Data processing system

Abstract

The system according to the present embodiment includes a collection unit, an analysis unit, and an alarm unit. The collection unit collects daily behavior data of a user. The analysis unit analyzes the data collected by the collection unit and predicts the likelihood that the user belongs to a cognitive disorder preparer. The alarm unit issues an alarm based on the prediction result obtained by the analysis unit.

Description

Technical Field

[0001] The technology disclosed herein relates to a system. Background Technology

[0002] Patent Document 1 discloses a personalized chatbot control method executed by at least one processor, the method comprising: receiving user speech; adding the user speech to a prompt containing instructions related to a chatbot role; encoding the prompt; and inputting the encoded prompt into a language model to generate chatbot speech in response to the user speech.

[0003] Patent document 1: Japanese Patent Application Publication No. 2022-180282.

[0004] With current technology, it is difficult to detect dementia in a timely manner, which may lead to delays in taking appropriate treatment measures to slow its progression. Summary of the Invention

[0005] The system described in this embodiment includes a collection unit, an analysis unit, and an alarm unit. The collection unit collects users' daily behavioral data. The analysis unit analyzes the data collected by the collection unit and predicts the likelihood that a user belongs to the dementia predisposition group. The alarm unit issues an alarm based on the prediction results obtained by the analysis unit. Attached Figure Description

[0006] Figure 1 This is a conceptual diagram illustrating an example of the configuration of a data processing system according to the first embodiment.

[0007] Figure 2 This is a conceptual diagram illustrating an example of the main functions of the data processing apparatus and smart device according to the first embodiment.

[0008] Figure 3 This is a conceptual diagram illustrating an example of the data processing system configuration in the second embodiment.

[0009] Figure 4 This is a conceptual diagram illustrating an example of the main functions of the data processing device and smart glasses according to the second embodiment.

[0010] Figure 5 This is a conceptual diagram illustrating an example of the data processing system configuration in the third embodiment.

[0011] Figure 6 This is a conceptual diagram illustrating an example of the main functions of the data processing device and head-mounted terminal according to the third embodiment.

[0012] Figure 7 This is a conceptual diagram illustrating an example of the data processing system configuration in the fourth embodiment.

[0013] Figure 8 This is a conceptual diagram illustrating an example of the functions of the main parts of the data processing device and robot according to the fourth embodiment.

[0014] Figure 9 It represents an emotion graph that maps multiple emotions.

[0015] Figure 10 It represents an emotion graph that maps multiple emotions.

[0016] Explanation of reference numerals in the attached figures Data processing systems 10, 210, 310, and 410 12 Data processing device 14 Smart devices 214 Smart Glasses 314 Head-mounted terminal 414 Robot. Detailed Implementation

[0017] Hereinafter, an example of an implementation of the system involved in this disclosure will be described with reference to the accompanying drawings.

[0018] First, let's explain the terms used in the following description.

[0019] In the following embodiments, the processor (hereinafter referred to as "processor") can be a single computing device or a combination of multiple computing devices. Furthermore, a processor can be a single computing device or a combination of multiple computing devices. Examples of computing devices include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose computing on Graphics Processing Units), APU (Accelerated Processing Unit), or TPU (Tensor Processing Unit), etc.

[0020] In the following implementation, the labeled RAM (Random Access Memory) is a memory that temporarily stores information and is used by the processor as working memory.

[0021] In the following embodiments, the labeled memory is one or more non-volatile storage devices used to store various programs and various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), disk (e.g., hard disk) or magnetic tape, etc.

[0022] In the following implementation, the labeled Communication I / F (Interface) is an interface that includes a communication processor and an antenna, etc. The Communication I / F is responsible for communication between multiple computers. Examples of communication standards applicable to the Communication I / F include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0023] In the following implementation, "A and / or B" is synonymous with "at least one of A and B". That is, "A and / or B" means that it can be only A, only B, or a combination of A and B. Furthermore, in this specification, when "and / or" connects more than three items, the same approach as "A and / or B" applies.

[0024] [First Implementation] Figure 1 An example of the configuration of the data processing system 10 according to the first embodiment is shown.

[0025] like Figure 1 As shown, the data processing system 10 includes a data processing device 12 and an intelligent device 14. An example of the data processing device 12 is a server.

[0026] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0027] The smart device 14 includes a computer 36, a receiver 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. In addition, the receiver 38, output device 40, and camera 42 are also connected to the bus 52.

[0028] The receiving device 38 includes a touchscreen 38A and a microphone 38B, etc., for receiving user input. The touchscreen 38A receives user input generated by contact with an indicator (e.g., a pen or finger). The microphone 38B receives user input generated by sound by detecting the user's voice. The control unit 46A sends data representing user input received via the touchscreen 38A and microphone 38B to the data processing unit 12. In the data processing unit 12, a specific processing unit 290 (see...) Figure 2 Get the data that represents the user input.

[0029] The output device 40 includes a display 40A and a speaker 40B, etc., and presents data to the user by outputting data in a user-perceptible form (e.g., sound and / or text). The display 40A displays visual information such as text and images according to instructions from the processor 46. The speaker 40B outputs sound according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0030] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for sending and receiving various information between processor 46 and processor 28 via network 54.

[0031] Figure 2 An example of the main functions of the data processing device 12 and the smart device 14 is shown.

[0032] like Figure 2 As shown, in the data processing apparatus 12, specific processing is performed by the processor 28. A specific processing program 56 is stored in the memory 32. The specific processing program 56 is an example of a "program" as understood in this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 based on the specific processing program 56 executed on the RAM 30.

[0033] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. The emotion inference function using the emotion-specific model 59 (emotion-specific function) includes various inferences and predictions related to the user's emotions, but is not limited to this example. Furthermore, emotion inference and prediction also include, for example, emotion analysis (analysis).

[0034] In the smart device 14, specific processing is performed by the processor 46. A specific processing program 60 is stored in the memory 50. The specific processing program 60 is used in conjunction with the data processing system 10. The processor 46 reads the specific processing program 60 from the memory 50 and executes the read specific processing program 60 on the RAM 48. Specific processing is implemented by the processor 46 operating as a control unit 46A based on the specific processing program 60 executed on the RAM 48. Furthermore, the smart device 14 may also have the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and use these models to perform the same processing as the specific processing unit 290.

[0035] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device (e.g., a generation server) may have the data generation model 58. In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with the server device that has the data generation model 58. In addition, the data processing device 12 may be a server device or a user-owned terminal device (e.g., a mobile phone, robot, home appliance, etc.). Next, an example of the processing performed by the data processing system 10 of the first embodiment will be described.

[0036] (Example) The system described in this invention is a system that collects behavioral data and combines it with AI prediction data to issue alerts regarding the likelihood of being in the dementia predisposition group, thereby enabling early treatment. The system collects users' daily behavioral data, which is then analyzed by AI to predict the user's likelihood of belonging to the dementia predisposition group. Based on the prediction results, when the likelihood of being in the dementia predisposition group is high, the system issues an alert. This alert allows the user to seek medical attention at a medical institution as early as possible and receive appropriate treatment. For example, when collecting users' daily behavioral data, various devices are used to collect data such as the user's walking patterns, conversation content, and eating frequency, thereby gaining a detailed understanding of the user's daily behavior. Next, the collected behavioral data is analyzed by AI. Based on the collected data, the AI ​​analyzes the user's behavioral patterns and predicts the likelihood of belonging to the dementia predisposition group. For example, it analyzes changes in walking patterns, conversation content, and eating frequency to detect signs of dementia. Thus, the likelihood of belonging to the dementia predisposition group can be predicted based on the user's behavioral data. Based on the prediction results, when the likelihood of being in the dementia predisposition group is high, the system issues an alert. For example, an alert is issued to the user using a notification function. The alert content includes a message suggesting that the user may belong to the dementia predisposition group and recommends seeking medical attention at a medical institution as soon as possible. In this way, users can seek medical attention and receive appropriate treatment as early as possible, achieving early detection and treatment of dementia. Users can grasp the signs of dementia based on their own behavioral data and take countermeasures early. For example, noticing changes in walking patterns or conversation content and seeking medical attention can slow the progression of dementia. In addition, family members or caregivers can also detect signs of dementia early and take appropriate measures by sharing user behavioral data. Thus, by collecting behavioral data and combining it with AI prediction data, it is possible to issue alerts to the possibility of dementia predisposition and achieve early treatment. Specifically, this system collects walking patterns (e.g., daily steps, walking speed, and temporal acceleration data of walking rhythm: a one-dimensional array, sampling frequency of 50Hz, approximately 4.32 million samples per day), conversation content (e.g., voice data recorded at 16kHz, spoken sentences converted into text by a speech recognition engine, an average of 100 sentences per day), and eating frequency (e.g., three daily eating events from a food tracking app, each event accompanied by food content, time, and calorie information), etc. This data is automatically acquired by the collection department through various sensor devices or applications and stored in a database in chronological order. The AI ​​analysis department uses this multidimensional data as input, and employs, for example, convolutional neural networks (CNN), recurrent neural networks (RNN), or time-series analysis models based on Transformers, to automatically extract abnormal walking patterns (e.g., decreased walking speed, disordered rhythm), changes in conversation content (e.g., reduced vocabulary, delayed speech, decreased semantic consistency), and changes in eating frequency (e.g., reduced number of meals, irregular meal times).Examples of AI inputs include: (1) a day's walking acceleration data (a one-dimensional array of 4.32 million points, 50Hz × 86400 seconds), (2) a day's spoken text (e.g., 100 sentences such as "The weather is nice today" and "I've eaten"), and (3) a week's dietary events (each event is accompanied by structured data of dietary content, calories, and time). The AI ​​outputs are: (a) a dementia predisposition risk score (a continuous value from 0.0 to 1.0), (b) risk element labels (e.g., "abnormal walking pattern" and "decreased conversational content"), and (c) recommended actions (e.g., "recommendation to a medical institution" and "continuous observation"). For example, if a user's data for a week is input, the AI ​​will generate outputs such as "risk score 0.82", "elements: decreased walking speed, reduced conversational vocabulary", and "recommendation: recommendation to a medical institution". After receiving this output, the alert department uses smartphone notification functions, email, voice notification APIs, etc., to send specific alerts to the user and their family such as "high probability of dementia predisposition, please seek medical attention as soon as possible". Subsequent processing includes recording user responses after an alert is issued (e.g., whether they sought medical attention, whether they shared the data with family members), and using this data for future AI learning. In terms of technical effectiveness, this system eliminates the need for manual observation or consultation, automatically parsing massive amounts of time-series data in a high-dimensional space. It applies unconventional feature extraction rules (such as CNNs for detecting local patterns and RNNs for understanding long-term dependencies), achieving higher accuracy and earlier detection of dementia symptoms than ever before. This reduces the diagnostic burden on medical institutions, enabling users and their families to proactively take early intervention measures. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and corporate health management. Furthermore, by aggregating anonymized multi-user data, it can also be used for creating regional dementia risk maps and for public health policy applications.

[0037] The dementia predisposition detection system according to this embodiment includes a collection unit, an analysis unit, and an alarm unit. The collection unit collects daily behavioral data of the user. This data includes, but is not limited to, walking patterns, conversation content, and eating frequency. The collection unit can collect the user's walking patterns using a pedometer or accelerometer. Furthermore, the collection unit can also collect the user's conversation content using speech recognition technology. Further, the collection unit can collect the user's eating frequency through a diet tracking app. For example, the collection unit can use a pedometer to record the user's daily steps and an accelerometer to analyze walking rhythm and speed. Speech recognition technology converts the user's conversation content into text data and uses natural language processing technology to analyze the conversation content. The diet tracking app records the eating frequency based on the user's input of food content and frequency. The analysis unit analyzes the data collected by the collection unit to predict the likelihood that the user belongs to the dementia predisposition group. The analysis unit can utilize AI to analyze the collected data. Based on the collected data, the AI ​​analyzes the user's behavioral patterns and detects signs of dementia. For example, the AI ​​analyzes changes in walking patterns, conversation content, and eating frequency to predict the likelihood that the user belongs to the dementia predisposition group. AI can detect changes in walking patterns by analyzing changes in walking rhythm and speed, changes in conversation content by analyzing changes in conversation content and tone of voice, and changes in eating frequency by analyzing changes in the number and content of meals. The alert unit issues an alert based on the predictions obtained by the analysis unit. The alert unit can issue alerts via smartphone notifications, email, or voice notifications. The alert content includes a suggestion that the user may belong to the dementia predisposition group and recommends seeking medical attention as soon as possible. For example, the alert unit can issue alerts to users via smartphone notifications, email, or voice notifications. Therefore, the dementia predisposition group detection system according to this embodiment achieves early detection and early treatment of dementia predisposition group by collecting and analyzing users' daily behavioral data and issuing alerts. Specifically, this dementia predisposition detection system automatically acquires multidimensional data from the collection unit, including daily steps (e.g., 10,000 steps), walking speed (e.g., 1.2 m / s), temporal acceleration data of walking rhythm (e.g., 50Hz sampling, approximately 4.32 million samples daily), spoken sentences converted to text by a speech recognition engine (e.g., 100 sentences daily), and dietary events from a diet tracking app (e.g., 3 times daily, each event including dietary content, time, and calorie information). This data is then stored in a database in chronological order. The analysis unit uses this data as input and employs AI architectures such as Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Transformer-based temporal analysis models to automatically extract abnormal walking patterns (e.g., decreased walking speed, disordered rhythm), changes in conversation content (e.g., reduced vocabulary, delayed speech, decreased semantic consistency), and changes in dietary frequency (e.g., reduced frequency of meals, irregular meal times).Examples of AI inputs include a day's walking acceleration data (a one-dimensional array of 4.32 million points, 50Hz × 86400 seconds), a day's spoken text (e.g., 100 sentences such as "The weather is nice today" and "I've eaten"), and a week's dietary events (each event accompanied by structured data on food content, calories, and time). The AI ​​outputs a dementia predisposition risk score (a continuous value from 0.0 to 1.0), risk factor labels (e.g., "abnormal walking pattern," "decreased conversational vocabulary"), and recommended actions (e.g., "recommend seeking medical attention" and "continued observation"). For example, inputting a user's data for a week, the AI ​​will generate outputs such as "Risk score 0.82," "Factors: decreased walking speed, reduced conversational vocabulary," and "Recommendation: Recommend seeking medical attention." Upon receiving this output, the alert department uses smartphone notifications, email, and voice notification APIs to send specific alerts to the user and their family, such as "High likelihood of dementia predisposition; please seek medical attention as soon as possible." Subsequent processing includes recording the user's reaction after the alert is issued (e.g., whether they sought medical attention, whether they shared the information with family members), and using this data for future AI learning. In terms of technical effectiveness, this system eliminates the need for manual observation or consultation. It can automatically analyze massive amounts of time-series data in a high-dimensional space and apply unconventional feature extraction rules (such as CNNs to detect local patterns and RNNs to grasp long-term dependencies), achieving higher accuracy and earlier detection of dementia symptoms than ever before. This reduces the diagnostic burden on medical institutions, enabling users and their families to proactively take early intervention measures. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and corporate health management. Furthermore, by aggregating anonymized multi-user data, it can also be used for regional dementia risk mapping and public health policy applications.

[0038] The data collection unit can gather data such as users' walking patterns, conversation content, and eating frequency. It can collect walking patterns using a pedometer or accelerometer. For example, the pedometer records the user's daily steps, and the accelerometer analyzes walking rhythm and speed. Furthermore, it can use speech recognition technology to collect conversation content. For instance, it converts the user's conversation content into text data using speech recognition technology and then analyzes the content using natural language processing. Additionally, it can collect the user's eating frequency through a food tracking app. For example, the app allows users to input their food intake and frequency, recording the eating frequency. Thus, the data collection unit can gather detailed daily behavioral data, enabling a more accurate assessment of the signs of dementia. Specifically, the data collection unit automatically acquires daily step counts from a pedometer (e.g., 10,000 steps), walking rhythm time-series data from an accelerometer (e.g., 50Hz sampling, 4.32 million samples daily), spoken sentences converted to text by a speech recognition engine (e.g., 100 sentences daily), and dietary events from a diet tracking app (e.g., 3 times daily, each event including dietary content, time, and calorie information), and stores them in a database chronologically. While collecting this data, the collection unit can automatically detect missing or outliers and perform outlier removal or normalization as preprocessing. Furthermore, the collection unit can integrate data from multiple sensor devices or applications and manage it by assigning a unique ID to each user, ensuring data consistency and reliability. In terms of technical effectiveness, compared to manual recording or reliance on memory, the collection unit can automatically collect massive amounts of time-series data with high frequency and high accuracy, thus enabling earlier and more objective assessment of dementia symptoms. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and corporate health management.

[0039] The analysis unit can analyze user behavior patterns based on collected data to detect signs of dementia. The analysis unit can utilize AI to analyze the collected data. AI analyzes user behavior patterns based on the collected data to detect signs of dementia. For example, AI analyzes changes in walking patterns, conversation content, and eating frequency to predict the likelihood of a user belonging to the dementia predisposition group. AI can detect changes in walking patterns by analyzing changes in walking rhythm and speed, changes in conversation content by analyzing changes in conversation content and tone of voice, and changes in eating frequency by analyzing changes in the number and content of meals. Thus, the analysis unit can detect early signs of dementia by analyzing the collected data. Specifically, the analysis unit inputs temporal acceleration data of walking patterns (e.g., 50Hz sampling, 4.32 million samples per day), spoken text (e.g., 100 sentences per day), and eating event data (e.g., structured data of weekly food content, calories, and time) into the AI ​​model. The analysis unit uses convolutional neural networks (CNNs) to extract local abnormal patterns from walking data (e.g., decreased walking speed, rhythm disorder), and uses recurrent neural networks (RNNs) or Transformer-based models to detect long-term behavioral changes. The analysis unit utilizes natural language processing algorithms (such as BERT and LSTM) to quantitatively assess the vocabulary, semantic consistency, and speech delay of conversational content. For dietary data, the analysis unit automatically extracts indicators such as reduced meal frequency, irregular meal times, and nutritional imbalances. AI input examples include a day's walking acceleration data (one-dimensional array), a day's spoken text (100 sentences), and a week's dietary events (structured data). The AI ​​output consists of structured data such as a dementia predisposition risk score (0.0–1.0), risk factor labels (e.g., "abnormal walking pattern," "decreased conversation content"), and recommended actions (e.g., "suggest visiting a medical institution"). The analysis unit performs threshold determinations or branching processes based on these outputs to notify the alert unit or for subsequent learning data accumulation. In terms of technical effectiveness, the analysis unit does not rely on subjective human observation or consultation; it can automatically analyze massive amounts of multidimensional data in a high-dimensional space, applying unconventional feature extraction rules to detect dementia symptoms with higher accuracy and earlier than before. Applicable areas include home-based elderly care, health management in nursing homes, and telemedicine support.

[0040] The Alert Department can issue alerts via smartphone notifications, email, and voice notifications. For example, it can send alerts via smartphone notifications, email, or voice notifications. This allows the Alert Department to prompt users to seek medical attention promptly. Specifically, the Alert Department automatically generates alert information based on structured data received by the Analysis Department, including dementia pre-existing condition risk scores (e.g., 0.82), risk factor tags (e.g., decreased walking speed, reduced conversational vocabulary), and recommended actions (e.g., "suggest seeking medical attention"). The Alert Department considers user attributes (e.g., age, smartphone usage) and current location (e.g., at home, away) to select the optimal notification method (e.g., smartphone notifications, email, voice notification API). After issuing an alert, the Alert Department records the user's reaction (e.g., notification confirmation, whether to seek medical attention) and uses this data for future AI learning. The Alert Department can also simplify or refine the alert content based on the user's emotional state (e.g., stress, relaxation) or situation (e.g., in a hurry). In terms of technical effectiveness, the alarm system can quickly and appropriately convey important health information to users and their families, promoting early medical visits or proactive health behaviors, which helps to inhibit the progression of dementia and reduce the burden on healthcare. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and corporate health management.

[0041] The collection department anonymizes the collected data and will not provide it to any third party. The collection department can anonymize the collected data. For example, it can remove personal information from the collected data and mask the data. Thus, the collected data is anonymized and will not be provided to any third party. The collection department can ensure that the data is not provided to third parties by setting data management methods and access restrictions. Anonymizing the collected data protects user privacy. Specifically, the collection department removes personally identifiable information such as user ID, name, address, and contact information from the database and assigns each user a randomly generated unique identifier (e.g., UUID). The collection department associates behavioral data such as walking patterns, conversation content, and eating events with the anonymized ID to ensure that individuals cannot be identified. The collection department strictly controls database access permissions; third parties cannot access the data except for administrators and the parsing department. Furthermore, the collection department uses encrypted communication (e.g., TLS) during data transmission to reduce the risk of unauthorized external access or leakage. In terms of technical effectiveness, while highly protecting user privacy, the collection department can securely implement AI analysis and group statistical processing based on anonymized data, which helps improve user trust and compliance (e.g., GDPR). Applicable areas include data utilization in the medical and nursing field, creation of regional health risk maps, and application of public health policies.

[0042] The data collection unit can infer a user's emotions and adjust the timing of behavioral data collection based on these inferred emotions. For example, the collection unit can use facial expression recognition technology to infer a user's emotions. The collection unit can adjust the timing of behavioral data collection according to the user's emotions. For example, when a user is stressed, the collection unit reduces the number of data collection sessions to alleviate the user's burden. When the user is relaxed, the collection unit increases the number of data collection sessions to collect more detailed data. When the user is in a hurry, the collection unit adjusts the collection timing to collect only important data. Thus, by adjusting the collection timing based on the user's emotions, the collection unit can reduce the user's burden. Specifically, to infer user emotions, the collection unit inputs facial image data (e.g., a 128×128 pixel RGB image, 1 frame per second), voice data (e.g., a 1-second voice waveform sampled at 16kHz), and biosensor data (e.g., one-dimensional time-series data such as heart rate and skin conductance). The data collection unit inputs this data into a Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), or Multimodal Transformer model, outputting emotion classification labels (e.g., "stressed," "relaxed," "hurried") and emotion intensity scores (continuous values ​​from 0.0 to 1.0). For example, when facial images and voice waveforms are input simultaneously, the collection unit outputs probability distributions such as "stress: 0.78," "relaxed: 0.12," and "hurried: 0.10." Based on these output values, the collection timing control module automatically applies rule-based scheduling, such as "collecting every 10 minutes when stressed, every minute when relaxed, and only collecting important data immediately when hurried." The collection unit monitors the emotion inference results in real time, automatically reducing the collection frequency when thresholds are exceeded (e.g., stress score above 0.7). To minimize user burden and ensure necessary data, the collection unit can also implement emotion state-based collection timing optimization algorithms (e.g., automatic parameter adjustment using reinforcement learning). In terms of technical effectiveness, the collection unit does not rely on subjective human judgment or manual settings; through AI-powered high-precision emotion inference and automatic scheduling, it maximizes data collection efficiency without compromising user experience. This significantly reduces user stress and burden while consistently acquiring high-quality behavioral data required for dementia predisposition assessment. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and health management related to accompanying stress. Furthermore, statistical analysis of the relationship between multi-user emotional states and collection frequency can be applied to group optimization or personalized health support.

[0043] The data collection department can analyze users' past behavioral data and select the optimal collection method. For example, it can analyze past records or historical data. It can customize collection methods based on users' frequently performed behaviors. It can select the most effective collection method from users' past behavioral data. It can analyze users' past behavioral patterns and optimize collection methods. Therefore, by analyzing past behavioral data, the collection department can select the optimal collection method. Specifically, the collection department takes as input each user's time-series accumulated behavioral data (e.g., daily steps, number of conversations, number of eating events, sleep time, etc., from one week to one year of history). The collection department inputs this historical data into time-series analytical models (e.g., LSTM, GRU, autoregressive models) or clustering algorithms (e.g., k-means, DBSCAN) to automatically extract user behavioral patterns (e.g., differences in activity between weekdays and weekends, seasonal changes, changes during specific events). Based on the extracted patterns, the collection department personalizes and optimizes the collection frequency and target data types. For example, if user A walks at the same time every day for the past month, the collection department increases the sampling frequency of walking data during that time period. Conversely, User B only increased their conversations on rest days, so the data collection department intensified its collection of rest-day conversation data. Based on AI pattern recognition results, the data collection department's collection method selection module automatically sets "adjustment of collection frequency, target sensors, and data granularity by weekday, time period, and event type." The data collection department uses optimization algorithms (such as Bayesian optimization and reinforcement learning) to continuously update parameters that maximize collection efficiency (such as the data volume / user burden ratio). In terms of technical effectiveness, the data collection department does not rely on human experience or uniform settings; through AI historical analysis and automatic optimization, it achieves the optimal data collection strategy for each user. This reduces invalid data collection and user burden, efficiently obtaining the information needed for dementia pre-diagnosis testing. Applicable areas include individual health management, personalized medicine, behavior change support, and improving the efficiency of on-site nursing services. Furthermore, by aggregating historical data from multiple users and deriving the optimal collection strategy for the group, it can be applied to large-scale data collection optimization in regions or institutions.

[0044] When collecting behavioral data, the data collection department can filter data based on the user's current health status and living environment. The department can assess the user's health status, for example, based on medical data or self-reported health conditions. It can also assess the user's living environment, for example, based on their residential environment or lifestyle habits. Furthermore, the department can filter collected data based on the user's health status and living environment. For example, if a user's health is poor, the department can reduce the amount of data collected, alleviating the user's burden. When a user's living environment changes, the department collects data on adapting to the new environment. The department can filter collected data based on the user's health status and living environment. Therefore, by filtering data based on the user's health status and living environment, the department can reduce the user's burden. Specifically, the department uses medical data on the user's health status (e.g., electronic medical record diagnosis history, medication information, self-reported health status scores) and living environment data (e.g., structured data such as housing type, number of cohabitants, and lifestyle habit questionnaire results) as input. The data collection department inputs this data into a rule engine or decision tree model, automatically applying dynamic collection control rules, such as "halving the frequency of walking data collection when health is poor" and "prioritizing the collection of new behavioral patterns when the living environment changes." For example, when a user is hospitalized, the collection department stops collecting walking or outdoor data and instead strengthens indoor activity or sleep data collection. When a user moves or their cohabitants change, the collection department focuses on collecting data related to the new environment (e.g., new lifestyle, dietary patterns). The collection department monitors health status and changes in the living environment in real time, automatically adjusting collection targets, frequency, and granularity based on threshold judgments (e.g., health score below a certain value) or event triggers (e.g., signs of changes in the living environment). In terms of technical effectiveness, the collection department can flexibly control data collection according to individual user circumstances, minimizing invalid data collection and user burden, while ensuring the acquisition of the information needed for dementia pre-diagnosis detection. Thus, compared to traditional uniform collection methods, it achieves a balance between data quality and user experience. Applicable areas include chronic disease management, home healthcare, efficiency improvement of nursing field operations, and personalized health support. In addition, by analyzing health status and living environment change patterns in groups, it can be applied to public health policies or optimization of regional medical resource allocation.

[0045] The data collection unit can infer a user's emotions and prioritize the collected behavioral data based on these inferred emotions. For example, the collection unit can use facial expression recognition technology to infer user emotions. It can also prioritize the collected behavioral data based on the user's emotions. For instance, when a user is stressed, the collection unit prioritizes collecting important data. When a user is relaxed, the collection unit prioritizes collecting detailed data. When a user is in a hurry, the collection unit adjusts the priority of the collected data. Thus, by prioritizing data collection based on user emotions, the collection unit can prioritize collecting important data. Specifically, to infer user emotions, the collection unit inputs facial image data (e.g., a 128×128 pixel RGB image), voice data (e.g., a 16kHz sampled voice waveform), and biosensor data (e.g., heart rate, skin conductance). The collection unit then inputs this data into a multimodal AI model (e.g., a combination of CNN and RNN or a Transformer-based emotion inference model) and outputs emotion labels (e.g., "stressed," "relaxed," "in a hurry") and emotion intensity scores. Based on the sentiment inference results, the collection department implements a priority decision module to determine the priority of collected target data (such as walking patterns, conversation content, eating events, sleep data, etc.). For example, when stress is high, "only walking patterns and sleep data are collected"; when relaxed, "all data are collected in detail"; and when rushed, "only conversation content and eating events are collected." The collection department uses priority decision algorithms, such as weighted scoring or decision trees, to automatically adjust the collection targets, frequency, and granularity based on sentiment status. The collection department records the priority results in real time for subsequent AI analysis or user experience optimization. In terms of technical effectiveness, the collection department prioritizes the collection of important data based on the user's sentiment status, minimizing the user's burden while ensuring the acquisition of the information needed for dementia pre-diagnosis detection. Thus, compared to traditional uniform collection methods, it achieves a balance between data quality and user experience. Applicable areas include home-based elderly care, health management in nursing homes, health management with accompanying stress, and personalized health support. In addition, by analyzing the relationship between sentiment status and data priority in a group setting, it can be applied to group optimization or behavior change support.

[0046] When collecting behavioral data, the data collection department can consider the user's geographic location information and prioritize the collection of highly relevant data. The data collection department can collect the user's geographic location information. For example, the data collection department can use GPS data or location information services to collect the user's geographic location information. The data collection department can prioritize the collection of highly relevant data based on the user's geographic location information. For example, when the user is in a specific location, the data collection department prioritizes the collection of data related to that location. When the user moves, the data collection department prioritizes the collection of data related to the destination. Therefore, by considering the user's geographic location information, the data collection department can prioritize the collection of highly relevant data. Specifically, the data collection department uses GPS data (e.g., time-series data with latitude, longitude, altitude, and timestamps, spaced 24 hours apart) and Wi-Fi / Bluetooth beacon information obtained from the user's smartphone or wearable device as input. The data collection department compares this location information with a map database and automatically assigns location labels such as "home," "workplace," "park," and "medical institution." The data collection department prioritizes the data collection targets for each location tag, implementing rule engines or decision trees for dynamic control, such as "prioritizing sleep and dietary data at home, walking patterns in parks, and health status data in medical institutions." When users move, the department analyzes their movement paths and speeds, focusing on collecting data related to their destination (e.g., conversation content and walking patterns). The department uses AI models (e.g., random forests, gradient boosting) to score the relevance between location information and behavioral data, prioritizing the collection of highly relevant data. Technically, the department can automatically apply optimal data collection strategies based on user geographic location, suppressing invalid data collection and efficiently acquiring the information needed for dementia predisposition detection. This achieves a balance between improved data quality and reduced user burden. Applicable areas include home-based elderly care, out-of-home health management, behavioral monitoring in nursing homes, and regional medical collaboration. Furthermore, by analyzing the correlation patterns between location information and behavioral data across groups, it can be applied to regional health risk mapping and public health policy.

[0047] When collecting behavioral data, the data collection department can analyze users' social media activities and collect relevant data. For example, the department can analyze the frequency of posted content or activities. The department can select data to collect based on information shared by users on social media. It can collect highly relevant data from users' social media activities. Therefore, by analyzing users' social media activities, the department can collect highly relevant data. Specifically, the department automatically obtains publicly posted data from various social media platforms used by users (e.g., text-based, image-sharing, live-streaming, etc.) via API. The department then inputs the obtained posting data (e.g., structured data such as text content, posting time, posting frequency, image metadata, video playback counts, etc.) into a natural language processing engine or image analysis AI model. For example, when posting text, the department uses BERT or a Transformer-based language model to extract features such as sentiment (e.g., "positive," "negative," "neutral"), topic category (e.g., "health," "interests," "family"), vocabulary size, and grammatical complexity. When images are posted, the collection department uses a CNN image classification model to automatically determine the image content (e.g., "outdoor activities", "food", "group activities"), color tone, number of subjects, etc. Regarding activity frequency, the collection department uses a time series analysis model (e.g., LSTM) to extract the number of posts per day and the changing patterns of weekdays and time periods. AI input examples include: (1) weekly text posts (each piece of content, posting time, sentiment score), (2) image posting metadata (image feature vector, posting time), (3) time series data of posting frequency (daily posting frequency vector), etc. AI outputs are: (a) social activity score (0.0 to 1.0 continuous values, such as 0.85 for active, 0.20 for low activity), (b) activity tendency labels (e.g., "decreased activity", "increased health topics", "isolation tendency"), (c) collection priority (e.g., "outing data priority", "conversation data priority") and other structured data. For example, if a user's posting frequency is halved within a month and the content is biased towards negative tendencies such as "loneliness" and "fatigue", the collection department assigns a "decreased activity" label and strengthens the collection of outing frequency or conversation data. The data collection department automatically adjusts the goals, frequency, and granularity of behavioral data collection based on AI output. In terms of subsequent processing, when changes in social media activity are detected, the analysis or alert department is notified for risk assessment of the dementia predisposition group or to trigger an alert. Technically, the collection department does not rely on subjective human observation or simple statistical analysis; through high-dimensional feature extraction and multi-angle AI analysis, it can accurately and early detect changes in users' social activities or isolation tendencies. Therefore, it can grasp the social isolation or decreased activity risk factors of the dementia predisposition group based on objective data, enabling appropriate data collection and intervention.Applicable areas include home-based elderly care, monitoring of social activities in nursing homes, telemedicine support, corporate health management, and the creation of regional isolation risk maps. Furthermore, by aggregating anonymized social activity data from multiple users, it can be applied to group social activity trends or public health policies.

[0048] The analysis unit can infer a user's emotions and adjust the expression of the analysis based on the inferred emotions. For example, the analysis unit can use facial expression recognition technology to infer the user's emotions. The analysis unit can adjust the expression of the analysis according to the user's emotions. For example, when the user is relaxed, the analysis unit provides detailed analysis results. When the user feels stressed, the analysis unit provides concise analysis results. When the user is in a hurry, the analysis unit provides analysis results that highlight the key points. Thus, by adjusting the expression of the analysis according to the user's emotions, the analysis unit can provide users with easily understandable analysis results. Specifically, to infer user emotions, the analysis unit inputs facial image data (e.g., 128×128 pixel RGB images, acquired per second), voice data (e.g., 1-second voice waveform sampled at 16kHz), and biosensor data (e.g., one-dimensional time-series data of heart rate and skin conductance). The analysis unit inputs this data into a multimodal AI model (e.g., a combination of CNN+RNN or a Transformer-based emotion inference model) and outputs emotion labels (e.g., "stressed," "relaxed," "in a hurry") and emotion intensity scores (continuous values ​​from 0.0 to 1.0). For example, when facial images and voice waveforms are input simultaneously, the analysis unit outputs probability distributions such as "stress: 0.72", "relaxation: 0.18", and "hurry: 0.10". Based on the emotion inference results, the analysis unit determines the automatic application rules of the module according to the expression method of the analysis results, such as "only briefly prompting the key points when stress is high, attaching detailed charts and explanations when relaxed, and prompting a list of key points when hurried". The analysis unit automatically selects the presentation format of the analysis results (such as: text summary, detailed charts, infographics, voice explanations) to adapt to the user's emotional state. AI output examples include: (1) "stress state: only prompting key points (such as: risk score 0.82, recommending medical institutions for treatment)", (2) "relaxation state: detailed analysis (such as: walking speed change chart, conversation vocabulary change, detailed description of diet pattern)", (3) "hurry state: list of key points (such as: risk factors, recommended actions)", etc. In terms of subsequent processing, the user's reaction (such as: content confirmation, detailed display request) is recorded after the analysis result is prompted, which is used for the optimization of the expression method of the next analysis. In terms of technical effectiveness, the analysis unit eliminates the need for subjective human judgment or standardized output. Through AI-powered high-precision emotion inference and automatic output control, it maximizes the comprehensibility and utilization efficiency of the analysis results without compromising user experience. This reduces user stress and burden, ensuring the effective delivery of crucial information from dementia predisposition detection. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and health management related to accompanying stress. Furthermore, statistical analysis of the relationship between multi-user emotional states and analysis expression methods can be applied to group optimization or personalized health support.

[0049] During analysis, the analysis unit can adjust the level of detail based on the importance of behavioral data. The analysis unit can assess the importance of behavioral data. For example, it can assess the importance of behavioral data based on data impact or analysis priority. The analysis unit can adjust the level of detail based on the importance of behavioral data. For example, it can perform detailed analysis on important behavioral data and brief analysis on behavioral data with low importance. The analysis unit can adjust the level of detail based on the importance of behavioral data, thus enabling efficient analysis. Specifically, the analysis unit takes multidimensional behavioral data (such as walking pattern temporal acceleration data, conversation content text data, structured data of eating events, sleep data, social media activity data, etc.) received by the collection unit as input. The analysis unit uses AI models (such as random forest, gradient boosting, neural networks with attention mechanisms) to calculate the importance score (a continuous value from 0.0 to 1.0) for each data point. For example, if abnormal walking patterns contribute significantly to the risk of dementia, the "walking data importance score is 0.92," while if the conversation content changes little, the "conversation data importance score is 0.35." The analysis department automatically applies rules to the detailed analysis module based on the importance score, such as "detailed analysis for importance scores above 0.8 (e.g., time series decomposition, anomaly detection, chart generation), and brief analysis for importance scores below 0.5 (e.g., only mean and trend)". AI input examples include: (1) a week's walking acceleration data (one-dimensional array), (2) a week's conversation text (100 sentences × 7 days), (3) a week's dietary events (structured data), (4) sleep data (sleep time, quality, number of times turning over), and (5) social media activity data (posting frequency, content), etc. The AI ​​output is: (a) the importance score of each data type, (b) the detailed analysis target list, (c) the brief analysis target list, etc., structured data. For example, "walking data importance 0.92 → detailed analysis" and "conversation data importance 0.35 → brief analysis". The analysis department generates charts or anomaly detection reports from the detailed analysis results, while the brief analysis results only provide a text summary or mean. In terms of subsequent processing, the analysis results are sent to the alarm department or user prompt module to issue alarms or detailed explanations based on the important data. In terms of technical effectiveness, the analysis department does not rely on human experience or standardized analysis. Through AI-powered data importance assessment and automatic detail adjustment, it achieves optimal allocation of computing resources and improved analysis efficiency. This enables the high-precision and efficient extraction of key information needed for dementia pre-diagnosis detection from massive datasets. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and health management accompanied by big data analysis. Furthermore, by optimizing the importance assessment algorithm parameters for each user, it can be applied to personalized analysis or group optimization.

[0050] During parsing, the parsing unit can apply different parsing algorithms based on the category of behavioral data. The parsing unit can define the categories of behavioral data. For example, it can define categories based on the type of behavioral data or the object being parsed. The parsing unit can apply different parsing algorithms based on the category of behavioral data. For example, a walking parsing algorithm can be applied to walking pattern data. A natural language processing algorithm can be applied to conversation content data. A diet parsing algorithm can be applied to dietary frequency data. Therefore, by applying appropriate parsing algorithms based on the category of behavioral data, the parsing unit can improve parsing accuracy. Specifically, the parsing unit takes diverse behavioral data received by the collection unit (such as walking pattern temporal acceleration data, conversation content text data, structured dietary event data, sleep data, social media activity data, etc.) as input. The parsing unit implements a category classification module for each data type, automatically assigning category labels such as "walking," "conversation," "diet," "sleep," and "social activity." The parsing unit automatically selects the optimal AI algorithm for each category. For example, for walking patterns, a convolutional neural network (CNN) or a temporal anomaly detection algorithm (such as LSTM, autoregressive model) is used to extract walking speed or rhythm anomalies. The conversation content uses natural language processing models (such as BERT, LSTM, Transformer) to quantitatively assess vocabulary size, semantic consistency, and speech delay. Dietary data uses decision trees or rule engines to extract the frequency of meals, nutritional balance, and changes in calorie intake. Sleep data uses time-series analysis models (such as GRU, autoregressive models) to detect abnormalities in sleep time, quality, and number of times the user turns over. Social media activity is analyzed using a combination of natural language processing and image analysis AI to extract the sentiment of the published content and changes in activity frequency. Examples of AI inputs include: (1) daily walking acceleration data (one-dimensional array), (2) weekly conversation text (100 sentences × 7 days), (3) weekly dietary events (structured data), (4) sleep data (sleep time, quality, number of times the user turns over), and (5) social media posting data (text, image features), etc. The AI ​​outputs are: (a) anomaly detection scores for each category, (b) risk factor labels (such as "abnormal walking pattern", "decreased conversation content", "poor dietary balance"), and (c) recommended actions (such as "suggesting medical treatment" and "improving dietary habits"), etc. The analysis department integrates the analysis results from various categories to conduct a comprehensive risk assessment of pre-dementia groups. In terms of subsequent processing, the analysis results are sent to the alert department or user prompt module for detailed category descriptions or alerts. Technically, the analysis department does not rely on human experience or a uniform algorithm; by automatically selecting the optimal AI model for each data type, it significantly improves analysis accuracy and efficiency. This enables high-precision, early detection of pre-dementia signs from complex and diverse behavioral data. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and health management accompanied by big data analysis. Furthermore, by optimizing the category classification algorithm or analysis model parameters for each user, personalized analysis or group optimization can be applied.

[0051] The analysis unit can infer the user's emotions and adjust the length of the analysis based on the inferred emotions. For example, the analysis unit can use facial expression recognition technology to infer the user's emotions. The analysis unit can adjust the length of the analysis according to the user's emotions. For example, when the user is relaxed, the analysis unit performs a detailed analysis; when the user is stressed, the analysis unit performs a brief analysis; when the user is in a hurry, the analysis unit performs a summary analysis. Thus, by adjusting the length of the analysis according to the user's emotions, the analysis unit can provide the user with appropriate analysis results. Specifically, to infer the user's emotions, the analysis unit inputs facial image data (e.g., a 128×128 pixel RGB image, acquired per second), voice data (e.g., a 1-second voice waveform sampled at 16kHz), and biosensor data (e.g., one-dimensional time-series data of heart rate and skin conductance). The analysis unit inputs this data into a multimodal AI model (e.g., a combination of CNN and RNN or a Transformer-based emotion inference model) and outputs emotion labels (e.g., "stressed," "relaxed," "in a hurry") and emotion intensity scores (a continuous value from 0.0 to 1.0). For example, when facial images and voice waveforms are input simultaneously, the parsing unit outputs probability distributions such as "stress: 0.80", "relaxation: 0.10", and "hurry: 0.10". Based on the emotion inference results, the parsing unit automatically applies rules by the parsing result length control module, such as "only briefly prompting key points when stress is high, attaching detailed charts and explanations when relaxed, and prompting a list of key points when hurried". The parsing unit automatically adjusts the length of the parsing results (e.g., number of words in the text summary, number of charts, and level of detail in the explanatory text) to adapt to the user's emotional state. Examples of AI output include: (1) "stress state: only prompting key points (e.g., risk score 0.82, recommending medical institution for treatment)", (2) "relaxation state: detailed analysis (e.g., walking speed change chart, conversational vocabulary change, detailed explanation of dietary patterns)", (3) "hurry state: list of key points (e.g., risk factors, recommended actions)", etc. In terms of subsequent processing, the user's reaction (e.g., content confirmation, request for detailed display) is recorded after the parsing result is prompted, which is used for the next parsing length optimization. In terms of technical effectiveness, the analysis unit eliminates the need for subjective human judgment or standardized output. Through AI-powered high-precision emotion inference and automatic output control, it maximizes the comprehensibility and utilization efficiency of the analysis results without compromising user experience. This reduces user stress and burden, ensuring the effective delivery of crucial information from dementia predisposition detection. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and health management related to accompanying stress. Furthermore, statistical analysis of the relationship between multi-user emotional states and analysis length can be applied to group optimization or personalized health support.

[0052] During parsing, the parsing unit can determine the parsing priority based on the timing of behavioral data collection. The parsing unit can evaluate the timing of behavioral data collection. For example, it can assess the timing of behavioral data collection based on data freshness or collection time. The parsing unit can determine the parsing priority based on the timing of behavioral data collection. For example, it can prioritize parsing the latest behavioral data. It can refer to past behavioral data while also emphasizing the latest data. Therefore, the parsing unit can prioritize parsing the latest data by determining the parsing priority based on the timing of behavioral data collection. Specifically, the parsing unit takes as input the timestamp information (e.g., UNIX epoch seconds, date and time labels) attached to the behavioral data (e.g., walking patterns, conversation content, eating events, sleep data, social media activities, etc.) received by the collection unit. The parsing unit automatically calculates the data freshness (e.g., difference from the current time, elapsed time) and calculates a priority score (a continuous value from 0.0 to 1.0, with the latest data receiving a higher score). Based on the priority score, the parsing unit automatically applies rules by the priority parsing module, such as "priority above 0.8 is parsed immediately, and below 0.5 is processed later." Examples of AI inputs include: (1) daily walking acceleration data (with timestamps), (2) weekly conversation text (each sentence with recorded time), (3) weekly dietary events (event occurrence time), (4) sleep data (sleep onset and wake-up times), and (5) social media posting data (posting time). The AI ​​output consists of structured data such as: (a) parsing priority scores for each data type, (b) priority parsing target list, and (c) delayed parsing target list. For example, "walking data (today) priority 0.95 → immediate parsing" and "conversation data (one week ago) priority 0.40 → delayed processing". The parsing department performs detailed AI parsing according to the priority parsing target order, while delayed targets are processed in batches or parsed at low frequency. In terms of subsequent processing, the parsing results are sent to the alarm department or user prompt module, which issues alarms or explanations based on the latest data. In terms of technical effectiveness, the parsing department does not need to rely on human experience or unified parsing. Through AI data freshness assessment and automatic priority control, it achieves both real-time performance and parsing efficiency. As a result, it can quickly grasp the latest behavioral changes in the dementia pre-diagnosis group detection and achieve early intervention and appropriate response. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and real-time health monitoring. Furthermore, by optimizing the priority evaluation algorithm parameters for each user, it can be applied to personalized analysis or group optimization.

[0053] During parsing, the parsing unit can adjust the parsing order based on the relevance of behavioral data. The parsing unit can assess the relevance of behavioral data. For example, it can evaluate the relevance of behavioral data based on the correlation or relevance score. The parsing unit can adjust the parsing order based on the relevance of behavioral data. For example, it can prioritize parsing behavioral data with high relevance and process behavioral data with low relevance later. Therefore, by adjusting the parsing order based on the relevance of behavioral data, the parsing unit can perform parsing efficiently. Specifically, the parsing unit takes multidimensional behavioral data (such as walking patterns, conversation content, eating events, sleep data, social media activity, etc.) received by the collection unit as input. The parsing unit calculates the correlation coefficients (such as Pearson correlation, Spearman rank correlation) or covariance matrices between these data and calculates a relevance score (a continuous value from 0.0 to 1.0). Based on the relevance score, the parsing unit automatically applies rules determined by the parsing order module, such as "data pairs with a relevance of 0.8 or higher are parsed simultaneously, and those with a relevance of less than 0.5 are processed later." Examples of AI inputs include: (1) weekly walking acceleration data and sleep data (correlation analysis), (2) conversation content and social media activities (topic and sentiment commonality analysis), and (3) dietary events and health status data (correlation analysis of nutritional intake and changes in physical condition). The AI ​​output is structured data such as: (a) correlation score of data pairs, (b) priority parsing list, and (c) delayed parsing list. For example, "correlation between walking data and sleep data 0.85 → simultaneous parsing" and "correlation between conversation data and dietary data 0.30 → delayed processing". The parsing department prioritizes the integration of highly correlated data for comprehensive analysis to extract composite risk factors or behavioral change patterns. In terms of subsequent processing, the parsing results are sent to the alarm department or user prompt module to issue alarms or explanations based on relevant data. In terms of technical effectiveness, the parsing department does not need to rely on human experience or unified analysis. Through AI data correlation assessment and automatic sequence control, it can achieve the discovery and parsing efficiency improvement of complex behavioral patterns. As a result, it can extract composite risk factors with high accuracy and efficiency in the detection of dementia pre-diagnosis groups. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and health management accompanied by big data analysis. Furthermore, by optimizing the relevance assessment algorithm parameters for each user, it can be applied to personalized analysis or group optimization.

[0054] The alarm unit can infer a user's emotions and adjust the alarm's delivery method based on these inferences. For example, the alarm unit can use facial expression recognition technology to infer the user's emotions. The alarm unit can adjust the alarm's delivery method according to the user's emotions. For example, when the user is relaxed, the alarm unit issues a gentle alarm. When the user is stressed, the alarm unit issues a concise and clear alarm. When the user is in a hurry, the alarm unit issues a quick and easy alarm. Thus, by adjusting the alarm's delivery method based on the user's emotions, the alarm unit can issue appropriate alarms to the user. Specifically, to infer the user's emotions, the alarm unit inputs facial image data (e.g., a 128×128 pixel RGB image, acquired per second), voice data (e.g., a 1-second voice waveform sampled at 16kHz), and biosensor data (e.g., one-dimensional time-series data of heart rate and skin conductance). The alarm unit inputs this data into a multimodal AI model (e.g., a combination of CNN and RNN or a Transformer-based emotion inference model) and outputs emotion labels (e.g., "stressed," "relaxed," "in a hurry") and emotion intensity scores (a continuous value from 0.0 to 1.0). For example, when facial images and voice waveforms are input simultaneously, the alarm department outputs probability distributions such as "stress: 0.75", "relaxation: 0.15", and "urgency: 0.10". Based on the emotion inference results, the alarm department determines the automatic application rules of the alarm issuance method, such as "only a brief notification of key points when stress is high, a detailed and gentle notification with explanations when relaxed, and a quick notification with a list of key points when in a hurry". The alarm department automatically selects the notification method (e.g., smartphone notification, voice notification, email) and the tone of the notification text (e.g., polite, concise, and emphasizing key points) to adapt to the emotional state. AI input examples include: (1) facial image + voice waveform (acquired simultaneously), (2) heart rate, skin conductance response time series data, and (3) multimodal combination data. AI output examples include: "stress: 0.80 → brief notification", "relaxation: 0.60 → detailed notification", and "urgency: 0.70 → key point notification". After the alarm is issued, the user's reaction (e.g., notification confirmation, content comprehension) is recorded for optimization of the alarm issuance method in the next alarm. In terms of technical effectiveness, the alarm system eliminates the need for subjective human judgment or uniform notifications. Through AI-powered high-precision emotion inference and automated notification control, it maximizes the efficiency and comprehensibility of delivering important information without compromising user experience. This reduces user stress and burden, ensuring that crucial alarms for dementia predisposition detection are effectively delivered in an appropriate manner. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and health management related to accompanying stress. Furthermore, statistical analysis of the relationship between multi-user emotional states and notification methods can be applied to group optimization or personalized health support.

[0055] When issuing an alert, the alert department can adjust the level of detail based on the importance of the predicted outcome. For example, the alert department assesses the importance of the predicted outcome. This assessment can be based on the impact of the predicted outcome and the alert's priority. The alert department can also adjust the level of detail based on the importance of the predicted outcome. For instance, a detailed alert will be issued for important predicted outcomes, while a concise alert will be issued for less important ones. The alert department can also automatically adjust the level of detail based on the importance of the predicted outcome. In this way, the alert department can adjust the level of detail based on the importance of the predicted outcome, thereby appropriately conveying important information. Specifically, the alert department takes structured data received from the analysis department, such as a dementia pre-existing condition risk score (a continuous value from 0.0 to 1.0), risk factor labels (e.g., "abnormal walking pattern," "decreased conversation content"), and recommended actions (e.g., "recommend seeking medical attention"), as input. The alert department uses AI models (e.g., random forest, gradient boosting, neural networks with attention mechanisms) to calculate the importance score (0.0 to 1.0) for each predicted outcome. For example, a risk score above 0.9 is rated as "importance 0.95", and a risk score below 0.5 is rated as "importance 0.30". Based on the importance score, the alarm department automatically applies rules such as "importance above 0.8 is a detailed alarm, and below 0.5 is a concise alarm". AI input examples include: (1) risk score 0.92 + element "decreased walking speed" + recommendation "seek medical treatment", (2) risk score 0.45 + element "reduced eating frequency" + recommendation "observed", etc. AI output examples include: "importance 0.92 → detailed alarm (including risk element, recommended action, chart)", "importance 0.30 → concise alarm (only key points)", etc. The alarm department adds a risk change chart and element description when issuing a detailed alarm, and only prompts the risk score and recommended action when issuing a concise alarm. After the alarm is issued, the user's reaction (such as: content confirmation, detailed display request) is recorded for the next detail optimization. The technological benefits are that the alarm department no longer needs to rely on manual rules of thumb or uniform notifications. Through AI-driven importance assessment and automatic detail control, it achieves optimal allocation of computing resources and improved notification efficiency. This allows for the delivery of essential information to users without omission, promoting rapid response and comprehension. Applicable areas include home-based elderly care, health management in elderly care institutions, telemedicine support, and health management projects accompanied by big data analysis. Furthermore, by optimizing the importance assessment algorithm parameters for each user, it can also be applied to personalized notifications and group optimization.

[0056] When issuing an alert, the alarm department can employ different alert methods based on the user's attribute information. For example, the alarm department collects user attribute information such as age, gender, and occupation. Based on this attribute information, the alarm department can select the optimal alert method. For instance, for the elderly, the alarm department prioritizes voice notifications; for younger users, it uses smartphone notifications. The alarm department can automatically select the optimal alert method based on user attribute information. In this way, the alarm department can choose the most suitable alert method based on user attribute information, thus sending alerts to users appropriately. Specifically, the alarm department uses structured data such as user attribute information (e.g., age, gender, occupation, smartphone usage, presence of hearing / visual impairment, and residential status) as input. Based on this attribute information, the alarm department uses an AI model (e.g., decision tree, rule engine) to automatically select the optimal notification method (e.g., voice notification, smartphone notification, email, family proxy notification). For example, for the elderly with low smartphone usage, "voice notification + family proxy notification" is prioritized; for younger users with high smartphone usage, "smartphone notification" is selected. AI input examples include: (1) Age 75, low smartphone usage frequency, no hearing impairment → voice notification; (2) Age 35, high smartphone usage frequency → smartphone notification; (3) Age 80, visual impairment → voice notification + family proxy notification, etc. AI output is a notification method label such as "voice notification", "smartphone notification", "email notification", "family proxy notification" etc. The alarm department records the notification method selection results and uses them for the next optimization based on user feedback (such as notification confirmation rate, content understanding). The technical effect is that the alarm department does not need to rely on human subjective judgment or uniform notification. Through AI attribute information analysis and automatic notification method selection, it can achieve optimal information delivery for each user. This can prevent notification omissions or misunderstandings and promote a fast and reliable response. Applicable fields include home care for the elderly, health management of elderly care institutions, remote medical support, personalized health management projects, etc. In addition, by analyzing the relationship between group attribute information and notification methods, it can also be applied to group optimization and support for socially vulnerable groups.

[0057] The alarm unit can infer the user's emotions and determine the priority of alarms based on the inferred emotions. For example, the alarm unit can infer user emotions using facial expression recognition technology. The alarm unit can determine the priority of alarms based on the user's emotions. For example, when the user is relaxed, important alarms are prioritized; when the user is stressed, concise alarms are prioritized; when the user needs to deal with something urgently, alarms that can be dealt with quickly are prioritized. In this way, the alarm unit can determine the alarm priority based on the user's emotions, thus prioritizing important alarms. Specifically, the alarm unit takes facial image data (128×128 pixel RGB image), audio data (16kHz sampled audio waveform), and biosensor data (heart rate, skin conductance response) as input, and uses a multimodal AI model (CNN+RNN or Transformer-based) to output emotion labels ("stress", "relaxed", "urgent need to deal with") and emotion intensity scores (0.0–1.0). The alarm department inputs the sentiment inference results and alarm content (such as risk score, element label, recommended action) into the priority decision module, and automatically applies rules such as "prioritize important alarms when stress is high, prioritize detailed alarms when relaxed, and prioritize immediate response alarms when urgent." AI input examples include: (1) sentiment score "stress 0.85" + multiple alarm candidates, (2) sentiment score "relaxation 0.70" + multiple alarm candidates, etc. AI output is priority labels such as "prioritize important alarms", "prioritize concise alarms", and "prioritize immediate alarms". The alarm department records the priority results and uses them for the next optimization based on user reactions (such as notification confirmation and response speed). The technical effect is that the alarm department does not need to rely on human subjective judgment or uniform notification. Through AI sentiment inference and automatic priority control, it maximizes the efficiency of important information transmission and response speed without harming the user experience. This can reduce user stress and burden and ensure that important alarms for dementia pre-diagnosis are reliably delivered in the appropriate order. Applicable fields include home care for the elderly, health management of elderly care institutions, remote medical support, and health management projects with accompanying stress. Furthermore, statistical analysis of the relationship between multi-user emotional states and alarm priorities can also be applied to group optimization and personalized health support.

[0058] When issuing an alert, the alarm department can consider the user's geographic location information to select the optimal alarm delivery method. The alarm department collects the user's geographic location information, such as using GPS data or location information services. Based on this information, the alarm department can select the optimal alarm delivery method. For example, when the user is at home, an alert is issued using a smartphone notification function; when the user is away, a voice notification is prioritized. The alarm department can automatically select the optimal alarm delivery method based on the user's geographic location information. In this way, the alarm department can select the most suitable alarm delivery method based on the user's geographic location information, thus delivering the alert to the user in an appropriate manner. Specifically, the alarm department takes GPS data (including latitude, longitude, altitude, and timestamp time-series data, with 1-minute intervals over 24 hours) and Wi-Fi / Bluetooth beacon information obtained from the user's smartphone or wearable device as input. The alarm department compares this location information with a map database and automatically assigns location labels such as "home," "workplace," "park," and "medical institution." The alarm department applies notification method selection rules for each location label (e.g., smartphone notification at home, voice notification when away, family proxy notification at medical institutions) and implements a rule engine. AI input examples include: (1) GPS data + location tag "home", (2) GPS data + location tag "out", (3) Wi-Fi beacon + location tag "medical institution", etc. AI output is a notification method tag such as "smartphone notification", "voice notification", "family representative notification" etc. The alarm department records the notification method selection results and uses them for the next optimization based on user feedback (such as notification confirmation rate, content understanding). The technical effect is that the alarm department does not need to rely on human subjective judgment or uniform notification. Through AI location information analysis and automatic notification method selection, it can achieve optimal information delivery based on the user's situation. This can prevent notification omissions or misunderstandings and promote a fast and reliable response. Applicable fields include home care for the elderly, out-of-home health management, behavior monitoring of elderly care institutions, regional medical collaboration, etc. In addition, by analyzing the relationship between location information and notification methods through group analysis, it can also be applied to regional optimal notification strategies and public health policies.

[0059] When issuing an alert, the alert department can analyze the user's social media activity and adjust the content of the alert. For example, the alert department analyzes the user's social media activity, such as parsing the content posted and analyzing the frequency of activity. The alert department can issue relevant alert content based on the user's social media activity. The alert department can adjust the alert content according to the information shared by the user on social media and select the optimal alert content from the social media activity. In this way, the alert department can analyze the user's social media activity and issue highly relevant alert content. Specifically, the alert department automatically obtains the user's publicly posted data (text content, posting time, posting frequency, image metadata, video playback count, and other structured data) on multiple social media platforms through API. The alert department inputs the obtained posting data into a natural language processing engine and an image analysis AI model to extract the sentiment tendency of the posted content (e.g., "positive", "negative", "neutral"), topic classification (e.g., "health", "interest", "family"), activity frequency change patterns, etc. AI input examples include: (1) a week's text posts (text, posting time, sentiment score), (2) metadata of image posts (image feature vector, posting time), (3) time series data of posting frequency (daily posting count vector), etc. The AI ​​output consists of structured data such as (a) a social activity score (0.0–1.0), (b) activity tendency tags (e.g., “declining activity,” “increased health topics,” “isolation tendency”), and (c) alert content adjustment instructions (e.g., “suggest going out,” “suggest increasing communication”). Based on these AI outputs, the alert department automatically applies rules such as “issue an alert suggesting going out or communicating when activity declines, issue a health maintenance alert when health topics increase, and issue family or community support information when there is an isolation tendency” through the alert content generation module. The alert department records the alert content adjustment results and uses them for future optimization based on user feedback (e.g., notification confirmation, behavioral changes). The technical effect is that the alert department does not need to rely on subjective human observation or uniform notifications. Through AI social activity analysis and automatic content adjustment, it achieves optimal information delivery based on the user's social situation. This can promote the early detection and appropriate intervention of social isolation and declining activity. Applicable areas include home-based elderly care, social activity monitoring in elderly care institutions, telemedicine support, health management projects, and the creation of regional isolation risk maps. In addition, by analyzing the relationship between social activity and alert content through group analysis, it can also be applied to public health policies and regional support strategies.

[0060] The system described in this implementation is not limited to the examples above; for instance, various modifications can be made. Specifically, the AI ​​model architecture can be switched to various structures such as convolutional neural networks, recurrent neural networks, Transformer-based models, decision trees, and gradient boosting. Data collection objects can also be expanded to include walking patterns, conversation content, dietary events, sleep data, social media activity, physiological data (heart rate, blood pressure, body temperature), and environmental sensor data (room temperature, humidity, illuminance). Furthermore, the input data format for AI can also be various variations such as one-dimensional time-series arrays, two-dimensional image tensors, structured tables, natural language text, and multimodal data. AI output can also support various forms such as risk scores, anomaly detection labels, recommended actions, detailed analysis reports, charts, infographics, and voice descriptions. Subsequent processing includes issuing alarms, optimizing notification methods / content / timing based on user attributes / emotions / location information / social activities, recording user reactions and feeding them back into AI learning data, creating group-level risk maps, and applying them to public health policies. The technical benefits are that this system breaks away from the traditional reliance on single data / single models. By combining multiple data sources, multiple AI models, diverse outputs, and flexible post-processing, it achieves a comprehensive improvement in parsing accuracy, notification efficiency, user experience, and social value. Applicable areas include home-based elderly care, health management in elderly care institutions, telemedicine support, health management projects, regional health risk map creation, public health policy, and personalized medical / nursing support.

[0061] In addition to collecting user behavioral data, the collection department can also collect users' physiological data. For example, by collecting physiological data such as heart rate, blood pressure, and body temperature, a more detailed understanding of the user's health status can be obtained. The analysis department analyzes the collected physiological data and combines it with behavioral data to predict the likelihood that the user belongs to the dementia predisposition group. For example, analyzing fluctuations in heart rate, changes in blood pressure, and abnormalities in body temperature can detect signs of dementia. The alarm department can issue alarms about the user's health status based on the analysis results. For example, an alarm can be issued suggesting medical attention when the heart rate is abnormally high or blood pressure fluctuates sharply. Thus, users can not only take measures against dementia but also other health problems early. Specifically, the collection department automatically acquires physiological data such as heart rate (e.g., time-series data at 1-minute intervals), blood pressure (e.g., twice-daily measurements), and body temperature (e.g., four times-daily measurements) through wearable devices or home biosensors and stores them in a time-series format in the database. The analysis unit inputs these physiological data into a time series analysis model (e.g., LSTM, autoregressive model) or anomaly detection algorithm (e.g., Isolation Forest, change point detection) to automatically extract rapid fluctuations in heart rate, abnormal blood pressure patterns, and sustained increases / decreases in body temperature. Examples of AI inputs include: (1) a week's heart rate time series data (1-minute intervals), (2) a month's blood pressure measurements (morning and evening), and (3) a week's body temperature measurements (4 times daily). The AI ​​output consists of structured data such as (a) a physiological data anomaly score (0.0–1.0), (b) anomaly element labels (e.g., "heart rate fluctuations", "rapid changes in blood pressure", "abnormal body temperature"), and (c) recommended actions (e.g., "recommend seeking medical attention" and "observation"). Based on these outputs, the alarm unit issues alarms to users or their families such as "abnormally high heart rate, recommend seeking medical attention" and "rapid fluctuations in blood pressure require attention." Subsequent processing involves recording the user's reaction after the alarm is issued (e.g., whether they seek medical attention or share the data with their family), and using this data as learning data for the next AI session. The technological benefits are that the collection, analysis, and alarm departments do not rely on manual observation or recording. Through AI-powered high-frequency, high-precision physiological data analysis and automatic alarm issuance, it can not only detect pre-dementia groups early but also identify and address various health risks at an early stage. This significantly improves user health maintenance, prevents serious illnesses, and reduces the burden of healthcare. Applicable areas include home-based elderly care, chronic disease management, telemedicine support, health management projects, and personalized medicine. Furthermore, by aggregating anonymized physiological data from multiple users, it can also be used for creating population health risk maps and for public health policy.

[0062] The analysis department can detect changes in users' social activities based on their behavioral data. For example, when users go out less frequently or communicate less with friends, signs of social isolation can be detected. The analysis department can evaluate these changes as signs of dementia and reflect them in the prediction results. When the alarm department detects signs of social isolation, it can issue an alarm to users to increase their social activities. For example, it can issue an alarm suggesting participation in local or community activities. In this way, users can prevent social isolation and reduce the risk of dementia. Specifically, the analysis department takes multidimensional behavioral data (such as time series data of outings, frequency of conversations with friends, frequency of social media posts, activity participation records, and other structured data) received by the collection department as input. The analysis department inputs this data into a time series analysis model (such as LSTM, autoregressive model) or a clustering algorithm (such as k-means, DBSCAN) to automatically extract patterns such as decreased outing frequency, decreased communication frequency, and reduced activity posts. Examples of AI inputs include: (1) number of outings per month (daily numerical vector), (2) number of conversation events per week (including event time and object information), (3) number of social media posts per month (daily vector), etc. The AI ​​output consists of structured data such as (a) a social activity score (a continuous value from 0.0 to 1.0, e.g., 0.20 indicates a decrease in activity), (b) isolation tendency labels (e.g., "decreased activity," "reduced communication"), and (c) recommended actions (e.g., "suggest going out," "suggest increasing communication"). For example, if the number of outings halve within a month and the number of conversation events also decreases, the analysis department assigns a "decreased activity" label and increases the risk score. Based on these outputs, the alert department sends alerts to users or their families such as "Recently, outings and communication have decreased; it is recommended to participate in local or community activities." Subsequent processing involves recording the user's response after the alert is issued (e.g., whether they participated in activities, whether they increased communication), and using this data for the next AI learning iteration. The technical advantage is that the analysis department does not need to rely on manual observation or simple counting; through AI high-dimensional feature extraction and multi-angle analysis, it can detect signs of social isolation with high accuracy and early detection. This allows for the understanding of social isolation risk factors in the dementia predisposition group based on objective data, enabling appropriate data collection and intervention. Applicable areas include home-based elderly care, social activity monitoring in elderly care institutions, telemedicine support, health management projects, and the creation of regional isolation risk maps. Furthermore, by aggregating anonymized social activity data from multiple users, it can also be used to analyze group social activity tendencies and public health policies.

[0063] In addition to collecting user behavior data, the collection department can also collect user sleep data. For example, by collecting data such as sleep duration, sleep quality, and number of times the user turns over, a detailed understanding of the user's sleep state can be obtained. The analysis department analyzes the collected sleep data and combines it with behavioral data to predict the likelihood of the user belonging to the pre-dementia group. For example, analyzing reduced sleep duration, decreased sleep quality, and increased number of times the user turns over can detect signs of dementia. The alarm department can issue alarms about the user's sleep state based on the analysis results. For example, when sleep duration is short or sleep quality is poor, an alarm can be issued suggesting improvements to the sleep environment. As a result, users can improve their sleep state and reduce the risk of dementia. Specifically, the collection department automatically acquires time-series data such as sleep duration (e.g., daily sleep onset / wake-up time), sleep quality (e.g., deep sleep / light sleep ratio, sleep efficiency), and number of times the user turns over (e.g., number of times per night) through wearable devices or bed sensors and stores it in a database. The analysis department inputs this data into a time series analysis model (e.g., LSTM, autoregressive model) or anomaly detection algorithm (e.g., Isolation Forest, change point detection) to automatically extract abnormal patterns such as reduced sleep time, decreased sleep efficiency, and increased number of times tossing and turning. Examples of AI inputs include: (1) weekly sleep time data (daily numerical vector), (2) weekly sleep efficiency data (time series of deep sleep ratio), and (3) weekly number of times tossing and turning (numerical value per night). The AI ​​output consists of structured data such as (a) sleep abnormality score (0.0-1.0), (b) abnormal element labels (e.g., "reduced sleep time", "decreased sleep efficiency", "increased tossing and turning"), and (c) recommended actions (e.g., "suggest improving sleep environment" and "suggest seeking medical treatment"). For example, if sleep time drops to below 5 hours on average within a week and the number of times tossing and turning increases, the analysis department assigns a "sleep abnormality" label and increases the risk score. Based on these outputs, the alarm department issues alarms to users or their families such as "recently, sleep time has become shorter; it is recommended to improve the sleep environment or seek medical treatment." The subsequent processing involves recording the user's response after the alarm is issued (e.g., whether the sleep environment was improved, whether medical attention was sought), and using this data for future AI learning. The technical benefits are that the collection, analysis, and alarm departments no longer rely on manual observation or recording. Through high-frequency, high-precision AI-powered sleep data analysis and automatic alarm issuance, not only can early detection of dementia predisposition groups be achieved, but also early detection and response to various health risks. This significantly improves user health maintenance, serious illness prevention, and reduces the burden on healthcare. Applicable areas include home-based elderly care, sleep disorder management, telemedicine support, health management projects, and personalized medicine. Furthermore, by aggregating anonymous sleep data from multiple users, it can also be used for creating population health risk maps and public health policies.

[0064] The analysis unit can detect changes in a user's cognitive function based on their behavioral data. For example, by analyzing the completion time and error frequency of daily tasks, a decline in cognitive function can be detected. The analysis unit can assess these changes as signs of dementia and reflect them in the prediction results. When a decline in cognitive function is detected, the alert unit can issue alerts to the user suggesting training or activities to maintain cognitive function. For example, alerts can be issued suggesting the use of brain training apps, reading, puzzles, and other activities to stimulate cognitive function. In this way, users can maintain cognitive function and reduce the risk of dementia. Specifically, the analysis unit takes as input the daily task execution data (such as shopping list completion time, medication management app operation records, household chores execution records, error occurrence frequency, and other structured data) received by the collection unit. The analysis unit inputs this data into a time series analysis model (such as LSTM, autoregressive model) or anomaly detection algorithm (such as Isolation Forest, change point detection) to automatically extract abnormal patterns such as delayed task completion time, increased error frequency, and disordered operation steps. Examples of AI inputs include: (1) weekly task completion time (time vector for each task), (2) monthly error occurrences (daily values), and (3) application operation records (time-series data of operation steps). The AI ​​output consists of structured data such as (a) cognitive function decline score (0.0–1.0), (b) abnormal element labels (e.g., “task delay”, “increased errors”, “operational confusion”), and (c) recommended actions (e.g., “brain training”, “reading”, “puzzle”). For example, if the average task completion time is delayed by 30% within a week and the number of errors increases, the analysis department assigns a “cognitive function decline” label and increases the risk score. Based on these outputs, the alarm department issues alarms to users or their families such as “recent task completion time has been extended; brain training, reading, or puzzle activities are recommended.” The subsequent processing involves recording the user’s reaction after the alarm is issued (e.g., whether training was conducted, activity content recorded) and using it as learning data for the next AI session. The technical effect is that the analysis department does not need to rely on manual observation or simple recording; through AI high-dimensional feature extraction and multi-angle analysis, it can detect signs of cognitive function decline with high accuracy and early detection. This allows for the identification of risk factors for cognitive decline in individuals at risk of dementia based on objective data, facilitating appropriate data collection and intervention. Applicable areas include home-based elderly care, cognitive function monitoring in elderly care institutions, telemedicine support, health management projects, and personalized cognitive support. Furthermore, by aggregating anonymized cognitive function data from multiple users, it can also be used for assessing group cognitive tendencies and public health policies.

[0065] In addition to user behavioral data, the collection department can also collect users' dietary data. For example, by collecting data on dietary content, calorie intake, and nutritional balance, it can gain a detailed understanding of users' eating habits. The analysis department analyzes the collected dietary data and combines it with behavioral data to predict the likelihood of a user belonging to the dementia predisposition group. For example, it analyzes nutritional imbalances or excessive or insufficient calorie intake to detect signs of dementia. The alert department can issue alerts about users' eating habits based on the analysis results. For example, it can issue alerts suggesting improvements to eating habits when there are nutritional imbalances or excessive or insufficient calorie intake. As a result, users can improve their eating habits and reduce the risk of dementia. Specifically, the collection department automatically acquires structured data such as dietary content (e.g., main dishes, side dishes, types of staple foods), calorie intake (e.g., calorie value per meal), and nutritional balance (e.g., protein, fat, carbohydrate, vitamin, and mineral intake) through diet recording applications or wearable devices, and stores it in a time-series database. The analysis department inputs this data into decision trees or rule-based diet analysis engines and time-series analysis models (e.g., LSTM) to automatically extract nutritional imbalances, excessive or insufficient calorie intake, and changes in eating patterns. Examples of AI inputs include: (1) weekly dietary data (category, calories, and nutrient vectors for each meal), (2) monthly calorie intake (daily values), and (3) weekly nutritional balance data (vectors of nutrient intake). AI outputs are structured data such as (a) a dietary anomaly score (0.0–1.0), (b) anomaly element labels (e.g., “nutritional balance deviation,” “excessive calories,” “insufficient calories”), and (c) recommended actions (e.g., “suggest improving dietary habits,” “suggest consulting a nutritionist”). For example, if protein intake is less than half the recommended value for a week, and calorie intake is excessive, the analysis department assigns a “nutritional balance deviation” label and increases the risk score. Based on these outputs, the alert department sends alerts to users or their families such as “Recent nutritional balance deviation; it is recommended to improve dietary habits or consult a nutritionist.” Subsequent processing involves recording the user's reaction after the alert is issued (e.g., whether they improved their dietary habits or consulted a nutritionist) and using this data for the next AI learning iteration. The technological benefits are that the collection, analysis, and alerting departments no longer rely on manual observation or recording. Through AI-powered high-frequency, high-precision dietary data analysis and automatic alerts, it can not only identify pre-dementia groups early but also detect and address various health risks at an early stage. This significantly improves user health maintenance, prevents serious illnesses, and reduces the burden on healthcare. Applicable areas include home-based elderly care, nutrition management, telemedicine support, health management projects, and personalized medicine. Furthermore, by aggregating anonymous dietary data from multiple users, it can also be used for creating population health risk maps and for public health policy.

[0066] The analysis unit can infer a user's emotions and adjust the presentation of the analysis results based on the inferred emotions. For example, when a user is stressed, the analysis results can be presented concisely; when the user is relaxed, detailed analysis results can be provided; when the user is eager to deal with something, analysis results highlighting the key points can be provided. Thus, users can obtain appropriate information based on their emotional state and utilize the analysis results more effectively. Specifically, to infer user emotions, the analysis unit inputs facial image data (e.g., 128×128 pixel RGB images, acquired at 1-second intervals), audio data (e.g., 16kHz sampled 1-second audio waveform), and biosensor data (e.g., one-dimensional time-series data of heart rate and skin conductance). The analysis unit inputs this data into a multimodal AI model (e.g., a combination of CNN and RNN or a Transformer-based emotion inference model) and outputs emotion labels (e.g., "stressed," "relaxed," "eager to deal with something") and emotion intensity scores (continuous values ​​from 0.0 to 1.0). For example, when a facial image and an audio waveform are input simultaneously, the analysis unit outputs probability distributions such as "stress: 0.72", "relaxation: 0.18", and "urgent need: 0.10". Based on the emotion inference results, the analysis unit automatically applies rules such as "only brief key points when stress is high, detailed charts and explanations when relaxed, and a list of key points when urgent need" by the analysis result presentation method decision module. The analysis unit automatically selects the presentation format of the analysis results according to the user's emotional state (e.g., text summary, detailed charts, infographics, voice explanations, etc.). AI output examples include: (1) "stress state: only key points are presented (e.g., risk score 0.82, medical advice recommended)"; (2) "relaxation state: detailed analysis (e.g., walking speed change chart, change in the number of words in the conversation content, detailed explanation of the diet pattern)"; (3) "urgent need: list of key points (e.g., risk factors, recommended actions)" etc. Subsequent processing records the user's reaction after the analysis results are presented (e.g., content confirmation, request for detailed display) and uses it for the next analysis expression optimization. The technological advantage lies in the fact that the analysis unit does not rely on subjective human judgment or uniform output. Through AI-powered high-precision emotion inference and automatic output control, it maximizes the comprehensibility and utilization efficiency of the analysis results without compromising user experience. This reduces user stress and burden, ensuring the reliable delivery of crucial information from dementia predisposition detection. Applicable areas include home-based elderly care, health management in elderly care institutions, telemedicine support, and health management projects related to accompanying stress. Furthermore, by statistically analyzing the relationship between multi-user emotional states and analyzed expressions, it can also be applied to group optimization and personalized health support.

[0067] The alarm unit can infer a user's emotions and adjust the alarm content based on the inferred emotions. For example, when a user is stressed, the alarm content can be simplified, providing only essential information; when the user is relaxed, detailed alarm content can be provided; when the user is in a hurry to deal with something, an alarm highlighting key points can be issued. Thus, users can receive appropriate alarms based on their emotional state and respond quickly and effectively. Specifically, to infer user emotions, the alarm unit inputs facial image data (e.g., a 128×128 pixel RGB image, acquired at 1-second intervals), audio data (e.g., a 16kHz sampled 1-second audio waveform), and biosensor data (e.g., one-dimensional time-series data of heart rate and skin conductance). The alarm unit inputs this data into a multimodal AI model (e.g., a combination of CNN and RNN or a Transformer-based emotion inference model) and outputs emotion labels (e.g., "stressed," "relaxed," "hurry to deal with something") and emotion intensity scores (continuous values ​​from 0.0 to 1.0). For example, when both facial images and audio waveforms are input simultaneously, the alarm unit outputs probability distributions such as "stress: 0.75," "relaxed: 0.15," and "hurry to deal with something": 0.10. Based on sentiment inference, the alert department automatically applies rules such as "brief notification only when stressed, detailed notification with explanation when relaxed, and notification with a list of key points when urgent" through the alert content generation module. The alert department automatically selects the tone (e.g., polite, concise, emphasizing key points) and level of detail (e.g., with charts, only key points) of the notification text based on the user's emotional state. AI output examples include: "Stressed state: brief notification (e.g., risk score 0.82, medical advice recommended)", "Relaxed state: detailed notification (e.g., risk factors, recommended actions, charts)", "Urgent state: key point notification (e.g., recommended actions only)", etc. After the alert is issued, the user's reaction (e.g., notification confirmation, content comprehension level) is recorded and used for future alert content optimization. The technical effect is that the alert department does not rely on subjective human judgment or uniform notifications. Through AI's high-precision sentiment inference and automatic notification content control, it maximizes the efficiency and comprehension of important information delivery without compromising user experience. This reduces user stress and burden, ensuring that important alerts for dementia predisposition detection are reliably delivered in an appropriate manner. Applicable areas include home-based elderly care, health management in elderly care institutions, telemedicine support, and health management projects related to stress management. Furthermore, by statistically analyzing the relationship between multi-user emotional states and notification content, it can also be applied to group optimization and personalized health support.

[0068] The data collection unit can infer a user's emotions and adjust the data collection frequency based on these inferences. For example, when a user is stressed, the data collection frequency can be reduced to lessen their burden; when a user is relaxed, the data collection frequency can be increased to collect more detailed data; and when a user is in a hurry to process data, important data can be collected first. Thus, users can receive appropriate data collection based on their emotional state, reducing their burden while ensuring they receive the necessary data. Specifically, to infer user emotions, the collection unit inputs facial image data (e.g., 128×128 pixel RGB images, acquired at 1-second intervals), audio data (e.g., 16kHz sampled 1-second audio waveforms), and biosensor data (e.g., one-dimensional time-series data of heart rate and skin conductance). The collection unit then inputs this data into a convolutional neural network (CNN), recurrent neural network (RNN), or multimodal Transformer model, outputting emotion classification labels (e.g., "stressed," "relaxed," "in a hurry to process data") and emotion intensity scores (continuous values ​​from 0.0 to 1.0). For example, when facial images and audio waveforms are input simultaneously, the collection unit outputs probability distributions such as "Stress: 0.78," "Relaxation: 0.12," and "Urgent Need for Processing: 0.10." Based on these output values, the collection unit automatically applies rules such as "10-minute intervals when stress is high, 1-minute intervals when relaxed, and immediate collection of only important data when urgent need for processing" through the collection timing control module. The collection unit monitors the emotion inference results in real time and automatically reduces the collection frequency when the score exceeds a threshold (e.g., stress score above 0.7). To minimize user burden and ensure necessary data, the collection unit can also implement an emotion-state-based collection timing optimization algorithm (e.g., automatic parameter adjustment through reinforcement learning). The technical effect is that the collection unit does not rely on human subjective judgment or manual settings. Through AI's high-precision emotion inference and automatic scheduling, it maximizes data collection efficiency without compromising user experience. This significantly reduces user stress and burden, and stably obtains high-quality behavioral data for dementia predisposition detection. Applicable fields include home-based elderly care, health management in elderly care institutions, telemedicine support, and health management projects with accompanying stress. Furthermore, statistical analysis of the relationship between multi-user emotional states and collection frequency can also be applied to group optimization and personalized health support.

[0069] The analysis unit can infer the user's emotions and determine the analysis priority based on the inferred emotions. For example, when the user is stressed, important data can be analyzed first and results provided quickly; when the user is relaxed, detailed analysis can be performed and comprehensive results provided; when the user is in a hurry to deal with something, key points can be analyzed for a quick response. Thus, the user can obtain appropriate analysis results based on their emotional state and utilize them effectively. Specifically, to infer the user's emotions, the analysis unit inputs facial image data (e.g., 128×128 pixel RGB images, acquired at 1-second intervals), audio data (e.g., 16kHz sampled 1-second audio waveform), and biosensor data (e.g., one-dimensional time-series data of heart rate and skin conductance). The analysis unit inputs this data into a multimodal AI model (e.g., a combination of CNN and RNN or a Transformer-based emotion inference model) and outputs emotion labels (e.g., "stressed," "relaxed," "in a hurry to deal with something") and emotion intensity scores (continuous values ​​from 0.0 to 1.0). For example, when facial images and audio waveforms are input simultaneously, the analysis unit outputs probability distributions such as "Stress: 0.80", "Relaxation: 0.10", and "Urgent Need: 0.10". Based on the emotion inference results, the analysis unit automatically applies rules such as "prioritizing the analysis of important data when stress is high, analyzing all data in detail when relaxed, and analyzing only key points when urgent need is urgent" through the analysis priority decision module. The analysis unit automatically sets the priority of the data to be analyzed (such as walking patterns, conversation content, eating events, sleep data, etc.) according to the emotional state, achieving optimal allocation of computing resources and improved analysis efficiency. AI output examples include: "Stress state: prioritize the analysis of important data", "Relaxed state: analyze all data in detail", "Urgent need: analyze key points", etc. Subsequent processing involves sending the analysis results to the alarm unit or user presentation module and recording user reactions (such as content confirmation, detailed display requests) for future analysis priority optimization. The technical effect is that the analysis unit does not need to rely on human subjective judgment or perform analysis indiscriminately. Through AI's high-precision emotion inference and automatic priority control, it maximizes the comprehensibility and utilization efficiency of the analysis results without compromising the user experience. This reduces user stress and burden, ensuring the reliable delivery of crucial information from dementia predisposition assessment. Applicable areas include home-based elderly care, health management in nursing homes, telemedicine support, and health management programs that address accompanying stress. Furthermore, by statistically analyzing the relationship between multi-user emotional states and parsing priorities, it can also be applied to group optimization and personalized health support.

[0070] The alarm unit can infer a user's emotions and adjust the timing of alarm issuance based on these inferences. For example, when a user is stressed, the alarm can be delayed until the user relaxes; when the user is relaxed, the alarm can be issued immediately to facilitate a rapid response; when the user is in a hurry to deal with something, important alarms can be prioritized for swift action. Thus, users can receive alarms at the appropriate time and respond effectively based on their emotional state. Specifically, to infer user emotions, the alarm unit inputs facial image data (e.g., 128×128 pixel RGB images, acquired at 1-second intervals), audio data (e.g., 16kHz sampled 1-second audio waveform), and biosensor data (e.g., one-dimensional time-series data of heart rate and skin conductance). The alarm unit then inputs this data into a multimodal AI model (e.g., a combination of CNN and RNN or a Transformer-based emotion inference model) and outputs emotion labels (e.g., "stressed," "relaxed," "in a hurry to deal with something") and emotion intensity scores (continuous values ​​from 0.0 to 1.0). For example, when facial images and audio waveforms are input simultaneously, the alarm department outputs probability distributions such as "Stress: 0.80", "Relaxation: 0.10", and "Urgent Need for Action: 0.10". Based on the emotion inference results, the alarm department automatically applies rules such as "delaying the alarm when stress is high, issuing it immediately when relaxed, and issuing only an important alarm immediately when there is an urgent need for action" through the alarm timing control module. The alarm department implements an optimal timing algorithm (e.g., automatic parameter adjustment through reinforcement learning) and uses user reactions (e.g., notification confirmation rate, response speed) for future optimization. AI output examples include: "Stress state: delayed alarm", "Relaxed state: immediate alarm", "Urgent need for action: important alarm issued immediately", etc. The technical effect is that the alarm department does not need to rely on human subjective judgment or uniform notification. Through AI's high-precision emotion inference and automatic timing control, it maximizes the efficiency and response speed of important information transmission without compromising user experience. This reduces user stress and burden, ensuring that important alarms for dementia predisposition detection are reliably delivered at the appropriate time. Applicable areas include home-based elderly care, health management in elderly care institutions, telemedicine support, and health management projects with accompanying stress. Furthermore, statistical analysis of the relationship between multiple users' emotional states and the timing of their emotional responses can also be applied to group optimization and personalized health support.

[0071] The following is a brief description of the processing flow of the implementation method. Specifically, the system's collection unit automatically collects users' daily behavioral data (such as temporal acceleration data of walking patterns, text data of conversation content, structured data of dietary events, sleep data, social media activity data, physiological data, etc.) from various sensors and applications, and stores it in a database in a time-series manner. The collection unit automatically detects missing data and outliers, performing outlier removal and normalization as preprocessing. In addition, the collection unit also simultaneously collects various attribute information such as users' emotional inferences (e.g., inputting facial images, audio, and biosensor data into an AI model to output emotional labels and intensity scores), health status, living environment, geographical location information, and social activities. The analysis unit inputs the multidimensional data received by the collection unit into AI architectures such as convolutional neural networks (CNN), recurrent neural networks (RNN), Transformer-based models, decision trees, and gradient boosting, automatically extracting abnormal walking patterns, changes in conversation content, abnormal dietary / sleep / physiological data, decline in social activity, and changes in cognitive function. The analysis department uses AI model outputs (such as risk scores, anomaly tagging, recommended actions, and detailed analysis reports) to determine thresholds and perform branching processes, which are then used to notify the alert department and accumulate subsequent learning data. The alert department automatically selects the optimal notification method, content, and timing based on the prediction results received by the analysis department, as well as user attributes, emotions, location information, and social activities, and sends an alert to the user or their family. After the alert is sent, the alert department records the user's reaction (such as notification confirmation, whether they sought medical attention, and changes in behavior) and uses this data for the next AI learning iteration. The technical benefits are that this system eliminates the need for manual observation or recording. Through AI high-dimensional data analysis and automatic alert issuance, it can not only identify early-stage dementia predisposition groups but also detect and address various health risks early. This significantly promotes user health maintenance, severe illness prevention, and reduces the burden of medical care. Applicable areas include home-based elderly care, health management in elderly care institutions, telemedicine support, health management projects, and personalized medicine. Furthermore, by aggregating anonymized multi-user data, it can also be used for creating population health risk maps and public health policies.

[0072] Step 1: The Collection Department collects users' daily behavioral data. This data includes walking patterns, conversation content, and eating frequency. The Collection Department collects walking patterns using pedometers or accelerometers, conversation content using voice recognition technology, and eating frequency using a diet tracking application. Step 2: The Analysis Department analyzes the data collected by the Collection Department to predict the likelihood of the user belonging to the dementia predisposition group. The Analysis Department uses AI to analyze the collected data, analyze user behavior patterns, and detect signs of dementia. For example, it analyzes changes in walking patterns, conversation content, and eating frequency. Step 3: The Alert Department issues an alert based on the prediction results obtained by the Analysis Department. The Alert Department issues alerts via smartphone notifications, emails, or voice notifications, stating that the user may belong to the dementia predisposition group and recommending early medical attention. Specifically, in step 1, the collection unit automatically acquires multi-dimensional data from pedometers or accelerometers, including daily steps (e.g., 10,000 steps), walking speed (e.g., 1.2 m / s), temporal acceleration data of walking rhythm (e.g., 50Hz sampling, approximately 4.32 million samples daily), spoken text transcribed by a speech recognition engine (e.g., 100 sentences daily), and dietary events from dietary record applications (e.g., 3 times daily, each including dietary content, time, and calorie information), and stores this data in a time-series format in the database. In step 2, the parsing unit inputs this data into AI architectures such as convolutional neural networks (CNN), recurrent neural networks (RNN), and Transformer-based temporal parsing models to automatically extract abnormal walking patterns (e.g., decreased walking speed, disordered rhythm), changes in conversation content (e.g., reduced vocabulary, speech delay, decreased semantic consistency), and changes in dietary frequency (e.g., reduced frequency of meals, irregular meal times). Examples of AI inputs include: one day of walking acceleration data (50Hz × 86400 seconds = 4.32 million one-dimensional arrays), one day of spoken text (e.g., 100 sentences such as "The weather is nice today" and "I've eaten"), and one week of dietary events (structured data including food content, calories, and time for each event). AI outputs include a dementia predisposition risk score (a continuous value from 0.0 to 1.0), risk factor tags (e.g., "Abnormal walking pattern," "Decreased conversation content"), and recommended actions (e.g., "Seek medical attention," "Based on observation"), among other structured data. In step 3, the alert unit receives these outputs and sends specific alerts to the user or their family via smartphone notifications, email, and voice notification APIs, such as "High likelihood of dementia predisposition; please seek medical attention as soon as possible." Subsequent processing involves recording the user's reaction after the alert is issued (e.g., whether they sought medical attention, whether they shared the information with family members), and using this data for the next AI learning iteration. The technical advantage is that this system does not rely on manual observation or consultation. By automatically parsing massive amounts of time-series data and applying unconventional feature extraction rules (such as CNN to detect local patterns and RNN to grasp long-term dependencies), it can detect signs of dementia with higher accuracy and earlier than before.This can reduce the diagnostic burden on medical institutions, enabling users and their families to proactively take early action. Applicable areas include home-based elderly care, health management in elderly care institutions, telemedicine support, and corporate health management projects. Furthermore, by aggregating anonymized multi-user data, it can also be used for creating regional dementia risk maps and for public health policy.

[0073] The specific processing unit 290 sends the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires voice representing the user's input to the result of the specific processing. The control unit 46A sends the voice data representing the user's input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0074] Data generation model 58 is what is known as generative AI (Artificial Intelligence). An example of data generation model 58 includes ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> Generative AI, such as data generation model 58, is obtained by deep learning through a neural network. The data generation model 58 is input with a prompt containing instructions, and with inference data such as speech data representing speech, text data representing text, and image data representing images (e.g., data of still images or data of moving images). The data generation model 58 infers the inference data based on the instructions shown in the prompt and outputs the inference result in one or more data forms such as speech data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the above-described specific processing while using the data generation model 58. The data generation model 58 can also be a fine-tuned model to output inference results from prompts without instructions; in this case, the data generation model 58 can output inference results from prompts without instructions. The data processing device 12, etc., includes multiple data generation models 58, including AI other than generative AI. AI other than generative AI includes, but is not limited to, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes. Furthermore, AI can also act as an AI agent. Additionally, when the processing described above is performed by AI, this processing may be partially or entirely performed by AI, but is not limited to this example. Moreover, processing performed by AI, including generative AI, can be replaced by rule-based processing, and vice versa.

[0075] Furthermore, the processing performed by the aforementioned data processing system 10 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart device 14, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart device 14. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the smart device 14 or external devices, and the smart device 14 acquires or collects information required for processing from the data processing device 12 or external devices.

[0076] Each of the aforementioned elements, including the collection unit, analysis unit, and alarm unit, can be implemented, for example, in at least one of the smart device 14 and the data processing device 12. For instance, the collection unit may use the pedometer and accelerometer sensor of the smart device 14 to collect the user's walking patterns and use voice recognition technology to collect conversation content. Furthermore, the collection unit may also use a specific processing unit 290 of the data processing device 12 to collect eating frequency using a diet tracking application. The analysis unit, for example, analyzes the collected data using the specific processing unit 290 of the data processing device 12 and uses AI to detect signs of dementia. The alarm unit may issue alarms, for example, through the notification function of the smart device 14, email, or voice notification. The correspondence between the various units and the device or control unit is not limited to the above examples and can be modified in various ways.

[0077] [Second Implementation] Figure 3 An example of the configuration of the data processing system 210 according to the second embodiment is shown.

[0078] like Figure 3 As shown, the data processing system 210 includes a data processing device 12 and smart glasses 214. One example of the data processing device 12 is a server.

[0079] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN and / or LAN, etc.

[0080] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0081] Microphone 238 receives user commands by receiving the user's voice. Microphone 238 captures the user's voice, converts the captured sound into speech data, and outputs it to processor 46. Speaker 240 outputs sound according to commands from processor 46.

[0082] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, used to capture the user's surroundings (e.g., the shooting range defined by an angle of view equivalent to the field of vision of an average healthy person).

[0083] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F 44 and 26 is performed in a secure state.

[0084] Figure 4 An example of the main functions of the data processing device 12 and the smart glasses 214 is shown. Figure 4 As shown, in the data processing device 12, specific processing is performed by the processor 28. The memory 32 stores a specific processing program 56.

[0085] The processor 28 reads a specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is implemented by the processor 28 operating as a specific processing unit 290 based on the specific processing program 56 executed on the RAM 30.

[0086] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. The emotion inference function using the emotion-specific model 59 (emotion-specific function) includes various inferences and predictions related to the user's emotions, but is not limited to this example. Furthermore, emotion inference and prediction also include, for example, emotion analysis (analysis).

[0087] In the smart glasses 214, specific processing is performed by the processor 46. A specific processing program 60 is stored in the memory 50. The processor 46 reads the specific processing program 60 from the memory 50 and executes the read specific processing program 60 on the RAM 48. Specific processing is implemented by the processor 46 operating as a control unit 46A based on the specific processing program 60 executed on the RAM 48. Furthermore, the smart glasses 214 may also have the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and use these models to perform the same processing as the specific processing unit 290.

[0088] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with the server device that has the data generation model 58. In addition, the data processing device 12 may be a server device or a user-owned terminal device (e.g., a mobile phone, robot, home appliance, etc.).

[0089] The specific processing unit 290 sends the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires voice input representing the user's input to the specific processing result. The control unit 46A sends the voice data representing the user's input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0090] The data generation model 58 is a so-called generative AI. An example of the data generation model 58 includes generative AIs such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with a prompt containing instructions, and also with inference data such as speech data representing speech, text data representing text, and image data representing images (e.g., data of still images or data of moving images). The data generation model 58 infers the inference data based on the instructions shown in the prompt and outputs the inference result in one or more data forms such as speech data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the aforementioned specific processing while using the data generation model 58. The data generation model 58 can also be a fine-tuned model to output inference results from prompts that do not contain instructions; in this case, the data generation model 58 is able to output inference results from prompts that do not contain instructions. The data processing device 12, etc., includes various data generation models 58, which include AI other than generative AI. These AIs include, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and are capable of various processing methods, but are not limited to these examples. Furthermore, AI can also be an AI agent. Additionally, when the processing described above is performed by AI, this processing may be partially or entirely performed by AI, but is not limited to these examples. Moreover, processing performed by AI including generative AI can be replaced by rule-based processing, and vice versa.

[0091] The data processing system 210 of the second embodiment performs the same processing as the data processing system 10 of the first embodiment. The processing performed by the data processing system 210 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart glasses 214, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart glasses 214. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the smart glasses 214 or external devices, and the smart glasses 214 acquires or collects information required for processing from the data processing device 12 or external devices.

[0092] Each of the aforementioned elements, including the collection unit, analysis unit, and alarm unit, can be implemented, for example, in at least one of the smart glasses 214 and the data processing device 12. For instance, the collection unit can collect the user's walking patterns using the pedometer and accelerometer sensor of the smart glasses 214, and collect conversation content using voice recognition technology. Furthermore, the collection unit can collect eating frequency using a diet tracking application via a specific processing unit 290 of the data processing device 12. The analysis unit, for example, analyzes the collected data via the specific processing unit 290 of the data processing device 12, and uses AI to detect signs of dementia. The alarm unit can issue alarms, for example, via the notification function of the smart glasses 214, email, or voice notification. The correspondence between the various units and the device or control unit is not limited to the above examples and can be modified in various ways.

[0093] [Third Implementation] Figure 5 An example of the configuration of the data processing system 310 according to the third embodiment is shown.

[0094] like Figure 5 As shown, the data processing system 310 includes a data processing device 12 and a head-mounted terminal 314. An example of the data processing device 12 is a server.

[0095] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN and / or LAN, etc.

[0096] The head-mounted terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0097] Microphone 238 receives user commands by receiving the user's voice. Microphone 238 captures the user's voice, converts the captured sound into speech data, and outputs it to processor 46. Speaker 240 outputs sound according to commands from processor 46.

[0098] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, used to capture the user's surroundings (e.g., the shooting range defined by an angle of view equivalent to the field of vision of an average healthy person).

[0099] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F 44 and 26 is performed in a secure state.

[0100] Figure 6 An example of the main functions of the data processing device 12 and the head-mounted terminal 314 is shown. Figure 6 As shown, in the data processing device 12, specific processing is performed by the processor 28. The memory 32 stores a specific processing program 56.

[0101] The processor 28 reads a specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is implemented by the processor 28 operating as a specific processing unit 290 based on the specific processing program 56 executed on the RAM 30.

[0102] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. The emotion inference function using the emotion-specific model 59 (emotion-specific function) includes various inferences and predictions related to the user's emotions, but is not limited to this example. Furthermore, emotion inference and prediction also include, for example, emotion analysis (analysis).

[0103] In the head-mounted terminal 314, specific processing is performed by the processor 46. A specific program 60 is stored in the memory 50. The processor 46 reads the specific program 60 from the memory 50 and executes the read specific program 60 on the RAM 48. Specific processing is implemented by the processor 46 operating as a control unit 46A based on the specific program 60 executed on the RAM 48. Furthermore, the head-mounted terminal 314 may also have the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and use these models to perform the same processing as the specific processing unit 290.

[0104] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with the server device that has the data generation model 58. In addition, the data processing device 12 may be a server device or a user-owned terminal device (e.g., a mobile phone, robot, home appliance, etc.).

[0105] The specific processing unit 290 sends the result of the specific processing to the head-mounted terminal 314. In the head-mounted terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires voice representing the user's input to the specific processing result. The control unit 46A sends the voice data representing the user's input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0106] The data generation model 58 is a so-called generative AI. An example of the data generation model 58 includes generative AIs such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with a prompt containing instructions, and also with inference data such as speech data representing speech, text data representing text, and image data representing images (e.g., data of still images or data of moving images). The data generation model 58 infers the inference data based on the instructions shown in the prompt and outputs the inference result in one or more data forms such as speech data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the aforementioned specific processing while using the data generation model 58. The data generation model 58 can also be a fine-tuned model to output inference results from prompts that do not contain instructions; in this case, the data generation model 58 is able to output inference results from prompts that do not contain instructions. The data processing device 12, etc., includes various data generation models 58, which include AI other than generative AI. These AIs include, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and are capable of various processing methods, but are not limited to these examples. Furthermore, AI can also be an AI agent. Additionally, when the processing described above is performed by AI, this processing may be partially or entirely performed by AI, but is not limited to these examples. Moreover, processing performed by AI including generative AI can be replaced by rule-based processing, and vice versa.

[0107] The data processing system 310 of the third embodiment performs the same processing as the data processing system 10 of the first embodiment. The processing performed by the data processing system 310 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the head-mounted terminal 314, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the head-mounted terminal 314. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the head-mounted terminal 314 or external devices, and the head-mounted terminal 314 acquires or collects information required for processing from the data processing device 12 or external devices.

[0108] Each of the aforementioned elements, including the collection unit, analysis unit, and alarm unit, can be implemented, for example, in at least one of the head-mounted terminal 314 and the data processing device 12. For example, the collection unit can collect the user's walking patterns using the pedometer and accelerometer sensor of the head-mounted terminal 314, and collect conversation content using voice recognition technology. Furthermore, the collection unit can collect eating frequency using a diet tracking application via a specific processing unit 290 of the data processing device 12. The analysis unit, for example, analyzes the collected data via the specific processing unit 290 of the data processing device 12, and uses AI to detect signs of dementia. The alarm unit can issue alarms, for example, via the notification function of the head-mounted terminal 314, email, or voice notification. The correspondence between the various units and the device or control unit is not limited to the above examples and can be modified in various ways.

[0109] [Fourth Implementation] Figure 7 An example of the configuration of the data processing system 410 according to the fourth embodiment is shown.

[0110] like Figure 7 As shown, the data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[0111] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN and / or LAN, etc.

[0112] Robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control object 443. Computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, and control object 443 are also connected to the bus 52.

[0113] Microphone 238 receives user commands by receiving the user's voice. Microphone 238 captures the user's voice, converts the captured sound into speech data, and outputs it to processor 46. Speaker 240 outputs sound according to commands from processor 46.

[0114] The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and an imaging element such as a CMOS image sensor or a CCD image sensor, used to photograph the user's surroundings (e.g., the shooting range defined by an angle of view equivalent to the field of vision of an average healthy person).

[0115] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F 44 and 26 is performed in a secure state.

[0116] The controlled object 443 includes a display device, LEDs for the eyes, and motors for driving the arms, hands, and feet. The posture and movements of the robot 414 are controlled by controlling the motors for the arms, hands, and feet. Some of the robot 414's emotions can be expressed by controlling these motors. In addition, facial expressions of the robot 414 can also be expressed by controlling the illumination state of the LEDs for the robot 414's eyes.

[0117] Figure 8 An example of the main functions of the data processing device 12 and the robot 414 is shown. For example... Figure 8 As shown, in the data processing device 12, specific processing is performed by the processor 28. The memory 32 stores a specific processing program 56.

[0118] The processor 28 reads a specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is implemented by the processor 28 operating as a specific processing unit 290 based on the specific processing program 56 executed on the RAM 30.

[0119] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. The emotion inference function using the emotion-specific model 59 (emotion-specific function) includes various inferences and predictions related to the user's emotions, but is not limited to this example. Furthermore, emotion inference and prediction also include, for example, emotion analysis (analysis).

[0120] In robot 414, specific processing is performed by processor 46. A specific program 60 is stored in memory 50. Processor 46 reads the specific program 60 from memory 50 and executes the read specific program 60 on RAM 48. Specific processing is achieved by processor 46 acting as control unit 46A based on the specific program 60 executed on RAM 48. Furthermore, robot 414 may also have the same data generation model and emotion-specific model as data generation model 58 and emotion-specific model 59, and use these models to perform the same processing as specific processing unit 290.

[0121] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with the server device that has the data generation model 58. In addition, the data processing device 12 may be a server device or a user-owned terminal device (e.g., a mobile phone, robot, home appliance, etc.).

[0122] The specific processing unit 290 sends the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the controlled object 443 to output the result of the specific processing. The microphone 238 acquires voice representing the user's input regarding the result of the specific processing. The control unit 46A sends the voice data representing the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0123] The data generation model 58 is a so-called generative AI. An example of the data generation model 58 includes generative AIs such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with a prompt containing instructions, and also with inference data such as speech data representing speech, text data representing text, and image data representing images (e.g., data of still images or data of moving images). The data generation model 58 infers the inference data based on the instructions shown in the prompt and outputs the inference result in one or more data forms such as speech data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the aforementioned specific processing while using the data generation model 58. The data generation model 58 can also be a fine-tuned model to output inference results from prompts that do not contain instructions; in this case, the data generation model 58 is able to output inference results from prompts that do not contain instructions. The data processing device 12, etc., includes various data generation models 58, which include AI other than generative AI. These AIs include, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and are capable of various processing methods, but are not limited to these examples. Furthermore, AI can also be an AI agent. Additionally, when the processing described above is performed by AI, this processing may be partially or entirely performed by AI, but is not limited to these examples. Moreover, processing performed by AI including generative AI can be replaced by rule-based processing, and vice versa.

[0124] The data processing system 410 of the fourth embodiment performs the same processing as the data processing system 10 of the first embodiment. The processing performed by the data processing system 410 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the robot 414, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the robot 414. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the robot 414 or external devices, and the robot 414 acquires or collects information required for processing from the data processing device 12 or external devices.

[0125] Each of the aforementioned elements, including the collection unit, analysis unit, and alarm unit, can be implemented, for example, in at least one of the robot 414 and the data processing device 12. For instance, the collection unit can collect the user's walking patterns using the robot 414's pedometer and accelerometer, and collect conversation content using voice recognition technology. Furthermore, the collection unit can collect eating frequency using a diet tracking application via a specific processing unit 290 of the data processing device 12. The analysis unit, for example, analyzes the collected data via the specific processing unit 290 of the data processing device 12 and uses AI to detect signs of dementia. The alarm unit can issue alarms, for example, via the robot 414's notification function, email, or voice notification. The correspondence between the various units and the devices or control units is not limited to the above examples and can be modified in various ways.

[0126] Furthermore, the emotion-specific model 59, serving as an emotion engine, can determine the user's emotion based on a specific mapping. Specifically, the emotion-specific model 59 can determine the user's emotion based on an emotion graph that serves as a specific mapping (see...). Figure 9 The system determines the user's emotions. In addition, the emotion-specific model 59 can also determine the robot's emotions in the same way, and the specific processing unit 290 can also perform specific processing using the robot's emotions.

[0127] Figure 9 This is a diagram representing an emotion map 400 that maps various emotions. In the emotion map 400, emotions are arranged radially from the center in concentric circles. The closer to the center of the concentric circles, the more primitive the emotion is. Further out on the concentric circles, emotions are arranged representing states or actions arising from mood. Emotion is a concept that includes both feelings and mental states. To the left of the concentric circles, emotions generated by reactions occurring in the brain are arranged roughly. To the right of the concentric circles, emotions guided by situational judgments are arranged roughly. Above and below the concentric circles, emotions generated by reactions occurring in the brain and guided by situational judgments are arranged roughly. Furthermore, the emotion of "pleasure" is arranged above the concentric circles, and the emotion of "unpleasantness" is arranged below. Thus, in the emotion map 400, various emotions are mapped according to the structure of emotion generation, while easily generated emotions are mapped nearby.

[0128] These emotions are distributed at the 3 o'clock position on the Emotion Chart 400, and usually fluctuate between peace and unease. In the right half of the Emotion Chart 400, because situational awareness is more dominant than internal feelings, it gives a sense of calm.

[0129] The inner side of the emotion diagram 400 represents the mind, and the outer side of the emotion diagram 400 represents actions. Therefore, the further you go to the outer side of the emotion diagram 400, the more the emotion can be seen (manifested in actions).

[0130] Here, human emotions are based on a balance of various factors such as posture and blood sugar levels. When these balances deviate from the ideal, it indicates unhappiness; when they approach the ideal, it indicates pleasure. In robots, cars, and motorcycles, emotions can also be created based on a balance of factors such as posture and remaining battery power. When these balances deviate from the ideal, it indicates unhappiness; when they approach the ideal, it indicates pleasure. Emotion maps can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Speech Emotion Recognition and Brain Physiological Signal Analysis Systems for Emotion, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). In the left half of the emotion map, emotions belonging to the "Reaction" domain, where sensation is dominant, are arranged. Furthermore, in the right half of the emotion map, emotions belonging to the "Situation" domain, where situational cognition is dominant, are arranged.

[0131] The emotion map defines two types of emotions that promote learning. One is a negative emotion located near the middle of "repentance" or "reflection" on the situation side. That is, when the robot experiences negative emotions such as "I never want to feel this way again" or "I never want to be scolded again." The other is a positive emotion located near "desire" on the response side. That is, when the robot experiences positive feelings such as "wanting more" or "wanting to know more."

[0132] The emotion-specific model 59 feeds user input into a pre-learned neural network to obtain emotion values ​​representing each emotion shown in the emotion graph 400, and determines the user's emotion. This neural network is pre-learned based on multiple learning data sets that combine user input with emotion values ​​representing each emotion shown in the emotion graph 400. Furthermore, this neural network is learned to... Figure 10 As shown in sentiment graph 900, sentiment values ​​in nearby configurations are similar to each other. Figure 10 Examples show how emotions such as "peace of mind," "stability," and "reassurance" can result in similar emotional values.

[0133] In the above embodiments, a specific processing is described by a single computer 22, but the technology disclosed herein is not limited to this, and distributed processing by multiple computers, including computer 22, is also possible.

[0134] In the above embodiments, an example of storing a specific processing program 56 in memory 32 is illustrated, but the technology disclosed herein is not limited thereto. For example, the specific processing program 56 may also be stored in a portable computer-readable non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 performs specific processing according to the specific processing program 56.

[0135] Alternatively, the specific processing program 56 can be stored in a storage device such as a server connected to the data processing device 12 via a network 54, and the specific processing program 56 can be downloaded and installed into the computer 22 upon request from the data processing device 12.

[0136] Furthermore, it is not necessary to store the entire specific process 56 in a storage device such as a server connected to the data processing device 12 via the network 54, nor is it necessary to store the entire specific process 56 in the memory 32; a portion of the specific process 56 may also be stored.

[0137] As a hardware resource for performing specific processing, various processors can be used. For example, a CPU is a general-purpose processor that functions as a hardware resource for performing specific processing by executing software, i.e., programs. Additionally, a dedicated circuit can be listed as a processor; it is a processor with a circuit structure specifically designed for performing specific processing, such as a FPGA (Field-Programmable Gate Array), PLD (Programmable Logic Device), or ASIC (Application Specific Integrated Circuit). Every processor has built-in or connected memory, and every processor executes specific processing by using that memory.

[0138] The hardware resources for performing a specific process can consist of one of these various processors, or a combination of two or more processors of the same or different types (e.g., a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Furthermore, the hardware resources for performing a specific process can also be a single processor.

[0139] As an example of a single processor, the first type consists of a combination of one or more CPUs and software, which functions as a hardware resource to perform specific processing. The second type uses a processor, such as a System-on-a-chip (SoC), which implements the entire system functionality, including multiple hardware resources performing specific processing, using a single IC chip. In this case, the specific processing is implemented using one or more of the aforementioned processors that serve as hardware resources.

[0140] Furthermore, as the hardware architecture of these various processors, more specifically, circuits combining semiconductor elements and other circuit components can be used. Moreover, the specific process described above is merely an example. Therefore, it goes without saying that, without departing from the main point, unnecessary steps can be removed, new steps can be added, or the processing order can be changed.

[0141] Furthermore, although the above examples have been described in terms of first to fourth embodiments, some or all of these embodiments can also be combined. Additionally, the smart device 14, smart glasses 214, head-mounted terminal 314, and robot 414 are just examples; they can be combined separately or are other devices.

[0142] The descriptions and illustrations above are detailed explanations of the parts involved in this disclosure and are merely one example of this disclosure. For example, the descriptions of the above-described structure, function, role, and effect are just one example of the structure, function, role, and effect of the parts involved in this disclosure. Therefore, it goes without saying that, without departing from the spirit of this disclosure, unnecessary parts can be deleted, new elements can be added, or replacements can be made to the descriptions and illustrations above. Furthermore, to avoid confusion and facilitate understanding of the parts involved in this disclosure, explanations of technical common sense that do not require special explanation for implementing this disclosure have been omitted from the descriptions and illustrations above.

[0143] All documents, patent applications and technical standards described in this specification are incorporated herein by reference as if they were specifically and individually described as incorporated by reference.

[0144] (Note 1) A system comprising: The data collection department is responsible for collecting users' daily behavioral data. The analysis unit is used to analyze the data collected by the collection unit and predict the likelihood that a user belongs to the dementia predisposition group; and An alarm unit is used to issue an alarm based on the prediction results obtained by the analysis unit.

[0145] (Note 2) The system as described in Appendix 1 is characterized in that, The collection unit collects data such as users' walking patterns, conversation content, and eating frequency.

[0146] (Note 3) The system as described in Appendix 1 is characterized in that, The analysis unit analyzes user behavior patterns based on the collected data and detects signs of dementia.

[0147] (Note 4) The system as described in Appendix 1 is characterized in that, The alarm unit issues alerts via smartphone notifications, emails, or voice notifications.

[0148] (Note 5) The system as described in Appendix 1 is characterized in that, The collection unit anonymizes the collected data and does not provide it to any third party.

[0149] (Note 6) The system as described in Appendix 1 is characterized in that, The data collection unit infers the user's emotions and adjusts the timing of behavioral data collection based on the inferred user emotions.

[0150] (Note 7) The system as described in Appendix 1 is characterized in that, The collection unit analyzes users' past behavior data and selects the optimal collection method.

[0151] (Note 8) The system as described in Appendix 1 is characterized in that, When collecting behavioral data, the collection unit filters data based on the user's current health status and living environment.

[0152] (Note 9) The system as described in Appendix 1 is characterized in that, The collection unit infers the user's emotions and determines the priority of the collected behavioral data based on the inferred user emotions.

[0153] (Postscript 10) The system as described in Appendix 1 is characterized in that, When collecting behavioral data, the collection unit considers the user's geographical location information and prioritizes collecting data with high relevance.

[0154] (Postscript 11) The system as described in Appendix 1 is characterized in that, When collecting behavioral data, the collection unit analyzes users' social media activities and collects relevant data.

[0155] (Postscript 12) The system as described in Appendix 1 is characterized in that, The parsing unit infers the user's emotions and adjusts the parsing expression based on the inferred user emotions.

[0156] (Postscript 13) The system as described in Appendix 1 is characterized in that, During the parsing process, the parsing unit adjusts the level of detail based on the importance of the behavioral data.

[0157] (Postscript 14) The system as described in Appendix 1 is characterized in that, During the parsing process, the parsing unit applies different parsing algorithms based on the category of the behavioral data.

[0158] (Postscript 15) The system as described in Appendix 1 is characterized in that, The parsing unit infers the user's sentiment and adjusts the parsing length based on the inferred user sentiment.

[0159] (Postscript 16) The system as described in Appendix 1 is characterized in that, During the parsing process, the parsing unit determines the parsing priority based on the timing of the collection of behavioral data.

[0160] (Postscript 17) The system as described in Appendix 1 is characterized in that, During the parsing process, the parsing unit adjusts the parsing order based on the relevance of the behavioral data.

[0161] (Postscript 18) The system as described in Appendix 1 is characterized in that, The alarm unit estimates the user's emotions and adjusts the way the alarm is issued based on the estimated user emotions.

[0162] (Postscript 19) The system as described in Appendix 1 is characterized in that, When issuing an alarm, the alarm unit adjusts the level of detail of the alarm based on the importance of the prediction results.

[0163] (Postscript 20) The system as described in Appendix 1 is characterized in that, When issuing an alarm, the alarm unit applies different alarm issuance methods based on the user's attribute information.

[0164] (Postscript 21) The system as described in Appendix 1 is characterized in that, The alarm unit estimates the user's emotions and determines the priority of the alarm based on the estimated user emotions.

[0165] (Postscript 22) The system as described in Appendix 1 is characterized in that, When issuing an alarm, the alarm unit considers the user's geographical location information and selects the optimal alarm issuance method.

[0166] (Postscript 23) The system as described in Appendix 1 is characterized in that, When issuing an alert, the alarm unit analyzes the user's social media activity and adjusts the content of the alert accordingly.

Claims

1. A system, characterized in that, include: The data collection department is responsible for collecting users' daily behavioral data. The analysis unit is used to analyze the data collected by the collection unit and predict the likelihood that a user belongs to the dementia predisposition group; as well as An alarm unit is used to issue an alarm based on the prediction results obtained by the analysis unit.

2. The system as described in claim 1, characterized in that, The collection unit collects data such as users' walking patterns, conversation content, and eating frequency.

3. The system as described in claim 1, characterized in that, The analysis unit analyzes user behavior patterns based on the collected data and detects signs of dementia.

4. The system as described in claim 1, characterized in that, The alarm unit issues alerts via smartphone notifications, emails, or voice notifications.

5. The system as described in claim 1, characterized in that, The collection unit anonymizes the collected data and does not provide it to any third party.

6. The system as described in claim 1, characterized in that, The data collection unit infers the user's emotions and adjusts the timing of behavioral data collection based on the inferred user emotions.

7. The system as described in claim 1, characterized in that, The collection unit analyzes users' past behavior data and selects the optimal collection method.

8. The system as described in claim 1, characterized in that, When collecting behavioral data, the collection unit filters data based on the user's current health status and living environment.

Images