Device and method for generating interactive content linked with ai doll based on infant speech recognition

The AI doll-linked system addresses the challenge of unilateral content delivery by using infant speech recognition to analyze and predict reactions, providing personalized educational experiences through an AI doll.

KR103044516B1Active Publication Date: 2026-11-06CARTIUS INC
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
KR1020260012046
Authority / Receiving Office
KR · KR
Patent Type
Patents
Current Assignee / Owner
Priority Date
2025-10-20
Filing Date
2026-01-21
Publication Date
2026-11-06
Estimated Expiration
2046-01-21

AI Technical Summary

Technical Problem

Existing educational systems for children deliver content unilaterally without considering their reactions, leading to low engagement and limited personalization, and lack the ability to predict and respond to children's expected reactions to new material.

Method used

An AI doll-linked system that uses infant speech recognition to collect and analyze responses, separating phonemes, predicting expected reactions, and generating personalized content using a large-scale language model to provide appropriate feedback.

Benefits of technology

Enables personalized conversational content by accurately identifying infant speech intent, predicting responses, and recommending subsequent content based on learning patterns and preferences, enhancing engagement and educational effectiveness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 112026008676447-PAT00006_ABST
    Figure 112026008676447-PAT00006_ABST
Patent Text Reader

Abstract

The embodiments provide an apparatus and method for generating interactive content linked to an AI doll based on infant speech recognition. The method according to the embodiments may include: a step of performing user authentication of a parent user by the server electronic device; a step of selecting a first content based on the parent user's choice and checking whether an AI doll for providing the first content is registered by the server electronic device when the user authentication is completed; a step of playing the first content by the server electronic device and outputting a first speech of the AI ​​doll proceeding in at least one of playing a fairy tale story or a thematic conversation; a step of collecting a first response of an infant user to the first speech by the server electronic device; and a step of selecting at least one answer from a plurality of answers pre-stored in a database based on the first response and outputting it to the infant user as a second speech by the server electronic device.
Need to check novelty before this filing date? Find Prior Art

Description

Technology Field

[0001] Embodiments of the present invention relate to an AI doll-linked interactive content generation device and method based on infant speech recognition. Background Technology

[0002] The limitations of one-way content delivery methods are continuously being raised in the field of early childhood education. Existing systems for playing stories or providing quizzes deliver content unilaterally without considering the children's reactions, resulting in low engagement and learning effectiveness. In particular, there is a lack of systems that collect and analyze the various reactions children show to stories or quizzes to provide appropriate feedback. Furthermore, the absence of a function to predict and prepare for children's expected reactions to new content leads to trial and error when introducing new material. Due to these issues, the effectiveness of AI-based educational services is limited, and it is currently difficult to provide personalized learning experiences. The problem to be solved

[0003] Embodiments of the present invention provide an apparatus and method for generating interactive content linked to an AI doll based on infant speech recognition to address the aforementioned problems.

[0004] The technical problems to be solved in the embodiments are not limited to those mentioned above, and other unmentioned technical problems may be considered by those skilled in the art from the various embodiments described below. means of solving the problem

[0005] A method for generating interactive content linked to an AI doll based on infant speech recognition according to an embodiment of the present invention may include: a step of performing user authentication of a parent user by a server electronic device; a step of selecting a first content based on the parent user's choice and checking whether an AI doll for providing the first content is registered by a server electronic device when the user authentication is completed; a step of playing the first content by a server electronic device and outputting a first speech of the AI ​​doll proceeding in at least one of playing a fairy tale story or a thematic conversation; a step of collecting a first response of an infant user to the first speech by a server electronic device; and a step of selecting at least one answer from a plurality of answers pre-stored in a database based on the first response and outputting it to the infant user as a second speech by a server electronic device.

[0006] The step of collecting the first response of the infant user may include: a step of recognizing the infant user's voice by a server electronic device and extracting a word related to at least one of a subject, an object, and a verb from the recognized voice; a step of separating the extracted word into phoneme units including consonants and vowels by a server electronic device; and a step of analyzing at least one of the duration, rhythm, volume, speech rate, and stress of each separated phoneme by a server electronic device, and converting and recognizing it into a standard pronunciation if the infant user's pronunciation accuracy is below a threshold.

[0007] If the first content is new content for which there is no prior selection history, the method may further include: a step of predicting a first response expected to be provided by the infant user before collecting the first response to the first utterance of the AI ​​doll by the server electronic device; a step of storing the predicted first response of the infant user in a database by the server electronic device; and a step of comparing the stored expected first response of the infant user with the first response of the infant user collected for the first utterance of the AI ​​doll in a second content different from the first content by the server electronic device.

[0008] The method may further include the step of, by means of a server electronic device, comparing a first reaction of an infant user expected for the first content with a first reaction of an infant user in the second content, and then extracting a reaction in which the relevance of the utterance between the first content and the second content is determined to be greater than or equal to a reference value.

[0009] The step of predicting a first response expected to be provided by the infant user may calculate a probability distribution of possible responses by analyzing at least one of the content provision history, age, gender, region, and preference selected by the infant user by a server electronic device.

[0010] If it is difficult to provide an appropriate response to the first reaction of the infant user by the server electronic device, the method may further include the step of generating and outputting the second utterance by utilizing a large-scale language model. The large-scale language model may generate the second utterance using at least one of vocabulary, expressions, and sentence structures suitable for the age group of the infant user.

[0011] The first content above includes at least one of a fairy tale, a quiz, and a thematic conversation, and the first utterance may include at least one of a question regarding the actions or ending of a character within the fairy tale story and a question that elicits a correct answer in the quiz.

[0012] The method may further include: a step of analyzing at least one of the learning pattern, preferred answer type, and topic of interest of the infant user based on the first response of the infant user by a server electronic device; and a step of recommending subsequent content to be provided to the infant user based on the analyzed result by a server electronic device.

[0013] Multiple answers stored in advance in the above database are classified and stored according to the type of the first utterance, and the similarity between the first response of the infant user and each of the multiple answers is calculated by the server electronic device, and the answer with a calculated similarity greater than or equal to a threshold value can be selected as the second utterance. Effects of the invention

[0014] According to embodiments of the present invention, personalized conversational content can be provided by collecting responses from infant users in real time and accumulating learning through an AI doll. In particular, the infant user's voice is separated into phoneme units to analyze duration, rhythm, volume, speech rate, and stress; if the pronunciation accuracy falls below a threshold, it is converted to standard pronunciation for recognition, thereby accurately identifying the speech intent despite the infant's inaccurate pronunciation. Furthermore, regarding new content, expected responses are predicted in advance by analyzing the infant user's content provision history, age, gender, region, and preferences. By comparing these responses with those from existing content and extracting responses with high relevance to the utterance, appropriate responses are possible even when new content is introduced. Among multiple answers stored in advance in a database, an answer with a similarity level above a threshold is selected as the second utterance; however, if there is no appropriate answer, a large-scale language model is utilized to generate a second utterance using vocabulary, expressions, and sentence structures suitable for the infant user's age group, thereby enabling flexible responses to unexpected reactions. Additionally, by analyzing the infant user's learning patterns, preferred response types, and topics of interest to recommend subsequent content, continuous and systematic customized educational services can be provided. Brief explanation of the drawing

[0015] The accompanying drawings, included as part of the detailed description to aid in understanding the embodiments, provide various embodiments and explain the technical features of the various embodiments together with the detailed description. FIG. 1 is a schematic diagram illustrating an AI-based early childhood education system (10) according to one embodiment of the present invention. FIG. 2 is a block diagram for explaining the structure of the server electronic device (100) of FIG. 1. Figure 3 is a diagram illustrating the multilayer neural network (122) of Figure 2. FIG. 4 is a block diagram for explaining the structure of the user terminal (200) of FIG. 1. FIG. 5 is a block diagram for explaining the structure of the AI ​​doll (300) of FIG. 1. FIG. 6 is a flowchart illustrating a method for providing AI doll-based interactive content according to an embodiment of the present invention. Figure 7 is a flowchart for explaining step S640 of Figure 6 in detail. Figure 8 is a flowchart illustrating the reaction prediction process for new content in step S630 of Figure 6. Figure 9 is a flowchart for explaining step S650 of Figure 6 in detail. Specific details for implementing the invention

[0016] The following embodiments are combinations of the components and features of the embodiments in a predetermined form. Each component or feature may be considered optional unless otherwise explicitly stated. Each component or feature may be implemented in a form not combined with other components or features. Additionally, various embodiments may be constructed by combining some components and / or features. The order of operations described in various embodiments may be changed. Some components or features of one embodiment may be included in another embodiment, or may be replaced with corresponding components or features of another embodiment.

[0017] In the description of the drawings, procedures or steps that could obscure the essence of the various embodiments were not described, nor were procedures or steps that can be understood by a person of ordinary knowledge in the relevant technical field described.

[0018] Throughout the specification, when a part is described as "comprising" or "including" a component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, terms such as "...part," "...unit," and "module" as used in the specification refer to a unit that performs at least one function or operation, and this may be implemented in hardware, software, or a combination of hardware and software. Additionally, "one (a or an)," "one," "the," and similar related terms may be used in the context describing various embodiments (particularly in the context of the following claims) in both singular and plural forms, unless otherwise indicated in the specification or clearly contradicted by the context.

[0019] Hereinafter, embodiments according to various examples will be described in detail with reference to the accompanying drawings. The detailed description disclosed below, together with the accompanying drawings, is intended to describe exemplary embodiments of various examples and is not intended to represent the only embodiment.

[0020] In addition, specific terms used in various embodiments are provided to aid in understanding the various embodiments, and the use of such specific terms may be modified in other forms within the scope of not departing from the technical concept of the various embodiments.

[0021] In this specification, "infant" or "infant user" may comprehensively refer to a user who requires assistance with speech recognition and communication due to unclear pronunciation or difficulty in sentence construction. For example, "infant user" may include, in addition to infants and toddlers, the elderly, people with developmental disabilities, people with hearing impairments, people with speech impairments, or other users who require support for communication. In this case, "parent" or "parent user" may be interpreted as the guardian, caregiver, or legal representative of the user. Additionally, "AI doll" may be interpreted as an AI-based assistive device or smart device that interacts with the user, "fairy tale content" as multimedia content or learning content suitable for the user's cognitive level and interests, and "art workbook" as materials for user-customized creative activities. Furthermore, terms used in this specification in relation to infants may be interpreted with meanings corresponding to the characteristics of the user group. FIG. 1 is a schematic diagram illustrating an AI-based early childhood education system (10) according to an embodiment of the present invention.

[0022] Referring to FIG. 1, an AI-based early childhood education system (10) may include a server electronic device (100), a user terminal (200), and an AI doll (300). The server electronic device (100), the user terminal (200), and the AI ​​doll (300) may be interconnected via a network to transmit and receive data.

[0023] The server electronic device (100) is implemented in the form of a cloud server and can handle core processing of an AI-based early childhood education service. The server electronic device (100) can analyze the voice data of the infant, process conversation content, manage content, and perform user authentication. The server electronic device (100) can store and manage large-capacity pronunciation datasets, infant conversation pattern data, educational content, etc.

[0024] The user terminal (200) may be a mobile device, such as a smartphone or tablet PC, used by a parent. An AI-based early childhood education application is installed on the user terminal (200), allowing the parent to monitor the learning status of the child, purchase content, and control the AI ​​doll (300). The user terminal (200) can perform content authentication by scanning the QR code of the art workbook through a camera.

[0025] In this specification, the AI ​​doll (300) may refer to a smart toy that interacts directly with an infant, such as a doll or device with built-in AI. It can converse with the infant through a built-in speaker and microphone. The AI ​​doll (300) can collect the infant's voice and transmit it to a server electronic device (100), and output a response generated by the server electronic device (100) as voice. The AI ​​doll (300) may be manufactured with a soft appearance and a friendly design so that the infant can form an emotional bond.

[0026] In one embodiment, the user terminal (200) can be directly connected to the AI ​​doll (300) via Bluetooth or Wi-Fi Direct. Through this, the user terminal (200) can control the volume, voice speed, LED brightness, etc. of the AI ​​doll (300) in real time.

[0027] In one embodiment, the AI ​​doll (300) can communicate in real time with the server electronic device (100) to process the infant's speech. When the AI ​​doll (300) collects the infant's voice, it transmits it to the voice processing unit (120) of the server electronic device (100), and the voice processing unit (120) can convert the voice into text and transmit it to the dialogue processing unit (140). The dialogue processing unit (140) generates a contextually appropriate response and transmits it back to the AI ​​doll (300), and the AI ​​doll (300) can synthesize it into speech and play it for the infant.

[0028] In one embodiment, the server electronic device (100) can perform parent authentication through the user terminal (200) and then transmit selected content to the AI ​​doll (300) for playback. For example, if a parent selects a specific fairy tale from the user terminal (200), the server electronic device (100) can stream the corresponding fairy tale content to the AI ​​doll (300) to be read to the infant.

[0030] FIG. 2 is a block diagram for explaining the structure of the server electronic device (100) of FIG. 1.

[0031] Referring to FIG. 2, the server electronic device (100) may include a user management unit (110), a voice processing unit (120), a content management unit (130), a dialogue processing unit (140), a database (150), a picture generation unit (160), an authentication processing unit (170), and a learning processing unit (180).

[0032] The user management unit (110) can register users to the AI-based early childhood education system through a membership registration process and collect and manage user information. The user management unit (110) can perform user registration and user information input processes. The user management unit (110) can provide a membership registration interface to the user terminal (200) and receive user information from the user terminal (200). Here, the user information may include at least one of the child's name, age, gender, region, developmental level, and parent's contact information. The user management unit (110) can provide services such as content purchase, AI doll linkage, and art workbook authentication to users registered as members.

[0033] The user management unit (110) may further include a learning management module (111) that manages the learning progress and achievement of the infant, and a user DB (112) which is a database built by collecting information on input user information, the infant's learning status, preferred content, conversation history, etc.

[0034] The learning management module (111) can calculate the learning achievement based on the completion rate, correct answer rate, and response time of the content performed by the child, and provide a learning report to the parent according to a predetermined period.

[0035] The voice processing unit (120) can process the voice data of the infant collected through the AI ​​doll (300) and analyze the accuracy of the infant's pronunciation. The voice processing unit (120) can determine the accuracy by comparing the infant's unstable pronunciation with standard pronunciation and, if necessary, convert the voice into text.

[0036] The voice processing unit (120) may further include a voice preprocessing unit (121), a multilayer neural network (122), and a learning engine (123).

[0037] The voice preprocessing unit (121) can preprocess the voice data of an infant received from the AI ​​doll (300). Here, preprocessing may refer to operations such as noise removal, volume normalization, and phoneme-unit segmentation of the voice data. The voice preprocessing unit (121) can separate the infant's voice into phonemes and extract the features of each phoneme to convert it into a form that can be compared with standard pronunciation.

[0038] In one embodiment, the voice preprocessing unit (121) can perform customized preprocessing that reflects the pronunciation characteristics of the corresponding infant group based on the infant's age, gender, and region information. For example, the preprocessing parameters can be adjusted considering that the pronunciation of certain consonants may be incomplete for a 3-year-old infant.

[0039] The multilayer neural network (122) can receive preprocessed voice data, calculate a pronunciation accuracy score, and perform text conversion. The multilayer neural network (122) is pre-trained with a large-scale infant voice dataset, so it can recognize even the infant's unclear pronunciation with high accuracy.

[0040] The learning engine (123) can supervise the multilayer neural network (122) using infant pronunciation data by age, gender, and region as learning data so that the multilayer neural network (122) can accurately recognize and analyze the infant's voice.

[0042] Figure 3 is a diagram illustrating the multilayer neural network (122) of Figure 2.

[0043] Referring to FIG. 3, the multilayer neural network (122) may include an input layer, one or more hidden layers, and an output layer.

[0044] In one embodiment, the multilayer neural network (122) may include an input layer having nodes corresponding to a plurality of grids of one or more frames that receive training data, a first hidden layer that outputs by multiplying each output value of the input layer by a weight and adding a bias, a second hidden layer that outputs by multiplying each output value of the first hidden layer by a weight and adding a bias, and an output layer that outputs by multiplying each output value of the second hidden layer by a weight and outputting the result using an activation function. Although only two hidden layers are shown in FIG. 3, one or more hidden layers may include a greater number of hidden layers in addition to the first hidden layer and the second hidden layer.

[0045] For example, the activation function may be a Softmax function, but embodiments of the present invention are not limited thereto, and the activation function may be various other functions such as a LeRU function. Weights and biases may be continuously updated by supervised learning. Specifically, the output vector may be input to a loss function layer connected to the output layer. The loss function layer may output a loss value using a loss function that compares the output vector with the correct answer vector for each training data. The parameters of the multilayer neural network (122) may be supervised learning in a direction that reduces the loss value.

[0046] A learning model may include an input layer, one or more hidden layers, and an output layer. Training data is input into the input layer, passes through one or more hidden layers and the output layer to produce an output vector, and the output vector may be input into a loss function layer connected to the output layer. The loss function layer may output a loss value using a loss function that compares the output vector with the correct answer vector for each training data. The parameters of the learning model may be trained in a direction that reduces the loss value.

[0047] To this end, the voice processing unit (120) may include a hardware structure specialized for processing an artificial intelligence model. The artificial intelligence model may be generated through machine learning. Such learning may be performed, for example, on the electronic device (100) itself where the artificial intelligence model is executed, or through a separate server.

[0048] Learning algorithms may include, for example, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but are not limited to the examples described above. An artificial intelligence model may include multiple artificial neural network layers. An artificial neural network may be a deep neural network (DNN), a convolutional neural network (CNN), a recurrent neural network (RNN), a restricted Boltzmann machine (RBM), a deep belief network (DBN), a bidirectional recurrent deep neural network (BRDNN), deep Q-networks, or a combination of two or more of the above, but is not limited to the examples described above. In addition to the hardware structure, the artificial intelligence model may include a software structure, either additionally or substantially.

[0049] Referring again to FIG. 2, the content management unit (130) can create, store, and manage various educational content provided to infants. The content management unit (130) can systematically classify content such as fairy tales, quizzes, and art workbook templates, and recommend content suitable for the age and developmental level of infants.

[0050] The content management unit (130) may include a content classification module (131), a content recommendation engine (132), and a content DB (133).

[0051] The content classification module (131) can classify content by age, topic, and difficulty level, and generate metadata for each piece of content. The metadata may include information such as the educational goals, time required, and required skills of the content.

[0052] The content recommendation engine (132) can recommend personalized content by analyzing the learning history, preferences, and current developmental level of the infant. The content recommendation engine (132) can utilize recommendation algorithms such as collaborative filtering and content-based filtering.

[0053] The content DB (133) can store and manage various forms of content, such as fairy tale text, quiz questions, art workbook templates, and audio files.

[0054] The conversation processing unit (140) is a key component that enables natural conversation between the AI ​​doll (300) and the infant, and may include a conversation engine based on a Large Language Model (LM). The conversation processing unit (140) can understand the infant's speech and generate an appropriate response that fits the context.

[0055] The conversation processing unit (140) may include a context analysis module (141), a response generation module (142), and a conversation history management unit (143).

[0056] The context analysis module (141) can identify the topic of the current conversation and analyze the relationship with the previous conversation content. The context analysis module (141) can identify the meaning of the words used by the infant and understand the flow of the conversation to determine an appropriate response direction.

[0057] The response generation module (142) can generate a response suitable for the age and language development level of the infant based on the results of context analysis. The response generation module (142) can generate a response using simple sentence structures and vocabulary that are easy for the infant to understand.

[0058] The conversation history management unit (143) can store and manage all conversation content between the AI ​​doll (300) and the infant. The conversation history management unit (143) can establish a personalized conversation strategy by analyzing the infant's interests, preferred topics, and frequently used expressions that appear in the conversation.

[0059] The database (150) can comprehensively store and manage various data necessary for AI-based early childhood education services provided by the server electronic device (100). The database (150) can build a data foundation to provide customized services tailored to the characteristics of individual children by linking with the user DB (112) and the content DB (133).

[0060] In one embodiment, the database (150) can be constructed by storing a dataset including a standard pronunciation dataset, a pronunciation characteristic dataset by age, a pronunciation variation dataset by region, a toddler conversation pattern dataset, and an educational effect measurement dataset.

[0061] Here, the standard pronunciation dataset may include reference data that analyzes the standard pronunciation of Korean at the phoneme level. The age-specific pronunciation characteristics dataset may include pronunciation development characteristics and general pronunciation error patterns of infants for each age group from 2 to 7 years.

[0062] The picture generation unit (160) can generate personalized art workbook content based on the conversation with the infant. The picture generation unit (160) can generate customized coloring pictures by identifying characters, animals, objects, etc. that the infant likes.

[0063] The image generation unit (160) may include an image generation engine (161), a difficulty adjustment module (162), and a template management unit (163).

[0064] The image generation engine (161) can automatically generate drawings that match the interests of the infant by utilizing a generative AI model. The image generation engine (161) can generate a prompt based on the infant's preferred topics identified by the conversation processing unit (140), and input this into the image generation model to generate line art for coloring.

[0065] The difficulty adjustment module (162) can adjust the complexity of the picture by taking into account the age and fine motor development level of the child. For example, a picture composed of simple shapes can be provided to a 3-year-old child, and a picture containing more fine details can be provided to a 6-year-old child.

[0066] The template management unit (163) manages pre-made art workbook templates and can create completed workbook pages by combining the generated drawings and templates.

[0067] The authentication processing unit (170) can be responsible for user authentication using a QR code and content access control. The authentication processing unit (170) can verify genuine users through the QR code included in the art workbook and manage access rights to the content.

[0068] The authentication processing unit (170) may include a QR code generation module (171), an authentication verification module (172), and an authorization management unit (173).

[0069] The QR code generation module (171) can generate a unique conversation start QR code and unlock QR code for each art workbook. The QR code may include encrypted authentication information and content identification information.

[0070] The authentication verification module (172) can verify the validity of a QR code scanned by a user terminal (200) and check whether it is genuine by comparing it with the user's purchase history.

[0071] The authority management unit (173) can grant access rights to the content to authenticated users and manage restrictions such as usage period and usage frequency.

[0072] The learning processing unit (180) can accumulate and analyze the infant's response and learning data and continuously improve the performance of the AI ​​model. The learning processing unit (180) can provide a better educational experience by comprehensively analyzing the infant's speech patterns, preferences, learning performance, etc.

[0073] The learning processing unit (180) may include a data collection module (181), a pattern analysis engine (182), and a model optimization unit (183).

[0074] The data collection module (181) can collect and store data such as the infant's voice, reaction time, and correct answer rate in real time through the AI ​​doll (300).

[0075] The pattern analysis engine (182) can analyze the learning patterns, areas of difficulty, and areas of interest of the infant from the collected data. The pattern analysis engine (182) can establish a personalized learning strategy based on these analysis results.

[0076] The model optimization unit (183) can continuously improve speech recognition models, dialogue generation models, content recommendation models, etc. based on accumulated training data. The model optimization unit (183) can improve model performance while protecting personal information by utilizing techniques such as transfer learning and federated learning.

[0077] The server electronic device (100) may have the same configuration as a conventional web server or cloud server in terms of hardware. However, in terms of software, it may include program modules that perform various functions and are implemented through any language such as Python, Java, Node.js, etc. Furthermore, the server electronic device (100) generally refers to a computer system that is connected to an AI doll (300) and a user terminal (200) through an open computer network such as the Internet, receives requests for task execution from these devices, derives and provides task results, and provides computer software (server program) installed for this purpose.

[0079] FIG. 4 is a block diagram for explaining the structure of the user terminal (200) of FIG. 1.

[0080] Referring to FIG. 4, the user terminal (200) may include a processor (210), memory (220), communication module (230), antenna module (240), touch display (250), and camera module (260). In one embodiment, at least one of these components may be omitted from the user terminal (200), or one or more other components may be added. In another embodiment, some of these components may be integrated into a single component.

[0081] The processor (210) can, for example, execute software to control at least one other component of a user terminal (200) connected to the processor (210) and perform various data processing or operations. In one embodiment, as at least part of the data processing or operations, the processor (210) can store commands or data received from other components in volatile memory (221), process the commands or data stored in volatile memory (221), and store the result data in non-volatile memory (222).

[0082] In one embodiment, the processor (210) may include a main processor (211) (e.g., a central processing unit or an application processor) or an auxiliary processor (212) that can operate independently or together with it (e.g., a graphics processing unit, a neural processing unit (NPU), or a sensor hub processor). For example, if the user terminal (200) includes a main processor (211) and an auxiliary processor (212), the auxiliary processor (212) may be configured to use less power than the main processor (211) or to be specialized for a designated function. The auxiliary processor (212) may be implemented separately from the main processor (211) or as part thereof.

[0083] An application running on a user terminal (200) can perform the following functions under the control of a processor (210). The application may include user authentication and login functions, AI doll (300) registration and linkage functions, content search and purchase functions, art workbook authentication functions via QR code scanning, functions to check the learning progress and achievements of the infant, and AI doll (300) control and setting functions.

[0084] The memory (220) can store various data used by at least one component of the user terminal (200). The data may include, for example, input data or output data for software and related commands. The memory (220) may include volatile memory (221) or non-volatile memory (222). The memory (220) can store AI-based early childhood education applications, user authentication information, downloaded content, AI doll (300) linkage information, etc.

[0085] The communication module (230) can support the establishment of a wired or wireless communication channel between a user terminal (200) and an external electronic device (e.g., a server electronic device (100), an AI doll (300)), and the performance of communication through the established communication channel. The communication module (230) operates independently of the processor (210) and may include one or more communication processors that support wired or wireless communication.

[0086] In one embodiment, the communication module (230) may include a wireless communication module or a wired communication module. The corresponding communication module among these may communicate with the server electronic device (100) via a network and may communicate directly with the AI ​​doll (300) via Bluetooth or Wi-Fi. These various types of communication modules may be integrated into a single component or implemented as multiple separate components.

[0087] The wireless communication module can support 5G networks after 4G networks and next-generation communication technologies, such as NR access technology (new radio access technology). In addition, the wireless communication module can support technologies such as Bluetooth, BLE (Bluetooth Low Energy), and Wi-Fi Direct for short-range communication to provide a stable connection with the AI ​​doll (300).

[0088] The antenna module (240) can transmit a signal or power to an external electronic device or receive it from the outside. In one embodiment, the antenna module (240) may include an antenna comprising a radiator made of a conductor or a conductive pattern formed on a substrate. In one embodiment, the antenna module (240) may include a plurality of antennas. In this case, at least one antenna suitable for a communication method used in a communication network, such as a network, may be selected from the plurality of antennas, for example, by a communication module (230). The signal or power may be transmitted or received between the communication module (230) and the external electronic device through the selected at least one antenna.

[0089] The touch display (250) may be an LCD (liquid crystal display) panel, an OLED (organic light emitting diode) panel, etc. The touch display may also include a touch screen panel (TSP) that includes a touch electrode for detecting a touch on the display panel. In one embodiment, the touch display may be an in-cell type in which the display panel and the touch panel are combined integrally.

[0090] The touch display (250) can display the user interface of an AI-based early childhood education application. The touch display (250) can display a list of content, the learning progress of the child, the connection status of the AI ​​doll (300), a QR code scan screen, a generated art workbook image, etc. Additionally, the touch display (250) can provide a dashboard where parents can check the child's learning performance, a payment screen for purchasing content, etc.

[0091] The camera module (260) can capture still images and video. In one embodiment, the camera module (260) may include one or more cameras. Each of the one or more cameras may include one or more lenses, image sensors, image signal processors, or flashes.

[0092] The camera module (260) can perform the primary function of scanning QR codes printed on the art workbook. The camera module (260) can accurately recognize conversation start QR codes and unlock QR codes and transmit the scanned information to the processor (210). Additionally, the camera module (260) can perform the function of photographing the art workbook pages colored by the child and transmitting them to the server electronic device (100), thereby allowing the child's creative activities to be recorded and shared.

[0093] In one embodiment, the user terminal (200) may further include a speaker (not shown) and a microphone (not shown). The speaker may output notification sounds such as the connection status with the AI ​​doll (300), completion of content purchase, and successful QR code authentication. The microphone may receive voice commands from a parent and support voice control of the application.

[0094] In one embodiment, an application running on a user terminal (200) may provide control functions such as volume control, conversation speed setting, and night mode setting for the AI ​​doll (300). Additionally, the application may visualize statistics such as the type of content preferred by the infant, learning time zone, and concentration level, and provide them to the parents.

[0095] In one embodiment, commands or data may be transmitted or received between a server electronic device (100) connected to a network and a user terminal (200). All or part of the operations executed on the user terminal (200) may be executed on the server electronic device (100). For example, complex operations such as QR code authentication, content recommendation, and learning data analysis may be performed on the server electronic device (100), and only the results may be transmitted to the user terminal (200).

[0097] FIG. 5 is a block diagram for explaining the structure of the AI ​​doll (300) of FIG. 1.

[0098] Referring to FIG. 5, the AI ​​doll (300) may include a control unit (310), a voice input / output unit (320), a communication module (330), a motion driving unit (340), a memory (350), a power supply unit (360), an LED display unit (370), and a sensor unit (380).

[0099] The control unit (310) is a central processing unit that controls the overall operation of the AI ​​doll (300) and can be implemented as a microprocessor or an embedded system-on-chip (SoC). The control unit (310) processes the voice of an infant collected through the voice input / output unit (320), transmits and receives data with the server electronic device (100) through the communication module (330), and can generate the movement of the doll by controlling the motion drive unit (340).

[0100] In one embodiment, the control unit (310) is implemented with a low-power ARM Cortex series processor to maximize battery life. The control unit (310) is equipped with a real-time operating system (RTOS) to perform voice processing and motion control simultaneously.

[0101] The voice input / output unit (320) may include a microphone (321), a speaker (322), and a voice processing chip (323).

[0102] The microphone (321) is an input device for collecting the infant's voice and can be implemented as a MEMS (Micro-Electro-Mechanical Systems) microphone. The microphone (321) is designed as a directional microphone to minimize ambient noise and clearly collect the infant's voice. In one embodiment, the microphone (321) is composed of multiple units and beamforming technology is applied to improve voice recognition accuracy.

[0103] The speaker (322) is a device that outputs AI response voice, fairy tales, music, etc., and can provide clear sound quality at a sound pressure level safe for the hearing of infants by using a high-quality neodymium driver. The speaker (322) is located inside the body of the doll and can be designed together with a resonance chamber for acoustic optimization.

[0104] In one embodiment, the voice input / output unit (320) can convert an analog voice signal into digital and perform preprocessing such as noise canceling and echo removal. The voice input / output unit (320) can have a Text-to-Speech (TTS) engine built in to convert text received from the server electronic device (100) into natural speech.

[0105] The communication module (330) may include a Wi-Fi module (331) and a Bluetooth module (332).

[0106] The Wi-Fi module (331) supports the IEEE 802.11n / ac standard and can provide a stable internet connection with the server electronic device (100). The Wi-Fi module (331) supports 2.4GHz and 5GHz dual bands and can ensure smooth communication even in network congestion situations.

[0107] The Bluetooth module (332) supports BLE (Bluetooth Low Energy) 5.0 or higher and can be connected to the user terminal (200) with low power. The Bluetooth module (332) can be used for initial setup of the AI ​​doll (300), firmware updates, receiving real-time control commands, etc.

[0108] The motion drive unit (340) may include a servo motor, a motor driver, and a position sensor.

[0109] The servo motor is a driving device that moves the head, arms, or tail of the AI ​​doll (300), and can be implemented as a digital servo motor capable of precise angle control. In one embodiment, a 2-axis servo motor is mounted on the head portion to perform actions such as nodding or turning left and right. Each arm portion is equipped with a 1-axis servo motor to implement actions such as waving hands or clapping.

[0110] The motor driver is a circuit that supplies appropriate current and voltage to the servo motor according to the command of the control unit (310), and may include an overcurrent protection function and a temperature monitoring function.

[0111] The position sensor can detect the current position of the servo motor and provide feedback to the control unit (310). Through this, the AI ​​doll (300) can achieve natural and smooth movement.

[0112] In one embodiment, the AI ​​doll (300) can perform lip-sync actions synchronized with voice output. For example, when the AI ​​says "hello," the LED on the mouth part may blink, or it may take a corresponding gesture when expressing a specific emotion.

[0113] The memory (350) may include flash memory and RAM.

[0114] The flash memory can store firmware, basic voice files, content for offline mode, etc. of the AI ​​doll (300). In one embodiment, frequently used responses or sound effects are cached in the flash memory to provide an immediate response even when the network connection is unstable.

[0115] RAM is a volatile memory that temporarily stores real-time voice processing, communication buffers, motion control data, etc., and can ensure fast data access.

[0116] The power supply unit (360) may include a rechargeable battery and a charging circuit.

[0117] The rechargeable battery can be implemented as a lithium polymer battery, which is lightweight and safe, and can be designed to be used continuously for more than 8 hours on a single charge. The battery (361) can be securely fixed inside the doll to protect it from access by infants.

[0118] The charging circuit can support USB-C or wireless charging to provide a convenient charging environment. The charging circuit can ensure safety by including overcharge protection and temperature protection functions.

[0119] In one embodiment, the LED display (370) may include a status LED and an emotion expression LED.

[0120] The status LED can display system status, such as Wi-Fi connection status, battery level, and voice recognition in progress, using colors and flashing patterns. For example, a solid blue light can indicate normal operation, while a flashing red light indicates low battery.

[0121] The emotion expression LED can be positioned on the eyes or cheeks of the AI ​​doll (300) to visually express various emotions. Using RGB LEDs, emotions such as joy, sadness, and surprise can be displayed through changes in color and brightness.

[0122] In one embodiment, the sensor unit (380) may include a touch sensor, an acceleration sensor, and a proximity sensor.

[0123] The touch sensor is implemented as a capacitive touch sensor and can detect when an infant strokes or hugs the doll. The touch sensors are distributed across key areas such as the head, belly, and hands to recognize various interactions.

[0124] The acceleration sensor is a 3-axis MEMS accelerometer and can detect movements, tilts, impacts, etc. of the AI ​​doll (300). For example, it can detect when an infant shakes or flips the doll and show an appropriate response.

[0125] The proximity sensor is implemented as an infrared sensor to detect whether an infant is nearby and can automatically switch to power-saving mode when not in use for a long time.

[0126] FIG. 6 is a flowchart illustrating a method for providing AI doll-based interactive content according to an embodiment of the present invention.

[0127] Referring to FIG. 6, in S610, the user management unit (110) of the server electronic device (100) can perform user authentication of the parent user. The user management unit (110) can perform authentication by receiving the parent's ID and password through an application installed on the user terminal (200). In one embodiment, the user management unit (110) can perform user authentication through at least one of biometric authentication, OTP authentication, and social login. The user management unit (110) can retrieve the account information of the authenticated parent user from the user DB (112) and load the personal information of the infant user registered to the account. The personal information of the infant user may include at least one of age, gender, region, and preference.

[0128] In S620, when user authentication is completed, the content management unit (130) can select a first content based on the parent user's selection and check whether the AI ​​doll (300) for providing the first content is registered. The first content may include at least one of a fairy tale, a quiz, and a thematic conversation. The content classification module (131) of the content management unit (130) can classify the content by age, topic, and difficulty level and present it to the parent user. The content recommendation engine (132) can recommend personalized content by analyzing the learning history, preferences, and current developmental level of the infant user. The authentication processing unit (170) can verify whether the AI ​​doll (300) is properly registered to the parent user's account by checking the unique identification information of the AI ​​doll (300). In one embodiment, if the AI ​​doll (300) is not registered, the AI ​​doll (300) registration guide screen can be displayed on the touch display (250) of the user terminal (200).

[0129] In S630, the server electronic device (100) can play the first content and output the first utterance of the AI ​​doll (300) that proceeds in at least one of the playback of a fairy tale story or a thematic conversation. The content management unit (130) can load at least one of the fairy tale text, quiz questions, and conversation scripts corresponding to the first content from the content DB (133). The conversation processing unit (140) can generate the first utterance of the AI ​​doll (300) based on the loaded content and output it through the speaker (322) of the AI ​​doll (300). The first utterance may include at least one of a question regarding the actions or ending of a character within the fairy tale story, and a question that elicits the correct answer in a quiz. For example, the first utterance may be a quiz-type utterance such as, "The rabbit asked the turtle to race. Who will win?" In another embodiment, the first utterance may be a thematic conversation-type utterance such as, "Which animal friend do you want to meet today?"

[0130] In S640, the server electronic device (100) can collect the first response of the infant user to the first utterance. The voice input / output unit (320) of the AI ​​doll (300) can collect the infant user's voice through the microphone (321) and transmit it to the server electronic device (100). When the infant answers, such as "The turtle wins!" or "The rabbit is fast," the voice processing unit (120) can convert the collected voice into text. In one embodiment, the voice processing unit (120) can analyze the infant user's pronunciation accuracy and, if the pronunciation accuracy is below a threshold, convert it into a standard pronunciation for recognition. A detailed explanation of this will be provided later with reference to FIG. 7.

[0131] In S650, the server electronic device (100) can learn the first response of the infant user cumulatively, select at least one answer from a plurality of answers pre-stored in the database (150) based on the first response, and output it to the infant user as a second utterance. The data collection module (181) of the learning processing unit (180) can store the first response of the infant user in the database (150). The pattern analysis engine (182) can analyze the cumulatively stored first responses to identify at least one of the infant user's learning pattern, preferred answer type, and topic of interest. Based on the analyzed pattern, the conversation processing unit (140) can select at least one answer from a plurality of answers pre-stored in the database (150) that has a similarity to the infant user's first response above a threshold value as a second utterance. In one embodiment, if it is difficult to provide an appropriate answer to the infant user's first response, the response generation module (142) of the conversation processing unit (140) can generate a second utterance by utilizing a large-scale language model (LLM). The above-described large-scale language model can generate a second utterance using at least one of vocabulary, expressions, and sentence structures suitable for the age group of the infant user. A detailed explanation thereof will be provided later with reference to FIG. 9.

[0132] In one embodiment, the learning processing unit (180) can recommend subsequent content to be provided to the infant user by linking with the content recommendation engine (132) based on at least one of the learning pattern, preferred answer type, and topic of interest analyzed based on the first response of the infant user. For example, if the infant user shows a high correct answer rate in an animal-related quiz, the content recommendation engine (132) can recommend advanced content on the animal topic.

[0133] Figure 7 is a flowchart for explaining step S640 of Figure 6 in detail.

[0134] Referring to FIG. 7, in S641, the voice processing unit (120) of the server electronic device (100) can recognize the voice of an infant user collected through the microphone (321) of the AI ​​doll (300). The voice processing unit (120) can convert the collected analog voice signal into a digital signal and perform at least one preprocessing of noise canceling and echo removal. In one embodiment, the microphone (321) of the AI ​​doll (300) is composed of multiple units, and the voice recognition accuracy can be improved by applying beamforming technology.

[0135] In S642, the voice processing unit (120) can extract words related to at least one of a subject, an object, and a verb from the recognized infant user's voice. The voice processing unit (120) can analyze the sentence structure within the infant's utterance through natural language processing. Specifically, the voice processing unit (120) can separate the words present in the analyzed sentence into nouns, verbs, adjectives, and adverbs, and select the subject and object from the separated nouns based on the analyzed sentence structure. In one embodiment, words separated into adjectives and adverbs may be excluded from phoneme analysis. This may be to accurately identify the utterance intention by focusing on the subject, object, and verb that convey the core meaning in the infant's utterance.

[0136] In S643, the voice processing unit (120) can separate the extracted words into phoneme units including consonants and vowels. In one embodiment, the voice processing unit (120) can separate the words into initial consonant, medial vowel, and final consonant, and then extract each consonant and vowel. For example, the word "rabbit" can be separated into phoneme units of "t-o-kk-i".

[0137] In S644, the voice processing unit (120) can analyze at least one of the duration, rhythm, volume, speech rate, and stress of each separated phoneme. The voice processing unit (120) can analyze each phoneme after separating it with an individual filter. The analyzed result can be compared with a pronunciation dataset corresponding to the personal information of the infant user stored in the database (150). The pronunciation dataset may include at least one of a standard pronunciation dataset, an age-specific pronunciation characteristic dataset, and a regional pronunciation variation dataset. In one embodiment, the voice processing unit (120) can extract information about the phonemes that can be pronounced from the infant user's age information, extract individual voice range information for the phonemes from gender information, and extract rhythm, speech rate, and stress information of regional pronunciation phonemes from regional information and compare them with the pronunciation dataset.

[0138] In S645, the voice processing unit (120) can recognize the pronunciation of a child user by converting it to a standard pronunciation if the pronunciation accuracy is below a threshold based on the comparison result. For example, if a 3-year-old child pronounces "rabbit" as "tokki," the voice processing unit (120) can recognize it by converting it to "rabbit" by referring to an age-specific pronunciation characteristic dataset. Through this, the intention of speech can be accurately identified despite the child's inaccurate pronunciation. In another embodiment, even if there is a difference from the standard pronunciation due to a dialect of a specific region, the pronunciation can be converted to a standard pronunciation by referring to a regional pronunciation variation dataset.

[0139] Figure 8 is a flowchart illustrating the reaction prediction process for new content in step S630 of Figure 6.

[0140] Referring to FIG. 8, at S631, the content management unit (130) of the server electronic device (100) can determine whether the first content is new content that does not have a history of being previously selected. The content management unit (130) can check whether the first content is new content or previously used content by querying the content usage history of the corresponding infant user in the database (150). If the first content is new content, the process can proceed to step S632. If the first content is previously used content, the process can proceed directly to step S640 of FIG. 6.

[0141] In S632, the learning processing unit (180) can predict the first response expected to be provided by the infant user after outputting the first utterance of the AI ​​doll (300) and before collecting the first response to it from the infant user. The pattern analysis engine (182) can calculate the probability distribution of possible responses by analyzing at least one of the content provision history, age, gender, region, and preference selected by the infant user. For example, if the infant user previously preferred a specific pattern of response in an animal-related quiz, a similar pattern of response can be predicted in a new animal quiz as well. In one embodiment, the pattern analysis engine (182) can improve prediction accuracy by utilizing collaborative filtering to refer to response data of other infant users with similar profiles.

[0142] In S633, the data collection module (181) may store a predicted first response of an infant user in a database (150). The predicted first response stored may include at least one of the predicted response content, the prediction reliability, and the underlying data used for the prediction.

[0143] In S634, the pattern analysis engine (182) can compare the first response of a stored infant user with the first response of an infant user that has been collected with respect to the first utterance of an AI doll (300) in second content that is different from the first content. The second content may be content that is similar to the first content in at least one of the subject, difficulty, or format. The pattern analysis engine (182) can calculate the similarity between the first utterance of the first content and the first utterance of the second content, and select utterances with a similarity greater than or equal to a threshold value as comparison targets.

[0144] In S635, the pattern analysis engine (182) can compare the first response of the infant user expected for the first content with the first response of the infant user in the second content, and then extract a response in which the utterance correlation between the first content and the second content is determined to be greater than or equal to a threshold. The extracted response is stored in the database (150) and can be used to select an appropriate second utterance when collecting the actual first response of the infant user to the first content. In one embodiment, based on the extracted response, the conversation processing unit (140) can minimize the response delay time by preparing a group of second utterance candidates in advance.

[0145] Figure 9 is a flowchart for explaining step S650 of Figure 6 in detail.

[0146] Referring to FIG. 9, in S651, the data collection module (181) of the learning processing unit (180) can accumulate and store the first response of the infant user in the database (150). The data collection module (181) can store at least one of the text content, voice characteristics, response time, and emotional state of the first response together. The accumulated first responses can be classified and managed by infant user.

[0147] In S652, the pattern analysis engine (182) can identify at least one of the learning pattern, preferred answer type, and topic of interest of the infant user by analyzing the accumulated stored first responses. The learning pattern may include the reaction tendency shown by the infant user to a specific type of question. The preferred answer type may include at least one of the length, tone, and content style of the answer preferred by the infant user. The topic of interest may include content topics in which the infant user shows high engagement. In one embodiment, the pattern analysis engine (182) may recommend subsequent content to be provided to the infant user by linking with the content recommendation engine (132) based on the analyzed results.

[0148] In S653, the context analysis module (141) of the conversation processing unit (140) can calculate the similarity between each of the first response of the infant user and a plurality of answers stored in advance in the database (150). The plurality of answers can be classified and stored according to the type of the first utterance. For example, answers to the first utterance in the form of a quiz and answers to the first utterance in the form of a topical conversation can be classified and stored, respectively. The context analysis module (141) can calculate the similarity based on at least one of the semantic similarity of the first response, the keyword matching score, and the contextual appropriateness.

[0149] In S654, the conversation processing unit (140) can determine whether there is an answer with a calculated similarity greater than or equal to a threshold. The threshold can be dynamically adjusted according to the content type, conversation context, and the age of the infant user. If there is an answer with a similarity greater than or equal to the threshold, the process can proceed to step S655. If there is no answer with a similarity greater than or equal to the threshold, the process can proceed to step S656.

[0150] In S655, the conversation processing unit (140) can select an answer with a similarity level greater than or equal to a threshold value as a second utterance and output it through the speaker (322) of the AI ​​doll (300). In one embodiment, if there are multiple answers with a similarity level greater than or equal to a threshold value, the conversation processing unit (140) can select the answer with the highest similarity level as the second utterance. In another embodiment, the conversation processing unit (140) can select a suitable answer among multiple candidates as the second utterance by referring to the preferred answer type of the infant user.

[0151] In S656, if it is difficult to provide an appropriate response to the first response of the infant user, the response generation module (142) of the conversation processing unit (140) can generate a second utterance by utilizing a large-scale language model (LLM). The large-scale language model can generate the second utterance by using at least one of vocabulary, expressions, and sentence structures suitable for the age group of the infant user. For example, the second utterance can be generated using simple words and short sentences for a 3-year-old infant, and using a more complex sentence structure for a 6-year-old infant. The response generation module (142) can generate a natural second utterance that fits the context by referring to previous conversation content stored in the conversation history management unit (143). In one embodiment, the second utterance generated through the large-scale language model is stored in the database (150) and can be used as a candidate for a response to a similar first response in the future. The model optimization unit (183) can continuously improve the response quality of the large-scale language model by analyzing the infant user's response to the generated second utterance.

[0152] According to embodiments of the present invention, personalized conversational content can be provided by collecting responses from infant users in real time and accumulating learning through an AI doll. In particular, the infant user's voice is separated into phoneme units to analyze duration, rhythm, volume, speech rate, and stress; if the pronunciation accuracy falls below a threshold, it is converted to standard pronunciation for recognition, thereby accurately identifying the speech intent despite the infant's inaccurate pronunciation. Furthermore, regarding new content, expected responses are predicted in advance by analyzing the infant user's content provision history, age, gender, region, and preferences. By comparing these responses with those from existing content and extracting responses with high relevance to the utterance, appropriate responses are possible even when new content is introduced. Among multiple answers stored in advance in a database, an answer with a similarity level above a threshold is selected as the second utterance; however, if there is no appropriate answer, a large-scale language model is utilized to generate a second utterance using vocabulary, expressions, and sentence structures suitable for the infant user's age group, thereby enabling flexible responses to unexpected reactions. Additionally, by analyzing the infant user's learning patterns, preferred response types, and topics of interest to recommend subsequent content, continuous and systematic customized educational services can be provided.

[0153] The method according to the embodiments described above may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either individually or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiments, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiments, and vice versa.

[0154] Although the embodiments have been described above with reference to the limited drawings, those skilled in the art can apply various technical modifications and variations based on the above. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0155] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

Claim 1 A step of performing user authentication of a parent user by a server electronic device, and querying the account information of the parent user whose user authentication is complete to load personal information including the age, gender, and region of the infant user registered in the account; a step of selecting a first content based on the parent user's choice when user authentication is complete by the server electronic device, and checking whether an AI doll for providing the first content is registered; a step of playing the first content by the server electronic device and outputting a first utterance of the AI ​​doll proceeding in at least one of playing a fairy tale story or a thematic conversation; a step of collecting a first response of the infant user to the first utterance by the server electronic device. The method comprises the step of selecting at least one answer from a plurality of answers pre-stored in a database based on the first response by the server electronic device and outputting it to the infant user as a second utterance; and the step of collecting the first response of the infant user comprises the step of recognizing the infant user's voice by the server electronic device, analyzing the sentence structure within the recognized voice to separate words existing in the sentence into nouns, verbs, adjectives, and adverbs, and extracting words related to at least one of the subject, object, and verb by selecting the subject and object among the separated nouns based on the analyzed sentence structure, wherein the words separated into adjectives and adverbs are excluded from the phoneme analysis; and the step of separating the extracted words into initial consonants, medial vowels, and final consonants by the server electronic device, and then extracting each consonant and vowel to separate them into phoneme units including consonants and vowels.A method comprising the step of analyzing at least one of the duration, rhythm, volume, speech rate, and stress of each separated phoneme by the server electronic device, extracting information on pronounceable phonemes from age information among the loaded personal information of the infant user, extracting individual voice range information for phonemes from gender information, and extracting rhythm, speech rate, and stress information of regional pronunciation phonemes from region information, comparing the analyzed result with a pronunciation dataset including at least one of a standard pronunciation dataset, an age-specific pronunciation characteristic dataset, and a regional pronunciation variation dataset, and, based on the comparison result, converting and recognizing the infant user's pronunciation accuracy to a standard pronunciation if it is below a threshold value. Claim 2 delete Claim 3 A method according to claim 1, wherein, if the first content is new content for which no prior selection history exists, the server electronic device outputs a first utterance of an AI doll and, prior to collecting a first response from a child user thereto, predicts a first response expected to be provided by the child user; the server electronic device stores the predicted first response of the child user in a database; and the server electronic device compares the stored predicted first response of the child user with a first response from a child user already collected for a first utterance of an AI doll in a second content different from the first content. Claim 4 A method further comprising, in paragraph 3, a step of comparing a first reaction of an infant user expected for the first content with a first reaction of an infant user in the second content by the server electronic device, and then extracting a reaction in which the utterance correlation between the first content and the second content is determined to be greater than or equal to a reference value. Claim 5 In paragraph 3, the step of predicting a first response expected to be provided by the infant user comprises calculating a probability distribution of possible answers by analyzing at least one of the content provision history, age, gender, region, and preference selected by the infant user by the server electronic device. Claim 6 A method according to claim 1, further comprising the step of generating and outputting the second utterance by utilizing a large-scale language model when it is difficult to provide an appropriate response to the first response of the infant user by the server electronic device. Claim 7 In claim 6, the large-scale language model generates the second utterance using at least one of vocabulary, expressions, and sentence structures suitable for the age group of the infant user. Claim 8 A method according to claim 1, wherein the first content comprises at least one of a fairy tale, a quiz, and a thematic conversation, and the first utterance comprises at least one of a question regarding the action or ending of a character within the fairy tale story and a question that elicits a correct answer in the quiz. Claim 9 A method according to claim 1, further comprising: a step of analyzing at least one of the learning pattern, preferred answer type, and topic of interest of the infant user based on the first response of the infant user by the server electronic device; and a step of recommending subsequent content to be provided to the infant user based on the analyzed result by the server electronic device. Claim 10 A method according to claim 1, wherein a plurality of answers stored in advance in the database are classified and stored according to the type of the first utterance, and the server electronic device calculates the similarity between the first response of the infant user and each of the plurality of answers, and selects the answer with a calculated similarity greater than or equal to a threshold value as the second utterance.

Citation Information

Patent Citations

  • Targeting automation device and method

    KR101584687B1

  • Infant Learning Apparatus and Method Using The Same

    KR1020180012192A

  • An ai robot for child and a system for child social-emotional intelligence development comprising the same

    KR1020250102302A

  • Personalized Korean Speaking Learning System Based on Generative AI

    KR1020250146485A

  • Interactive friend linked cloud-based toy

    US20160361663A1