Virtual classroom teaching device and teaching method thereof

Abstract

This invention provides a virtual classroom teaching device and its teaching method, the device of which is described below. A textbook format recognition module can input multiple digital textbooks and automatically recognize and convert their formats to generate multiple modular lesson plan units with the same file format that can be disassembled. A modular reorganization engine module can adjust, supplement, and optimize the order of these modular lesson plan units based on the learner's learning behavior and learning outcomes. A virtual teacher control module can drive a virtual teacher image to interact with the learner and provide teaching responses; this virtual teacher image has voice recognition, semantic understanding, and non-verbal instruction recognition functions. A feedback detection and behavior analysis module can collect the learner's real-time learning status data and use this real-time learning status data as a basis for push recommendations.

Description

Technical Field

[0001] This invention relates to the field of virtual teaching and smart learning, and more particularly to a virtual classroom teaching device and its teaching method that supports XR formats (such as VR, AR, MR), modular reconfiguration, AI-optimized recommendation, and virtual teacher interactive control. Prior Technology

[0002] A virtual classroom simulates a real classroom in an online space. Courses can usually be conducted synchronously, with instructors and learners appearing in the same online space and interacting in real time. However, due to the busy lifestyles of many people, many virtual courses incorporate pre-recorded content based on the needs of instructors or learners. Generally speaking, like a real classroom, a virtual classroom can flexibly adapt to the learning styles and needs of all students. For example, it can provide video conferencing, digital whiteboards, instant messaging, group discussions, and more.

[0003] However, pre-recorded virtual courses may also have the drawbacks of traditional online courses, such as a lack of emotional connection. Furthermore, the content of these virtual courses may remain unchanged, and learners' learning effectiveness may be greatly reduced if they encounter boring and uninteresting virtual courses.

[0004] Therefore, in summary, current virtual teaching is mostly limited to online videos and non-interactive teaching materials, unable to adjust lesson plans according to learners' situations, and lacks real-time feedback and intelligent adaptation mechanisms for human-computer interaction. With the advancement of XR technology and AI interactive systems, providing a virtual classroom device that can be flexibly reconfigured, offers real-time interaction, provides immersive teaching, and has feedback and adjustment capabilities has become an important technical issue.

[0005] In summary, the inventors of this invention conceived and designed a virtual classroom teaching device and its teaching method in order to improve upon the deficiencies of conventional technologies and thereby enhance their industrial application. Summary of the Invention

[0006] To achieve the above objectives, this invention provides a virtual classroom teaching device, comprising a textbook format recognition module, a modular reconfiguration engine module, a virtual teacher control module, and a feedback detection and behavior analysis module. The textbook format recognition module can input multiple digital textbooks and automatically recognize and convert their formats to generate multiple modular lesson plan units with the same file format that can be disassembled. The modular reconfiguration engine module can adjust, supplement, and optimize the order of these modular lesson plan units based on the learner's learning behavior and learning outcomes. The virtual teacher control module can drive a virtual teacher image to interact with the learner and provide teaching responses; this virtual teacher image has voice recognition, semantic understanding, and non-verbal instruction recognition functions. The feedback detection and behavior analysis module can collect the learner's real-time learning status data and use this real-time learning status data as a basis for push recommendations.

[0007] Preferably, this modular refactoring engine module includes a version control unit, which can record the process of lesson plan refactoring and provide version annotation, comparison and restoration functions for teachers to switch interfaces and track progress.

[0008] Preferably, this feedback detection and behavior analysis module includes a voice emotion analysis unit and an eye movement tracking unit to instantly identify the learner's emotions and concentration state, and adjust or enhance the content of the modular lesson plan units accordingly.

[0009] Preferably, this virtual teacher control module includes a database of virtual characters that can automatically switch the voice, appearance, and interaction style of the virtual teacher image according to the course topic and the learner's language.

[0010] Preferably, the virtual classroom teaching device of the present invention further includes an AI learning analysis module, wherein this AI learning analysis module includes a behavioral feature extraction unit, a multi-dimensional classification and clustering model unit, a module fit calculation unit, and a recommendation and ranking unit. The behavioral feature extraction unit can be used to analyze learners' answer paths, operation clicks, visual dwell times, and interactive responses. The multi-dimensional classification and clustering model unit can be used to automatically identify the learner's type. The module fit calculation unit can be used to predict the learner's expected performance under different lesson plan modules. The recommendation and ranking unit can automatically calculate reinforcement and reorganization strategies based on real-time detected behavioral data to produce a personalized lesson plan recommendation module list and confidence matrix based on process data.

[0011] To achieve the above objectives, the present invention further provides a virtual classroom teaching method, which is applied to a virtual classroom teaching device. This virtual classroom teaching device may include a textbook format recognition module, a modular reconstruction engine module, a virtual teacher control module, and a feedback detection and behavior analysis module. This virtual classroom teaching method includes the following steps:

[0012] (a) Textbook Upload and Modular Conversion Steps: Upload the digital textbook to the textbook format identification module to parse its type and convert it into a modular lesson plan unit that can be disassembled.

[0013] (b) Lesson plan distribution and initialization steps: The modular lesson plan unit is automatically loaded according to the learner's terminal device, and a learning task process is initialized.

[0014] (c) Virtual teaching interaction steps: The virtual teacher controls the module to guide the teaching, supports voice input, facial feedback, gesture or click interaction, and switches the teaching style according to the context of the modular lesson plan unit.

[0015] (d) Behavioral feedback acquisition steps: The learning process record file is formed by acquiring the learner's operation data, speech recognition results, facial expressions and eye tracking information through the feedback detection and behavior analysis module.

[0016] (e) AI Analysis and Recommendation Steps: Input the learning process record file into the AI ​​learning analysis module to perform behavioral model inference, calculate a module fit and a teaching confidence value, and generate a recommendation list and restructuring suggestions.

[0017] (f) Teacher review and fine-tuning steps: The reorganization results and system suggestions are presented in the lesson plan editing interface on the teacher's end, and the lesson plan editing interface is used to make fine-tuning, replace modules or modify the process sequence.

[0018] (g) Push and process archiving steps: Push the confirmed modular teaching plan unit to the learner's device to carry out interactive teaching, and automatically save the complete learning process record file in the cloud resume database after class for future diagnosis and teaching optimization reference.

[0019] Preferably, the virtual classroom teaching method of the present invention further includes the following steps.

[0020] Detect the prerequisite dependencies and capability progress annotations between each modular lesson plan unit.

[0021] Does the automatic comparison of module order violate the logic of cognitive progression in teaching?

[0022] Detect whether there are duplicate teaching sessions or missing modules.

[0023] The lesson plan editing interface displays flags for non-standard or duplicate modules and provides adjustment suggestions.

[0024] Preferably, this learning process record can be presented in real-time by a visual dashboard, which may include a module progress tracking unit, an error and pause hotspot analysis chart, an emotion and focus trend graph, and a real-time teacher interaction toolbar. The module progress tracking unit can display the learner's completion percentage and operation record in each lesson plan module. The error and pause hotspot analysis chart can generate key warning nodes based on operation time and repeated error locations. The emotion and focus trend graph can be a visual curve constructed by combining voice tone, facial expressions, and eye movement analysis. The real-time teacher interaction toolbar allows teachers to mark, group, and distribute reinforcement modules for abnormal nodes.

[0025] Preferably, in asynchronous learning processes, this virtual classroom teaching device further includes an emotional feedback reinforcement module, which comprises a voice emotion recognition unit, an facial expression and body language sensing unit, a reinforcement lesson plan recommendation engine unit, and a learning interface response unit. The voice emotion recognition unit continuously analyzes the learner's speech rate, tone, and volume changes to determine their focus and emotional stability. The facial expression and body language sensing unit integrates video footage to identify the learner's confused, tired, or unresponsive behavioral characteristics. The reinforcement lesson plan recommendation engine compares the behavioral assessment results with a lesson plan database, selects reinforcement modules, and pushes them to the user. The learning interface response unit dynamically adjusts the teaching interface colors, speech rate, interaction density, or prompts to improve learner engagement and comprehension.

[0026] Preferably, this virtual teacher control module may further include a multimodal interactive response unit, which includes a voice command parsing submodule, a gesture sensing submodule, a teaching instruction response submodule, and a virtual teacher synchronous simulation engine submodule. The voice command parsing submodule can receive and parse the learner's voice input. The gesture sensing submodule can receive specific actions detected by the XR device, such as raising a hand, shaking the head, or clicking. The teaching instruction response submodule can determine the operation content based on the voice or gesture input, such as playback control, replaying the previous segment, providing examples, switching speech rate, or language mode. The virtual teacher synchronous simulation engine submodule can immediately trigger the corresponding voice, facial expression, and visual response actions of the virtual character based on the determination result.

[0027] Preferably, this AI learning analysis module can perform cross-lesson learning performance modeling, and may further include a learning feature extraction unit, a feature vector normalization and comparison unit, a machine learning algorithm unit, and a lesson plan adjustment prompt interface unit. The learning feature extraction unit can be used to extract the learner's behavioral data across different course modules. The feature vector normalization and comparison unit can be used to compare and measure the similarity of operational patterns between modules. The machine learning algorithm unit can construct individual learning behavior models and predict their subsequent module learning effectiveness and risks. The lesson plan adjustment prompt interface unit can convert the model output into interface suggestions for teachers to review and then use for reinforcement or skipping topics.

[0028] Preferably, this virtual classroom teaching device supports four input channels: voice, facial expression, gesture, and eye tracking. It may include a multimodal input synchronization management module, an input conflict detection module, a multimodal input recording and archiving mechanism module, and a teacher-side retrospective review tool module. The input conflict detection module can be used to determine the priority and consistency among multiple input sources, the multimodal input recording and archiving mechanism module can track learning behaviors and strategy preferences, and the teacher-side retrospective review tool module can assist teachers in reconstructing learning situations and emotional trajectories with unstructured input.

[0029] Preferably, this virtual teacher control module supports automatic switching between language and subtitles, and may include a language recognition engine unit, a speech simulation switching unit, a subtitle synchronization processing unit, and a translation accuracy correction unit. The language recognition engine unit can detect the user's preset language and speech output feedback. The speech simulation switching unit can switch the virtual teacher's accent and speech rate model according to the lesson plan's language. The subtitle synchronization processing unit supports bilingual subtitle switching and automatic time synchronization. The translation accuracy correction unit can adjust the real-time subtitle content based on the context to ensure semantic integrity.

[0030] Preferably, the virtual classroom teaching device may further include a diagnostic question recommendation module, which may include a dynamic question bank management unit, a difficulty and knowledge point mapping calculation unit, an AI-recommended diagnostic question unit, and a response result analysis unit. The difficulty and knowledge point mapping calculation unit can be used to compare the learner's current module comprehension errors. The AI-recommended diagnostic question unit can generate targeted diagnostic questions and embed them into the lesson plan process in real time. The response result analysis unit can calculate the accuracy rate, reaction time, and error distribution to recommend corresponding reinforcement modules and content extensions.

[0031] Preferably, the cloud-based resume database may include a behavior trajectory recording module, a multi-modal input process organization engine module, an achievement marking and learning milestone module, and a cloud-based resume query and retrospective interface module to support multi-level analysis and process review functions for teachers and students.

[0032] Preferably, this learning process record file can be automatically synchronized to a school learning management system (LMS) platform through the virtual classroom teaching device. This virtual classroom teaching device may further include a data interface transcoding module, a student data mapping module, a grade backfilling and process upload synchronization module, and an export report tool and assessment conversion rule engine module. The data interface transcoding module can support LTI, SCORM, or xAPI format conversion. The student data mapping module can correspond to the school roster and student identification.

[0033] Preferably, teachers can use this virtual classroom teaching device to make horizontal comparisons of the learning processes of different students in different modules. The virtual classroom teaching device also includes a learning process visualization parallel coordinate module, a differentiated heat map presentation tool module, and a group behavior analysis module to support teachers in designing different teaching diversion strategies based on behavior groups.

[0034] Preferably, this AI learning analytics module can analyze the correlation between different lesson plan styles (such as gamified, narrative, and task-oriented) and learning outcomes, and use it to produce a lesson plan style classification database, a learning effectiveness trend model, and an adaptive correspondence graph and teacher feedback loop module.

[0035] Preferably, all learning process and behavioral data can be uploaded to a central model training platform after anonymization. This central model training platform may include a data stripping module, a hash mapping and anonymized ID recompilation engine module, a hierarchical behavioral feature extraction module and data pool module, and a lesson plan design error prediction algorithm module.

[0036] Preferably, the tone, speech, speech rate, and supplementation method of the virtual teacher image in this virtual teacher control module can be adjusted in real time according to the user's experience and feedback data. This virtual teacher control module may also include a tone adaptation algorithm unit, a perception friendliness correction parameter setting unit, a contextual guidance logic decision unit, and an expandable speech feature library unit. Simple Explanation of the Diagram

[0037] Figure 1 is a block diagram of the virtual classroom teaching device according to an embodiment of the present invention.

[0038] Figure 2 is a block diagram of a virtual classroom teaching device according to another embodiment of the present invention.

[0039] Figure 3 is a flowchart of the virtual classroom teaching method according to an embodiment of the present invention.

[0040] Figure 4 is another flowchart of the virtual classroom teaching method according to an embodiment of the present invention.

[0041] Figure 5 is a block diagram of the visualization dashboard according to an embodiment of the present invention.

[0042] Figure 6 is a block diagram of the emotion feedback enhancement module in an embodiment of the present invention.

[0043] Figure 7 is a block diagram of the multimodal interactive response unit according to an embodiment of the present invention.

[0044] Figure 8 is a block diagram of the AI ​​learning and analysis module according to an embodiment of the present invention.

[0045] Figure 9 is a block diagram of the virtual classroom teaching device according to the second embodiment of the present invention.

[0046] Figure 10 is a block diagram of a virtual teacher control module according to another embodiment of the present invention.

[0047] Figure 11 is a block diagram of the diagnostic question recommendation process module according to an embodiment of the present invention.

[0048] Figure 12 is a block diagram of the cloud-based resume database according to an embodiment of the present invention.

[0049] Figure 13 is a block diagram of the virtual classroom teaching device according to the third embodiment of the present invention.

[0050] Figure 14 is a block diagram of the virtual classroom teaching device according to the fourth embodiment of the present invention.

[0051] Figure 15 is a block diagram of the central model training platform according to an embodiment of the present invention.

[0052] Figure 16 is a block diagram of the virtual teacher control module according to the second embodiment of the present invention. Implementation

[0053] To enable the examiner to understand the inventive features, content, advantages, and effects of this invention, the invention is described in detail below with reference to the accompanying drawings and in the form of embodiments. The drawings used are for illustrative purposes only and to assist in the description. They may not represent the actual proportions and precise configurations of the invention after implementation. Therefore, the proportions and configurations of the accompanying drawings should not be used to interpret or limit the scope of the invention in actual implementation.

[0054] The advantages, features, and technical methods of this invention will be more readily understood by referring to the exemplary embodiments and accompanying drawings. This invention may be implemented in different forms and should not be construed as limited to the embodiments set forth herein. Rather, the embodiments provided will make this disclosure more thorough, complete, and fully convey the scope of the invention to those skilled in the art. This invention will be defined only by the appended claims.

[0055] Please refer to Figure 1, which is a block diagram of the virtual classroom teaching device according to an embodiment of the present invention. In this embodiment, the virtual classroom teaching device 100 can be applied to a learner's learning environment. It may include a teaching material format recognition module 10, a modular reconstruction engine module 20, a virtual teacher control module 30, and a feedback detection and behavior analysis module 40. The teaching material format recognition module 10, the modular reconstruction engine module 20, and the virtual teacher control module 30 can be software applications. The former two can be installed on a cloud server, while the latter can be installed on the learner's computer host, laptop, or VR / AR device. The computer host, laptop, or VR / AR device can be connected to the cloud server via the Internet. The feedback detection and behavior analysis module 40 can be a hardware or software device that monitors learner behavior, such as a camera, a microphone, a keyboard, a mouse, or any combination thereof, and is operated and monitored by a software application. This feedback detection and behavior analysis module 40 can be electrically connected to the learner's computer host or laptop.

[0056] In this embodiment, the teaching material format recognition module 10 in the cloud server allows users (e.g., teachers) to input multiple digital teaching materials A and automatically recognize and convert the format of these digital teaching materials A to generate multiple modular lesson plan units 11 with the same file format and which can be disassembled. This teaching material format recognition module 10 can automatically recognize 360-degree images, 3D interactive models, and immersive video formats.

[0057] This teaching material format recognition module 10 can effectively parse various immersive teaching material source formats, including 360-degree panoramic photography, 3D model data such as FBX / GLTF, interactive WebXR scenes, and immersive video files that conform to international standards (such as 8K videos encoded in H.265). Through metadata tagging and format conversion processes, it can automatically classify and convert original XR teaching materials (such as VR, AR, or MR digital data) into teaching material formats acceptable to the teaching material center database. This module also supports dynamic detection of teaching material changes, ensuring seamless integration into virtual teacher scenes, interactive learning sections, and AI assessment modules in the subsequent teaching process, providing highly accurate and efficient automated support for integrating XR teaching content.

[0058] Furthermore, this modular lesson plan unit 11 can be labeled with learning objectives, knowledge points, and competency indicators, facilitating subsequent AI inference. This device supports adding tagging functionality to each segment of the modular lesson plan unit 11. Teachers can manually or use templates to assign learning outcomes, corresponding syllabus knowledge points, and competency indicators to each teaching segment. This labeled information will serve as crucial parameters for AI recommendations and learning progress tracking, and will also facilitate subsequent comparison and evaluation of teaching outcomes. This allows the system to implement individualized learning progress control and module suggestions for each learner, truly achieving refined, structured, and goal-oriented smart teaching.

[0059] Furthermore, the modular lesson plan unit 11 of this invention is detachable, with each paragraph or task independently editable and replaceable. The lesson plan design adopts a modular, segmented structure, breaking down a complete lesson plan into several teaching nodes (units), such as concept explanations, practical activities, interactive exercises, and assessment tasks. Each node is an independently loadable and executable module, supporting real-time replacement, version updates, and parallel testing of multiple versions. This structure facilitates teachers' rapid reorganization of lesson plan processes for different classes, teaching strategies, or textbook revision needs. It also benefits AI by enabling refined effectiveness analysis and individualized recommendations for individual modules, significantly improving the adaptability and flexibility of teaching materials.

[0060] In one embodiment, the textbook format recognition module 10 may include a lesson plan parser, which supports parsing XR content from SCORM, HTML5 textbooks, and custom JSON-structured lesson plans. The lesson plan parser proposed in this embodiment has diverse content format support capabilities, covering mainstream e-learning standards (such as SCORM and xAPI) and modern web textbook standards (such as HTML5 animations, Canvas modules, and WebXR scenarios), and also supports the JSON-structured textbook description syntax commonly used in educational technology development. Through the parser's built-in syntax recognizer and structure mapping engine, it can automatically extract textbook titles, paragraphs, interactive objects, and teaching node annotations, and combine them with the XR module's reorganization engine to achieve integrated management and conversion of cross-format textbooks. This improves the efficiency of textbook updates and lowers the barrier to reusing teaching resources.

[0061] Furthermore, the modular reorganization engine module 20 within the cloud server can adjust, reinforce, and optimize the order of these modular lesson plan units 11 based on learners' learning behavior and learning outcomes. Learning outcomes can be assessed through a satisfaction survey or online quiz conducted by learners. Order adjustment means adjusting the sequence of teaching content in the modular lesson plan unit 11. For example, when learning English inverted sentence grammar, if learning behavior and outcomes indicate that following the order of modules 12345 is difficult for learners, it can be adjusted to "0"1"32"45"67, reinforcing the prior knowledge of "0" and reinforcing the subsequent usage of inverted sentence grammar in section 67 or other teaching details.

[0062] In a preferred embodiment, this modular reorganization engine module 20 includes a version control unit 21, which records the reorganization process of the modular lesson plan unit 11 and provides version labeling, comparison, and restoration functions for teachers to switch interfaces and track progress. Furthermore, each reorganization, update, or adjustment of the modular lesson plan unit 11 is automatically saved as an independent version, and the operation time, teacher account, and modification summary are recorded. Teachers can revert to a previous version at any time or compare versions to observe the impact of different designs on learning outcomes. The version control mechanism supports visual representation of differences and allows teachers to label "stable version" or "experimental version." This design responds to the need for "content as an evolutionary asset" in educational settings, ensuring that lesson plan content can continuously evolve, be reviewed, and accumulated, further enhancing its technological advancement and educational record value.

[0063] In a preferred embodiment, the virtual classroom teaching device 100 provides an AI-automated reorganization module that reorganizes the lesson plan sequence according to the learner's level and preferences. It can analyze the learner's past behavioral records, interaction habits, learning styles (such as visual or hands-on), and past test performance through a deep neural network architecture to automatically predict the optimal learning sequence and content absorption pace. By combining the AI ​​module with a modular lesson plan structure, the order of modules can be rearranged, supplementary modules added, or familiar units skipped without affecting the teaching logic, forming a personalized lesson plan version. This dynamic lesson plan process can provide real-time feedback to the teacher interface for adjustment and approval, significantly improving learning efficiency and the ability to address individual differences.

[0064] After lesson plans are restructured, an automatic logical check is performed to prevent breaks in teaching or misalignment of sequences. To ensure the consistency of the logical flow of lesson plans generated by AI or restructured by teachers, this device has a built-in "Teaching Path Consistency Check Module." This module compares the dependencies, knowledge progression order, and prerequisite skill requirements between lesson plan modules to check for module misalignment (such as teaching advanced concepts before basic ones) or logical breaks (such as missing transitional explanations). It also provides warnings and correction suggestions for teachers to refer to, ensuring the logical smoothness of the teaching process and the continuity of learners' cognition. This function is particularly important for large-scale deployment and multi-version maintenance, improving the quality and maintainability of lesson plan content.

[0065] On the other hand, the virtual teacher control module 30 can drive a virtual teacher image 31 to interact with learners and provide teaching responses. The virtual teacher image 31 has a speech recognition system, supports commands such as "speak again" and "slow down." This virtual teacher control module 30 incorporates speech recognition and semantic understanding technology, enabling it to listen to students' speech input in real time and respond to common commands. For example, students can naturally say phrases such as "speak again," "replay the previous segment," and "speak slower." The device will analyze the speech content and drive corresponding teaching responses, such as replaying, adjusting speed, providing supplementary explanations, or offering examples. This module supports multilingual input and accent tolerance optimization, and combines contextual semantic interpretation technology to improve the accuracy and fluency of speech interaction. Its aim is to allow students to enjoy a "response upon speaking" teaching experience, just like in a physical classroom, even when learning independently.

[0066] Furthermore, the virtual teacher control module 30 supports gesture commands, allowing users to wave, nod, or frown as interactive inputs. This module can be equipped with image recognition and skeleton tracking algorithms, which can detect the user's gestures, head movements, and facial expressions through the camera lens and translate them into interactive teaching inputs. For example, students can wave to access interactive menus, nod to indicate "I understand," shake their heads to trigger supplementary explanations, or frown to indicate confusion and request further explanation. This non-verbal interactive design not only enhances immersion but is also particularly suitable for beginners in language learning, children, or learners with special educational needs, making teaching communication more intuitive and natural. In addition, this device can also analyze gesture frequency and emotional fluctuations to detect learning fatigue and adjust the teaching pace.

[0067] The virtual teacher image 31 allows users to switch voice tone and virtual face style, enhancing approachability. In this invention, the virtual teacher image 31 is a crucial interface element of the device. Its design allows teachers to pre-select or learners to dynamically switch the virtual teacher's voice style (lively, steady, friendly, etc.), tone speed, gender-specific voice type, and facial appearance (such as cartoon style, realistic style, or culturally localized style). This flexible switching of appearance and voice styles effectively enhances learners' emotional engagement and acceptance, especially for young children, language learners, and students with emotional disorders; an approachable interactive character is key to boosting learning motivation. The virtual teacher image also supports synchronized facial expressions and voice-over, enhancing its realism and immersive experience.

[0068] This invention supports role switching in the classroom, allowing virtual teachers to simulate scenarios such as interviews and debates. The virtual teacher image 31 of this invention supports multi-role switching and interactive story simulation, enabling the configuration of multiple virtual characters within a single lesson plan for interactive presentations, such as simulating an interview, a historical debate, or a scientific dialogue experiment. Each character has independent voice and action profiles, and automatically triggers appearances and interactions according to script nodes, enhancing the immersive learning environment. Teachers can also customize character personalities, dialogue logic, and interaction rhythm through a scene editor. Students can choose to embody one of the characters, engaging in role-playing and strengthening their expression, logic, and pragmatic abilities through interactive learning.

[0069] This invention supports multilingual virtual teachers who can automatically switch between subtitle and voice modules. It features a built-in multilingual synthesized voice and subtitle synchronization module, supporting over ten languages ​​including English, Japanese, Korean, French, and German. Users can preset their language before use, or the device can automatically determine the language environment based on the logged-in region and switch the interface and voice accordingly. The virtual teacher will also switch its voice model and lip-sync animation according to language changes, ensuring consistent voice and subtitle synchronization and enhancing the user experience for non-native language learners. Technically, this function integrates a TTS module, NLP semantic prediction, and a cross-language subtitle engine, showcasing the platform's innovative breakthroughs in international and multilingual teaching.

[0070] The feedback detection and behavior analysis module 40 can be used to collect real-time learning status data 41 of one of the learners and use this real-time learning status data 41 as a basis for push recommendations. In one embodiment, a deep learning model can be used to track learners' operational behaviors, interaction frequency, and dwell time. This virtual classroom teaching device can deploy a deep learning network model (such as LSTM or Transformer architecture) based on time-series data analysis to continuously monitor and analyze every behavioral record of students on the teaching platform, including click order, number of interactions, mouse trajectory, material dwell time, answer time difference, and error rate distribution. After feature transformation, this behavioral vector can be used as a basis for predicting students' learning status, supporting real-time intervention reminders, dynamic adjustment of module difficulty, and personalized reinforcement suggestions. Through this model, group learning behavior patterns and material hotspots can also be analyzed, providing feedback to teachers for content optimization and teaching strategy adjustment.

[0071] Additionally, the feedback detection and behavior analysis module 40 can utilize an AI push notification algorithm, taking into account individual learning speed, historical error patterns, and style preferences. The system's AI push notification recommendation module integrates various learning behavior data sources, including past quiz results, reaction speed distribution, learning time preferences (e.g., nighttime vs. daytime), frequently used devices (phones, tablets), content dwell hotspots, and learning style models (e.g., Felder-Silverman or VARK scales). This data is aggregated, feature-embedded, and input into a deep recommendation network architecture to generate a highly adaptive sequence of lesson plan recommendations. The push notification mechanism can also consider teachers' pre-set teaching topics, enabling real-time content matching and synchronous course notifications, effectively enhancing student engagement and participation.

[0072] Furthermore, the recommendation system can simultaneously compare the behavior of other students in the same domain to reinforce recommendations. This feedback detection and behavior analysis module 40 not only personalizes content delivery based on a single student's learning journey but also introduces collaborative filtering technology across user groups to analyze and compare the behavioral patterns of multiple learners within the same lesson plan module. For example, for a student practicing English conversation, the recommendation system will examine the performance data of other students in the same module (such as interaction time, answer accuracy, replay frequency, etc.) to identify the behavioral trends that best improve learning outcomes, and then adjust the delivery content and pace accordingly. This mechanism reduces the cold start problem, improves the accuracy of the recommendation system and the adaptability of teaching content, thereby helping students keep up with their learning progress and strengthen their weaknesses.

[0073] In a preferred embodiment, the device may include a learning process analysis module that can instantly identify learners' knowledge comprehension bottlenecks based on behavioral data such as error types, repeated playback segments, and excessive dwell time, and automatically activate corresponding supplementary teaching material modules or design reinforcement tasks. For example, if a student makes too many errors on a specific type of math problem, the device will recommend supplementary animated explanations or provide interactive question banks for practice. These reinforcement modules are all independent components with loading trigger conditions and exit strategies, and can be seamlessly integrated with the main teaching plan to form a targeted and precise diagnostic teaching mechanism.

[0074] Learners' learning progress can be compared across different lesson plans. This virtual classroom teaching device 100 features cross-lesson plan progress tracking and comparison capabilities, integrating learners' behavioral and performance data across different lesson plan modules, teaching topics, and stages, such as interaction frequency, error types, completion time, and replay rate. Through standardized learning performance indicators, this device can cross-reference student performance across different lesson plans, helping teachers identify learners' cognitive gaps and enabling AI modules to predict potential performance risks in future modules. Compared to traditional single-lesson-plan assessments, this invention represents a significant improvement, effectively addressing the problems of fragmented learning processes, tracking difficulties, and untimely remediation.

[0075] In this embodiment, lesson plan recommendations can incorporate multi-level confidence metrics, allowing teachers to decide whether to adopt them based on reliability. When generating recommended lesson plans, the AI-powered push system simultaneously adds multi-level confidence scores, calculated based on model matching, historical effectiveness, learner characteristic fit, and group usage effectiveness. Confidence metrics include: personalized fit rate, module matching accuracy, recommendation model stability, and peer success rate. These metrics are displayed in the teacher's console, helping teachers quickly identify whether recommendations are worth adopting. Teachers can also set confidence thresholds for automatic adoption or prompt for manual review. Furthermore, teachers can continuously train the model based on feedback results, improving the reliability and accuracy of future recommendations.

[0076] In one embodiment, the feedback detection and behavior analysis module 40 may include a voice emotion analysis unit 42 and an eye movement tracking unit 43 to instantly identify the learner's emotions and concentration state, and adjust or enhance the content of the modular lesson plan units 11 accordingly.

[0077] In one embodiment, the virtual teacher control module 30 may include a virtual character database 32 and may automatically switch the voice, appearance and interaction style of the virtual teacher image 31 according to the course theme and the learner's language.

[0078] In another embodiment, the virtual classroom teaching device 100 may include an AI learning analysis module 50, which may further include a behavioral feature extraction unit 51, a multi-dimensional classification and clustering model unit 52, a module fit calculation unit 53, and a recommendation and ranking unit 54. The multi-dimensional classification and clustering model unit 52 can be used to analyze learners' answer paths, operation clicks, visual dwell times, and interactive responses. The multi-dimensional classification and clustering model unit 52 can be used to automatically identify the learner's type. The module fit calculation unit 53 can be used to predict the learner's expected performance under different lesson plan modules. The recommendation and ranking unit 54 can automatically calculate reinforcement and reorganization strategies based on real-time detected behavioral data to produce a personalized lesson plan recommendation module list and confidence matrix based on process data. A schematic diagram is shown in Figure 2.

[0079] In one embodiment, the present invention can automatically generate a preview thumbnail of an XR lesson plan for teachers to compare differences in teaching materials. To effectively assist teachers in understanding and selecting appropriate teaching materials, the present invention can include a built-in XR content visual transcoding module that can automatically extract core images, interactive nodes, or virtual scene images from each XR lesson plan, generating static thumbnails and interactive timeline markers. These thumbnails can be displayed instantly in the lesson plan editing interface and recommendation module results, allowing teachers to clearly understand the key differences, visual styles, and levels of interaction among various teaching materials, thus easily selecting the lesson plan that best meets learners' needs. This module also supports multilingual previews and version markings, enhancing cross-regional teaching material management and comparative analysis.

[0080] In one embodiment, the present invention provides a teacher editing interface that allows for semi-automatic and manual adjustment of module combinations. The invention designs a user-experience-oriented lesson plan editing interface where teachers can drag and drop teaching modules, set jump conditions, add notes, and insert external teaching materials. An embedded AI module provides real-time content recommendations and sequence optimization suggestions, and displays the relationships and recommendation scores between modules. However, teachers can still manually override decisions, preserving the flexibility of professional teaching judgment. This teacher editing interface also features version management, preview simulation, and process comparison functions, helping teachers track the potential impact of lesson plan modifications on learning outcomes, achieving an optimized instructional design model of "professional-led, AI-assisted."

[0081] In a preferred embodiment, the present invention can design a set of "recommendation correction control panels," allowing teachers to adjust the push priority and weight values ​​of specific lesson plan modules, or directly exclude certain lesson plan modules, based on student satisfaction, completion rate, and effectiveness evaluation data collected by the system after teaching. This function, combined with a human-machine co-creation mechanism, preserves the adjustment space of teachers' subjective teaching experience, while simultaneously feeding back into the AI ​​recommendation system for retraining. This achieves collaborative evolution between machine learning and practical teaching, effectively enhancing the practical usability and technical flexibility of the device.

[0082] During the teaching process, student interactions can be recorded in real time for synchronous feedback. This device provides a teaching process recording module that automatically records students' interactive behavior data while they are taking virtual lessons, including voice commands, responses, eye movements, gestures, and virtual object operations. This data is instantly converted into a behavior log file and provides synchronous prompts with a feedback engine, such as immediately displaying the correct answer when a student answers incorrectly, adjusting the speaking speed when a student shows confusion, and sending encouraging messages when there is no interaction. All behavior processes are also stored in the learning portfolio, allowing teachers to diagnose, provide feedback, and make individualized adjustments after class, enabling learners to correct mistakes in real time and enhancing their learning motivation.

[0083] It is worth mentioning that in this invention, all learners' interaction records are anonymously uploaded to the cloud for group behavior modeling. To enhance learning analysis without infringing on personal data privacy, this invention introduces an anonymization mechanism, separating and storing personal identification information and interaction behavior data on the student's end, and transmitting them to the cloud data lake platform using hash encoding. The AI ​​model can perform group behavior modeling in the cloud computing environment, including learning strategy classification, interaction hotspot prediction, and learning error pattern recognition. This anonymized modeling method ensures the legal use of educational materials on the one hand, and enhances the depth and breadth of big data analysis on the other, laying a technical foundation for future international platformization and educational research.

[0084] In this embodiment, switching between synchronous teaching (online live streaming) and asynchronous learning (pre-recorded XR lesson plans) is supported. This device supports a dual-track teaching mode: synchronous mode uses XR classroom live streaming technology to allow teachers and students to simultaneously enter a virtual space for real-time teaching and interaction; asynchronous mode primarily uses pre-recorded XR lesson plans, allowing learners to determine their own time and pace. This device allows teachers to save the live stream as pre-recorded lesson plans for later playback and reuse. Students can also ask questions at any time during asynchronous sessions, which will be compiled synchronously when the teacher logs in. This mode combines flexibility and immediacy, making it suitable for flipped classroom, distance learning, and differentiated learning scenarios.

[0085] In one embodiment, during an asynchronous learning process, learning records are integrated and an analytics dashboard is provided. All student behavioral data during asynchronous learning, such as video playback counts, interactive clicks, question-answering processes, and voice response records, are automatically integrated into their individual learning portfolios. The device provides a visual dashboard displaying student participation, accuracy, repetition frequency, and progress trends across different modules, helping teachers quickly grasp learning outcomes. Students can also reflect on their learning through self-process maps, creating a virtuous cycle of self-adjustment and goal setting. The dashboard also supports access from parent or tutor accounts, facilitating learning support and feedback.

[0086] In one embodiment, the virtual classroom teaching device 100 can automatically establish a horizontal comparison matrix to analyze which types of lesson plans trigger higher interaction. This module can automatically establish a lesson plan interaction comparison matrix based on learners' interactive behaviors across multiple lesson plans, such as voice input frequency, gesture response frequency, answering speed, and eye focus time, and use a clustering algorithm to distinguish between high-interaction lesson plan groups and low-response lesson plan groups. This horizontal analysis technology has implicit innovation, identifying the correlation between teaching material interactivity and learning effectiveness, assisting teachers in adjusting lesson plan design or prioritizing the distribution of high-interaction modules. Furthermore, through long-term accumulated matrix data, the AI ​​model can iteratively adjust recommendation strategies, forming a database of optimal teaching material interaction strategies.

[0087] It is worth mentioning that all modules of this invention have extensible APIs, facilitating integration with existing campus systems and learning platforms. To enhance the platform's scalability and ease of campus implementation, this invention features a modular architecture with RESTful API and WebSocket support. Each lesson plan module, interactive interface, and analytics engine provides API interfaces that can be connected to third-party systems such as school management platforms, digital whiteboards, teaching projectors, or IoT devices. Through this open architecture, educational institutions can customize and integrate projects according to their internal needs, and it can also be extended to blockchain authentication, credit exchange, or educational big data platforms in the future, demonstrating the device's high scalability and innovative potential in future educational infrastructure.

[0088] The following three embodiments illustrate the virtual classroom teaching device proposed in this invention.

[0089] [Implementation Case 1: Junior High School Science XR Interactive Teaching] This embodiment illustrates a junior high school science XR interactive teaching method. In this embodiment, a junior high school science teacher can create a modular XR course on "Volcanic Eruptions and Plate Tectonics" using this virtual classroom teaching device. This course plan may include: a 360-degree panoramic video of a volcano (Module A), a virtual teacher's guided explanation (Module B), interactive model observation (Module C), and an earthquake simulation experiment (Module D). Through student operation records and answer performance in different modules, the AI ​​learning analysis module of this invention can perform behavioral model deduction and determine that some students have a weak understanding of plate tectonics. Therefore, it pushes supplementary lesson plan module E and automatically incorporates module E into the student's personalized learning process. In this way, the teacher can simultaneously monitor all students' behavioral hotspots and error types, and provide immediate reinforcement in the next lesson. This embodiment demonstrates the advantages of this device in "real-time interaction," "diagnostic reinforcement," and "XR immersive teaching."

[0090] [Implementation Case 2: Application of Electrical Engineering Practice Module Reconfiguration in Vocational Colleges] This embodiment illustrates the reconfiguration of an electrical engineering practice module in a vocational college. In this embodiment, vocational colleges can integrate this virtual classroom teaching device into their electrical engineering practice courses and establish a modular XR teaching plan for "Circuit Testing and Safe Operation" for students. In this embodiment, students can use AR glasses to enter the experimental field to simulate wiring devices and receive real-time voice prompts and instructions through the virtual teacher image of this invention. On the other hand, the AI ​​learning and analysis module can automatically determine the student's repeated errors in the "Grounding Principle" section through operation path analysis and repeated error comparison, and then push relevant auxiliary modules, which may include: animated breakdown diagrams (module X), error demonstration cases (module Y), and voice interactive Q&A (module Z). This hybrid learning process is integrated into the LMS for teacher review and provides students with precise diagnostic feedback, significantly improving learning effectiveness and practice safety.

[0091] [Implementation Case 3: International Chinese Language Teaching and Cross-Language XR Synchronous Interaction] This embodiment illustrates an example of international Chinese language teaching and cross-language XR synchronous interaction. In this embodiment, a multilingual XR virtual interactive teaching system for language learning can be designed. The design supports student-side devices with automatic language switching and synchronous subtitle / voice control modules. Students can wear XR devices with speech recognition and subtitle matching. In this XR device, the virtual teacher image in the virtual teacher control module can synchronously output Chinese and English sentences to the student. While the student practices speaking, this virtual teacher control module can simultaneously record the pronunciation accuracy and correction suggestions, and display a historical practice record and an evaluation score. Furthermore, this XR device can provide different interactive learning based on the connection status.

[0092] This embodiment combines innovative modules such as speech recognition, multilingual subtitles and speech modules, asynchronous learning record integration, and anonymized feedback analysis. With synchronous and asynchronous interactive learning as the dual core mode, it effectively solves the pain points of traditional language learning, such as lack of immediacy, interactivity, and internationalization.

[0093] Specifically, an overseas Chinese language learning center has implemented a lesson plan on "Everyday Chinese - Ordering Food" using the virtual classroom teaching device of this invention. After logging into the virtual classroom teaching device, learners automatically switch between virtual teacher's voice and subtitles according to a preset language system. The lesson plan seen after logging in includes an AR scan of a real scene, a menu selection task, and a simulated ordering of food in spoken Chinese. The feedback detection and behavior analysis module can analyze the learner's pronunciation accuracy and speech rhythm, and, in conjunction with the AI ​​learning analysis module, recommend highly interactive lesson plans from other regions as supplementary resources. For learners, they can practice asynchronously at any time. The virtual classroom teaching device of this invention will automatically record speech parameters and correction processes, which will be converted into references for teacher evaluation. This case demonstrates the innovative application value of this device in areas such as "multilingual interaction," "voice feedback," and "cross-border synchronous recommendation."

[0094] Please refer to Figure 3, which is a flowchart of the virtual classroom teaching method according to an embodiment of the present invention, and also refer to Figures 1 and 2. The virtual classroom teaching method of the present invention can be applied to the aforementioned virtual classroom teaching device 100, wherein the virtual classroom teaching device 100 may include a textbook format recognition module 10, a modular reorganization engine module 20, a virtual teacher control module 30, and a feedback detection and behavior analysis module 40. This virtual classroom teaching method includes the following steps.

[0095] Step (a) is a textbook uploading and modularization conversion step. The teacher can upload a digital textbook A to the textbook format recognition module 10 to parse its type and convert it into a modular lesson plan unit 11 that can be broken down.

[0096] Step (b) is a lesson plan distribution and initialization step. The adapted modular lesson plan unit 11 is automatically loaded according to a learner's terminal device, and a learning task flow is initialized.

[0097] Step (c) is a virtual teaching interaction step. The virtual teacher controls module 30 to guide the teaching, supporting voice input, facial feedback, gestures or clicks for interaction, and switching the teaching style according to the context of the modular lesson plan unit 11.

[0098] Step (d) is a behavioral feedback acquisition step. The feedback detection and behavior analysis module 40 acquires the learner's operation data, speech recognition results, facial expressions and eye tracking information to form a learning process record file.

[0099] Step (e) is an AI analysis and recommendation step. The learning process record file mentioned above is input into an AI learning analysis module 50 to perform behavioral model inference, calculate a module fit and a teaching confidence score, and generate a recommendation list and restructuring suggestions.

[0100] Step (f) is a teacher review and fine-tuning step. The reorganization results and system suggestions are presented in a lesson plan editing interface on the teacher's end, and the teacher can use this lesson plan editing interface to fine-tune, replace modules, or modify the workflow sequence.

[0101] Step (g) is a push and process archiving step: the confirmed modular teaching plan unit 11 is pushed to the learner's device to perform interactive teaching, and the complete learning process record file is automatically saved in a cloud resume database 202 after class for future diagnosis and teaching optimization reference.

[0102] The virtual classroom teaching method disclosed in this invention can effectively assist lesson plan designers in arranging and reviewing course units through a series of automated steps, ensuring the logic, systematicity, and completeness of the teaching content. Referring to Figure 4, in a preferred embodiment, steps (e) to (f) may further include the following four steps:

[0103] Step (e1) examines the prerequisite dependencies and competency progression labels between each modular lesson plan unit 11. In this step, it analyzes whether each modular lesson plan unit 11 depends on the basic knowledge or skills of other teaching units and labels its corresponding competency level. The purpose of this step is to establish the logical framework and learning ladder between the modular lesson plan units 11, ensuring that learners can progressively construct knowledge and skills.

[0104] Step (e2) automatically compares whether the module order violates the logic of cognitive progression in instruction. For example, if a modular lesson plan unit 11 requires prior "understanding" ability before proceeding to the "application" stage, the invention will check whether the arrangement order violates this logic. If students enter a higher-level application module before completing basic learning, the device will determine it as unreasonable. Through this mechanism, the lesson plan can better align with learners' learning curves and cognitive load theory, avoiding comprehension barriers caused by skipping steps in learning.

[0105] Step (e3) detects whether there is duplicate teaching or missing modules. Duplicate teaching refers to the same content or skills being repeatedly arranged in different modular lesson plan units 11, which may lead to a waste of course resources and student learning fatigue. Missing modules mean that certain necessary aspects of ability development are not covered in the overall lesson plan, which will affect learners' learning outcomes and the overall integrity of teaching. Through this detection function, this device can remind designers to adjust the content to make the course more efficient and complete.

[0106] Step (e4) displays non-compliant or duplicate modules in the lesson plan editing interface and provides adjustment suggestions. These modules will be clearly displayed in the lesson plan editing interface through visual methods such as color coding and icon reminders, allowing users to immediately identify problems. In addition, the device will also provide adjustment suggestions, which may include suggestions for reordering modules, adding teaching content, or merging duplicate content, etc. These suggestions help designers quickly optimize, save modification time, and improve teaching quality.

[0107] In summary, this virtual classroom teaching method balances the logical flow of instruction with the completeness of content. Through intelligent and automated analysis and reminder mechanisms, it makes the lesson plan design process more efficient and scientific. It not only enhances teachers' instructional design capabilities but also ensures students' learning continuity and overall effectiveness, demonstrating high practical value and educational promotion potential. In the future, combining it with more diverse learning process data and artificial intelligence technology will further enable the realization of individualized instruction and intelligent curriculum design goals.

[0108] Please refer to Figure 5 for further explanation. The learning process record file of the present invention can be presented in real time by a visualization dashboard 203, wherein the visualization dashboard 203 can be implemented by a software application. The visualization dashboard 203 may include: a module progress tracking unit 2031, an error and dwell hotspot analysis chart 2032, an emotion and focus trend curve chart 2033, and a teacher real-time interaction toolbar 2034.

[0109] The module progress tracking unit 2031 can be used to display the learner's completion percentage and operation records in each lesson plan module, including clicks, interactions, submissions, and feedback. Teachers can use this unit to understand whether learners are completing their learning on schedule, whether there are delays or skipped modules, and make timely adjustments.

[0110] The Error and Dwell Time Hotspot Analysis Chart 2032 can generate key warning nodes based on operation time and repeated error locations. It analyzes learners' operation time on the device and the frequency of errors to identify key nodes where learning difficulties or excessive dwell time frequently occur; these nodes are called "learning bottlenecks." The device automatically marks these locations as "warning nodes," reminding teachers of potential teaching blind spots or learning difficulties, facilitating targeted teaching or remedial measures.

[0111] The Emotion and Attention Trend Chart 2033 can be visualized by combining voice tone, facial expression and eye movement analysis. After analysis, this data will generate a visual chart, allowing teachers to clearly understand whether students are inattentive, depressed or anxious during specific learning periods, and thus understand the fluctuations in students' learning motivation and state.

[0112] The Teacher Real-Time Interaction Toolbar 2034 allows teachers to mark, group, and distribute reinforcement modules for abnormal nodes. Teachers can annotate and explain the abnormal nodes marked by the system based on the data provided, and can directly group students or distribute reinforcement modules to specific students to improve learning outcomes. These real-time intervention and feedback mechanisms help achieve differentiated teaching and individualized tutoring, making overall teaching more precise and effective.

[0113] In summary, these modules and tools together constitute an intelligent, data-driven teaching monitoring and support system. It not only allows teachers to more accurately grasp the learning status of each student, but also provides a solid foundation for individualized learning and intelligent teaching, representing an important direction for the future development of digital education.

[0114] In this invention, an asynchronous learning process may also be included. In this asynchronous learning process, the virtual classroom teaching device 100 may include an emotion feedback enhancement module 60, which includes a voice emotion recognition unit 61, an expression and body sensing unit 62, an enhancement lesson plan recommendation engine unit 63, and a learning interface response unit 64, the block diagram of which is shown in Figure 6.

[0115] In this embodiment, the speech emotion recognition unit 61 can continuously analyze the learner's speech rate, tone, and volume changes to determine their focus and emotional stability. These speech signals can reflect the learner's emotions and mental state; for example, speaking too fast may indicate anxiety, while a monotonous tone or low volume may indicate fatigue or lack of motivation. Through continuous speech analysis, this device can instantly determine the learner's level of focus and emotional stability, thus providing a basis for subsequent teaching adjustments.

[0116] The facial expression and body sensing unit 62 can integrate video camera images to identify the learner's confused, fatigued, or unresponsive behavioral manifestations. This teaching device captures and identifies common signs of learning fatigue or cognitive impairment, such as confusion, fatigue, and unresponsiveness. This information further supplements the results of speech analysis, making the overall emotion assessment more accurate and comprehensive. When a learner exhibits continuous inefficient behavior, this device can be considered a signal of a learning bottleneck and activate subsequent remedial mechanisms.

[0117] The supplementary lesson plan recommendation engine unit 63 compares the behavioral judgment results with the lesson plan database, selects supplementary modules, and pushes them to the user's device. When the aforementioned facial expression and body sensing unit 62 detects that the learner has comprehension difficulties or low attention, this supplementary lesson plan recommendation engine unit 63 will automatically select suitable supplementary teaching modules from the lesson plan database based on the analysis results and push them to the learner's device in real time. These supplementary modules may include extended explanations, interactive exercises, or key point reviews, which help to address learning problems in a targeted manner.

[0118] The learning interface response unit 64 can dynamically adjust the teaching interface colors, speech rate, interaction density, or prompts to improve the learner's attention span and comprehension. Adjustments may include softening interface colors, changing the speed of speech delivery, adjusting interaction frequency, and even providing timely prompts and reminders. This design aims to enhance learners' psychological comfort and cognitive acceptance, allowing them to maintain attention for longer periods and effectively absorb the teaching content.

[0119] In summary, this invention combines voice, video and other sensing technologies with artificial intelligence analysis, emphasizing not only a personalized learning experience but also shifting education from "passive provision" to "active adaptation." It has high application potential in scenarios such as improving distance learning, self-directed learning, and special education support.

[0120] Please refer to Figure 7. In this invention, the virtual teacher control module 30 may further include a multi-modal interactive response unit 33, which includes a voice command parsing sub-module 331, a gesture sensing sub-module 332, a teaching command response sub-module 333, and a virtual teacher synchronous simulation engine sub-module 334. The technical features of these sub-modules are described below.

[0121] The voice command parsing submodule 331 can receive and parse the voice input content issued by the learner. The learner can interact with the system through natural language, such as saying "Please say it again," "Switch to English," or "Play the next segment." The device will recognize the voice content in real time and further perform semantic understanding. This function can effectively reduce the burden on the learner's operating system and enhance the intuitiveness and user-friendliness of the teaching device.

[0122] The gesture sensing submodule 332 can be used to receive specific actions detected by the XR device, such as raising a hand, shaking a head, or tapping a command. For example, raising a hand represents asking a question, shaking a head represents disagreement, giving a thumbs-up indicates understanding, or directly tapping a virtual object to trigger a function. This non-verbal input mode allows learners to interact with content more naturally in an immersive environment, and is especially useful for learners who have difficulty using voice or are in noisy environments.

[0123] The teaching instruction response submodule 333 can determine the operation content based on voice or gesture input, such as playback control, replaying the previous segment, providing examples, and switching speech speed or language mode. For example, if a learner says, "I don't understand, please give me an example," the device will understand it as a request for supplementary examples; if the learner waves their hand to indicate stop, the playback will be paused. In addition, the device can also adjust the speech speed, switch language modes, and even automatically provide additional explanations or key points according to personalized needs, making the teaching process more in line with the learner's real-time needs.

[0124] The virtual teacher synchronous simulation engine sub-module 334 can immediately trigger corresponding voice, facial expression, and visual responses from the virtual character based on the judgment results. These behaviors include voice responses, facial expression changes (such as smiling and frowning), eye movements, and body movements, thereby creating a realistic interactive scenario. This real-time situational simulation can significantly enhance learners' sense of participation and learning immersion, and is particularly helpful for applications such as language learning, situational training, and simulated practice.

[0125] In conclusion, this interactive module design fully embodies the future trend of human-computer interaction. Through natural input methods such as voice and gestures, coupled with real-time feedback from virtual characters, it establishes a highly responsive and human-centered smart teaching environment. This not only enhances the efficiency and engagement of teaching but also encourages learners to actively participate, contributing to the achievement of individualized, contextualized, and immersive modern digital learning goals. Future integration of semantic understanding and learning process data analysis could potentially create a truly capable AI teacher assistant with comprehension and instructional decision-making abilities.

[0126] Please refer to Figure 8. In this embodiment, the AI ​​learning analysis module 50 can perform cross-lesson plan learning performance modeling. It may include a learning feature extraction unit 55, a feature vector normalization and comparison unit 56, a machine learning algorithm unit 57, and a lesson plan adjustment prompt interface unit 58.

[0127] The learning feature extraction unit 55 can capture learners' behavioral data across different course modules. This behavioral data may include learning time, operation sequence, selection frequency, number of replays, error rate, etc. These micro-learning behaviors can reveal learners' learning strategies and cognitive states. For example, frequent replays may indicate difficulty in understanding, while quickly skipping may show a lack of patience or prior familiarity with the content. By fully recording and extracting these features, a systematic foundation of data is provided for subsequent model building.

[0128] The feature vector normalization and comparison unit 56 can be used to compare and measure the similarity of operational patterns between modules. After normalization, the device can measure and compare the similarity of learning behaviors between different modules or different students, further identifying which modules may cause common learning difficulties, or which types of students have similar learning styles and potential risks. This process helps to identify abnormal learning patterns and high-risk learners.

[0129] The machine learning algorithm unit 57 can be used to construct individual learning behavior models and predict the learning effectiveness and risks of subsequent modules. It trains individualized learning behavior models based on the standardized feature vector data mentioned above. Through continuous learning and updating, this device can predict students' performance in future modules, including learning effectiveness, completion probability, and comprehension risk. When the model predicts a student has a high probability of failure or error in an upcoming module, the device can proactively prompt the teacher to intervene, such as providing reinforcement content or changing teaching strategies.

[0130] The lesson plan adjustment prompt interface unit 58 can convert model output into interface suggestions for teachers to review and then use for reinforcement or skipping topics. These suggestions may include recommendations for supplementary teaching materials for specific learners, timely skipping of already mastered content, and the arrangement of review modules or group teaching strategies. Unlike traditional adjustment methods based on teachers' subjective judgment, this interface transforms AI analysis into specific, real-time, and objective lesson plan optimization suggestions, improving the scientific nature and efficiency of teaching decisions.

[0131] In summary, this embodiment provides teachers with a data-driven and artificial intelligence-assisted teaching tool through behavioral data extraction, data modeling, and real-time interface suggestions. It not only enhances the visualization and predictability of students' learning processes but also promotes precise and individualized instruction, making it highly valuable for improving learning outcomes and teacher productivity. Future integration with modules such as sentiment analysis and semantic understanding could potentially develop into a more comprehensive intelligent teaching decision-making system.

[0132] Please refer to Figure 9. In one embodiment, this virtual classroom teaching device 100 can support four input channels: voice, facial expression, gesture and eye movement, and includes a multimodal input synchronization management module 81, an input conflict detection module 82, a multimodal input record archiving mechanism module 83 and a teacher-side retrospective review tool module 84, which are described below.

[0133] The multimodal input synchronization management module 81 is responsible for integrating sensor data from learner U, including voice input (such as answering questions and making requests), facial expressions (such as smiling and frowning), gestures (such as raising a hand and pointing), and eye movement patterns (such as attention points and gaze duration). This module performs real-time synchronization management, integrating signals from different channels into a unified sequence of input events, enabling the device to accurately understand the learner's current intentions and responses. This synchronization processing greatly helps avoid signal delays or misunderstandings, and is especially important in immersive teaching or highly interactive scenarios.

[0134] The input conflict detection module 82 is used to determine the priority and consistency of inputs from multiple sources. For example, if a student nods (indicating agreement) but verbally replies "I don't understand," the system needs to be able to prioritize and select the more meaningful input for processing. This module is responsible for determining the consistency between input signals and prioritizing them according to preset logic (such as chronological order, reliability of the input source, etc.) to ensure the accuracy and coherence of teaching responses and avoid misinterpretations that could lead to confusion in the teaching process.

[0135] The multimodal input recording and archiving module 83 can be used to track learning behaviors and strategy preferences. This module can track students' behavioral preferences and strategies in different learning activities over a long period of time. For example, a student may habitually use eye movements to quickly browse key areas, or frequently interact with gestures while using less voice. Through the analysis of these behavioral patterns, teachers and this device can gradually build personalized learning models for learners, which is helpful for subsequent allocation of teaching resources and identification of learning styles.

[0136] The teacher-side retrospective review tool module 84 helps teachers reconstruct learning contexts and emotional trajectories from unstructured input. Teachers can use this module to trace learners' unstructured input trajectories, such as a student's confused facial expressions, frequent shifting eye movements, or hesitant speech during responses. These emotional and behavioral trajectories can be visualized, making it easier for teachers to understand students' learning bottlenecks and emotional reactions, thereby adjusting teaching strategies and content to achieve individualized instructional intervention.

[0137] In summary, the virtual classroom teaching device of this embodiment combines four input methods: voice, facial expressions, gestures, and eye tracking. Through a multimodal input management and analysis module, it creates a smart teaching platform that can understand in real time, respond accurately, and track the learning process over the long term. It not only enhances the naturalness and efficiency of human-computer interaction but also provides teachers with more in-depth learning diagnostic tools, demonstrating high application potential for promoting immersive teaching, distance education, and smart teaching decisions.

[0138] Please refer to Figure 10. In this embodiment, the virtual teacher control module 30 can support automatic switching between voice system and subtitles. It may further include a language recognition engine unit 34, a voice simulation switching unit 35, a subtitle synchronization processing unit 36, and a translation accuracy correction unit 37, which are described below.

[0139] The language recognition engine unit 34 can be used to detect the user's preset language and voice output feedback. When a user enters the teaching platform for the first time or opens a course module, this unit will automatically detect the language used (possibly through account settings, voice samples, or regional information) to ensure that subsequent interactions and content presentation conform to the user's language habits. In addition, this language recognition engine unit 34 also supports the recognition and processing of real-time voice input, so that teaching interactions are not limited by language barriers.

[0140] The voice simulation switching unit 35 can switch the virtual teacher's accent and speaking speed model according to the language family of the lesson plan. Based on the language family of the selected lesson plan (e.g., English, Chinese, Spanish, etc.), this module can switch the virtual teacher's voice model, including different accents (e.g., British, American, Indian English, etc.) and speaking speed adjustments, allowing learners to learn in an environment that matches their language preferences. This is particularly helpful for language learning and cross-cultural teaching situations. It also makes the virtual teacher more realistic and approachable.

[0141] The subtitle synchronization processing unit 36 ​​supports bilingual subtitle switching and automatic time synchronization. This unit not only supports real-time subtitle generation and language switching, but also automatically synchronizes subtitles with the teacher's audio or video playback progress, ensuring subtitles appear at the correct time, which helps learners understand both audio and text content simultaneously. The bilingual subtitle design is particularly suitable for language learners, assisting them in comparing vocabulary, grammar, and sentence structure, further enhancing language absorption.

[0142] The translation accuracy correction unit 37 can adjust the subtitle content in real time based on the context to ensure semantic integrity. Its core function is to perform real-time correction based on the context, avoiding grammatical inconsistencies or semantic deviations caused by literal translation. For example, for polysemous words or culturally relevant vocabulary, this unit can adjust the meaning according to the dialogue and teaching content, ensuring that the final subtitles match the actual meaning. This not only improves translation accuracy but also significantly enhances the clarity of teaching communication.

[0143] In summary, this language processing module, through technologies such as speech recognition, virtual teacher voice simulation, synchronized subtitles, and semantic translation correction, creates a flexible, accurate, and user-friendly multilingual teaching support system. Its functions can be widely applied to language learning courses, cross-border distance learning, and virtual classrooms with multilingual participants, demonstrating high practicality and scalability potential, and making a significant contribution to improving learning efficiency and the quality of internationalized teaching.

[0144] Please refer to Figure 11. In this embodiment, the virtual classroom teaching device 100 may further include a diagnostic question recommendation process module 70, which includes a dynamic question bank management unit 71, a difficulty and knowledge point mapping calculation unit 72, an AI-recommended diagnostic question unit 73, and a question result analysis unit 74. Through the collaborative operation of these units, a smart teaching process with real-time response capability and personalized teaching support is formed. The technical features of these units are described below.

[0145] The dynamic question bank management unit 71 can flexibly schedule and update the question bank content based on learners' learning progress, knowledge structure, and past performance. This design allows the device to adjust the question type, difficulty, order, and frequency according to the needs of different learners, achieving truly individualized instruction. Each question in the question bank is labeled with the corresponding knowledge point, difficulty level, and related modules to ensure accurate matching in subsequent calculations.

[0146] The difficulty and knowledge point mapping calculation unit 72 can be used to compare the learner's current module comprehension errors. This calculation unit dynamically compares the learner's previous interactive behaviors (such as learning time, repeated playback, incorrect answers, etc.) with their performance in previous modules, combined with the knowledge points and difficulty indicators of the questions. For example, if a student repeatedly makes mistakes in the "Fraction Arithmetic" module, this unit will review their understanding in the "Least Common Multiple" or "Common Denominator" modules to confirm whether there are any fundamental knowledge gaps. This calculation logic helps to build a knowledge structure map for students and identify learning bottlenecks.

[0147] The AI-recommended diagnostic question unit 73 generates targeted diagnostic questions and integrates them instantly into the lesson plan process. When the device identifies potential comprehension errors, the AI-recommended diagnostic question unit 73 automatically selects the most suitable diagnostic questions from a dynamic question bank based on the calculation results. These questions are not only targeted but also instantly integrated into the current lesson plan process, maximizing learning effectiveness with minimal teaching interference. For example, when a student is learning a new module, if it detects that they haven't fully grasped a basic concept, the device can immediately insert a relevant diagnostic question to test their understanding of that concept. This design not only improves learning efficiency but also prevents students from continuing to progress on incorrect foundations, thus avoiding a greater cognitive burden. Furthermore, the recommended questions consider diverse question types and difficulty levels, and can use charts, animations, multiple-choice questions, fill-in-the-blank questions, or open-ended questions based on different subject characteristics to more comprehensively assess students' depth of understanding and application ability.

[0148] The answer result analysis unit 74 can calculate accuracy, reaction time, and error distribution to recommend corresponding reinforcement modules and content extensions. This unit can perform multi-dimensional analysis of students' answer performance. In addition to traditional accuracy calculation, this unit also analyzes response time, error distribution, and recurring error patterns to assess whether students are simply careless, lack concepts, or have deep misunderstandings about certain question types. For example, if a student answers the same type of question incorrectly multiple times but reacts very quickly, it may indicate a flawed mindset; conversely, if the reaction time is too long but the answer is correct, it may be a tentative answer, showing an unstable concept. After the analysis, the device automatically recommends corresponding reinforcement modules based on the results, such as providing further explanatory videos, interactive exercises, or review units, and can even guide students back to specific lesson plan modules for further learning. In addition, for students with good learning outcomes, the device can also recommend extended learning content, challenging them with higher-level knowledge or cross-module integrated questions to promote higher-level cognitive development.

[0149] Overall, this implementation can dynamically adjust the teaching process and content based on learners' real-time performance. It emphasizes a deep understanding of each student's learning status through knowledge point comparison and response behavior tracking, embedding diagnostic questions and reinforcement mechanisms in real time during the learning process to avoid the accumulation of errors. It not only addresses weaknesses but also provides advanced learning paths for students with good learning progress. Teachers can fine-tune lesson plans based on system suggestions to improve overall teaching effectiveness.

[0150] Please refer to Figure 12. In this embodiment, the cloud-based resume database 202 may include a behavior trajectory recording module 2021, a multi-modal input process processing engine module 2022, an achievement marking and learning milestone module 2023, and a cloud-based resume query and backtracking interface module 2024, which are described below:

[0151] The 2021 Behavioral Tracking Module continuously tracks and records learners' various interactive behaviors within the learning device, including operation clicks, entry and exit times from the learning module, number of repeated learning sessions, response process, and resource usage frequency. This detailed behavioral data can not only be used to assess students' learning engagement but also helps in subsequent analysis of their learning habits, learning bottlenecks, and concept mastery.

[0152] The Multimodal Input Process Processing Engine Module 2022 is responsible for processing and integrating these heterogeneous and unstructured input data, converting them into a standard format that can be analyzed by the system. This allows the device to accurately present the student's complete interaction process and assists teachers or AI systems in making more accurate diagnoses and suggestions.

[0153] The Achievement Tagging and Learning Milestones module 2023 automatically tags students' achievements and key milestones (such as completing units, passing diagnostic tests, and mastering specific knowledge points) based on their progress and performance. These learning milestones are not only visually presented but also serve as important summaries of the learning process, helping students review their growth trajectory and facilitating teachers to quickly assess individual students' learning curves and key outcomes.

[0154] The Cloud-based Resume Query and Retrospective Interface Module 2024 allows teachers to quickly retrieve the learning process, performance changes, and behavioral records of individual or entire classes, enabling them to design precise teaching and reinforcement strategies. Students, on the other hand, can independently view their own learning records, achievements, and areas for improvement, further enhancing their self-reflection and learning autonomy.

[0155] In summary, this cloud-based resume database 202 integrates learning behavior records, input organization, achievement tagging, and visual querying. Through the integration of behavioral data and multimodal input data, it can comprehensively depict students' learning patterns and trends, providing a solid basis for precise teaching and individualized guidance.

[0156] Please refer to Figure 13. In this embodiment of the invention, the learning process record file can be automatically synchronized to a school learning management system (LMS) platform through the virtual classroom teaching device 100. In this embodiment, the virtual classroom teaching device 100 may further include a data interface transcoding module 85, a student data mapping module 86, a grade backfilling and process upload synchronization module 87, and an export report tool and assessment conversion rule engine module 88. These are explained below.

[0157] The data interface transcoding module 85 supports various formats, including common ones such as LTI (Learning Tools Interoperability), SCORM (Sharable Content Object Reference Model), and xAPI (Experience API). Transcoding using these standard protocols ensures good compatibility for data exchange between different schools or platforms, further enabling the integration of cross-platform teaching tools and data. For example, when a learner completes a unit module, their operation records and achievement information are instantly converted into SCORM-encapsulated format and sent back to the school's LMS.

[0158] The student data mapping module 86 can be used to map student information to school rosters. To address the inconsistency between learner IDs within the virtual classroom and data in the school's student registration system, this module is responsible for mapping student identification information on the platform to the school roster on a one-to-one basis. This ensures that after data synchronization, grades and progress are correctly mapped to each student, avoiding mismatches or data loss. Its design also considers multi-school systems, class management, and student movement tracking, improving flexibility and accuracy.

[0159] The Grade Regression and Progress Tracking Module 87 automatically synchronizes learners' performance in the virtual classroom (such as module completion, answer results, and interactive behavior analysis) to the LMS system and regresses it into the school's transcripts or progress logs. Teachers no longer need to manually enter various assessment grades, significantly reducing administrative burden and human error. The synchronization function also supports batch updates and real-time synchronization, allowing teachers and learners to monitor learning progress in real time.

[0160] The Export Reporting Tool and Assessment Transformation Rule Engine Module 88 can provide teachers or system administrators with tools to export various learning reports, including student performance overviews, behavioral analysis summaries, milestone achievement statistics, etc.

[0161] Overall, this virtual classroom teaching device 100 demonstrates a high degree of integration and standards compatibility. Lesson plans can be integrated into the school's LMS system in the cloud, and teaching records can be directly updated to student assessments. The device provides API integration modules with mainstream learning management systems (such as Moodle, Canvas, and Google Classroom), automatically connecting student interaction records, assessment scores, self-assessment feedback, and progress reports on this platform to the school's LMS grading system, achieving data interoperability between the teaching platform and the learning device. This function saves teachers' manual work time and enhances the integration of the teaching system, substantially demonstrating the advantages of system integration innovation and data interoperability between educational platforms.

[0162] Furthermore, this virtual classroom teaching device 100 can be integrated with educational cloud accounts to provide a personalized startup environment. Ideally, this device is compatible with the OpenID Connect protocol, enabling identity verification with educational cloud services provided by schools at all levels (such as Taiwan Education Cloud and Google Workspace for Education). After logging in, students can automatically load their historical lesson plans, learning records, preference settings, and customized avatars, forming a personalized learning environment. This startup mechanism not only improves ease of use and data continuity but also serves as a unified identity verification center across learning platforms, effectively reducing management costs and enabling innovative applications such as one-click access to multiple scenarios.

[0163] Please refer to Figure 14. In this embodiment, teachers can use the virtual classroom teaching device 100 to make horizontal comparisons of the learning processes of different learners in different modules. The virtual classroom teaching device 100 may further include a learning process visualization parallel coordinate module 89, a differentiated heat map presentation tool module 90, and a group behavior analysis module 91, which are described below.

[0164] The learning process visualization parallel coordinate module 89 can visually present multiple behavioral indicators of learners in various teaching modules (such as completion rate, answer accuracy, number of retries, operation time, etc.). This visualization technology can effectively reveal the changing trends and comparative relationships between multi-dimensional data, enabling teachers to quickly identify students with high or low learning performance. For example, a student may perform well in the reading module but frequently make mistakes in the interactive operation module. This chart can clearly present such cross-module differences, helping to diagnose their learning preferences or difficulties.

[0165] The Differentiated Heatmap Radar Chart Presentation Tool Module 90 utilizes radar chart technology to integrate learners' performance across multiple core learning dimensions, such as focus, interaction frequency, error distribution, and emotional responses, highlighting strengths or weaknesses as heatmaps. This visualization tool is particularly suitable for comparing performance differences between different students, classes, or groups. Teachers can use this module to quickly identify learning hotspots or colds requiring special attention, enabling focused instruction or the design of remedial activities.

[0166] The grouping behavior analysis module 91 uses data mining techniques such as clustering to classify students' behavioral data and automatically group students with similar learning characteristics. For example, this device may categorize students with characteristics such as "high interaction but low accuracy," "low interaction but steady progress," or "high emotional fluctuation" into different groups. Teachers can then design differentiated teaching strategies based on these grouping results, such as providing individualized tutoring, adjusting the pace of instruction, arranging remedial modules, or conducting group discussions.

[0167] In summary, the three modules described above together constitute a powerful set of tools for visualizing learning behaviors and designing strategies, significantly enhancing teachers' depth of understanding of students' learning processes and the flexibility of their instructional responses. These modules not only improve teaching effectiveness but also facilitate learners' access to learning support that is more tailored to their individual needs.

[0168] In a preferred embodiment, the AI ​​learning analysis module 50 can be used to analyze the correlation between different lesson plan styles (such as gamified, narrative, and task-oriented) and learning outcomes, and to generate a lesson plan style classification database, a learning effectiveness trend model, and an adaptation mapping and teacher feedback loop module.

[0169] In a preferred embodiment, all learning process and behavioral data can be anonymized and uploaded to a central model training platform 204. This central model training platform 204 may include a data stripping module 2041, a hash mapping and anonymized ID refactoring engine module 2042, a hierarchical behavioral feature extraction module and data pool module 2043, and a lesson plan design error prediction algorithm module 2044. These are described below.

[0170] The 2041 data stripping module is used to handle the first step of data de-identification, which involves removing personally identifiable information, such as name, student ID, and IP address, from learners' original learning records to ensure privacy. This step is fundamental to the entire data anonymization process, complies with legal and ethical requirements for user data protection in digital learning systems, and provides a secure data source for subsequent data analysis and model training.

[0171] The Hash Mapping and Anonymous ID Refactoring Engine Module 2042, after stripping personal data, further generates a unique anonymous identifier for each piece of learning data through a hash algorithm. This ensures that the data retains the ability to be tracked and analyzed across modules without revealing the individual's identity. The anonymous identifier supports the reconstruction of correlations between data, enabling the device to observe learners' progress and behavioral evolution from a long-term, continuous perspective.

[0172] The hierarchical behavioral feature extraction module and the data pool module 2043 are responsible for extracting meaningful behavioral features from large-scale anonymized learning data. These features include learning time distribution, error patterns, interaction frequency, and reaction time, and are then hierarchically stored and managed according to classification indicators such as grade, subject, and learning style. This feature data will serve as key material for subsequent machine learning model training and can also be used for teaching decision support and system optimization.

[0173] The Lesson Plan Design Error Prediction Algorithm Module 2044 utilizes the aforementioned machine learning algorithms to evaluate and predict lesson plan structure and content design. When the device detects that learners generally spend too much time in a particular module, have a high error rate, or exhibit abnormal emotional reactions, it can preliminarily identify potential lesson plan design errors, such as overly rapid transitions between teaching steps, unclear explanations, or excessive cognitive load. Based on historical data, the device will propose possible optimization suggestions, such as adjusting the module order, adding guiding examples, or segmenting teaching steps, to improve overall learning effectiveness.

[0174] In summary, this central model training platform 204 balances learner privacy protection with the reuse of data value. Through a series of modules, from data stripping and anonymized ID construction to behavioral feature analysis and error prediction calculation, the platform can establish a dynamic, self-optimizing teaching system mechanism. Teachers and system developers can then precisely adjust lesson plans, while students benefit from a more tailored allocation of learning resources to their needs and abilities, ultimately achieving truly individualized, data-driven smart education.

[0175] Please refer to Figure 16. In this embodiment, the tone, pronunciation, speech rate, and supplementation method of the virtual teacher image 31 in the virtual teacher control module 30 can be adjusted in real time according to the user's experience and feedback data. This virtual teacher control module 30 may further include a tone adaptation algorithm unit 381, a perception friendliness correction parameter setting unit 382, ​​a contextual guidance logic decision unit 383, and an expandable speech feature library unit 384. The following is a description of these components.

[0176] The intonation adaptation algorithm unit 381 can automatically adjust the virtual teacher's tone, speed, and intonation based on learner feedback (such as comprehension difficulties, emotional fluctuations, and learning progress) through speech generation and modulation technology. For example, when the device detects that the learner is in a low-engagement state, the intonation algorithm can raise the pitch and speed to boost attention; if comprehension difficulties are detected, it adjusts to a slower and more emphatic tone to aid understanding. Such adjustments allow the virtual teacher to present a more human and flexible teaching interaction.

[0177] The perceived friendliness correction parameter setting unit 382 is responsible for setting the "friendliness" parameters of the virtual teacher's voice output based on the learner's characteristics (such as age, language ability, personality traits, etc.), such as tone of voice and level of emotional expression. This individualized setting helps build learners' positive feelings and trust in the virtual teacher, thereby enhancing learning motivation and willingness to interact. Teachers can also adjust these parameters according to the student group to create a more inclusive learning environment.

[0178] The Contextualized Decision-Making Unit 383 is responsible for determining the current context of a learning activity (e.g., teaching, prompting, feedback, encouragement, error correction), and selecting the most suitable language style and communication method based on decision-making logic. For example, when a student answers incorrectly, this unit can choose to use an encouraging tone to guide them, rather than directly pointing out the error, to reduce frustration and promote a positive learning cycle. This unit enables virtual teachers to possess contextual understanding and responsiveness similar to human teachers.

[0179] The expandable speech feature library unit 384 can serve as a database of virtual teacher speech styles and corpus samples, containing various speech feature modules that support different language families, accents, speech rates, and intonation patterns. Furthermore, this database is expandable, meaning that new speech models, dialect corpora, or individual teacher styles can be continuously added in the future, enhancing the diversity and customization capabilities of virtual teachers.

[0180] In summary, the Virtual Teacher Control Module 30, through the aforementioned units, achieves a high degree of humanization and contextual awareness in voice interaction, enabling virtual teachers to adjust their teaching style in real-time, naturally, and effectively according to the needs and learning processes of different students. This not only enhances the immersion and interactivity of digital learning systems but also extends individualized instruction beyond content to include language and emotional aspects of teaching, playing a crucial role in promoting AI-assisted teaching and smart classrooms.

[0181] As can be seen from the above, the virtual classroom teaching device and method proposed in this invention possess multiple technical advantages, effectively enhancing the intelligence, automation, and individualization of digital teaching. Firstly, through the textbook format recognition module, various digital textbook formats can be automatically identified and converted into modular lesson plan units with detachable characteristics. This design not only improves the efficiency of textbook integration and reuse but also facilitates the standardization of textbooks from different sources, thereby supporting subsequent dynamic teaching configurations. Secondly, the modular reconfiguration engine module can automatically adjust the teaching sequence and content of lesson plan units based on learner behavior data and learning outcomes, providing reinforcement and recommendations. This module has a learning process tracking and intelligent recommendation mechanism, providing optimized teaching paths based on individual differences, effectively strengthening learning outcomes and motivation, and demonstrating the technological advantages of highly personalized teaching. Furthermore, the virtual teacher control module integrates speech recognition, semantic understanding, and non-verbal command recognition technologies, enabling interaction between the virtual teacher image and learners. This not only enhances the sense of presence in distance learning but also overcomes the limitation of real teachers not being able to respond in real time, making teaching responses more immediate and interactions more natural. Finally, the feedback detection and behavior analysis module can collect and analyze learners' learning status data in real time, serving as the basis for dynamic push notifications and content recommendations, thus enabling real-time adjustments to teaching strategies. In summary, this invention significantly enhances the flexibility of textbook processing, the intelligence of teaching interaction, and the individualized support for the learning process in virtual teaching environments, demonstrating significant technical benefits for digital learning systems.

[0182] The embodiments described above are merely for illustrating the technical ideas and features of the present invention. Their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be used to limit the patent scope of the present invention. That is, all equivalent changes or modifications made in accordance with the spirit disclosed in the present invention should still be covered within the patent scope of the present invention.

[0183] 100: Virtual Classroom Teaching Device

[0184] 10: Textbook Format Recognition Module

[0185] 11: Modular Lesson Plan Unit

[0186] 20: Modular Recombination Engine Module

[0187] 21: Version Control Unit

[0188] 30: Virtual Teacher Control Module

[0189] 31: Virtual Teacher Image

[0190] 32: Virtual Character Database

[0191] 33: Multimodal Interactive Response Unit

[0192] 331: Voice Command Parsing Submodule

[0193] 332: Gesture Sensing Sub-module

[0194] 333: Teaching Instruction Response Submodule

[0195] 334: Virtual Teacher Synchronous Simulation Engine Sub-module

[0196] 34: Speech Recognition Engine Unit

[0197] 35: Voice Analog Switching Unit

[0198] 36: Subtitle Synchronization Processing Unit

[0199] 37: Translation Accuracy Correction Unit

[0200] 381: Intonation Adaptation Algorithm Unit

[0201] 382: Perception Friendliness Correction Parameter Setting Unit

[0202] 383: Context-Guided Logical Decision-Making Unit

[0203] 384: Expandable Speech Feature Library Unit

[0204] 40: Feedback Detection and Behavior Analysis Module

[0205] 41: Real-time learning status data

[0206] 42: Voice Emotion Analysis Unit

[0207] 43: Eye-tracking unit

[0208] 50: AI Learning and Analysis Module

[0209] 51: Behavioral Feature Extraction Unit

[0210] 52: Multidimensional classification and clustering model unit

[0211] 53: Module Adaptability Calculation Unit

[0212] 54: Recommendation Ranking Unit

[0213] 55: Learning Feature Extraction Unit

[0214] 56: Eigenvector Normalization and Comparison Unit

[0215] 57: Machine Learning Algorithm Unit

[0216] 58: Lesson Plan Adjustment Prompt Interface Unit

[0217] 202: Cloud-based resume database

[0218] 2021: Behavior Trajectory Recording Module

[0219] 2022: Multi-modal input process processing engine module

[0220] 2023: Achievement Markers and Learning Milestones Module

[0221] 2024: Cloud-based resume query and retrospection interface module

[0222] 203: Visual Dashboard

[0223] 2031: Module Progress Tracking Unit

[0224] 2032: Error and Dwell Time Hotspot Analysis Chart

[0225] 2033: Emotion and Focus Trend Chart

[0226] 2034: Teacher Real-Time Interaction Toolbar

[0227] 204: Central Model Training Platform

[0228] 2041: Personal Data Stripping Module

[0229] 2042: Hash Mapping and Anonymous ID Recompilation Engine Module

[0230] 2043: Hierarchical Behavioral Feature Extraction Module and Data Pool Module

[0231] 2044: Lesson Plan Design Error Prediction Algorithm Module

[0232] 60: Emotional Feedback Enhancement Module

[0233] 61: Voice Emotion Recognition Unit

[0234] 62: Facial Expression and Body Sensing Unit

[0235] 63: Enhance the Lesson Plan Recommendation Engine Unit

[0236] 64: Learning Interface Response Unit

[0237] 70: Diagnostic Question Recommendation Module

[0238] 71: Dynamic Question Bank Management Unit

[0239] 72: Difficulty and Knowledge Point Correspondence Calculation Unit

[0240] 73: AI Recommendation Diagnostic Questions Unit

[0241] 74: Answer Result Analysis Unit

[0242] 81: Multimodal Input Synchronization Management Module

[0243] 82: Input Conflict Detection Module

[0244] 83: Multi-mode input record archiving mechanism module

[0245] 84: Teacher-side Retrospective Review Tool Module

[0246] 85: Data Interconnection and Transcoding Module

[0247] 86: Student Data Mapping Module

[0248] 87: Synchronous Module for Grade Backfilling and Process Upload

[0249] 88: Export Report Tool and Assessment Transformation Rule Engine Module

[0250] 89: Learning Process Visualization Parallel Coordinate Module

[0251] 90: Differentiated Hot Zone Radar Map Presentation Tool Module

[0252] 91: Cluster Behavior Analysis Module

[0253] (a)~(g),(e1)~(e4): Process steps

Claims

1. A virtual classroom teaching device suitable for a learner's learning environment, comprising: a textbook format recognition module for inputting a plurality of digital textbooks and automatically recognizing and converting the formats of these digital textbooks to generate a plurality of modular lesson plan units with the same file format and decomposable; a modular reorganization engine module for adjusting, supplementing, and optimizing the order of the plurality of modular lesson plan units based on the learner's learning behavior and learning outcomes; a virtual teacher control module for driving a virtual teacher image to interact with the learner and provide teaching responses, wherein the virtual teacher image has the functions of speech recognition, semantic understanding, and non-verbal instruction recognition; and a feedback detection and behavior analysis module for collecting real-time learning status data of the learner and using the real-time learning status data as a basis for push recommendations.

2. The virtual classroom teaching device as described in claim 1, wherein the modular reorganization engine module includes a version control unit, which records the process of lesson plan reorganization and has version annotation, comparison and restoration functions, for teachers to switch interfaces and track.

3. The virtual classroom teaching device as described in claim 1, wherein the feedback detection and behavior analysis module includes a voice emotion analysis unit and an eye movement tracking unit to instantly identify the learner's emotions and concentration state, and adjust or enhance the content of the modular teaching plan units accordingly.

4. The virtual classroom teaching device as described in claim 1, wherein the virtual teacher control module includes a virtual character database, which automatically switches the voice, appearance and interaction style of the virtual teacher image according to the course topic and the learner's language.

5. The virtual classroom teaching device as described in claim 1 further includes an AI learning analysis module, which includes: a behavioral feature extraction unit for analyzing the learner's answer path, operation clicks, visual pauses and interactive responses; and a multi-dimensional classification and grouping model unit for automatically identifying the learner's type. A module fit calculation unit is used to predict the learner's expected performance under different lesson plan modules; It also includes a recommendation and ranking unit that automatically calculates reinforcement and reorganization strategies based on real-time detected behavioral data to generate a personalized lesson plan recommendation module list and confidence matrix based on process data.

6. A virtual classroom teaching method applied to a virtual classroom teaching device, the virtual classroom teaching device comprising a textbook format recognition module, a modular reconstruction engine module, a virtual teacher control module, and a feedback detection and behavior analysis module, the virtual classroom teaching method comprising: (a) a textbook uploading and modular conversion step: uploading a digital textbook to the textbook format recognition module to parse its type and converting it into a detachable modular lesson plan unit; (b) a lesson plan distribution and initialization step: automatically loading the adapted modular lesson plan unit according to a learner's terminal device and initializing a learning task flow; (c) a virtual teaching interaction step: the virtual teacher control module provides teaching guidance, supporting voice input, facial feedback, gestures, or click interactions, and switching teaching styles according to the context of the modular lesson plan unit; (d) (e) Behavioral Feedback Acquisition Step: The feedback detection and behavior analysis module acquires the learner's operation data, speech recognition results, facial expressions, and eye-tracking information to form a learning process record file; (f) AI Analysis and Recommendation Step: The learning process record file is input into an AI learning analysis module to perform behavioral model deduction, calculate a module fit and a teaching confidence score, and generate a recommendation list and restructuring suggestions; (g) Teacher Review and Fine-tuning Step: The restructuring results and system suggestions are presented in a lesson plan editing interface on the teacher's end, and fine-tuning, module replacement, or process sequence modification is performed using the lesson plan editing interface; and (g) Push and Process Archiving Step: The confirmed modular lesson plan unit is pushed to the learner's device to perform interactive teaching, and the complete learning process record file is automatically saved in a cloud resume database after class for future diagnosis and teaching optimization reference.

7. The virtual classroom teaching method as described in claim 6 further includes: detecting the prerequisite dependencies and ability progress labels between each modular lesson plan unit; automatically comparing whether the module order violates the teaching cognitive progression logic; detecting whether there is repeated teaching or missing modules; and displaying the violation or duplicate module label on the lesson plan editing interface and providing adjustment suggestions.

8. The virtual classroom teaching method as described in claim 6, wherein the learning process record file is presented in real time by a visual dashboard, the visual dashboard comprising: a module progress tracking unit for displaying the learner's completion percentage and operation record in each lesson plan module; an error and dwell hotspot analysis graph for generating key warning nodes based on operation time and repeated error locations; an emotion and focus trend curve graph, a visual curve constructed by combining voice tone, facial expression and eye movement trajectory analysis; and a real-time teacher interaction toolbar for providing teachers with the ability to mark, group and distribute reinforcement modules for abnormal nodes.

9. The virtual classroom teaching method as described in claim 6, wherein in the asynchronous learning process, the virtual classroom teaching device further includes an emotional feedback reinforcement module, comprising: a voice emotion recognition unit for continuously analyzing the learner's speech rate, tone, and volume changes to determine their focus and emotional stability; an expression and body sensing unit for integrating video camera images to identify the learner's confused, tired, or unresponsive behavioral representations; a reinforcement lesson plan recommendation engine unit for comparing the behavioral judgment results with a lesson plan database, selecting reinforcement modules, and pushing them to the user; and a learning interface response unit for dynamically adjusting the teaching interface colors, speech rate, interaction density, or reminder prompts to improve the learner's endurance and comprehension rate.

10. The virtual classroom teaching method as described in claim 6, wherein the virtual teacher control module further includes a multimodal interactive response unit, comprising: a voice command parsing submodule for receiving and parsing the voice input content issued by the learner; a gesture sensing submodule for receiving specific actions detected by the XR device, such as raising a hand, shaking the head, or selecting an instruction; a teaching instruction response submodule for determining the operation content based on the voice or gesture input, such as playback control, replaying the previous segment, providing examples, switching speech speed or language mode; and a virtual teacher synchronous simulation engine submodule for immediately triggering the corresponding voice, facial expression, and visual response actions of the virtual character based on the determination result.

11. The virtual classroom teaching method as described in claim 6, wherein the AI ​​learning analysis module performs cross-lesson learning performance modeling, and further includes: a learning feature extraction unit for extracting the learner's behavioral data across different course modules; a feature vector normalization and comparison unit for comparing and measuring the similarity of operation patterns between modules; a machine learning algorithm unit for constructing individual learning behavior models and predicting their subsequent module learning effectiveness and risks; and a lesson plan adjustment prompt interface unit for converting model output into interface suggestions for teachers to review and then use for reinforcement or skipping questions.

12. The virtual classroom teaching method as described in claim 6, wherein the virtual classroom teaching device supports four input channels: voice, facial expression, gesture, and eye tracking, and includes: a multimodal input synchronization management module; an input conflict detection module for determining the priority and consistency among multiple input sources; a multimodal input recording and archiving mechanism module for tracking learning behavior and strategy preferences; and a teacher-side retrospective review tool module to assist teachers in reconstructing learning situations and emotional trajectories using unstructured input.

13. The virtual classroom teaching method as described in claim 6, wherein the virtual teacher control module supports automatic switching of voice system and subtitles, and further includes: a language recognition engine unit for detecting the user's preset language and voice output feedback; a voice simulation switching unit that can switch the virtual teacher's accent and speech rate model according to the teaching plan language system; a subtitle synchronization processing unit that supports bilingual subtitle switching and automatic time synchronization; and a translation accuracy correction unit that adjusts the real-time subtitle content in combination with the context to ensure semantic integrity.

14. The virtual classroom teaching method as described in claim 6, wherein the virtual classroom teaching device further includes a diagnostic question recommendation process module, which includes: a dynamic question bank management unit; a difficulty and knowledge point mapping calculation unit for comparing the learner's current module comprehension errors; an AI-recommended diagnostic question unit for generating targeted diagnostic questions and embedding them into the lesson plan process in real time; and a question result analysis unit for calculating accuracy, reaction time, and error distribution to recommend corresponding reinforcement modules and content extensions.

15. The virtual classroom teaching method as described in claim 6, wherein the cloud-based resume database includes: a behavior trajectory recording module; a multi-modal input process processing engine module; an achievement marking and learning milestone module; and a cloud-based resume query and retrospective interface module, supporting multi-level analysis and process review functions for teachers and students.

16. The virtual classroom teaching method as described in claim 6, wherein the learning process record file is automatically synchronized to a school learning management system (LMS) platform through the virtual classroom teaching device, the virtual classroom teaching device further comprising: a data interface transcoding module supporting LTI, SCORM or xAPI format conversion; a student data mapping module corresponding to the school roster and student identification; a grade backfilling and process upload synchronization module; and an export report tool and assessment conversion rule engine module.

17. The virtual classroom teaching method as described in claim 6, wherein the teacher can use the virtual classroom teaching device to make a horizontal comparison of the module learning process of different students, and the virtual classroom teaching device further includes: a learning process visualization parallel coordinate module; a differentiated heat map presentation tool module; and a group behavior analysis module to support the teacher in designing different teaching diversion strategies based on behavior groups.

18. The virtual classroom teaching method as described in claim 6, wherein the AI ​​learning analysis module analyzes the correlation between different lesson plan styles (such as gamified, narrative, and task-oriented) and learning outcomes, and uses it to produce a lesson plan style classification database, a learning effectiveness trend model, and an adaptive correspondence graph and teacher feedback loop module.

19. The virtual classroom teaching method as described in claim 6, wherein all learning process and behavioral data are uploaded to a central model training platform after being anonymized, the central model training platform comprising: a data stripping module; a hash mapping and anonymized ID refactoring engine module; a hierarchical behavioral feature extraction module and a data pool module; and a lesson plan design error prediction algorithm module.

20. The virtual classroom teaching method as described in claim 6, wherein the tone, speech, speech rate, and supplementation method of the virtual teacher image in the virtual teacher control module are adjusted in real time according to the user's experience and feedback data, and the virtual teacher control module further includes: a tone adaptation algorithm unit; a perception friendliness correction parameter setting unit; a contextual guidance logic decision unit; and an expandable speech feature library unit.

Images