A multimodal virtual avatar AI agent interaction system and method

Abstract

This invention relates to the fields of artificial intelligence agents and multimodal human-computer interaction technology, and provides a multimodal virtual avatar AI agent interaction system, including a voice wake-up interaction module, a virtual avatar rendering module, an AI agent core scheduling module, an OpenClaw execution control module, a terminal document intelligent management module, a daily news generation module, a visualization linkage output module, and a multimodal data acquisition unit. The AI ​​agent core scheduling module establishes bidirectional data connections with other modules, issuing commands and receiving feedback. This invention achieves three major benefits: first, it proactively analyzes user preferences, automatically pushes information, and intelligently processes documents, improving efficiency; second, it dynamically adapts to personalized virtual avatars and supports voice control, enhancing the interactive experience; and third, it standardizes data display through the visualization module and supports review, forming a complete interactive closed loop.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence agents and multimodal human-computer interaction technology, specifically to a multimodal virtual avatar AI agent interaction system and method. Background Technology

[0002] While existing large language models, AI agents, and virtual avatar interaction technologies have gradually developed, they still have the following significant technical shortcomings in practical applications, making it difficult to meet the personalized, intelligent, and one-stop needs for daily office work and companionship: (1) Lack of proactive service and intelligent document processing capabilities. Most existing intelligent interactive systems adopt a passive response command mode, which cannot proactively combine users' historical data and daily information to analyze personalized needs. As a result, users need to manually collect and organize daily information and personal matters, resulting in a low level of intelligence. At the same time, the number of local documents on smart terminals is large and complex, and users need to manually browse and find key content. There is a lack of intelligent marking, summary extraction and quick search functions, resulting in low document management efficiency.

[0003] (2) The virtual avatars are monotonous and the interaction methods are rigid, lacking emotional appeal and convenience. Most interactive systems lack personalized virtual avatars, and only complete the interaction through text or simple pop-ups, resulting in a rigid human-computer interaction experience and a lack of emotional companionship. Existing virtual avatar interaction systems mostly use fixed avatars and cannot automatically adapt to personalized avatars based on user content platform preferences, resulting in poor interaction exclusivity and adaptability. In addition, the system does not have a voice wake-up function, and operation relies on manual input, which is not convenient enough.

[0004] (3) Lack of dedicated visualization and execution log review functions, resulting in fragmented data presentation. Existing technologies lack dedicated visualization modules, and the interaction results and task data are displayed in a fragmented manner. They do not support the retention of execution logs and subsequent review, making it inefficient for users to view and trace back data, and failing to form a complete interactive loop.

[0005] Therefore, there is an urgent need to provide a multimodal virtual avatar AI intelligent agent interaction system and method to solve the technical problems of passive response, lack of personalization, inefficient document processing and inability to review data in the existing technology. Summary of the Invention

[0006] This invention provides a multimodal virtual avatar AI intelligent agent interaction system and method, which addresses the problems of passive response, lack of personalization, inefficient document processing and fragmented data in existing systems, and aims to achieve proactive, personalized and closed-loop intelligent interaction.

[0007] A multimodal virtual avatar AI intelligent agent interaction system, comprising: The voice wake-up interaction module collects user voice commands and converts them into text data, receives feedback information from the core scheduling module of the AI ​​agent, and converts it into voice for broadcast. The virtual avatar rendering module matches and switches virtual avatars according to user preferences, and displays interactive content in sync with voice feedback. The core scheduling module of the AI ​​intelligent agent, as the central control unit of the system, parses voice commands, analyzes user historical data and preference profiles, generates execution commands, schedules the work of each module, and stores user interaction and task data; The OpenClaw execution control module receives instructions from the AI ​​agent core scheduling module, executes person recognition and user sentiment analysis tasks, and reports the task status and results back to the AI ​​agent core scheduling module. The terminal document intelligent management module automatically scans documents of various formats within smart terminals, extracts the core content of documents through semantic analysis, marks key paragraphs, and generates summaries and viewing links. The Daily News Generation Module generates personalized daily news reports based on users' historical data, daily activities, and areas of interest. The visualization and interactive output module displays the results of character emotional interactions, system execution logs, key document summaries, and daily news updates. The multimodal data acquisition unit collects raw image and voice data from users, providing the underlying data for emotion recognition and person recognition to the OpenClaw execution control module; The AI ​​intelligent agent core scheduling module establishes bidirectional data connections with the voice wake-up interaction module, the virtual image rendering module (200), the OpenClaw execution control module, the terminal document intelligent management module, the daily fresh news generation module, and the visualization linkage output module, respectively, and issues instructions and receives feedback.

[0008] Further preferably, the voice wake-up interaction module has a custom wake-up word trigger input terminal and a timed automatic trigger input terminal.

[0009] In a further preferred embodiment, after user authorization, the virtual avatar rendering module receives preference recognition results generated by the AI ​​intelligent agent core scheduling module based on the user's browsing, likes, and collection data from social and content platforms, and matches and switches virtual avatars from a virtual avatar library including real people, cute pets, anime, and cartoons according to the preference recognition results.

[0010] Further optimization involves the visualization linkage output module and the virtual avatar rendering module outputting data and logs independently; the visualization linkage output module outputs specific data and logs to local storage, while the virtual avatar rendering module outputs the dynamic display content of the virtual avatar.

[0011] Further optimization involves the AI ​​agent core scheduling module incorporating a natural language understanding submodule, an instruction generation submodule, a memory storage submodule, and a task scheduling submodule.

[0012] This invention also provides a multimodal virtual avatar AI agent interaction method, comprising: Obtain a startup command, which is triggered by a timed method or a voice wake-up method; According to the start command, the AI ​​intelligent agent core scheduling module reads the user's authorized historical preference data and analyzes the user's current preferred virtual avatar type based on the historical preference data; The AI ​​intelligent agent core scheduling module sends a switching instruction to the virtual image rendering module. Based on the switching instruction, the virtual image rendering module performs virtual image matching for the user according to the virtual image type to obtain the user's current virtual image. The AI ​​intelligent agent core scheduling module sends task instructions to the terminal document intelligent management module and the daily news generation module in parallel. The terminal document intelligent management module extracts key content from the document according to the task instructions; the daily news generation module generates personalized information reports according to the task instructions. The AI ​​agent core scheduling module receives the key content of the document and the personalized information report, integrates them, and generates an integrated result. The AI ​​intelligent agent core scheduling module sends a display instruction to the virtual image rendering module, and the virtual image rendering module outputs the integrated result with emotional voice based on the user's current virtual image according to the display instruction; The AI ​​agent core scheduling module sends output instructions to the visualization linkage output module in parallel, and the visualization linkage output module displays the integrated results according to the output instructions.

[0013] Further optimization involves the voice wake-up interaction module collecting user voice commands and converting them into text data, then transmitting the text data to the AI ​​agent core scheduling module.

[0014] Further preferably, the timed triggering method includes: automatically generating a start command according to a preset time point.

[0015] In a further preferred embodiment, while the AI ​​agent core scheduling module sends task instructions to the terminal document intelligent management module and the daily news generation module in parallel, the method also includes: The AI ​​agent core scheduling module sends an emotion recognition instruction to the OpenClaw execution control module; the OpenClaw execution control module performs user emotion recognition according to the emotion recognition instruction and feeds back the recognition result to the AI ​​agent core scheduling module. The AI ​​intelligent agent core scheduling module associates the recognition result with the user's current virtual image obtained by the virtual image rendering module.

[0016] Further preferably, the visualization linkage output module associates the emotional voice output content with the corresponding execution log and stores it in local memory; and / or The visualization and linkage output module associates the displayed data with the corresponding execution logs and stores them in local memory.

[0017] The beneficial effects of this invention are: (1) Achieving proactive intelligent services and efficient document management. This invention abandons the traditional passive response mode and proactively analyzes user preference profiles through the core scheduling module of the AI ​​intelligent agent to automatically generate personalized daily information reports without requiring users to manually collect and organize them; at the same time, the terminal document intelligent management module automatically scans and extracts the core content of multi-format documents and marks key paragraphs, significantly improving document processing and office efficiency, and solving the technical problems of low intelligence level and time-consuming and laborious document management in existing systems.

[0018] (2) Provide personalized virtual avatars and convenient voice interaction. This invention uses a virtual avatar rendering module to match and switch virtual avatars from a virtual avatar library, including real people, cute pets, anime, and cartoons, based on the preference data of social and content platforms and the user's authorization, to create an emotional and exclusive interactive experience. At the same time, it is equipped with a voice wake-up interaction module, which supports a dual mode of custom wake-up words and timed automatic triggering, to achieve full voice control, replace traditional manual input, and solve the problems of stiff interaction, lack of emotional companionship, and insufficient ease of operation of existing systems.

[0019] (3) Achieve standardized data display and closed-loop review functions. This invention adds a visual linkage output module, which adopts a panel mode to specifically display the results of human emotional interaction and system execution logs, and is compatible with displaying key document summaries and daily news content. It supports local saving of interactive reports and subsequent review and retrospection; combined with the global task scheduling and data storage capabilities of the AI ​​intelligent agent core scheduling module, a complete active interaction closed loop is formed, solving the technical problems of fragmented data presentation, inability to review, and lack of complete closed loop in the existing system. Attached Figure Description

[0020] Figure 1 This is a block diagram of the multimodal virtual avatar AI intelligent agent interaction system architecture of the present invention; Figure 2 This is a flowchart illustrating the multimodal virtual avatar AI agent interaction method in an embodiment of the present invention; In the diagram: 100, Voice wake-up interaction module; 200, Virtual image rendering module; 300, AI intelligent agent core scheduling module; 400, OpenClaw execution control module; 500, Terminal document intelligent management module; 600, Daily news generation module; 700, Visualized linkage output module; 800, Multimodal data acquisition unit. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1 This embodiment provides a multimodal virtual avatar AI intelligent agent interaction system, such as... Figure 1 The multimodal virtual avatar AI intelligent agent interaction system architecture diagram of this invention is shown, including: The voice wake-up interaction module 100 collects user voice commands and converts them into text data. It receives feedback information from the AI ​​intelligent agent core scheduling module 300 and converts it into voice for broadcast. The voice wake-up interaction module 100 has a custom wake-up word trigger input terminal and a timed automatic trigger input terminal.

[0023] The virtual avatar rendering module 200 matches and switches virtual avatars according to user preferences, and displays interactive content synchronously with voice feedback. After user authorization, the virtual avatar rendering module 200 receives preference recognition results generated by the AI ​​intelligent agent core scheduling module 300 based on the user's browsing, likes, and collection data from social and content platforms, and matches and switches virtual avatars from a virtual avatar library including real people, cute pets, anime, and cartoons according to the preference recognition results.

[0024] The AI ​​agent core scheduling module 300, as the central control unit of the system, parses voice commands, analyzes user historical data and preference profiles, generates execution commands, schedules the work of various modules, and stores user interaction and task data. The AI ​​agent core scheduling module 300 has built-in natural language understanding submodules, command generation submodules, memory storage submodules, and task scheduling submodules.

[0025] The OpenClaw execution control module 400 receives instructions from the AI ​​agent core scheduling module, executes person recognition and user sentiment analysis tasks, and reports the task status and results back to the AI ​​agent core scheduling module.

[0026] The Terminal Document Intelligent Management Module 500 automatically scans documents of various formats within smart terminals, extracts the core content of the documents through semantic analysis, marks key paragraphs, and generates summaries and viewing links.

[0027] The Daily News Generation Module 600 generates personalized daily news reports based on users' historical data, daily activities, and areas of interest.

[0028] The visualization-linked output module 700 displays the results of character emotional interactions, system execution logs, key document summaries, and daily news updates. The visualization-linked output module 700 and the virtual avatar rendering module 200 output independently; the visualization-linked output module 700 outputs specific data and logs to local storage, while the virtual avatar rendering module 200 outputs the dynamic display content of the virtual avatar.

[0029] The multimodal data acquisition unit 800 collects raw image and voice data from users and provides underlying data for emotion recognition and person recognition to the OpenClaw execution control module.

[0030] The AI ​​intelligent agent core scheduling module 300 establishes bidirectional data connections with the voice wake-up interaction module 100, the virtual image rendering module 200, the OpenClaw execution control module 400, the terminal document intelligent management module 500, the daily fresh news generation module 600, and the visualization linkage output module 700, respectively, and issues instructions and receives feedback.

[0031] This embodiment provides a multimodal virtual avatar AI intelligent agent interaction method, such as... Figure 2 The flowchart of the multimodal virtual avatar AI intelligent agent interaction method in this embodiment of the invention is shown. The method steps include: Step S1: Obtain a startup command, which may be triggered by a timed event or by voice wake-up. Voice wake-up includes: the voice wake-up interaction module 100 collecting the user's voice command and converting it into text data, then transmitting the text data to the AI ​​agent core scheduling module 300. Timed triggering includes: automatically generating a startup command according to a preset time point.

[0032] Step S2: According to the start command, the AI ​​intelligent agent core scheduling module 300 reads the historical preference data authorized by the user, and analyzes the user's current preferred virtual image type based on the historical preference data.

[0033] Step S3: The AI ​​intelligent agent core scheduling module 300 sends a switching instruction to the virtual image rendering module 200. Based on the switching instruction, the virtual image rendering module 200 performs virtual image matching for the user according to the virtual image type to obtain the user's current virtual image.

[0034] Step S4: The AI ​​agent core scheduling module 300 sends task instructions in parallel to the terminal document intelligent management module 500 and the daily news generation module 600, respectively. The terminal document intelligent management module 500 extracts key content from the documents according to the task instructions; the daily news generation module 600 generates personalized information reports according to the task instructions. While the AI ​​agent core scheduling module 300 sends task instructions in parallel to the terminal document intelligent management module 500 and the daily news generation module 600, it also includes: The AI ​​agent core scheduling module 300 sends an emotion recognition instruction to the OpenClaw execution control module 400; the OpenClaw execution control module 400 performs user emotion recognition according to the emotion recognition instruction and feeds back the recognition result to the AI ​​agent core scheduling module 300.

[0035] The AI ​​intelligent agent core scheduling module 300 associates the recognition result with the user's current virtual image obtained by the virtual image rendering module 200.

[0036] Step S5: The AI ​​intelligent agent core scheduling module 300 receives the key content of the document and the personalized information report, integrates them, and generates an integrated result.

[0037] Step S6: The AI ​​intelligent agent core scheduling module 300 sends a display command to the virtual image rendering module 200. The virtual image rendering module 200 outputs an emotional voice to the integrated result based on the display command and the user's current virtual image.

[0038] Step S7: The AI ​​intelligent agent core scheduling module 300 sends an output instruction to the visualization linkage output module 700 in parallel. The visualization linkage output module 700 displays the integrated result according to the output instruction.

[0039] In addition, the multimodal virtual avatar AI intelligent agent interaction method of this embodiment also includes: the visualization linkage output module 700 associating the emotional voice output content with the corresponding execution log and storing it in the local memory; and / or the visualization linkage output module 700 associating the data display content with the corresponding execution log and storing it in the local memory.

[0040] Example 2 This embodiment describes the complete process of the system automatically starting within a preset time period, without requiring manual user intervention, to complete virtual avatar adaptation, intelligent document processing, and personalized information push.

[0041] Step 101: Trigger system startup at regular intervals The system has a built-in timer module, and users can preset the daily start time in the system configuration interface (e.g., 8:00 AM on weekdays). When the preset time is reached, the timer module automatically generates a trigger signal, and the voice wake-up interaction module 100 receives the signal and activates the system without requiring user voice input or manual operation.

[0042] Step 102: User Preference Analysis and Virtual Avatar Switching With prior user authorization and in compliance with data security regulations, the AI ​​agent core scheduling module 300 reads the user's browsing history, likes, favorites, and other preference data from authorized social and content platforms. Simultaneously, the memory storage and analysis submodule retrieves stored historical chat logs and daily schedules. Based on this data, the AI ​​agent core scheduling module 300 analyzes the user's current preferred virtual avatar type (e.g., a simple cartoon avatar on weekdays, a cute pet avatar on weekends), generates standardized switching instructions, and sends them to the virtual avatar rendering module 200. The virtual avatar rendering module 200, according to the instructions, matches the corresponding avatar from the preset avatar library, completes the rendering switch, and simultaneously displays it on the terminal interface.

[0043] This step utilizes user-authorized social and content platform preference data to achieve dynamic and personalized matching of virtual avatars. Compared to existing systems using fixed avatars, this solution can adapt avatars to various styles, such as real people and cute pets, based on user preferences, and, combined with voice interaction, significantly enhances the sense of exclusivity and emotional companionship in human-computer interaction.

[0044] Step 103: Parallel execution of intelligent document processing and information generation The AI ​​intelligent agent core scheduling module 300 simultaneously sends task instructions to the terminal document intelligent management module 500 and the daily fresh news generation module 600.

[0045] The terminal document intelligent management module 500 automatically scans multi-format documents (including PDF, Word, TXT, etc.) in specified directories (such as "Work Documents" and "Recent Projects") within the smart terminal. It uses semantic analysis technology to extract documents that have been added or modified in the past two days, extracts the core content of the documents, marks key paragraphs, and generates concise summaries and quick access links.

[0046] The Daily News Generation Module 600, combined with the AI ​​Intelligent Agent Core Scheduling Module 300, stores user-focused areas (such as new energy policies and industry trends) and daily tasks (such as to-do meetings and project milestones). It automatically retrieves and integrates relevant information to generate structured and personalized daily news reports.

[0047] Step 104: Dual-end synchronous output and interactive reminders After receiving the feedback results from the two modules, the AI ​​intelligent agent core scheduling module 300 sends display instructions to the virtual image rendering module 200 and the visualization linkage output module 700, respectively.

[0048] The virtual avatar rendering module 200 displays the switched virtual avatar on the terminal interface and broadcasts reminders (such as "Good morning, there are 3 new documents and 2 key news items to pay attention to today") through the voice wake-up interaction module 100.

[0049] The visualization and interactive output module 700 displays a list of key document summaries, core content of daily news, and system execution logs for this interaction in a panel-based format.

[0050] Users can click to view specific content in the virtual avatar interface or visual panel, achieving a fully automated, personalized, and proactive intelligent service without human intervention.

[0051] The technical effect of this embodiment is that it realizes fully automatic intelligent interaction triggered by time, which can complete virtual image adaptation, document processing and information push without user operation, and significantly improve the level of office automation.

[0052] This embodiment describes a voice wake-up instant interaction scenario, where the user actively triggers the system by using a custom wake-up word to complete the interactive process of multi-task parallel processing and instant feedback.

[0053] Through steps 102 to 104 above, this system achieves proactive analysis and service of user preferences. The AI ​​intelligent agent core scheduling module abandons the traditional passive response mode and proactively triggers the generation of personalized information reports; at the same time, the terminal document intelligent management module automatically completes the summary extraction and key point marking of local multi-format documents. The entire process eliminates the need for users to manually collect and organize information or browse through a large number of documents, significantly improving the efficiency of daily office work and information acquisition, and effectively solving the technical problems of low intelligence and time-consuming and laborious document management in existing technologies.

[0054] Step 201: Voice wake-up trigger When a user speaks a preset wake-up word, the voice wake-up interaction module 100 collects the user's voice command through its built-in microphone, uses speech recognition technology to convert the speech into standard text data, and transmits it to the AI ​​intelligent agent core scheduling module 300 in real time.

[0055] This step supports custom wake-word voice triggering modes, which, combined with timed automatic triggering modes, provide users with a convenient entry point for full voice control. Users can start the system and issue complex commands without any manual input, solving the problems of cumbersome operation and insufficient convenience of traditional interaction methods.

[0056] Step 202: Instruction Parsing and Task Scheduling After receiving text commands, the AI ​​agent core scheduling module 300 parses the user's intent through the natural language understanding submodule. For example, the user command might be, "Analyze the emotions in the previous conversation and find the key points in the project document." The AI ​​agent core scheduling module 300 breaks down the task into two parallel subtasks: a sentiment analysis task and a document extraction task.

[0057] Step 203: Execute multi-module tasks in parallel The AI ​​intelligent agent core scheduling module 300 synchronously sends instructions to the OpenClaw execution control module 400 and the terminal document intelligent management module 500.

[0058] The OpenClaw execution control module 400 receives user facial images and voice feature data collected in real time by the multimodal data acquisition unit 800, loads the emotion analysis model, identifies the user's current emotional state (such as positive, calm or anxious), and feeds back the identification results to the AI ​​intelligent agent core scheduling module 300 in real time.

[0059] The terminal document intelligent management module 500 scans the specified document directory based on keywords in the instruction (such as "project document"), extracts the core paragraphs of the document and generates a summary, which is then fed back to the AI ​​intelligent agent core scheduling module 300.

[0060] Step 204: Results Integration and Emotional Feedback from Virtual Avatars After receiving feedback from the two modules, the AI ​​intelligent agent core scheduling module 300 integrates the data. On the one hand, it uses the emotion recognition results as the basis for the virtual image rendering module 200 to generate virtual image expressions and body movements that match the user's emotions (e.g., when the user is anxious, the virtual image displays reassuring expressions and movements). On the other hand, it prepares the document summary content as display data.

[0061] Subsequently, the AI ​​agent core scheduling module 300 sends display and voice broadcast commands to the virtual avatar rendering module 200, and sends specific data display commands to the visualization linkage output module 700. The virtual avatar provides feedback on the analysis results through voice broadcast (e.g., "We detected that you were a little nervous just now; the key points of the project document have been compiled"), and the visualization linkage output module simultaneously displays the emotion recognition confidence level, document summary, and execution log.

[0062] Step 205: Closed-loop response completed The complete data of this interaction (including user commands, recognition results, feedback content, and execution timestamps) is stored by the AI ​​agent core scheduling module 300 to the memory storage and analysis submodule. The visualization and linkage output module 700 saves the interaction report locally for users to review later.

[0063] Through the collaboration of the visualization-linked output module and the core scheduling module, this system presents the results of user-generated emotional interactions, system execution logs, document summaries, and information content in a standardized panel format, and stores them locally. After completing an interaction, users can review historical reports at any time for analysis and retrospection. Combined with the global task scheduling and data storage capabilities of the core module, a complete proactive interactive loop is formed, from user demand triggering, task scheduling and execution, multimodal presentation of results, to data tracking and reuse. This overcomes the technical shortcomings of existing systems, such as fragmented data presentation and inability to effectively review data.

[0064] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A multimodal virtual avatar AI intelligent agent interaction system, characterized in that, include: The voice wake-up interaction module (100) collects user terminal voice commands and converts them into text data, receives feedback information from the AI ​​intelligent agent core scheduling module (300), and converts it into voice for broadcast. The virtual avatar rendering module (200) matches and switches virtual avatars according to user preferences, and displays interactive content in sync with voice feedback. The AI ​​intelligent agent core scheduling module (300), as the central control unit of the system, parses voice commands, analyzes user historical data and preference profiles, generates execution commands, schedules the work of each module, and stores user interaction and task data; The OpenClaw execution control module (400) receives instructions from the AI ​​agent core scheduling module, executes person recognition and user sentiment analysis tasks, and reports the task status and results back to the AI ​​agent core scheduling module. The terminal document intelligent management module (500) automatically scans multi-format documents in smart terminals, extracts the core content of documents through semantic analysis, marks key paragraphs, and generates summaries and viewing links. The Daily News Generation Module (600) generates personalized daily news reports based on users' historical data, daily activities, and areas of interest. The visualization and linkage output module (700) displays the results of character emotional interactions, system execution logs, key document summaries, and daily news content. The multimodal data acquisition unit (800) acquires raw user image and voice data and provides the underlying data for emotion recognition and person recognition to the OpenClaw execution control module; The AI ​​intelligent agent core scheduling module (300) establishes bidirectional data connections with the voice wake-up interaction module (100), the virtual image rendering module (200), the OpenClaw execution control module (400), the terminal document intelligent management module (500), the daily fresh news generation module (600), and the visualization linkage output module (700), respectively, and issues instructions and receives feedback.

2. The multimodal virtual avatar AI intelligent agent interaction system according to claim 1, characterized in that, The voice wake-up interaction module (100) has a custom wake-up word trigger input terminal and a timed automatic trigger input terminal.

3. The multimodal virtual avatar AI intelligent agent interaction system according to claim 1, characterized in that, After user authorization, the virtual avatar rendering module (200) receives the preference recognition results generated by the AI ​​intelligent agent core scheduling module (300) based on the user's browsing, likes, and collection data from social and content platforms, and matches and switches virtual avatars from a virtual avatar library including real people, cute pets, anime, and cartoons according to the preference recognition results.

4. The multimodal virtual avatar AI intelligent agent interaction system according to claim 1, characterized in that, The visualization linkage output module (700) and the virtual image rendering module (200) output independently of each other; the visualization linkage output module (700) outputs special data and logs to local storage, and the virtual image rendering module (200) outputs the dynamic display content of the virtual image.

5. The multimodal virtual avatar AI intelligent agent interaction system according to claim 1, characterized in that, The AI ​​agent core scheduling module (300) includes a natural language understanding submodule, an instruction generation submodule, a memory storage submodule, and a task scheduling submodule.

6. A multimodal virtual avatar AI intelligent agent interaction method, characterized in that, include: Obtain a startup command, which is triggered by a timed method or a voice wake-up method; According to the start command, the AI ​​intelligent agent core scheduling module (300) reads the user's authorized historical preference data and analyzes the user's current preferred virtual image type based on the historical preference data; The AI ​​intelligent agent core scheduling module (300) sends a switching instruction to the virtual image rendering module (200). Based on the switching instruction, the virtual image rendering module (200) performs virtual image matching for the user according to the virtual image type to obtain the user's current virtual image. The AI ​​intelligent agent core scheduling module (300) sends task instructions to the terminal document intelligent management module (500) and the daily news generation module (600) in parallel. The terminal document intelligent management module (500) extracts key content from the document according to the task instructions. The daily news generation module (600) generates personalized information reports according to the task instructions. The AI ​​intelligent agent core scheduling module (300) receives the key content of the document and the personalized information report, integrates them, and generates an integration result; The AI ​​intelligent agent core scheduling module (300) sends a display instruction to the virtual image rendering module (200), and the virtual image rendering module (200) outputs the integrated result with emotional voice based on the user's current virtual image according to the display instruction; The AI ​​intelligent agent core scheduling module (300) sends output instructions to the visualization linkage output module (700) in parallel. The visualization linkage output module (700) displays the integrated results according to the output instructions.

7. The multimodal virtual avatar AI intelligent agent interaction method according to claim 6, characterized in that, The voice wake-up method includes: the voice wake-up interaction module (100) collects the user's voice command and converts it into text data, and transmits the text data to the AI ​​intelligent agent core scheduling module (300).

8. The multimodal virtual avatar AI intelligent agent interaction method according to claim 6, characterized in that, The timed triggering method includes: automatically generating a start command according to a preset time point.

9. The multimodal virtual avatar AI intelligent agent interaction method according to claim 6, characterized in that, While the AI ​​agent core scheduling module (300) issues task instructions to the terminal document intelligent management module (500) and the daily news generation module (600) in parallel, the method also includes: The AI ​​agent core scheduling module (300) sends an emotion recognition instruction to the OpenClaw execution control module (400); the OpenClaw execution control module (400) performs user emotion recognition according to the emotion recognition instruction and feeds back the recognition result to the AI ​​agent core scheduling module (300). The AI ​​intelligent agent core scheduling module (300) associates the recognition result with the current virtual image of the user obtained by the virtual image rendering module (200).

10. The multimodal virtual avatar AI agent interaction method according to claim 6, characterized in that, The method further includes: the visualization linkage output module (700) associating the emotional voice output content with the corresponding execution log and storing it in local memory; and / or The visualization linkage output module (700) associates the data display content with the corresponding execution log and stores it in the local memory.

Images