Artificial intelligence education integrated terminal equipment for teenagers and system architecture thereof

Abstract

The invention discloses a teenager-oriented artificial intelligence education integrated terminal device and a system architecture thereof. The terminal device comprises a perception interaction module; the local reasoning and core processing module is provided with a four-core processor and is internally provided with a light-weight large language model for local operation; the student learning portrait and path recommendation module is used for collecting multi-modal data such as student click behaviors, answer accuracy, learning duration and input keywords, and constructing a user behavior log; the multi-modal knowledge graph module is used for constructing a cross-modal knowledge graph; teaching resource scheduling and course package engine: teaching resources are packaged into containerized course packages, and classified and packaged according to grades / themes; the hardware expansion and external interface module is used for reserving a GPIO (General Purpose Input / Output) interface area and supporting the sensor module, the camera module and the voice interaction expansion module; and a software architecture and an operation mechanism. The adolescent education terminal integrates artificial intelligence course resources, a programming environment, a large language model, a knowledge graph and a personalized recommendation system.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence education and computing terminal equipment, and specifically relates to an integrated terminal device for artificial intelligence education for teenagers and its system architecture. Background Art

[0002] Existing educational technology tools and learning equipment still cannot meet the requirements of the integrated development of "curriculum-tools-personalized services".

[0003] Currently, the more common AI learning carriers are:

[0004] (1) General computing terminals (such as PCs, tablets, Raspberry Pi, Jetson Nano, etc.): They have basic programming and model running capabilities, but lack educational-specific functions such as course mapping, learning behavior analysis, and visual feedback systems. They are complex to operate and are not friendly to primary and secondary school students.

[0005] (2) Platform-based learning systems (such as Tencent Koding and Intel AIxBoard): Although they integrate certain learning resources, they are unable to provide closed-loop support for “localized model operation + path adaptive recommendation + knowledge structured learning”.

[0006] (3) AI educational teaching aids (such as intelligent robot kits and AI experiment boxes): They emphasize project practice but lack teaching content support and reasoning ability, and cannot form an intelligent feedback mechanism.

[0007] In summary, there is currently no AI learning terminal device that can cover course content, model deployment, behavior analysis, path recommendation, interactive feedback and other functions in one, and is suitable for the learning scenarios of primary and secondary school students.

[0008] Current AI education has the following shortcomings:

[0009] Disadvantage 1: Course resources are fragmented, the standard system is not unified, and teaching effectiveness is limited by the ability to integrate content;

[0010] Disadvantage 2: Most models are deployed on remote servers, which are subject to network constraints, resulting in high interaction latency and poor data security.

[0011] Disadvantage 3: The student learning process lacks behavioral data support, making it impossible to implement process evaluation and personalized feedback;

[0012] Disadvantage 4: The design of teaching terminals is out of line with primary and secondary school curriculum, and the configuration is complex, which is not conducive to large-scale deployment and use by teachers. Summary of the Invention

[0013] In order to solve the defects and shortcomings of the above-mentioned existing technologies, the present invention provides an integrated AI education platform that can build a deep integration of software and hardware, and provide a complete teaching support tool chain; realizes a large language model with localized operation, and has functions such as AI question and answer, code generation, and dialogue guidance, thereby improving the efficiency of student human-computer interaction; builds student portraits based on graph neural networks, collects learning behavior data, and dynamically adjusts resource recommendation paths; constructs a multimodal knowledge graph that integrates text, images, and audio to support visual teaching and dynamic knowledge expansion; and realizes teaching management functions such as task release, process monitoring, and learning evaluation on the teacher side, which is an integrated terminal device for artificial intelligence education for teenagers and its system architecture.

[0014] The present invention provides the following technical solution: an integrated terminal device for artificial intelligence education for teenagers and its system architecture, comprising:

[0015] Perception and Interaction Module: Installed on the front of the device, it includes a touch screen, a teaching start button, a microphone array, a wide-angle camera, a status indicator light, and an audio output.

[0016] Local reasoning and core processing module: equipped with a quad-core processor (x86 / ARM compatible), built-in lightweight large language models (such as MiniGPT or DeepSeekLite) running locally;

[0017] Student learning profile and path recommendation module: collects multimodal data such as student click behavior, answer accuracy, learning time, input keywords, etc., and builds a user behavior log;

[0018] Multimodal knowledge graph module: Builds a cross-modal knowledge graph with entity types including "course module", "task node", "knowledge point", "project resource", and "AI interaction portal";

[0019] Teaching resource scheduling and course package engine: Teaching resources are packaged into containerized course packages (supporting Docker / Flatpak formats) and packaged by grade / topic classification;

[0020] Hardware expansion and external interface module: The device reserves a GPIO interface area to support sensor modules (such as light / temperature and humidity), camera modules, and voice interaction expansion modules;

[0021] Software architecture and operation mechanism: The software of the entire teaching system consists of four parts: teaching operating system layer, model service layer, recommendation system layer, and interactive control layer.

[0022] Preferably, the voice input of the perception interaction module is completed through the local semantic recognition module of a domestic AI chip (such as Cambricon); the camera collects images for project task recognition, face check-in and expression status analysis, and the output is sent to the AI analysis module; the teaching start button has one-click wake-up and direct course access functions, and wakes up the system through GPIO interruption.

[0023] Preferably, the model inference engine of the local reasoning and core processing module is based on ONNX Runtime and OpenVINO toolkit and is deployed locally on the device. It has an internally integrated prompt engineering scheduling mechanism and adopts an education-customized Prompt template to guide students to conduct a "Socratic" learning dialogue. The device matches the model dialogue path through the preset teaching material structure index to achieve the recall of knowledge points and personalized question and answer recommendations.

[0024] Preferably, the student learning profile and path recommendation module performs node embedding calculation based on the graph attention network (GAT), matches the student status with the knowledge point graph, and outputs a "knowledge status graph". The system retrieves matching content from the graph database based on the current status graph, and feeds back the optimal task path. It supports multi-round adaptive recommendations, and the technology stack includes: Neo4j graph database + PyG graph learning library + Flask routing interface management.

[0025] Preferably, each node of the multimodal knowledge graph module stores modal information such as text descriptions, pictures, voice commands, etc. The edge weights represent the learning path dependencies. The graph supports real-time queries (Cypher statements) and is dynamically linked with the student portrait module. The display layer draws a dynamic graph view through Vis.js to achieve visual feedback on both the teacher and student sides.

[0026] Preferably, the teaching resource scheduling and course package engine system has a built-in course scheduler, which dynamically pulls the required resources according to the student portrait and task progress. The teacher assigns tasks through the teaching background (Web), and the system sends the tasks to the terminal device through the MQTT protocol. The student completes the task in the form of "conversational guidance + practical exercises + graphic feedback", and automatically records the operation log during the process.

[0027] Preferably, the hardware expansion and external interface module supports standard interfaces such as USB 3.0, HDMI, Ethernet, SD card slot, Type-C power supply, etc. The external interface has a hot-plug recognition mechanism, and the device automatically switches to the corresponding course mode (such as automatically switching to the image classification task when connected to the visual recognition module).

[0028] Preferably, the teaching operating system layer of the software architecture and operation mechanism: takes Ubuntu as the core, loads a graphical learning desktop, and has an embedded course launcher (supports self-starting according to weekly progress); the model service layer: encapsulates the ONNX model call service, provides local interaction through HTTP / Socket, and has a response time of <300ms; the recommendation system layer: based on the GAT graph model, updates the portrait of each round of learning process and recommends tasks according to weight changes; the interactive control layer: is responsible for task scheduling and data synchronization between the teacher platform and the student terminal.

[0029] Preferably, large model deployment can be switched to LAN edge inference services to reduce the terminal load;

[0030] The student portrait module can be replaced by a Transformer-based or BERT model;

[0031] The operating system can be adapted to domestic Kylin or Windows IoT to meet government procurement requirements;

[0032] Teaching resources can support SAAS mode with cloud synchronization backup and unified API scheduling;

[0033] The terminal form can be expanded into a card-type module, an all-in-one teaching machine or a teacher's console form

[0034] Compared with the existing technology, the beneficial effects achieved by the present invention are: the present invention can build an integrated AI education platform with deep integration of software and hardware, and provide a complete teaching support tool chain; realize localized operation of a large language model with functions such as AI question and answer, code generation, and dialogue guidance, thereby improving the efficiency of human-computer interaction among students; build student portraits based on graph neural networks, collect learning behavior data, and dynamically adjust resource recommendation paths; build a multimodal knowledge graph that integrates text, images, and audio to support visual teaching and dynamic knowledge expansion; the teacher side realizes teaching management functions such as task release, process monitoring, and learning evaluation to ensure that students can access the platform in a unified hardware environment. The platform can generate learning portraits based on their historical learning paths, answering performance, and behavior records, and push the most suitable courses, exercises, and project tasks to achieve a "thousand faces" of AI personalized learning experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a structural schematic diagram of the present invention;

[0036] Figure 2 This is a system architecture diagram of the present invention. DETAILED DESCRIPTION

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0038] The present invention provides the following technical solutions:

[0039] Example 1

[0040] See also Figure 1 and 2 , an integrated terminal device for artificial intelligence education for teenagers and its system architecture. The system structure and modules are divided as follows:

[0041] 1. Perception interaction module

[0042] Technical implementation:

[0043] Mounted on the front of the device, it includes a touchscreen display, a teaching start button, a microphone array, a wide-angle camera, a status indicator light, and an audio output.

[0044] Voice input is completed through the local semantic recognition module of domestic AI chips (such as Cambricon).

[0045] The camera collects images for project task recognition, face check-in and expression status analysis, and the output is sent to the AI analysis module.

[0046] The teaching start button has one-click wake-up and direct course access functions, and wakes up the system through GPIO interrupt.

[0047] 2. Local inference and core processing modules

[0048] Technical implementation:

[0049] It is equipped with a quad-core processor (x86 / ARM compatible) and has a built-in lightweight large language model (such as MiniGPT or DeepSeekLite) that runs locally.

[0050] The model inference engine is based on ONNX Runtime and OpenVINO toolkit and is deployed locally on the device.

[0051] An internally integrated prompt engineering scheduling mechanism uses customized prompt templates for education to guide students in "Socratic" learning dialogues.

[0052] The device matches the model dialogue path through the preset textbook structure index, realizing the process from knowledge point recall to personalized question and answer recommendation.

[0053] 3. Student learning profile and path recommendation module

[0054] Technical implementation:

[0055] The module collects multimodal data such as students' click behavior, answer accuracy, learning time, input keywords, etc. to build a user behavior log.

[0056] Based on the graph attention network (GAT), node embedding calculation is performed to match the student status with the knowledge point graph and output the "knowledge status graph".

[0057] Based on the current state graph, the system retrieves matching content from the graph database, feeds back the optimal task path, and supports multi-round adaptive recommendations.

[0058] The technology stack includes: Neo4j graph database + PyG graph learning library + Flask routing interface management.

[0059] 4. Multimodal Knowledge Graph Module

[0060] Technical implementation:

[0061] Build a cross-modal knowledge graph, with entity types including "course module", "task node", "knowledge point", "project resource", "AI interaction entrance", etc.

[0062] Each node stores modal information such as text descriptions, pictures, and voice commands, and the edge weights represent learning path dependencies.

[0063] The graph supports real-time queries (Cypher statements) and is dynamically linked with the student portrait module.

[0064] The display layer uses Vis.js to draw dynamic graph views to achieve visual feedback on both the teacher and student sides.

[0065] 5. Teaching resource scheduling and course package engine

[0066] Technical implementation:

[0067] Teaching resources are packaged as containerized course packages (supporting Docker / Flatpak formats) and packaged by grade / topic.

[0068] The system has a built-in course scheduler that dynamically pulls required resources based on student profiles and task progress.

[0069] The teacher assigns tasks through the teaching backend (Web), and the system sends the tasks to the terminal device through the MQTT protocol.

[0070] The student side completes the task through "dialogue guidance + practical exercises + graphic feedback", and automatically records the operation log during the process.

[0071] 6. Hardware expansion and external interface module

[0072] Technical implementation:

[0073] The device has a reserved GPIO interface area to support sensor modules (such as light / temperature and humidity), camera modules, and voice interaction expansion modules.

[0074] Supports standard interfaces such as USB 3.0, HDMI, Ethernet, SD card slot, and Type-C power supply.

[0075] The external interface has a hot-swap recognition mechanism, and the device automatically switches to the corresponding course mode (such as automatically switching to the image classification task when connected to the visual recognition module).

[0076] 7. Software architecture and operation mechanism

[0077] The software of the entire teaching system consists of four parts: teaching operating system layer, model service layer, recommendation system layer, and interactive control layer.

[0078] Teaching operating system layer: With Ubuntu as the core, it loads a graphical learning desktop and has a built-in course launcher (supports self-starting according to weekly progress).

[0079] Model service layer: encapsulates ONNX model call service, provides local interaction through HTTP / Socket, and has a response time of <300ms.

[0080] Recommendation system layer: Based on the GAT graph model, the profile of each round of learning process is updated and tasks are recommended based on weight changes.

[0081] Interactive control layer: responsible for task scheduling and data synchronization between the teacher platform and student terminals.

[0082] Example 2

[0083] An integrated terminal device for artificial intelligence education for teenagers and its system architecture, the method of which is implemented as follows:

[0084] The teacher issues task instructions;

[0085] The student device receives the task and wakes up the corresponding course package;

[0086] Local models begin to guide the conversation;

[0087] Real-time recording of students’ operation behaviors;

[0088] Behavioral data is fed into the graph neural network to generate updated portraits;

[0089] The recommendation system returns the next stage of tasks or content;

[0090] The teacher side obtains data feedback synchronously.

[0091] This invention discloses a youth education terminal that integrates artificial intelligence course resources, a programming environment, a large language model, a knowledge graph, and a personalized recommendation system. It is suitable for use in primary and secondary school "information technology" courses and artificial intelligence teaching scenarios. Its main innovations are as follows:

[0092] The ability to deploy and run the Large Learning Model (LLM) locally on the terminal;

[0093] The student portrait system uses graph neural network algorithms to achieve dynamic path reasoning and recommendation;

[0094] Multimodal knowledge graph structure construction method and calling logic (integrated with recommendation system);

[0095] Coupling logic between the teaching task scheduling mechanism and hardware;

[0096] Modular software and hardware collaborative architecture supports regional deployment and course customization.

[0097] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations may be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An integrated terminal device for artificial intelligence education for teenagers and its system architecture, characterized by: include: Perception and Interaction Module: Installed on the front of the device, it includes a touch screen, a teaching start button, a microphone array, a wide-angle camera, a status indicator light, and an audio output. Local inference and core processing module: equipped with a quad-core processor and built-in lightweight large language model for local operation; Student learning profile and path recommendation module: collects multimodal data such as student click behavior, answer accuracy, learning time, and input keywords to build a user behavior log; Multimodal knowledge graph module: Builds a cross-modal knowledge graph with entity types including "course module", "task node", "knowledge point", "project resource", and "AI interaction portal"; Teaching resource scheduling and course package engine: Teaching resources are packaged into containerized course packages, classified and packaged by grade / topic; Hardware expansion and external interface module: The device reserves a GPIO interface area to support sensor modules, camera modules, and voice interaction expansion modules; Software architecture and operation mechanism: The software of the entire teaching system consists of four parts: teaching operating system layer, model service layer, recommendation system layer, and interactive control layer.

2. The integrated terminal device for artificial intelligence education for teenagers and its system architecture according to claim 1 is characterized by: The voice input of the perception interaction module is completed through the local semantic recognition module of the domestic AI chip; the camera collects images for project task recognition, face check-in and expression status analysis, and the output is sent to the AI analysis module; the teaching start button has one-click wake-up and direct course access functions, and wakes up the system through GPIO interrupt.

3. The integrated terminal device for artificial intelligence education for teenagers and its system architecture according to claim 1 is characterized by: The model inference engine of the local reasoning and core processing module is based on ONNX Runtime and the OpenVINO toolkit and is deployed locally on the device. It has an integrated prompt engineering scheduling mechanism and uses customized educational prompt templates to guide students in "Socratic" learning dialogues. The device matches the model dialogue path through a preset textbook structure index, realizing the process from knowledge point recall to personalized question and answer recommendations.

4. The integrated terminal device for artificial intelligence education for teenagers and its system architecture according to claim 1 is characterized by: The student learning profile and path recommendation module performs node embedding calculations based on the graph attention network (GAT), matches student status with the knowledge point graph, and outputs a "knowledge state graph". Based on the current state graph, the system retrieves matching content from the graph database and provides feedback on the optimal task path. It supports multi-round adaptive recommendations and its technology stack includes: Neo4j graph database + PyG graph learning library + Flask routing interface management.

5. The integrated terminal device for artificial intelligence education for teenagers and its system architecture according to claim 1 is characterized by: Each node of the multimodal knowledge graph module stores text description, picture, and voice command modal information. The edge weight represents the learning path dependency. The graph supports real-time query (Cypher statement) and is dynamically linked with the student portrait module. The display layer draws a dynamic graph view through Vis.js to achieve visual feedback on both the teacher and student sides.

6. The integrated terminal device for artificial intelligence education for teenagers and its system architecture according to claim 1 is characterized by: The teaching resource scheduling and course package engine has a built-in course scheduler that dynamically pulls required resources based on student profiles and task progress. The teacher assigns tasks through the teaching backend (Web), and the system sends tasks to terminal devices through the MQTT protocol. The student completes the task in a "conversational guidance + practical exercises + graphic feedback" manner, and automatically records operation logs during the process.

7. The integrated terminal device for artificial intelligence education for teenagers and its system architecture according to claim 1 is characterized by: The hardware expansion and external interface module supports USB 3.0, HDMI, Ethernet, SD card slot, and Type-C power supply standard interface. The external interface has a hot-plug recognition mechanism, and the device automatically switches to the corresponding course mode (such as automatically switching to the image classification task when connected to the visual recognition module).

8. The integrated terminal device for artificial intelligence education for teenagers and its system architecture according to claim 1 is characterized by: The software architecture and operating mechanism include: a teaching operating system layer with Ubuntu as the core, loading a graphical learning desktop, and embedding a course launcher; a model service layer that encapsulates the ONNX model call service and provides local interaction through HTTP / Socket with a response time of <300ms; a recommendation system layer that updates the portrait of each round of learning process based on the GAT graph model and recommends tasks based on weight changes; and an interaction control layer that is responsible for task scheduling and data synchronization between the teacher platform and the student terminal.

9. The integrated terminal device for artificial intelligence education for teenagers and its system architecture according to claim 1 is characterized by: Large model deployment can be switched to LAN edge inference services to reduce terminal load; The student portrait module can be replaced by a Transformer-based or BERT model; The operating system can be adapted to domestic Kylin or Windows IoT to meet government procurement requirements; Teaching resources can support SAAS mode with cloud synchronization backup and unified API scheduling; The terminal form can be expanded into a card-type module, an all-in-one teaching machine or a teacher's version console form.

Images