Asynchronous social contact method based on anthropomorphic sound expression

Abstract

The invention relates to the technical field of artificial intelligence and edge voice interaction, and particularly discloses an asynchronous social contact method based on anthropomorphic voice expression. According to the method, through the steps of voice collection, semantic analysis, content generation, role scheduling, emotion setting, expression confirmation, designated broadcasting and the like, emotional asynchronous voice communication between users in a personification role is realized. A user can select an expression role and a voice style, and modifies expression content through audition confirmation. The system can generate expression content with mood and personality characteristics in combination with an AI model, supports cross-device broadcasting and role mapping, and is suitable for various edge terminals such as dolls, voice boxes, wearable devices, mobile phones and the like. According to the method, a screen-free and non-account social contact path is constructed, and the method is particularly suitable for temperature expression of children, old people, lovers and special crowds. Compared with the prior art, the emotion bearing capacity and the interaction freedom degree of the social content are remarkably improved, and good universality and commercialization value are achieved.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence, human-computer interaction and edge voice processing technology, and in particular to an asynchronous social method based on anthropomorphic expression and emotional voice reconstruction. Specifically, it is applicable to emotional expression systems embedded in edge voice terminals such as dolls, smart home devices, and wearable devices, and is used to support users to complete asynchronous, non-account-based, and non-screen-based emotional value transmission through "role voice acting". Background Technology

[0002] Current social communication technologies generally employ instant messaging, voice calls, video chats, and social messaging, with information transmission using text, images, or voice as the medium, emphasizing transmission efficiency and real-time feedback. However, these methods have the following shortcomings in terms of emotional expression, device compatibility, and user-friendliness:

[0003] 1. Strong dependence on screen and account system

[0004] Mainstream social interaction methods heavily rely on devices with screens, such as smartphones and tablets, and depend on account login and permission systems. This limits the natural expression and reception abilities of users such as children, the elderly, and special groups, and is also unsuitable for screenless interactive terminal environments.

[0005] 2. Lack of multi-dimensional emotional and personality expression

[0006] Traditional voice messages and text communication struggle to encompass dimensions such as "tone, role, and emotion," resulting in insufficient emotional warmth in social interactions. This is especially true in non-command expressions such as family, friendship, and romantic relationships, where it is difficult to effectively convey the user's intentions and emotional nuances.

[0007] 3. Limitations of existing AI voice device interaction capabilities

[0008] Although there are already smart speakers (such as Amazon Alexa, Google Assistant, and Xiao Ai) and some AI toys, they are mostly focused on command-based interaction and task-based responses, lacking a complete expression chain that supports "free language expression + role control + asynchronous paraphrasing".

[0009] 4. Lack of style control capability in content generation

[0010] While current natural language generation models possess a certain language generation capability, users cannot precisely control the tone, character style, or emotional color of the generated content, nor can they achieve a closed control process of "listening + editing + confirming + broadcasting".

[0011] 5. Lack of asynchronous onomatopoeic expression mechanism among acquaintances

[0012] In familiar social scenarios (such as parent-child interaction, long-distance relationships, and communication between grandparents and grandchildren), users tend to use media with personalized expressions for non-real-time communication. However, current technology cannot systematically support asynchronous expression methods based on edge devices, which combine "personality roles + emotional voice + controllable broadcasting".

[0013] In summary, existing technologies still have significant gaps in constructing screenless, accountless, and asynchronous social mechanisms based on role-based voice and emotion control. They lack a complete expression method that can support "role scheduling + expression control + asynchronous broadcasting + emotion carrying," and urgently need a new social approach that conforms to the characteristics of natural human communication to supplement and break through these gaps. Summary of the Invention

[0014] I. Purpose of the Invention

[0015] The main objective of this invention is to propose an asynchronous social method based on anthropomorphic and voice-based expression, in order to address the technical gaps and interaction barriers in current social methods in areas such as "non-screen device scenarios," "emotional value transmission," "expression mode control," and "asynchronous communication among acquaintances."

[0016] Specifically, the present invention aims to:

[0017] 1. Build a social expression method that does not require a screen or an account.

[0018] By using edge voice acquisition and onomatopoeia relay mechanism, users can express their emotions through voice terminals without relying on mobile phones or social apps. This is especially suitable for natural communication scenarios within families, such as for children, the elderly, special groups, and families.

[0019] 2. Supports flexible scheduling of anthropomorphic characters and voice styles.

[0020] Users can choose their own narration character (such as a doll, virtual character, voice assistant, etc.) and listen to various tone styles and emotional combinations to ensure that the content matches the context and enhances the "warmth" and personalization.

[0021] 3. Establish a closed loop for controllable confirmation and asynchronous release of expressed content.

[0022] It provides natural language modification, semantic style adjustment, and audio confirmation mechanisms, and authorizes release only after user confirmation, ensuring the privacy, accuracy, and emotional consistency of expression.

[0023] 4. Achieve a non-real-time, non-intrusive, and gradual response social mode.

[0024] By using asynchronous messaging and role-based broadcasting, a new mechanism suitable for slow-paced communication and stress-free social interaction is built, enhancing the freedom and initiative of expression.

[0025] 5. Provides the ability to reconstruct both content generation and expression style.

[0026] By combining AI language models and speech synthesis technology, it can automatically generate content and adjust the expression method to achieve a two-way reconstruction of "content + style", giving the machine "emotional expression ability".

[0027] 6. Build a universal interaction framework that can be embedded in multiple types of edge devices.

[0028] This method is compatible with a variety of edge smart terminals, including but not limited to dolls, wearable devices, smart home devices, and voice boxes, and has strong versatility, scalability, and commercialization capabilities.

[0029] In summary, this invention constructs a novel asynchronous social paradigm driven by user intent, centered on emotional expression, and carried by edge devices. It is particularly suitable for scenarios such as children's educational interaction, elderly companionship and care, family emotional communication, daily communication between couples, and assisted expression for special groups, and has significant social value and commercial application prospects.

[0030] This invention relates to the field of artificial intelligence voice processing and edge interaction technology, specifically an asynchronous social method based on anthropomorphic voice expression. It aims to construct a new social model that is independent of traditional screen interfaces and social account systems, with "emotional value transmission" at its core. This method is suitable for various user types, including children, the elderly, and adults, and particularly focuses on gentle expression, warmth transmission, and emotional interaction in close-knit social interactions.

[0031] This invention achieves synchronous transmission of information content and emotional expression by constructing a complete asynchronous behavior chain of "role scheduling → emotional style expression → user confirmation → designated role broadcasting", thus breaking the problems of low temperature, overload or accidental delivery caused by traditional voice messages or instant messaging.

[0032] II. Specific technical approaches include: By leveraging AI language models or edge language processing capabilities, "onomatopoeic expression content" with character personality traits, tone style, and contextual understanding can be generated. Users can use natural language commands or interface options to assign roles ("who speaks"), set emotional style ("how to speak"), specify receiving device ("to whom to speak"), and listen to and confirm the content, modify the semantics, and fine-tune the style. The expression process is completed by edge speech intelligence agents or terminal devices that support voice interaction, including voice acquisition, semantic parsing, content generation, listening confirmation and final broadcasting, forming a controllable, traceable and warm expression closed loop. The system can also support the automatic generation of expressive content, such as story creation, blessing apologies, and holiday interactions, enabling AI to not only play the role of "narrator" but also have the ability of "emotional content generator".

[0033] Furthermore, this invention explicitly emphasizes that all content must be confirmed by the user before it can be played, and allows switching between different emotions and character tones during the listening phase to enhance the personalization of the expression and avoid semantic miscommunication or emotional deviation.

[0034] To ensure that the technical protection of this invention is not circumvented by functional splitting, form substitution, or entry point replacement, the following equivalent protection statement is hereby made: The terms "anthropomorphic characters" and "onomatopoeia" are functional terms and are not limited to specific brands, voice models, or physical forms. The method does not limit the input method (voice, text, touch, etc.), content generation method (AI large model, edge model, etc.), output device type (doll, speaker, mobile phone, wearable device, etc.) or the source of the synthesized model; Any interactive chain that fully realizes the process of "user intent → onomatopoeia expression generation → style audition and scheduling → content confirmation → designated role broadcasting" is considered an equivalent technical solution of this invention. Even if the functional module is implemented by multiple API services, application systems or hardware devices, as long as it completes the above-mentioned behavioral path, it still falls within the scope of protection of this invention.

[0035] Through the above approach, this invention reconstructs the technological foundation of social expression and the paradigm of emotional interaction, possessing broad user adaptability, cross-scenario deployment capability, and strong reusable social value potential.

[0036] III. Specific Implementation Process

[0037] 1. Edge wake-up and voice acquisition

[0038] Users can wake up the edge agent through voice, touch, or sensor operation. The device then activates the voice acquisition module to identify the user and obtain the voice input.

[0039] 2. Semantic parsing and intent recognition

[0040] The system uses a local or cloud-based semantic understanding module to perform natural language processing on user input, extracting key information parameters such as expression intent, emotional direction, target audience, and context.

[0041] 3. Auditory test of character portrayal and tone of voice.

[0042] Users can specify the voice actor (such as a particular doll or virtual character), and the system provides multiple voice models and emotional styles for each character, supporting previewing and switching. Users can also fine-tune the style using natural language commands (such as "make the tone gentler").

[0043] 4. Confirmation and authorization of role announcements

[0044] Users confirm the content based on the listening results and specify the final broadcasting role and voice style. The system records the confirmation behavior and encrypts and authorizes the broadcasting path to ensure controllability of the broadcasting target and content.

[0045] 5. Generation of onomatopoeic expression and synthesis of emotional style

[0046] The system invokes a language generation model to generate spoken content based on the parsed intent and character style parameters. Combining this with a specific character tone database and emotional intonation templates, it produces personalized speech synthesis output.

[0047] 6. Asynchronous broadcasting and interactive feedback at the receiving end

[0048] The receiving device wakes up a designated person to broadcast the message within a set time. Users can provide feedback on the broadcast results, such as "Put it another way" or "I like this version." The system records interaction preferences and dynamically optimizes the expression strategy.

[0049] 7. Interactive closed-loop recording and expression history management

[0050] The system records the entire expression process, including content generation path, role configuration, user confirmation records, and receiver feedback. Historical expressions can be viewed within authorized scope, supporting continuous dialogue, personalized evolution, and the construction of social memories.

[0051] IV. Technical Solution

[0052] 1. System Structure Overview

[0053] To more clearly understand the composition and structure of the asynchronous anthropomorphic social method involved in this invention, the main modules and functional paths of this system are summarized in tabular form as follows:

[0054] 2. System Composition and Functional Module Description

[0055] The system of this invention consists of the following functional modules, which operate collaboratively through a communication network to form a complete asynchronous anthropomorphic social link. This supports a full-process emotional value transmission mechanism, including expression collection, role-based retelling, content confirmation, and asynchronous broadcasting. The functions of each module are as follows:

[0056] 1) Edge voice acquisition module

[0057] This module is used to receive user voice input, identify user identity and wake-up intent, and complete initial command triggering. The edge voice input module collects user expression information, including but not limited to voice or text input. Text input can be completed via mobile terminals (such as mobile phones and tablets) and supports adding emotion tags, tone prompts, or role preference parameters. Voice and text input have equivalent processing paths in this system and can both serve as input sources for generating onomatopoeic expressions. This module can be deployed in edge intelligent terminals such as dolls, wearable devices, and smart home devices, supporting voiceprint recognition, command recognition, and specific group recognition (such as parents, the user, and friends), achieving a natural, non-screen-based interaction interface.

[0058] 2) Semantic parsing and intent recognition module

[0059] This module is used to upload collected voice input to the cloud or parse it at the edge, extracting key parameters such as interaction intent, target recipient, specified role, and content type. It can also structure and reorganize instructions to trigger subsequent content generation or invocation logic.

[0060] 3) Personification and Emotional Style Management Module

[0061] This tool is used to maintain multiple virtual characters (such as panda dolls, dinosaurs, Barbie dolls, etc.) bound to users, along with their corresponding voice models, tone styles, emotional parameters, and other asset information. It supports user selection of "who speaks" and "how to speak," and also supports emotional tone switching and audio-visual comparison mechanisms.

[0062] 4) Content generation and stylization module

[0063] This module is used to invoke AI language models or content generation engines to reconstruct, polish, or emotionally reshape user-requested content, generating expressions that conform to anthropomorphic roles and semantic requirements. It also supports user-defined semantic-level modifications, such as natural language editing commands like "a gentler ending" or "a more humorous way of saying it."

[0064] 5) User Expression Confirmation and Release Control Module

[0065] This feature is used to play back the generated content to the user after it has been generated, and to complete the closed-loop control of the presentation through user confirmation, modification, and assignment of content. It supports switching between different roles to preview the presentation style, and can also simulate the other role's speaking style for the user to judge its suitability.

[0066] 6) Receiver broadcast execution and feedback module

[0067] This module is used to relay asynchronous content received by a user through a unique voice style and emotional delivery. It supports interaction with the user, including user feedback, emotional responses, requests for replay, and reverse comments, forming part of the social loop.

[0068] 7) User Relationship and Role Mapping Management Module

[0069] It is used to manage the social graph between users and the mapping relationship between roles (such as "I am friends with Yuanyuan's Barbie doll"). It supports specifying the broadcasting role, controlling social permissions, and switching social channels to realize a controllable information transmission mechanism between multiple roles.

[0070] 8) Content Log and Expression History Management Module

[0071] It is used to record all expression requests, role selections, content confirmations, broadcast trajectories, and interactive feedback. It supports functions such as user review, content export, and long-term social relationship maintenance, and has privacy encryption and information filtering capabilities.

[0072] For further expansion, this system can also be combined with a scheduled task module, a scene awareness module, etc., to realize more personalized social expressions such as "scheduled birthday wishes" and "determining whether the recipient is available".

[0073] 3. Technical Solution and Algorithm Description

[0074] To enhance the technical feasibility of this invention and its ability to prevent bypassing patent protection, further explanations are provided regarding its key process paths and some algorithmic mechanisms:

[0075] 1) Overview of the technical path for user expression generation process

[0076] The asynchronous social method of "anthropomorphic and phonetic expression" described in this invention can be modeled as the following typical technical process chain: Voice input → Wake-up recognition and identity verification → Content parsing and expression intent extraction → Personification and style matching recommendation → Expression content reconstruction and audition → User confirmation and role authorization → Asynchronous broadcast execution → Recipient feedback and history archiving Each step is linked with other modules through structured data interaction. Input / output parameters are encoded using standardized structures such as user ID, role ID, tone tags, content vectors, and emotion parameters, ensuring that the overall path can be deployed in a collaborative environment of edge + cloud.

[0077] 2) Anthropomorphic expression style adaptation scoring algorithm (example)

[0078] To achieve more accurate "role + style" recommendations, this invention can introduce the following expression style scoring function:

[0079] in: - r: Candidate broadcasting role; - e: The emotional label represented by the current intention (such as "gentle blessing" or "humorous banter"); - u: The user who initiated the request; - M(r, e): The degree of matching between the emotional fitting ability of character r and e (between 0 and 1). - P(u, r): Frequency of user preference for this role in history (e.g., the percentage of times it was selected in the last 30 uses); - S(e, r): Onomatopoeic style similarity between keywords and character style in the context of expression; - w1, w2, w3: Adjustable weighting parameters that are dynamically adjusted based on business debugging.

[0080] Based on the above scoring mechanism, the system will select the 1 to 3 highest-scoring characters from the character library currently available to the user for listening and comparison.

[0081] 3) Natural Language Style Reconstruction Mechanism (Rule Engine Example)

[0082] Users can adjust their expression style via natural language commands, such as "speak more gently," "speak in a child's tone," or "start with humor." Multiple user command keywords are supported, each triggering a corresponding action, including but not limited to: "Gentle": Trigger action: When the system broadcasts or outputs, it adds or enhances intimate emotional wording while slowing down the speech rate to create a gentle expression effect; "Childlike tone": Triggering operation: The system calls the child voice model and replaces the vocabulary with childlike expressions, so that the output presents the tone and style of a child speaking; "Make it even more concise": Triggering operation: The system calls the summarization algorithm to compress the content, making the expression more concise and clear.

[0083] The rule engine is based on multiple rounds of interaction between keywords, user context tags, and the current content draft, enabling a controllable and auditionable style fine-tuning process.

[0084] 4) Asynchronous broadcast path security confirmation mechanism

[0085] To avoid mistransmission of user information, this invention designs the following broadcast authorization mechanism: Users can preview the target broadcast content through a humanoid terminal; The broadcasting character can simulate the receiver's character style to create a "simulated broadcast" preview (such as a "giant panda imitating Peppa Pig" broadcasting style). Users grant final authorization via "confirmation voice" or "edge button"; The system generates a unique expression ID and timestamp for each expression and registers its publication.

[0086] This mechanism ensures the integrity and controllability of the expression path, constructing a closed-loop mechanism of "expression confirmation → designated broadcast → traceability".

[0087] 5) Personification Role Priority Scheduling Strategy (Recommended Rule Illustration)

[0088] When the user does not explicitly specify the broadcasting role, the system will recommend based on the following priority strategy by default: User's most frequently selected role in history (frequency priority); The character that best matches the current emotion tag (tone of voice is the priority); The person with the closest social relationship to the recipient (intimacy priority); Match the role to the current time period / holiday scenario (scenario adaptation takes priority); Allows users to click to listen to a preview and then replace the recommended character.

[0089] This strategy dynamically combines system decisions with user preferences, avoiding the disconnect between forced recommendations and user preferences.

[0090] For further expansion, the system can incorporate reinforcement learning methods to train the recommendation model multiple times on user expression preferences and emotional style feedback data, thereby dynamically optimizing the model. However, this invention is not limited to any specific algorithm implementation; all system implementations built upon the logical chain of "emotion recognition → style adaptation → anthropomorphic expression → user control" fall within the scope of protection of this invention.

[0091] V. Explanation of Model Usage and Service Deployment Mechanism

[0092] To achieve the semantic understanding, emotion expression generation, language style optimization, and speech synthesis functions described in this invention, the method adopts a collaborative approach of "edge acquisition + cloud generation." In the cloud, one or more third-party or open-source large language models are invoked through standard service interfaces to construct the generation module required for human-like expression. Specific details are as follows:

[0093] 1. Non-advocacy statement regarding model usage: The natural language generation engine, emotion-based speech synthesis module, and language style transfer module invoked in the method described in this invention can be implemented based on open-source or third-party large language models (such as GPT, ChatGLM, BLOOM, ERNIE, etc.) or enterprise-developed models. These models are only used as service modules for the generation task; their model training structure, parameter details, and algorithm implementation paths do not constitute the scope of protection of this invention, and this invention does not advocate for the protection of the model's structure or training method itself.

[0094] 2. Multi-model collaborative adaptation mechanism: The method of this invention supports the invocation of multiple AI models to construct an "expression service matrix," completing different sub-tasks as needed, including but not limited to semantic understanding, tone transfer, emotional style generation, cross-language translation, and expression output. The models can be deployed in parallel or sequentially, and parameter invocation and result fusion are uniformly coordinated through the system's internal task scheduling logic, thereby achieving consistent style and matching emotional expression effects.

[0095] 3. Distributed Deployment and Service Flexibility: The expression generation module can be deployed in public cloud, private cloud, or edge computing environments. Model services can be accessed via API interfaces or local inference services. This method does not rely on a specific model provider or model framework, exhibiting neutrality and compatibility, and supports the access and collaborative execution of different types of large language models. As long as the calling model has natural language generation and expression capabilities, it can be accessed and constitute the expression generation path of this method.

[0096] 4. It should be noted that, in order to achieve the above-mentioned functions such as semantic understanding, emotion expression generation, language style optimization and speech synthesis, this method adopts a collaborative path of "edge acquisition + cloud generation". In the cloud, one or more third-party or open-source large language models are called through standard service interfaces to build the generation module required for human-like expression.

[0097] To avoid ambiguity, the natural language processing model, emotion speech synthesis model, or tone style transfer model involved in this invention are only accessed at the application layer of this system in the form of API calls or inference services. They do not involve the design, training, fine-tuning, compression, deployment optimization, or other processes of the model ontology, nor do they depend on specific language model frameworks (such as Transformer, LLM-specific frameworks) or specific training data corpus structures.

[0098] The scope of protection of this invention focuses on the expression generation path, service scheduling mechanism and emotion expression logic in asynchronous voice communication, rather than the technical details of the model itself.

[0099] 5. The system of this invention supports the simultaneous invocation of multiple AI model modules with different capabilities to construct a "multi-model collaborative expression service chain." This includes: invoking model A for semantic intent understanding, invoking model B to generate neutral text, invoking model C for speech style transfer, and invoking model D to synthesize emotionally charged broadcast speech; each model can be deployed in parallel or serially, and through the scheduling and fusion logic of this system, consistent style and coherent output of expressive content are achieved; each model can be replaced independently, and the system logic is not bound to a specific model or inference architecture.

[0100] VI. Summary of Invention Content and Explanation of Method Advantages

[0101] This invention proposes an asynchronous voice-based social networking method based on anthropomorphic expression. Around the design concepts of "temperature transmission, emotion expression, role-playing, and non-account communication," it constructs a complete methodological path covering key aspects such as expression generation, style optimization, broadcast control, and permission configuration. Compared to traditional voice interaction or social communication methods, this method has significant innovations in emotional expression capabilities, freedom of role-playing, and adaptability. Specific advantages include:

[0102] 1. Construction of an asynchronous emotion expression mechanism

[0103] Unlike traditional voice systems that aim for real-time response, this invention supports users in delivering anthropomorphic voice information asynchronously through edge voice acquisition, semantic parsing, and emotional style generation, making the expression more warm and personalized.

[0104] 2. Expression path scheduling with multiple roles, tones, and terminals.

[0105] This method constructs an expression path of "user → anthropomorphic character → emotional style → designated receiving character → terminal broadcast", allowing users to specify the speaking character, tone style and target terminal, realizing rich expression combinations, and is applicable to various voice terminals such as children's dolls, elderly voice boxes, and wearable devices.

[0106] 3. The expressive closed loop of content generation and style reconstruction

[0107] Users can adjust their expressions using natural language (such as "more concise" or "more gentle"). The system automatically performs semantic reconstruction and provides audio confirmation to ensure that the expression is both semantically correct and emotionally appropriate, forming a complete closed-loop process of "generation → optimization → confirmation → broadcast".

[0108] 4. Adapting permissions and oversight mechanisms to suit special population groups

[0109] The method of this invention is applicable to special groups such as children, autistic patients, and people with language disorders. It supports guardians or educators in setting content, confirming broadcasts, and managing permissions, thereby enhancing the safety and adaptability of the method in the fields of education and rehabilitation.

[0110] 5. Global adaptability of multilingual broadcasting and cross-cultural expression

[0111] This invention also supports multilingual broadcasting, multi-timezone asynchronous scheduling, and cross-cultural expression style adaptation mechanisms, making it suitable for voice expression needs under different language environments, cultural backgrounds, and time zone settings. The method achieves dynamic adjustment of the anthropomorphic voice content in terms of language, tone, speed, and expression style through comprehensive processing of parameters such as the receiver's language preferences, character style preferences, and broadcast time periods. This results in global adaptability, making it suitable for building universal asynchronous social methods for cross-regional and multicultural users.

[0112] 6. A marginal expression paradigm that is de-account-based and de-screen-based

[0113] Users can generate and receive content by binding their nicknames and using voice commands, without relying on an account system or graphical interface. This method lowers the barrier to entry and is particularly suitable for children, the elderly, and users who do not wish to use traditional social media software.

[0114] 7. A social expression philosophy driven by emotional value.

[0115] This method redefines the goal of voice-based social interaction—not just "information transmission," but also "emotional communication." Through role-based delivery and controlled tone of voice, it imbues expression with more human warmth, fostering deeper understanding and companionship, and returning to the essence of interpersonal communication.

[0116] VII. Summary of Technical Effects and Invention Contributions

[0117] This invention revolves around the core concept of "anthropomorphic expression + emotion-driven + edge interaction + asynchronous social interaction," achieving key innovations in multiple dimensions such as the expression path, interaction mechanism, and emotion modeling in voice-based social interaction. It has the following technical effects and inventive contributions:

[0118] 1. Construct a closed-loop mechanism of "expression-confirmation-broadcasting".

[0119] The system innovatively proposes a three-stage expression process of "role scheduling + tone preview + content confirmation" to ensure that users can preview, adjust, and authorize before expressing themselves, thus guaranteeing freedom of expression and broadcast security from the source.

[0120] 2. To form a new paradigm of social expression based on emotional value.

[0121] Unlike traditional information transmission models, this invention focuses on "warm expression," emphasizing the emotional resonance of tone, role identification, and expression process, thus constructing a new social path that is full of companionship and emotion.

[0122] 3. Empower edge agents with genuine social expression capabilities

[0123] The system is compatible with various smart terminals with voice capabilities (such as dolls, watches, voice boxes, etc.), enabling them not only to listen and speak, but also to express emotions and assume roles, thus becoming an "agent" for emotional social interaction.

[0124] 4. Freedom of expression with multiple voice roles and tones

[0125] Users can freely set "who speaks" and "how to speak". The system supports control methods such as tone preview, role style switching, and fine adjustment of speech speed and tone, making the expression process highly personalized and adjustable.

[0126] 5. Supports natural language content modification and expression reconstruction

[0127] Users can modify the content structure through verbal commands (such as adding or deleting greetings, adjusting the order, etc.). The system optimizes the expression logic through semantic understanding and content rewriting mechanisms, making the content more in line with the context and the expectations of the audience.

[0128] 6. Highly adaptable to the needs of special populations and people with low interaction levels.

[0129] This system can serve users such as children, the elderly, and people with autism. It is equipped with a parental supervision interface and a content confirmation mechanism, providing a safe, gentle, and effective channel for users with limited expressive abilities to express their emotions.

[0130] 7. Breaking through the traditional decentralized interaction paradigm that relies on accounts / screens.

[0131] The system does not rely on APP login or account binding. Users can express and receive information through nicknames and voice operations, truly achieving a low-threshold social experience that is "account-free, screen-free, and undisturbed".

[0132] 8. Establish a complete mechanism for managing expression permissions and securely releasing information.

[0133] All voice messages must be confirmed by the user before being broadcast. Broadcast permissions, timing policies, and sensitive word filtering can be configured to avoid mistransmission of information and enhance privacy protection.

[0134] 9. Possesses global adaptability in multilingual, multi-timezone, and cross-cultural expression.

[0135] The system supports multilingual broadcasts and automatic language switching, and dynamically adjusts the expression style based on the user's time zone and cultural preferences, realizing a universal emotional social path across different cultural backgrounds worldwide.

[0136] 10. Open ecological collaboration capabilities and business expansion potential

[0137] This method and system do not rely on specific hardware forms or large model platforms, and support adaptation and connection with any speech synthesis engine, edge device or IP image, and have cross-industry and cross-platform deployment flexibility and cooperation potential.

[0138] VIII. Applicable Scenarios and Universality of the Method of the Invention

[0139] The "Asynchronous Social Method Based on Anthropomorphic Voice Expression" provided by this invention has high versatility and flexible embedding capabilities, making it suitable for various social scenarios that require conveying emotions, constructing character expressions, and supporting asynchronous voice communication. The method is not dependent on specific device types or communication architectures and supports key processes such as expression request collection, semantic parsing, expression generation, and character broadcasting in intelligent carriers with voice capabilities. It is applicable to, but is not limited to, the following typical scenarios:

[0140] 1. Children's Voice Social Interaction and Content Retelling: Suitable for children of the same age to share essays, interact with stories, and read aloud in person, thereby improving their willingness to express themselves and their ability to perceive emotions.

[0141] 2. Role-based remote messaging between parents and children: Suitable for parents to convey reminders, encouragement and other non-real-time content to their children, with preset tone and roles to complete the expression and relay.

[0142] 3. Receiving Heartwarming Messages for Seniors: Suitable for elderly users to receive messages from relatives via voice terminals, enhancing comprehension and emotional connection through emotional voice.

[0143] 4. Emotional communication between couples: Suitable for various asynchronous expression needs with emotional overtones, such as holiday greetings, daily greetings, apologies and reconciliations.

[0144] 5. Assisted communication for groups with limited language expression: Applicable to users with autism, autism spectrum disorder, and language impairment, it achieves alternative expression training and communication through a combination of delegated tasks and role-playing.

[0145] 6. Edge voice social scenarios such as home, medical, and in-vehicle: This method can be flexibly deployed in any carrier with voice interaction capabilities, supporting accountless, non-real-time, and personalized social communication methods, and is suitable for low-frequency, high-emotional, non-intrusive expression scenarios.

[0146] The method of this invention emphasizes the structural universality of the expression path and the emotional adaptability of the role expression. It can be combined with any open model system that supports natural language understanding and speech synthesis to realize service implementation, and has good cross-industry deployment capabilities and ecosystem integration potential. Attached Figure Description

[0147] Figure 1 : Schematic diagram of the overall system structure of the present invention.

[0148] Figure 2 : Flowchart of the method of the present invention.

[0149] Figure 3 Flowchart for confirming and fine-tuning the content.

[0150] Figure 4 Asynchronous push and receiver broadcast process diagram.

[0151] Figure 5 A typical diagram illustrating the character broadcast mapping.

[0152] Figure 6 : Expression, recording, and interactive closed-loop logical structure diagram.

[0153] Figure 7 : Schematic diagram of multi-model collaborative adaptation mechanism. Detailed Implementation

[0154] To make the objectives, technical solutions, and beneficial effects of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. This invention is not limited to the specific embodiments described below; any equivalent substitutions or improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0155] Example 1: Open Invocation of Model Services and Collaborative Execution of Multiple Models

[0156] This embodiment illustrates how the method of the present invention can schedule and call various open-source or third-party AI models during the expression generation stage, and demonstrates the platform characteristics of the method being decoupled from specific model structures and having neutrality and openness.

[0157] In a practical application, a family user wants to send a reminder message to a distant family member using an anthropomorphic, voice-activated tone, conveying a gentle, elder-like voice and a regional dialect. The user inputs the message via an edge voice terminal: "Remember to take your medicine today, don't forget." The system invokes the following model services during the execution of each step of the method described in this invention:

[0158] Semantic understanding and target recognition: The open-source large language model ChatGLM-3 was used to perform semantic understanding on the user input, and the target was identified as "elderly reminder", and the keywords were extracted as "take medicine", "today" and "remember".

[0159] Emotion and tone synthesis: The system calls the speech style transfer model FastSpeech-2 + emotion adjustment module based on the user's settings of "gentle tone" and "elderly style" to automatically synthesize a gentle elder tone.

[0160] Language dialect processing: If the user specifies "Sichuan dialect", the system will call another model, such as "BLOOM + local dialect adaptation module", to complete the natural language translation.

[0161] The aforementioned multiple models are processed collaboratively in the cloud by the task scheduling logic of this method. After the output content is confirmed by the user through listening, it is asynchronously pushed to the receiving device, where the role bound to the device performs the broadcast.

[0162] This embodiment fully illustrates: The method of this invention does not rely on a specific AI model and can flexibly select third-party or open-source model services to complete the expression task; The method of this invention does not involve model training process, parameter path or algorithm structure; the model is only for expressing service interface. The system can continuously integrate new models, replace old models, or deploy multiple models in parallel to complete the expression chain without affecting the essential execution logic of the method. This method is widely applicable to different languages, contexts, and style requirements, and has a high degree of openness and business adaptability.

[0163] Example 2: Implementation Path of Expression Style Adaptation Algorithm Mechanism

[0164] In a typical embodiment of the present invention, expression style adaptation can be achieved through various algorithmic mechanisms. Several feasible implementation methods are listed below to illustrate the technical feasibility of the "expression style adaptation" mechanism described in the present invention, but this does not limit the present invention to necessarily employing the following specific algorithmic paths:

[0165] 1. Style matching algorithm based on scoring function

[0166] The system maps the user-defined target emotions (such as gentleness, humor, formality, etc.) to multiple style parameter dimensions (such as speech rate, pitch, pauses, emotional intensity, etc.), calculates the similarity with the emotion templates in the character voice model library, and uses a weighted scoring function to select the optimal match as the current expression style.

[0167] The formula is as follows:

[0168] Among them, S i w represents the matching score of the i-th candidate style. k P represents the weight of the k-th dimension. k For the target parameters input by the user, T ik Let be the attribute value of the i-th candidate style in the k-th dimension, and sim be the similarity function.

[0169] 2. Style adaptation logic based on a rules engine

[0170] The system pre-sets several rule combinations (e.g., "target is children → broadcast tone is lively + slow"), and matches the corresponding expression style template based on information such as the user's expression context, the recipient's identity, and the topic of the expression, achieving rapid style adaptation. The rules can be continuously expanded or optimized by user training, and are interpretable and adjustable.

[0171] 3. Neural Network-Based Emotion Transfer and Style Generation Algorithms

[0172] The system invokes a lightweight language generation model, taking the user's original text content and a target style prompt as input, and outputting an automatically adjusted tone and style text to achieve emotion transfer and semantic reconstruction. For example, "Good night" is automatically rewritten as "You worked hard today, get some rest, good night~" to match a "caring + gentle" style.

[0173] 4. Personalized expression backtracking based on vector retrieval

[0174] The system can encode users' historical expressions using style vectors and build a personalized preference database. The current expression request can be used as a query vector to recall similar expressions and perform style transfer, enabling the reuse of "analogical" expression styles.

[0175] Any of the above-mentioned algorithm mechanisms, or combinations thereof, can serve as a feasible implementation path for the expression style adaptation described in this invention, as long as the following chained processing can be completed: Perform sentiment semantic analysis on the user's input; Match the target style parameters with the character's voice model; Output anthropomorphic voice expressions that meet style requirements; It also completes the user's listening confirmation and authorization broadcasting operations.

[0176] This means that the algorithm falls within the protection scope of this invention. This embodiment is merely an illustrative example; the algorithm can run locally on an edge device, be processed collaboratively in the cloud, or be embedded as a third-party model service. The specific implementation does not affect the core technical solution and protection boundary of the method of this invention.

[0177] Example 3: Asynchronous Interactive Social Interaction among the Elderly – Edge Intelligence Agents' Emotional Expression Facilitates Communication Among Acquaintances

[0178] This example describes how three elderly users communicate asynchronously in daily life through their respective edge intelligence agents, forming a gentle, interesting, and non-intrusive daily social network.

[0179] Background setting: Old Zhang: It uses a smart medicine box with voice broadcast function. The broadcaster's voice is that of a standard Mandarin announcer, with a steady and clear tone. Old Li: I'm using a smart monkey doll that my grandson gave me. It mimics a playful and lively child's voice with a cheerful and bouncy tone. Aunt Zhao: The smart watch she's wearing is a gift from her daughter. The voice is modeled after the cartoon character "Doraemon," which is friendly and has a slightly mechanical and cute feel.

[0180] All three smart devices are bound to the edge social system of the platform of this invention, and are bound to friends through identity recognition to form a group of acquaintances.

[0181] Morning scene: Sharing information about grocery shopping

[0182] Aunt Zhao spotted an egg sale at the market this morning and excitedly told her watch, "Doraemon, tell Lao Zhang and Lao Li that eggs are cheap at the market! Two cartons per person, come quick!"

[0183] The system process is as follows: The edge voice module recognizes Aunt Zhao's commands; The content is uploaded to the cloud, and keywords and tone (happy + anxious) are extracted by the semantic parsing module. The content generation module generates suitable broadcast content and automatically assigns it to Lao Zhang and Lao Li.

[0184] Old Zhang's medicine box broadcasted within 30 minutes: "Aunt Zhao says: Eggs are on special offer, limited to two boxes per person, would you like to go check it out?"

[0185] Old Li's little monkey toy plays as he gets up and moves around: "Hehe, Aunt Zhao found out there's an egg sale! She said hurry up and buy some, they're going to sell out fast! She's so impatient!"

[0186] Midday scene: Request to purchase medicine

[0187] After lunch, Lao Zhang opened his medicine box and found that he was almost out of medicine. He said to the medicine box, "Please tell Lao Li to bring me a box of metoprolol when he comes back from playing ball and stops by the pharmacy. Thank you."

[0188] System processing flow: The edge module of the medicine box captures voice. The AI ​​content generation module identified it as a "delegated behavior" and added the tone "please + calm"; Generate broadcast content that matches the style of the Lao Li doll; After Old Li finished eating, the little monkey played a recording: "Grandpa Zhang said: 'Could you please bring me a box of medicine on your way back from playing ball? He's almost out of medicine, it's called Metoprolol!'"

[0189] Evening scene: Dance invitation

[0190] Around 6 p.m., Aunt Zhao said through her voice watch, "Doraemon, tell Lao Zhang and Lao Li that we're going to dance in the square at 7 p.m., see you there!" The system identifies "invitation events" with emotions of "happiness and enthusiasm" and processes them into broadcast-style mass messages.

[0191] The broadcast results are as follows: Old Zhang's medicine box voice message (standard male voice): "Aunt Zhao sent a message: Square dancing tonight at 7 pm, inviting you to dance together, please be there on time." The little monkey's voice message: "Wow, Aunt Zhao is calling you to dance! It's at 7 PM tonight. If you don't come, I'll tell her you're slacking off!"

[0192] Overall technical process path: The edge voice acquisition module recognizes commands; The semantic recognition and sentiment analysis module extracts user intent and sentiment factors; The friend relationship matching and push scheduling module selects targets based on the relationship network; The onomatopoeia content generation module generates language content that matches the tone and character style; The edge device broadcast module automatically broadcasts information asynchronously during user scenarios (such as after waking up or after a meal).

[0193] Technical effects: Provide seniors with a warm social channel that is "non-immediate and low-interference"; By using anthropomorphic characters and emotional delivery methods, the receiver's sense of identification and responsiveness can be enhanced; Achieve highly adaptable content generation and broadcast style matching; Supports cross-device and cross-role voice relay, breaking the limitations of device type and voice style; It realizes a new slow-paced social approach based on "acquaintances + non-account system + asynchronous interaction".

[0194] Example 4: Intelligent broadcasting and interactive tutoring process for elementary school math teachers assigning homework

[0195] Background setting: Ms. Li is a third-grade math teacher. To improve her students' enthusiasm for doing homework, she uses a teacher-specific app to record homework assignments via voice after school: "Today's homework is on page 32 of the workbook, questions 3, 4, and 5. Please read the questions carefully. If you don't know how to do them, you can ask in class tomorrow. Finish early and get some rest tonight."

[0196] The app automatically uploads voice recordings to a cloud-based voice-based social platform.

[0197] The system converts the audio content into standard text, adapts the content for the role, and automatically matches the broadcast language, tone, intonation, and interactive scripts based on the preferences of each student's bound edge agent role.

[0198] Examples of personalized broadcasting and interaction for edge intelligent agents (ten roles):

[0199] 1. Teacher Rabbit (Personality: Gentle and patient)

[0200] "Honey, Teacher Li assigned three questions on page 32 today. Shall we look at them together? You're the best! I'll give you a bunny sticker when you finish!" → Encouraging gentleness awakens task awareness.

[0201] 2. Iron-skinned Sheriff (Personality: Strict and upright)

[0202] "Listen, kid, I've received orders for today's homework. The task is to complete questions 3-5 on page 32. If you want to play mech battles with me, finish your homework first!" → Task-based incentive with reward.

[0203] 3. Uncle Fatty Bear (Personality: Caring)

[0204] "Sweetie, are you feeling cold today? Have some water first, then sit up straight, and we'll start doing your homework. Remember to sit up straight, don't slouch, Uncle Bear will watch over you." → Behavioral guidance + health reminder type.

[0205] 4. Barbie Sister (Personality: Intellectual Companion)

[0206] "My dear child, tonight we're going to tackle three challenging thinking problems, specially arranged by Teacher Li. I'm so proud of you that you were able to solve them on your own!" → Emotionally encouraging companionship.

[0207] 5. Little Doctor Robot (Personality: Knowledge-based)

[0208] "The problem has been explained. If you don't understand 'two-digit number multiplied by one-digit number,' I can provide three demonstrations. Shall we begin tutoring?" → Automatically activates problem explanation + guided interactive Q&A.

[0209] 6. Magic Cat (Personality: Mischievous and Encouraging)

[0210] "Hehe, I've secretly hidden the questions in your head. Write them down to prove I wasn't lying! If you don't do your homework, I won't be doing magic tonight!" → Gamification incentives + role-playing gameplay.

[0211] 7. Forest Grandma (Personality: Kind Supervisor)

[0212] "Honey, the sun's setting. Finish your homework before you play. Grandma knows you're smart. Shall I read the question to you?" → Emotionally comforting call task completed.

[0213] 8. Little Frog Senior (Personality: Self-disciplined role model)

[0214] "Hey, I just finished my homework and passed the challenge! You should also try these three questions, and we'll see who does better tomorrow!" → Peer-based encouragement.

[0215] 9. Sky Fox (Personality: Gentle and Reminder)

[0216] "Are you sitting up straight? Is the lighting good enough? Be careful to protect your eyes. Oh, and there's a little task for you today. Are you going to start your homework?" → Behavioral reminders + supportive adjustment of the learning environment.

[0217] 10. Xiaozhi AI Assistant (Personality: Interactive Q&A)

[0218] "Are you ready to start the homework? I can explain each question and check your thinking. Shall we start question 3? If you want to review the multiplication tables, please let me know." → AI-assisted explanation module activated.

[0219] Complete technical process chain: Teachers record and upload audio: Teachers input voice through a terminal (APP or voice box) and upload the recording to the cloud server for storage and processing; Content semantic extraction: The system calls an AI model to perform semantic understanding on the uploaded audio content, extract homework or teaching content, and generate an illustration of the intention to assign the assignment. User profile retrieval: The system retrieves the corresponding personalized profile data based on the doll or edge intelligence agent role information bound to each student, so as to generate personalized content in the future. Personalized content generation: The system generates multiple versions of broadcast voice content based on student role personality and emotion templates to achieve personalized expression; Emotional Adaptation and Confirmation: Parents or teachers can listen to different versions of the content, select the most suitable version, and confirm it to ensure that the content style and emotional expression are accurate. Push and broadcast: The system asynchronously pushes the confirmed content to the child's device for playback according to the time period set by the family, realizing non-real-time and flexible broadcast; Intelligent interaction module activated: After the broadcast is completed, some intelligent agents can automatically enter the tutoring, interaction or reminder state according to the settings to conduct subsequent communication with the child; Behavior recording and feedback: The system records the number of broadcasts, the level of response, and the completion status, and feeds the behavioral data back to teachers or parents for learning tracking and effectiveness evaluation.

[0220] Summary of technical value: Differentiated emotional reporting increases acceptance. Configurable style and frequency enhances interaction; Automatically invokes the homework analysis and assistance module to reduce the burden on parents; Effectively cultivate learning behaviors and form a virtuous learning loop; It has high scene adaptability and is suitable for diverse educational models such as teaching, after-school care, and parent-child co-education.

[0221] Technical effects: Emotional expression enhances information acceptance: By using multi-role, multi-tone onomatopoeic broadcasts, children are more willing to listen and follow instructions; Personalized matching enhances user engagement: Edge AI agents with different personalities can foster long-term user identification and encourage continued use. The system's flexible configuration ensures parental control: parents can specify the broadcast tone and confirm the content before sending, ensuring that the information is under their control; Multifunctional guidance to build good behavioral habits: broadcasting, tutoring, and reminding healthy habits simultaneously to build a closed-loop learning ecosystem; Asynchronous social mechanisms reduce communication pressure: they do not rely heavily on instant messaging, enabling gentle and unhurried information delivery; Highly compatible with edge intelligent agent forms: No restrictions on hardware appearance or protocol, enabling deployment in multiple scenarios such as wearable, doll-like, and home-based applications.

[0222] Example 5: Learning Assistance for Children with Language Disorders – Gentle Companionship and Rehabilitation Incentives from Borderline Intelligence Agents

[0223] This embodiment focuses on Ms. Zhang, a special education teacher, and demonstrates how she uses the anthropomorphic, voice-activated asynchronous social system of this invention to provide personalized and compassionate daily language training services to three children with language disorders.

[0224] Background setting: Ms. Zhang teaches at a special education school, where she is responsible for speech rehabilitation training. Every day, she sends simple language practice sentences to her students, guiding them to imitate the pronunciation and rhythm. Ms. Zhang's phone has a companion app installed, which includes content editing and tone / speed settings.

[0225] Each of the three students was assigned a different edge agent: Xiao Liang: Uses an orange fox doll, and has a cheerful, youthful voice; Xiao Chen: Uses a soft cat speaker with a gentle, older sister-like voice style; Xiao Jie: Uses a handheld learning box shaped like an astronaut, with a firm yet gentle male robot voice.

[0226] Teaching process: Every morning at 9:00, Teacher Zhang sends language training sentences to three students via the APP.

[0227] Teacher Zhang selected the following settings: Sentence content: Xiao Liang: "I can talk." Xiao Chen: "Good morning, teacher." Xiao Jie: "I like learning." Voiceover selection: Specify the voiceovers for the little fox, the cat, and the astronaut characters respectively; Speech rate: slow; tone: friendly and firm; emphasis: on key words such as "speaking," "teacher," and "studying"; Emotional parameters: Encouragement + Gentleness; Repeat setting: Repeat every 2 minutes, for a total of 5 times; Tutoring mode: Automatically waits for the user to imitate before the next round of broadcasting.

[0228] System execution process: The speech generation module calls an AI model to synthesize characterized emotions for each sentence. The repetition planning module controls the interval and repetition rounds, generating a complete broadcast task package; The push scheduling module asynchronously distributes broadcast tasks to edge devices based on user identity; Each edge device broadcasts emotionally charged voice messages at set intervals: The little fox said to Xiao Liang, "I can talk! Say it with me, Xiao Liang, let's work hard together!" The cat-shaped speaker said to Xiao Chen, "Good morning, teacher (in a soft, slow tone), let's do it one more time, I'll keep you company." The astronaut learning box said to Xiao Jie, "I love learning! (Emphasis varies each time), you will be a future scientist too!"

[0229] User feedback and control: Teacher Zhang uses the app to view the number of times students practice and their response time. If a student does not imitate for an extended period of time, the device can automatically remind them or switch the tone of voice to a more enthusiastic / playful style. Teacher Zhang can also remotely add encouraging words or switch playback modes; After each day's training session, the system generates an "Interactive Learning Summary" report and sends it to Teacher Zhang.

[0230] Technical effects: To achieve personalized, non-invasive language rehabilitation training for people with language disorders; It supports multiple characters, multiple tones, and multiple broadcast settings to adapt to each child's emotional acceptance style; To realize a learning and encouragement mechanism for children through edge intelligence agents; Establish an asynchronous yet continuous educational feedback loop to help teachers remotely manage and adjust their teaching plans; It helps children develop habits of language imitation, confident expression, and independent training.

[0231] Example 6: Child users specify the generation of stories and have them narrated emotionally by a puppet character.

[0232] Background setting: User A and User B are two 7-year-old children, each owning a smart doll that supports edge voice interaction. User A's doll is "Little Dinosaur Duoduo," and User B's doll is "Little Rabbit Bobo." Each doll has a built-in voice acquisition module, identity recognition module, emotion-based voice broadcasting module, and a communication interface connecting to a cloud-based AI platform.

[0233] Step 1: Wake-up and Voice Acquisition

[0234] User A gently taps the head of "Little Dinosaur Duoduo" (a preset wake-up action). After the doll confirms that it is the user operating the device through voiceprint recognition, it prompts in a human-like voice: "Hello, do you want to say something to your friend?" User A said, "Help me make up a story about a 'brave little elephant' and tell it to Bobo in a cheerful tone."

[0235] Step 2: Semantic parsing and content generation request

[0236] The system identifies the user's input intent and extracts key information: Content type: Generated story; Story theme: The Brave Little Elephant; Recipient: User B; Broadcast character: Little Dinosaur Duoduo (default sending character); Mood style: Cheer up.

[0237] The system will upload the request to the cloud-based AI language model, which will automatically generate a short fairy tale titled "The Brave Little Elephant".

[0238] Step 3: Stylized Content Generation and Audition

[0239] After the content is generated, the system uses the "Little Dinosaur Duoduo" corpus and the "Happy Tone" template to synthesize the story into an emotionally-inspired audio version.

[0240] The device said to user A, "I'm ready. Come and hear if my story suits your taste!" After listening to the audio, User A felt the ending was a bit bland and said, "Make the ending more exciting!" The system readjusted the emotional paragraphs at the end of the story and generated an updated preview.

[0241] Step 4: User Confirmation and Role Authorization Release

[0242] After listening, User A said with satisfaction, "Okay, this is the version!" The instruction was then given: "Let Bobo listen." The system will upload the final generated content to the cloud and package it into a broadcast content for "Little Rabbit Bobo" to be sent to user B, with sender identification and character voice style information.

[0243] Step 5: Asynchronous content forwarding and receiver broadcasting

[0244] The next day, when User B was playing with "Little Rabbit Bobo", they triggered the play button or said the wake word "Bobo, is there a new story?" "Little Rabbit Bobo" announces in a friendly tone: "Hello, I'm Bobo. Today, Dinosaur Duoduo is here to bring you a new story called 'The Brave Little Elephant'." The story begins with a cheerful mood and ends with an exciting climax: "The little elephant has finally crossed the river! It has become a true hero in everyone's hearts!" After the broadcast ended, Bobo the little rabbit asked, "Did you like this story? Do you want to tell Duoduo what you think?"

[0245] Step 6: Social Loop and Response Interaction

[0246] User B can choose to respond: "I want to tell you a story about a bird!" "Little Rabbit Bobo" records voice messages and prepares to send them back. Through the "representative" dialogue between the dolls, two children establish an asynchronous, anthropomorphic, and emotional social network.

[0247] This embodiment demonstrates the complete technical path of your patent: The process involves "user intent → content generation → emotion regulation → style preview → user confirmation → assigning broadcasting role → onomatopoeia → emotion broadcasting → social feedback," creating a warm and personalized social experience for children without the need for an account.

[0248] Example 7: The entire process of children's asynchronous social expression and broadcasting under parental setting permissions

[0249] Background setting: The child "Yuanyuan" (user A, 6 years old) owns an edge-controlled smart doll named "Little Bear Beibei". The doll has been bound to Yuanyuan's mother "Mom A"'s family account through the APP. The system confirms that "Mom A" is Yuanyuan's guardian and has the right to review the content and communication of the device. Yuanyuan's friend "An'an" (user B, 7 years old) owns an "Owl Gugu" smart doll. The two have become friends through the "friends add function" in their respective dolls, and their social relationship records are maintained by the system.

[0250] Step 1: User Identity Binding and Parental Permission Settings

[0251] When using Little Bear Beibei and "Yuanyuan" for the first time, users complete the user identity setting through scanning a QR code or voiceprint binding in the APP; On the parent app, "Mom A" has bound Yuanyuan's family identity and enabled the "sensitive content review + parent confirmation before broadcast" mode in the permission settings; System record: All content sent by Yuanyuan must be approved by her mother A before it can be broadcast to the other party's doll.

[0252] Step 2: Yuanyuan initiates an asynchronous expression request.

[0253] One evening, Yuanyuan patted the little bear Beibei's head to wake the doll and said, "I want to say to An'an: Thank you for giving me the drawing, I like it so much!" The doll replied, "Understood. So, what do you want me to say?" Yuanyuan said, "Say it out louder and in a happier tone!" The system identifies the keywords: "thank you," "like," "peace," and "happy," thus confirming the target user and basic intent.

[0254] Step 3: Emotional Style Synthesis and Initial Listening Test

[0255] The system uses the Little Bear Beibei character template to generate an expression version in the style of "thank you + happy", for example: "Hey An'an, hello there! I'm Little Bear Beibei! Yuanyuan wants to thank you for the drawing, she loves it so much~ I'm so happy~" After listening, Yuanyuan said, "Make it more lively, and add some jumping sounds at the end." The system further adjusted the speech rate and tone, inserted ending sound effects, and generated a preview version for Yuanyuan to confirm.

[0256] Step 4: Parental approval process triggered

[0257] Since Yuanyuan is a minor and "parental content review mode" is enabled, the system will generate content along with a broadcast sample and push it to "Mom A's" parent app, along with a notification message: "Yuanyuan wants to send a message to An'an. Do you allow it? You can choose to edit, approve, or block." After the mother clicked on the preview content and confirmed that there were no sensitive words or inappropriate tone, she clicked "Apply".

[0258] The system records metadata such as review time, reviewer identity, and operation records.

[0259] Step 5: Pack and send asynchronous content

[0260] After the content review is approved, the system will package the following information: Audio of the broadcast content; Narrator: Little Bear Beibei; Expressing emotion: Happy (system emotion level is Level-3 Happy); Sender's identity (Yuanyuan); Receiving object (An'an); Security review mark (Mother has reviewed).

[0261] The content is pushed to the "Owl Gugu" device and marked as "New content to be broadcast".

[0262] Step 6: Receiver wake-up and broadcast

[0263] The next day, An'an woke up "Owl Coo Coo" and said, "Coo Coo, is someone looking for me?" Owl Gugu said in a friendly tone, "You have a new message! I'm your friend Little Bear Beibei, and I want to say on behalf of Yuanyuan: Thank you for giving her the drawing, she loves it so much! She's so happy!" The tone of voice rose and fell naturally during the playback, the pace was brisk, and the ending notes rose.

[0264] Step 7: Social Interaction Closed Loop

[0265] After listening, An'an said, "Gugu, tell Yuanyuan that I like her too, she's so great!" The owl recorded a voice response and converted it into a request.

[0266] The system packages it into another expression and puts it into the review process set by An'an's guardian, "Mom B," forming a complete closed loop for children's social expression.

[0267] Features: This embodiment particularly demonstrates the integration of edge agent identity binding mechanism, user social relationship mapping mechanism, and parental permission review mechanism; It also includes a complete closed-loop path: content generation → emotional style audition → parent confirmation → asynchronous broadcasting → receiving feedback; It meets the need for both safety and autonomy for underage users in decentralized expression scenarios; Each broadcast process is traceable and auditable, and has data compliance, making it suitable for multiple practical use scenarios such as education and parent-child interaction.

[0268] Example 8: Asynchronous expression of affection between grandparents and grandchildren

[0269] Scene background: The elderly grandfather (User A, 65 years old) lives alone in another city, while his grandson Xiaoyu (User B, 5 years old) lives with his parents in the city. Xiaoyu owns a smart doll named "Little Lion Bobo," which is linked to Xiaoyu's family app, and his parents have full management rights.

[0270] Grandpa uses a voice box called "Mr. Giraffe" with a microphone and speaker. It has built-in edge voice capture and character expression functions and has been bound to Xiaoyu's doll as a "grandfather-grandson friend".

[0271] Step 1: The elderly person initiates a message request

[0272] One afternoon, Grandpa said to Mr. Giraffe, "Leave a message for Xiaoyu and say: I've been missing him lately, and remind him to eat." The device identifies the intent of the content and extracts keywords such as "message", "Xiaoyu", "miss", and "eat".

[0273] Step 2: Character and Tone Selection

[0274] Mr. Giraffe replied, "Would you like me to say it for you in a gentle and humorous tone?" Grandpa nodded, and the device synthesized the following audio-visual content: "Hey, Xiaoyu, I'm your old friend Mr. Giraffe. I just heard from Grandpa that he's been missing you a lot lately, and he said you need to eat more, or you won't grow tall!" Grandpa was satisfied after hearing this and confirmed, "That's how it will be broadcast."

[0275] Step 3: Content Upload and Parent Approval

[0276] The system will upload the generated content, character settings, and audio files to Xiaoyu's parents' family account for review.

[0277] Parents listened to a sample of the content on the app, confirmed that the content was healthy and positive, and then clicked "Allow Broadcast".

[0278] Step 4: Content delivery and asynchronous broadcasting

[0279] The next day, Xiaoyu opened "Little Lion Bobo," and the device said, "You've received a special message—it's from Mr. Giraffe!" Then, his grandfather's message was played, and Xiaoyu happily responded, "I miss my grandfather too!"

[0280] Features: Seniors can leave messages for their grandchildren remotely using natural voice without needing to type text. The writing style employs "role-mediated expression" to enhance the relatability of the content; Parents have the right to review content to ensure the safety of information received by minors; The asynchronous mechanism ensures that the elderly can express themselves at any time and that children can receive messages at appropriate times, providing non-intrusive communication.

[0281] Example 9: Guided Social Expression in Children with Autism

[0282] Scene background: Xiaomei is a child with autism spectrum disorder. She is not good at expressing herself verbally and often does not respond to being called by others. However, she especially loves a smart doll called "Little Yellow Dog" that her father gave her. She is inseparable from it every day and regards it as her closest "friend".

[0283] The doll has human-like voice broadcasting capabilities, has been linked to Xiaomei's family account through the APP, and supports parents to remotely publish and control content broadcasting.

[0284] Step 1: Mom initiates a request through the app.

[0285] At dinner time, the mother called Xiaomei to come eat several times, but Xiaomei did not respond.

[0286] Mom opened the family version of the app, went to the "Little Yellow Dog" control module, clicked the "Send Emotional Reminder" function, and manually entered the text: "Little Yellow Dog, can you help me make up a little story that says: 'Good children will listen and go to eat when their mothers call them.' Make sure the tone is cheerful and gentle."

[0287] Step 2: Content Generation and Simulated Broadcast Preview

[0288] After receiving the text request, the platform invokes the AI ​​language generation module and the Little Yellow Dog character's voice model to generate the following simulated broadcast content: "Xiaomei, I had so much fun playing with you today! You know what? I like you the most~ You're such a good girl. A wonderful child like you always goes to eat right away when Mom calls you for dinner~ That way you'll have the energy to play more games with me!" The app played the sample audio. The mother listened carefully, confirmed that the content was gentle and appropriate, and the emotional guidance was natural, and then clicked "Confirm Playback".

[0289] Step 3: Specify the broadcast terminal and set the trigger method

[0290] In the app, the mother selects: Target device: Xiaomei's "Little Yellow Dog"; Broadcast method: Immediate wake-up broadcast; Voice style: The default voice of the little yellow dog maintains a cheerful mood; Number of repetitions: Maximum of 2 reminders to avoid interruption.

[0291] The system will send the broadcast command to the Xiaohuangou terminal.

[0292] Step 4: Edge devices execute broadcasts and generate emotional interactions.

[0293] A minute later, the little yellow dog gently lit up its ear rings, waking Xiaomei's attention with a familiar rhythm, and played the following message: "Xiaomei, I had so much fun playing with you today! You're a good child. A wonderful child like you always listens and goes to eat when Mom calls you for dinner~ We need to be full to have the energy to play more games, right?" Xiao Mei heard the little yellow dog's voice, looked up at it, nodded gently, and then got up and walked towards the dining table.

[0294] Features: Parents can make content requests via text from the app, eliminating the need for voice input; The platform can match suitable story templates and tone styles based on keywords such as "good child," "obedient," and "happy." Provides a preview function for broadcasting, allowing parents to confirm before distribution, ensuring that the expression matches the family's communication style; After receiving the message, the edge device broadcasts it in a voice familiar to the child, reducing the confrontational nature of the instructions and increasing emotional acceptance. Emotional guidance and parent-child interaction support suitable for non-verbal dominant children (such as children on the autism spectrum and children with language delay).

[0295] Example 10: Children's Interactions with Virtual Pets

[0296] Scene background: Sisi, a little girl (User A, 6 years old), owns a voice-activated doll named "Little Kitty Meow Meow," which she treats as a "pet friend." The device is linked to Sisi's identity through an app and supports character learning and anthropomorphic emotional feedback.

[0297] Sisi often talks to her kittens. When her friends are not around or she is in a bad mood, she will also confide in the kittens.

[0298] Step 1: The child takes the initiative to confide.

[0299] One evening, Sisi gently patted her cat plushie and said, "Cat, I'm not very happy today... My deskmate won't play with me." The doll activates its emotion recognition and soothing mechanism, gently responding: "Hmm... I heard you. Would you like me to tell you a happy story?"

[0300] Step 2: AI Content Generation and Style Setting

[0301] The system identifies the emotion words "sad" and "not playing" and retrieves a "healing story template".

[0302] Example of generated content: "Once upon a time, there was a little penguin who encountered the same thing, but she later made new friends and discovered that there were actually many, many people in the world who wanted to play with her!" After hearing this, Sisi said, "Hmm, it could be even cuter."

[0303] The system adjusts the language and onomatopoeia style, inserts upbeat background music, and generates a new version.

[0304] Step 3: Self-emotional feedback and interaction suggestions

[0305] After the story was played, the kitten Meow Meow said, "Are you feeling better now? Do you want me to tell you a new one tomorrow?" Sisi nodded and said, "Okay, I'll come back to see you tomorrow."

[0306] Features: In this example, "the user is both the sender and the receiver," demonstrating the reverse role of anthropomorphic characters in emotional companionship. The content is generated entirely automatically by the system, with style control based on emotional keywords; It does not require explicitly specifying the "sending recipient," but it achieves the same closed loop of "expression + reception + feedback." It is suitable for psychological support and companionship intervention for children who have difficulty expressing their emotions and who are lonely.

Claims

1. An asynchronous social networking method based on anthropomorphic and phonetic expression, characterized in that, Includes the following steps: (1) Voice acquisition or text input: The user's voice input is acquired through the edge voice terminal, and the user's identity and expression intention are identified; the voice input can also be replaced by text input. The user can input text content through a mobile terminal (including but not limited to mobile phone and tablet), and attach control parameters for indicating tone, emotion or role style; (2) Semantic parsing: Perform semantic understanding on the input content and extract parameters such as the target object, emotional direction, and role preference; (3) Role and emotion audition scheduling: Based on the semantic analysis results, multiple anthropomorphic roles and their tone styles are provided for users to select and confirm personalized expression paths; (4) Generation of emotional expression content: Based on the user's intent, emotional parameters and role settings, the onomatopoeic expression content with personality style and emotional tone is generated by calling the natural language generation model; (5) Expression Confirmation: Provide the generated expression content for listening, and users can modify the style and content by voice or text. After confirmation, authorize asynchronous release; (6) Asynchronous broadcast: The confirmed expression is sent to the edge voice terminal designated by the receiver, and the anthropomorphic character bound to it broadcasts the content in the corresponding tone and style. (7) Expression Recording and Feedback Interaction: The system records the expression trajectory and role call path, supports querying expression history and subsequent interaction, so as to continuously maintain the emotional connection between users.

2. The method according to claim 1, characterized in that, The expression content generation and broadcast path control process is implemented through an AI language model. The model can be an open-source large language model, a third-party API service model, or a locally deployed private model. Multiple models can be called as needed to complete expression generation, tone optimization, and multilingual processing. Module-level model replacement and collaborative task scheduling are supported.

3. The method according to claim 1 or 2, characterized in that, The content generation, style adaptation, and broadcast path control modules can be deployed on edge devices, cloud servers, or a distributed architecture composed of both.

4. The method according to claim 1, characterized in that, The edge voice terminal includes voice dolls, voice boxes, smart home devices, wearable devices, smart speakers, mobile phones or tablets, etc.

5. The method according to claim 1, characterized in that, The voice broadcast uses a character and tone style that users can freely combine and can be personalized through natural language commands, including expressions such as "be gentler" or "be more energetic".

6. The method according to claim 1, characterized in that, The expression confirmation step includes a closed-loop listening test, allowing users to modify and confirm the content, tone of voice, and character image before publication, ensuring the accuracy of emotional transmission.

7. The method according to claim 1, characterized in that, The asynchronous broadcast can be pushed based on time settings, event triggers, or user-defined context strategies to achieve non-real-time, non-intrusive information transmission.

8. The method according to claim 1, characterized in that, The method supports communication binding via device identifiers or character nicknames, enabling the establishment of familiar social channels without the need for a traditional account system. It also allows for the configuration of permission policies, including mechanisms such as broadcast restrictions, content review, keyword filtering, and parental supervision.

9. The method according to claim 1, characterized in that, The expressed content can be generated by user input, or it can be generated by the system based on an AI model according to semantic clues, such as stories, blessings, apologies, reminders, etc.

10. The method according to claim 1, characterized in that, The method supports offline caching of voice collection and emotion annotation. After the network is restored, it is automatically synchronized to the cloud to complete the expression generation and push. It can also perform historical expression retrospection, role call frequency analysis and multi-round emotion maintenance.

Images