy), and facial expressions (communication attitude), and generate a comprehensive score through a multi-task CNN model, with consistency ≥92% with the assessment of a human examiner.

[0028] 2.4 Optimization of Skills Assessment in Professional Qualification Examinations (Taking the Physician Qualification Examination as an Example) 2.4.1 High-definition medical image interactive module: 2.4.1.1 Ultra-high-definition image presentation: Supports 1:1 native resolution restoration of DICOM format medical images (chest X-ray, CT, MR), with a resolution of 10240×7680. It can be zoomed (up to 20x), rotated, and annotated (e.g., marking lesion locations) via XR handles (the basic form of scene-adaptive interactive tools), improving detail clarity by 600% compared to traditional methods; employing bicubic interpolation + DANCE network combined processing (ensuring 3300 PPI accuracy after magnification), it can clearly display even 0.1mm micronodules; 2.4.1.2 Dynamic Image Analysis: The system has built-in AI-assisted annotation (not answer hints), such as automatically selecting suspicious areas (but not determining their nature), recording the examinee's observation trajectory (such as the first fixation point and duration of fixation), as an additional dimension (weight 15%) for diagnostic ability assessment, and adding an optical flow field deformation algorithm to dynamically track lesion changes.

[0029] 2.4.2 Virtual Consultation and Scenario Simulation Module: 2.4.2.1 3D Virtual Patient: A virtual patient model is built based on real cases, supporting the dynamic evolution of symptoms (such as the increase in body temperature and decrease in blood pressure in patients with abdominal pain), and including real-time changes in physiological indicators (pulse and respiratory rate are simulated by sensors). 2.4.2.2 Consultation Interaction System: Candidates interact with virtual patients through voice (e.g., "Where do you feel unwell?") ​​or gestures (e.g., palpation actions, with scene-adaptive interactive tools simulating hand palpation trajectories). The system records the consultation logic (e.g., whether to inquire about medical history or allergies) and generates a communication ability score (with consistency ≥90% with the real examiner) by combining a standardized patient rating sheet.

[0030] 2.4.3 Skills Operation Assessment Module: 2.4.3.1 Virtual operation scenario: Simulate operation environments such as puncture and suturing. The physical feedback (weight, resistance) of scenario-adaptive interactive tools (such as virtual scalpels, needle holders, and hemostatic forceps) has an error of less than 5% compared with real instruments. 2.4.3.2 Data-driven assessment: Motion capture (accuracy ±0.05°) records operational details (e.g., a needle insertion angle of 30°±5° is acceptable, continuity and uniformity of the suture line), and force sensors collect operational force (e.g., a tension of 2-5N is appropriate for suturing and knotting). An objective scoring report (including the completeness and accuracy of the operational steps) is generated. An AI digital proctor monitors the examinee's operation process in real time. If obvious errors are detected that could lead to adverse consequences, a warning is issued and the operation is paused until the examinee confirms their understanding before continuing.

[0031] 3. Beneficial effects 3.1 The Core Value of the Anti-Cheating System 3.1.1 Enhanced Security Level: Through a multi-layered architecture of "virtual isolation (blocking visual / auditory contact) + biometrics (static + dynamic verification to prevent cheating) + full-behavior tracking (head / eye / hand / mouth monitoring without blind spots, including scene-adaptive interactive tool operation trajectories) + full-band detection (identifying smart wearable devices) + behavior-signal cross-verification," the system achieves: eliminating traditional cheating at its source (cheat sheets, whispering, peeking at others' answers); accurately identifying new types of cheating (high-tech devices, identity forgery, AI-assisted answering): terahertz radar combined with a distributed network can 100% identify smart wearable devices such as earphones and wrist-mounted question transmitters; and completely cracking down on cheating using "low-power concealed devices" through cross-verification of "abnormal signals + corresponding behaviors" (such as ear canal signals + frequent ear touching, hand signals + irregular movements). 3.1.2 Improved invigilation efficiency: AI digital human invigilation enables real-time, accurate monitoring and rapid response to examinees, reducing the workload of human invigilators and allowing them to focus more on handling complex and abnormal situations. This significantly improves the management efficiency of large-scale examinations and reduces examination organization costs by more than 20%. 3.1.3 Optimized candidate experience: The timely response and professional answers from the AI ​​digital proctor reduced candidates' confusion and anxiety during the exam. Issues such as equipment malfunctions were handled quickly, ensuring the candidates' normal exam rhythm and increasing their recognition of the fairness and standardization of the exam. 3.1.4 Active defense capability: The real-time eye and hand tracking system (including scene-adaptive interactive tool trajectory) can detect abnormal actions of candidates in a timely manner; the terahertz radar can identify cheating intentions (such as hand movements of operating concealed devices) 0.5 seconds in advance; the AI ​​digital human proctor can provide early warning and intervention at the first time, realizing "early warning before cheating". 3.1.5 Improved Data Credibility: Blockchain-based evidence storage ensures a near 100% success rate in proving cheating and reduces the candidate appeal rate by 80%; AI-powered digital human proctoring logs provide detailed evidence for the retrospective analysis of the examination process, facilitating the summarization of experiences and problems in examination organization.

[0032] 3.2 The benefits of contextualizing foreign language exams 3.2.1 Accuracy of Application Ability Assessment: The 3D scenario reduces the deviation between listening comprehension and actual application to within 5%, improving the consistency between candidates' "exam scores" and "actual abilities" by 40%; 3.2.2 Authenticity of Oral Interaction: The virtual human's random follow-up questions (such as "Please explain what 'flexibility' specifically refers to?") effectively break through template memorization, and the accuracy rate of identifying impromptu responses is ≥95%. 3.2.3 Improved Multi-Dimensional Assessment: The newly added non-verbal communication dimension (accounting for 20%) makes the language ability assessment closer to actual communication scenarios; the AI ​​digital human proctor can simulate unexpected situations in real communication during the oral exam, such as interrupting the conversation or raising objections, further testing the candidates' on-the-spot reaction ability.

[0033] 3.3 Benefits of Professional Qualification Examinations 3.3.1 Improved accuracy of skills assessment: High-definition imaging improves the accuracy of lesion identification by 35%, and the digitization of operation data of scene-adaptive interactive tools (such as suture lines and needle insertion angles) reduces the difference in examiner scores to <3 points. 3.3.2 Enhanced Scene Realism: The virtual patient employs dynamic disease simulation, unlike traditional static cases; the physical feedback of the scene-adaptive interactive tools is highly consistent with real instruments, making the candidate's operating experience closer to reality. After the candidate proposes a treatment plan, changes in the patient's symptoms can be observed, making the diagnosis and differential diagnosis process closer to clinical practice, comprehensively covering complex case assessments, and significantly improving assessment coverage and scene realism. The 3300 PPI screen completely eliminates the screen-door effect, improving the recognition rate of medical image details (such as 0.1mm micronodules) by 200%, solving the problem of lesion misjudgment caused by insufficient pixel density in traditional head-mounted displays; 3.3.3 Integration of Training and Assessment: The system can record candidates' operational weaknesses (such as deviations in puncture angles or irregular suture lines) and generate targeted improvement suggestions, thus combining assessment and training functions. The AI ​​digital proctor can recommend relevant learning resources and practice modules based on the candidate's performance, helping them to improve specifically after the exam.

[0034] 3.4 The supplementary value of the simplified system Lowering the technical barrier: The lightweight design makes the XR examination system more acceptable to candidates and invigilators, reducing the learning cost of operation; Flexible adaptation to needs: Provides options for exam scenarios with different budgets and scales, avoiding a "one-size-fits-all" approach to technology upgrades; Smooth transition guarantee: While retaining the core process of traditional examinations, XR elements are introduced to gradually cultivate user habits and lay the foundation for the promotion of the full version of the system.

[0035] 4. System Improvement Potential and Scalability 4.1 Technological Iteration Direction: With the continuous development of XR technology, artificial intelligence, and sensing technology, the system has significant room for improvement in terms of interaction accuracy, scene realism, and the intelligence of AI digital human proctoring. For example, more advanced brain-computer interface technology can be adopted in the future to achieve more natural and accurate human-computer interaction; deep learning algorithms can be used to optimize the anomaly detection and problem-solving capabilities of AI digital human proctoring, enabling it to handle more complex proctoring scenarios. 4.2 Multi-Scenario Expansion: The system's modular design allows it to easily adapt to more types of examination scenarios. In addition to existing large-scale selection exams, foreign language proficiency assessments, and professional qualification exams, it can be expanded to skills competitions, internal corporate assessments, military training assessments, and other fields. Simply adjust the parameters of the virtual scene, assessment dimensions, interactive tools, and AI digital human proctoring according to the needs of different scenarios. 4.3 Functional Expansion: More practical functions can be added to the existing system, such as an intelligent test paper generation system that automatically generates personalized test papers based on the examinee's knowledge level and examination objectives; a learning analysis system that conducts in-depth analysis of examinee's examination data and provides precise improvement suggestions for teaching; and a remote proctoring function that uses 5G technology to achieve real-time proctoring across regions and expand the coverage of examinations. Attached Figure Description

[0036] Figure 1 (System Architecture Diagram): Showing the three-tier architecture and interaction relationships of a dynamic examination system based on XR extended reality: Data support layer: includes the National Education Examination Blockchain Platform (for evidence storage and tamper-proofing), signal feature database (electromagnetic features of cheating devices), test question database (traditional / virtual scenario test questions), and candidate database (biometric information such as iris / fingerprint / avatar). System and Algorithm Layer: Connecting examination staff and examinees through AI digital humans to achieve data interaction and invigilation logic; Core hardware layer: integrates XR smart head-mounted display (biometric recognition, eye tracking, full-band detection, noise-canceling headphones) and interaction and motion tracking devices (written test kit, practical operation kit).

[0037] Figure 2 (Schematic diagram of XR smart headset structure): The following is a standardized explanation of the numbered labels in the attached figure, based on the technical features of the instruction manual, strictly corresponding to the core modules and functional parameters of the XR smart headset: 1 — Binocular full-color camera (built into the right side of the head-mounted display, 150° field of view, 8K resolution, used to capture real-time images of the real environment in perspective mode, eliminate distortion through image processing algorithms, help candidates familiarize themselves with the examination room layout before the exam, and provide a real-world image overlay carrier for virtual guidance information (such as seat signs); 2 — Status display panel (integrated into the function area, used to display status information such as "exam in progress" and "equipment normal" in real time, and synchronously output to the human proctoring terminal, supporting interactive information display with AI digital human proctoring, ensuring that proctors can keep track of the equipment's operating status in real time). 3 — The 3D structured light scanning module (including an infrared dot projector and an infrared camera) is integrated into the multimodal biometric unit. It constructs a three-dimensional facial model through infrared dot projection, which can resist the forgery of flat photos / videos and achieve identity verification in conjunction with dynamic face recognition. At the same time, it combines micro-expression analysis (such as concealed mouth sounds and abnormal blinking frequency) to help monitor facial movements related to cheating.

[0038] 4 — Noise-canceling headphone module (right channel, built-in bone conduction and active noise cancellation unit, noise reduction ≥40dB, physically isolates external sound while detecting abnormal sound waves through ear canal vibration sensor to help identify cheating devices in the ear canal). 5 — Fingerprint recognition module (500dpi resolution): Integrated on one side of the head-mounted display, it is pressed for verification before the exam, and the fingerprint features are compared with those in the registration database, with a matching accuracy of ≥99.8%; 6 — Distributed Cooperative Detection Unit: The head-mounted displays form a detection network through a low-power wireless protocol (latency <10ms). When a head-mounted display detects an abnormal signal, the surrounding head-mounted displays synchronously locate the signal source and accurately lock onto the specific examinee through a triangulation algorithm (positioning error ≤30cm). 7 — Environmental visual tracking cameras (4 cameras, SLAM cameras) are symmetrically distributed 4 environmental perception cameras with environmental perception function. They can scan the examination room space in real time to generate a 3D boundary model. When the examinee crosses the boundary, an alert is automatically triggered. At the same time, they support 6DoF spatial positioning of the head-mounted display to ensure stable presentation of the virtual scene.

[0039] 8 — Photoelectric pulse tracking module (fitted to the inner side of the left temple of the head-mounted display, corresponding to the superficial temporal artery region, using photoplethysmography with a sampling rate of 128Hz, used to record the baseline pulse before the exam (60-100 beats / minute), and to monitor abnormal fluctuations (≥20% deviation) in real time during the exam as a dynamic physiological verification indicator); 9 — Terahertz radar (integrated on the top left side of the head-mounted display, operating in the 0.3-3THz frequency band, capable of identifying smart wearable devices at the 0.5cm³ level (such as ear canal miniature earphones and wrist-mounted question transmitters), with a response time of <100ms, and is the core sensing component of the distributed collaborative detection unit). 10 — Eye Tracking and Iris Verification Unit (Internal View Core Module, eye tracking accuracy ±0.5°, real-time capture of eye movement trajectory to identify abnormal movements; iris recognition with 300 feature points, anti-forgery capability ≥99.9%, static identity verification completed by comparing with pre-exam entered information). 11 — Fresnel lens group and silicon-based Micro-OLED dual screen (internal view display core, dual screen single eye resolution 3840×3840, screen size 1.38 inches, PPI up to 3300, combined with Fresnel lens group to achieve the limit of human eye resolution, supports dynamic focus adjustment (0.3-5 meters adaptive), output 360° seamless virtual scene); 12 — Virtual content display area (achieved by the screen and lens group of 11 working together, it can display exam-related virtual scenes and content, completely blocking real visual input, retaining only the virtual information authorized by the system, and physically blocking traditional cheating channels). 13 — The 32-megapixel dual-lens depth camera is a multi-dimensional perception camera that can capture hand skeletal movements (accuracy 0.1°) and subtle facial expressions. It works in conjunction with a behavior analysis system to identify abnormal operations (such as hands obstructing the camera or transmitting hidden signals) and integrates AI semantic gesture recognition algorithms to support air gesture commands (such as virtual clicks and zooming).

[0040] Figure 3 (Schematic diagram of the entire examination process): Presenting the closed-loop management logic of "before the exam → during the exam → after the exam": Before the exam: Device initialization (perspective mode + AI digital human guidance), complete multi-dimensional identity verification of "face → iris → fingerprint → pulse"; During the test: Switch to immersive mode and generate a safety snapshot every 10 seconds (biometric fluctuations, abnormal behavior trajectory, signal environment monitoring), triggering a level 3 warning through AI cross-analysis; Post-exam: Package the security snapshot, biometrics, behavioral trajectory and invigilation log, generate hash value and upload to the National Education Examination Blockchain Platform for evidence storage (retained for 10 years).

[0041] Figure 4 (Flowchart of Dynamic Monitoring and Early Warning During Examinations): Refining the Monitoring Mechanism During Examinations: Data acquisition: Real-time acquisition of biometric features (iris / face matching degree, pulse fluctuation), behavioral trajectory (head / eye / hand / mouth movements), and signal environment (terahertz radar / distributed positioning); Cross-analysis: associating “abnormal signal + corresponding behavior” (e.g., ear canal signal → touching the ear); Tiered early warning: Level 1 Warning (Single Minor Anomaly): AI Text Reminder (e.g., "Standardize Posture"); Level 2 Warning (Double Anomaly): Lock the interface + Notify a live proctor; Level 3 Warning (Severe Anomaly): Stop the exam + Physical alarm + Switch perspective mode.

[0042] Figure 5 (Illustrative diagram of a virtual scenario for foreign language examination): An airport check-in interactive scenario is constructed (international check-in counter, baggage, flight information screen). Candidates can perform motion-sensing operations (such as waving to interrupt the conversation) through a scenario-adaptive interactive XR controller, simulating a real foreign language communication environment and supporting scenario-based assessment of "listening immersion + oral interaction". Detailed Implementation Example of anti-cheating system operation (scenarios of college entrance examination and university course final exams) 1.1 Exam Preparation: After entering the examination room, candidates wear XR devices. The system verifies their identity through face, iris, and fingerprint (compared with the registration database), while the photoelectric pulse tracking module records baseline values ​​(60-100 beats / minute). An AI digital proctor welcomes candidates on a virtual interface and provides detailed explanations of the examination rules and how to use the devices. A virtual fence (2m high, impenetrable) is generated, and a surrounding field sensor scans a 1m radius around the perimeter to remove any remaining electronic devices (such as miniature earphones); scene-adaptive interactive tools are automatically matched to a stylus form. 1.2 Exam-related defense: During the exam, a virtual scene is generated using XR devices to achieve the physical isolation effect of traditional cheating methods. On this basis, if the terahertz radar detects a 0.5cm³ millimeter-wave transmitter (30GHz signal) inside the examinee's collar, and the distributed network confirms that the signal source is the examinee, and detects that "the examinee frequently looks down at his collar + his pulse suddenly rises to 130 beats / minute", the system determines that it is a clear case of cheating. The AI ​​digital human proctor immediately freezes the exam paper interface, triggers a level three warning: "Cheating device detected, the exam has been suspended, please wait for the proctor to handle it", and simultaneously uploads the signal characteristics and behavioral trajectory to the blockchain. If a candidate wears a miniature earphone in their ear canal (a terahertz radar detects a 0.3THz signal), and the ear canal vibration sensor of the noise-canceling earphone captures abnormal sound waves, and their eyes are continuously focused on the lower left (suspected of receiving signals), the system will determine this as cheating, trigger a level-two warning, and notify a live proctor for investigation. During the exam, a candidate had a question about a question and called the AI ​​digital proctor via voice. The AI ​​digital proctor clearly explained the meaning of the question to the candidate, and the solution process was recorded in real time. 1.3 Post-exam processing: After the exam, all relevant data and AI digital human proctoring logs are uploaded to the blockchain platform, and candidates can check their exam records within the specified time.

[0043] Example of a foreign language exam scenario (CET-4 and CET-6) 2.1 Listening section: The virtual scenario is an "airport check-in counter." Candidates can see the anxious expressions of ground staff (due to flight delays) and hear background announcements and baggage cart noises. The listening content is synchronized with the scene's actions (e.g., when a ground staff member points to the flight information screen, the corresponding content is played simultaneously). Candidates can zoom in on the lip movements of ground staff using a controller (the basic form of a scene-adaptive interactive tool). The system records this interaction. If a candidate correctly understands and answers the related questions, they will receive an additional 2 points. If a candidate attempts to cheat during the listening test, such as by making eye contact with other candidates, the AI ​​digital proctor will immediately issue a warning. 2.2 Oral Section: A virtual human, acting as an "interviewer," asks, "Describe a time when you helped the astranger." As the candidate responds, the system simultaneously analyzes their speech (85% pronunciation accuracy), lip movements (correct / θ / tongue movement), and facial expressions (a smile indicating a positive attitude), giving an overall score of 8.5 / 10 (compared to a traditional score of only 7.2 / 10). If the candidate is wearing a translation earpiece, the terahertz radar detects an abnormal signal and simultaneously monitors "frequent head tilting + wandering eyes + regular sound waves captured by the earpiece vibration sensor." The system determines this as cheating, immediately stopping the oral exam and recording the evidence. The AI ​​digital proctor monitors the candidate's oral expression process, providing appropriate prompts if the candidate pauses for extended periods or speaks incoherently.

[0044] Example of a professional qualification examination scenario (Physician qualification skills examination) 3.1 Image Interpretation Section: Candidates can view a high-resolution chest X-ray (7680×4320 resolution) using an XR device, which can magnify and observe a small nodule (3mm in diameter) in the lower lobe of the right lung. The system simultaneously records the time when the nodule is first discovered (<30 seconds is considered excellent). If a candidate misjudges the nature of a nodule, the system will not directly provide an answer, but will increase the weight of similar images in subsequent tests. The AI-powered digital proctor will record the candidate's interpretation process, providing a basis for subsequent scoring and analysis. 3.2 Virtual Consultation Section: A virtual patient (who can be set as a middle-aged male) complains of right lower abdominal pain. The examiner first observes the patient's appearance for a preliminary assessment, then inquires about any history of migratory right lower abdominal pain, and performs a physical examination, including palpation at McBurney's point (simulating palpation using a scenario-adaptive interactive tool, with the system recording the palpation site and pressure). Simultaneously, auxiliary examinations such as blood tests and abdominal ultrasound are ordered. After the system provides feedback, the examiner diagnoses, differentiates, and treats the patient. The system then displays the corresponding changes in the virtual patient's condition based on the treatment plan. If the diagnosis and treatment are correct, the condition improves; if incorrect, it worsens. In this case, the examiner must re-diagnose and implement remedial treatment measures. The system generates a communication ability score (85 / 100) based on the completeness of the consultation (e.g., whether it covers the triggers and relief factors) and the communication attitude (e.g. whether it comforts the patient). 3.3 Skill Operation Section: Simulating a "thoracentesis" procedure, the examinee inserts a virtual puncture needle using a scenario-adaptive interactive tool. The system detects the needle insertion angle (30° is the standard) and depth (4cm is suitable) through motion capture, and the force sensor provides feedback on the pleural breakthrough sensation (sudden drop in resistance). During the simulated "suturing" operation, candidates use a virtual needle holder to perform suturing. The system records the continuity of the suture line and the uniformity of the stitch spacing (standard 5mm±1mm). After the operation is completed, a data report is generated: "Needle entry angle 32° (pass), depth 4.2cm (pass), continuous suture line (excellent), operation time 90 seconds (good)," assisting examiners in objective scoring. The AI ​​digital human proctor monitors the candidate's operation steps in real time, and promptly stops and corrects any violations.

[0045] Simplified system implementation example (school daily quizzes) 4.1 Pre-exam preparation: Candidates wear a lightweight XR headset, the system switches to "virtual desktop mode" and displays the traditional exam paper interface; dynamic facial recognition is completed through a monocular camera (compared with the school's student database), and the AI ​​digital human prompts in text form "Please confirm your personal information and wait for the exam to begin".

[0046] 4.2 In-exam monitoring: During the exam, if the head-mounted display detects that the examinee's head frequently turns to the next seat (120° at a time), the system will trigger a Level 1 warning: "Please keep your eyes on the screen and abide by the exam rules." If an abnormal signal (such as a Bluetooth device) is detected within the range of the exam room signal jammer at the same time, a Level 2 warning will be triggered, the answer interface will be frozen, and the invigilator will be notified.

[0047] 4.3 Post-exam processing: After the exam, candidates submit virtual exam papers (or fill out paper answer sheets). The system stores the answer records and behavior trajectory locally, supporting query and tracing within 7 days. The exam paper data is synchronously exported to the school's original marking system.

[0048] Through the collaborative design of the full and simplified versions, this invention not only meets the requirements of high-security and highly interactive examinations, but also provides a feasible solution for low-cost and transitional scenarios, realizing the layered implementation of XR technology in the field of education and examination.

Claims

1. A dynamic examination system based on XR extended reality, characterized in that, Includes both full and simplified versions of the system; The complete system includes: (1) an XR smart head-mounted display, which serves as the core carrier and supports dual-mode switching between "perspective mode" and "immersive mode". It integrates: a display module, which adopts a silicon-based Micro-OLED dual screen with a single-eye resolution of ≥3840×3840 and a pixel density of ≥3300PPI, supporting dynamic focal length adjustment of 0.3-5 meters and ≥60PPD pixels per degree; a multimodal biometric unit, including iris recognition (≥300 feature points, anti-counterfeiting capability ≥99.95%), 3D structured light scanning (accuracy ±0.1mm), fingerprint recognition (≥800dpi resolution) and photoelectric pulse tracking (sampling rate ≥256Hz); and a multi-dimensional perception module, including 3200 10,000-pixel binocular depth camera (hand bone tracking accuracy ±0.1°), eye tracking unit (accuracy ±0.1°) and noise-canceling headphones (noise reduction ≥50dB); full-band detection module, including terahertz radar (0.3-3THz, response time <50ms) and distributed collaborative detection unit (positioning error ≤15cm); (2) scene-adaptive interactive device, including: basic tracking unit, using a six-axis IMU sensor + UWB positioning module (spatial accuracy ≤0.5cm), equipped with an infrared sensor, supporting 6DoF tracking, combined with AI motion prediction algorithm; practical scenario kit, including virtual equipment (integrated 50-1000Hz wideband linear motor, force feedback accuracy ±0.05°) and mixed reality interaction enhancement module; (3) AI digital human proctoring engine, configured with cloud collaborative algorithm, to realize multi-dimensional data linkage analysis (recognition accuracy ≥99%), candidate interaction response and anti-tampering log generation; (4) dynamic super-resolution rendering system, for medical image processing, using bicubic interpolation + DANCE The network joint algorithm supports ≥20x stepless magnification while maintaining 3300PPI accuracy and can identify micro nodules ≤0.1mm; The simplified system includes: (1) a lightweight XR head-mounted display, which serves as the core device and supports 100% immersive mode (completely covering the field of vision and shielding real visual input). It integrates: a display module with a silicon-based LCD dual screen, a single-eye resolution of 2560×1440, a PPI of 1200, and supports 1080P full HD virtual desktop display; a basic biometric unit with dynamic face recognition (100 feature points, anti-forgery capability ≥98%); a monocular camera (8 million pixels) to capture head movements (accuracy ±1°) and eye movements; a simplified noise-canceling headphone (noise reduction ≥20dB, only retaining passive noise reduction function); and a traditional computer host connection interface (such as HDMI, USB-C) to support synchronous display with a traditional monitor; (2) a matching basic interactive input tool that relies on the original signal jammer in the examination room (covering 10MHz-8GHz) and cancels the distributed detection network.

2. The system according to claim 1, characterized in that, The full-version XR smart headset's "Perspective Mode" relies on a 150° field of view and 8K resolution dual full-color cameras to capture the real environment, while the "Immersive Mode" uses a 360° seamless virtual scene (peak brightness ≥20,000 nits) to shield real-world visual input. The simplified version's lightweight XR headset's 100% immersive mode completely blocks real-world visual input through screen display and light-blocking structure, displaying only virtual content, and the virtual desktop can simulate the layout of a traditional computer screen.

3. The system according to claim 1, characterized in that, The full version of the system uses a multimodal biometric unit to achieve identity verification through a process of "iris recognition → 3D structured light scanning → fingerprint verification → photoelectric pulse baseline recording", where dynamic face recognition is combined with micro-expression analysis (including hidden mouth sounds and abnormal blink frequency monitoring); the simplified version of the system only completes identity verification through dynamic face recognition.

4. The system according to claim 1, characterized in that, The full-band detection module of the complete system constructs an anti-cheating network of "terahertz radar real-time scanning + distributed network collaborative positioning + signal feature database comparison", which can identify smart wearable devices at the 0.5cm³ level and distinguish "head-mounted display signal" from "external cheating signal" with an accuracy of ≥99.8%.

5. The system according to claim 1, characterized in that, The complete system's scene-adaptive interactive devices also include a written test scenario kit, a stylus with infrared sensing and 4096 levels of pressure-sensitive electromagnetic induction (handwriting reproduction accuracy ≥99.6%), and a flexible digitizer with integrated UWB anchor points; a head motion tracking unit (accuracy ±0.05°) monitors the head turning amplitude (single > 60°) and frequency (≥3 times within 1 minute) to identify abnormal behavior.

6. The system according to claim 1, characterized in that, The full version of the system uses a "safety snapshot" mechanism for dynamic monitoring during the exam, generating comprehensive data and triggering a three-level warning every 5 seconds. The simplified version of the system focuses on detecting abnormal head rotation (a single time > 90° or ≥ 5 times within 1 minute) and continuous eye deviation from the screen (> 5 seconds), triggering only a first-level warning (text reminder) or a second-level warning (notifying the live proctor), and canceling the physical alarm of the third-level warning.

7. The system according to claim 1, characterized in that, The complete system's dynamic super-resolution rendering system supports native 10240×7680 resolution restoration for DICOM format medical images. It dynamically tracks lesion changes through an optical flow field deformation algorithm and integrates AI-assisted annotation and records the examinee's observation trajectory (weight ≥ 15%).

8. The system according to claim 1, characterized in that, The full version of the system also includes a foreign language exam scenario module, containing a 3D immersive listening unit (constructing a virtual scene with 5.1 channel environmental sound effects and a library of 100+ micro-expressions, with an error of <5%) and a virtual oral interaction unit (using ±0.05° precision motion capture, supporting lip-sync latency <30ms and multimodal evaluation); the simplified version of the system retains the 3D immersive listening module for foreign language exam adaptation (simplified to a single channel), cancels the virtual oral interaction, and uses recording for oral responses.

9. The system according to claim 1, characterized in that, The complete version of the system includes a professional qualification examination skills assessment module that includes a virtual operation scenario (the error between the virtual instrument's physical feedback and the real instrument is < 3%) and a data-driven evaluation unit (which records operation details and force through ±0.01° precision motion capture).

10. The system according to claim 1, characterized in that, The post-exam data of the full version system is stored on the blockchain (retained for 10 years) and includes security snapshots, biometrics, behavioral trajectories, and AI proctoring logs; the post-exam data of the simplified version system is stored locally in encrypted form, supports 7-day traceability, and can be upgraded to the full version system by adding infrared sensors and terahertz radar to the hardware and unlocking the full-band detection function in the software.

Images