System

Abstract

A system is provided.SOLUTION: A system including means for collecting behavior data of a user, means for preprocessing the collected behavior data, means for training a local AI model using the preprocessed data, means for transmitting the trained local AI model to a server, means for aggregating a plurality of local AI models and performing federated learning, means for delivering an improved AI model to a device of a user, and means for supporting a daily behavior of the user based on the delivered improved AI model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] In modern society, users are faced with a vast number of choices, and they spend a lot of time and effort managing and deciding on them. As a result, users find it difficult to concentrate on important decisions, which can lead to a decline in their quality of life. A particular disadvantage is the amount of time spent on small daily decisions such as what to wear, what to eat, shopping, and schedule management. The purpose of this invention is to reduce the burden on users and support more efficient lifestyles by handling these complicated daily decisions on their behalf. [Means for solving the problem]

[0005] The present invention solves the above-mentioned problems by providing a system including means for collecting user behavioral data, means for preprocessing the collected behavioral data, means for training a local AI model using the preprocessed data, means for transmitting the trained local AI model to a server, means for aggregating multiple local AI models and performing federated learning, means for delivering an improved AI model to a user's device, and means for supporting the user's daily activities based on the delivered improved AI model, thereby enabling the user to focus on important decision-making and improving their quality of life.

[0006] "Behavioral Data" refers to information about the activities you engage in using your device, including app usage history, location information, click patterns, and web browsing history.

[0007] "Preprocessing" is the process of cleansing collected raw data, filling in missing data, and standardizing formats to convert it into a format suitable for AI models.

[0008] A "local AI model" is a dynamic machine learning model that is trained on an individual device using pre-processed data.

[0009] "Federated learning" is a machine learning method that aggregates multiple local AI models on a server and uses them to improve a global model.

[0010] A "server" is a central computer that aggregates data, performs federated learning, and distributes improved AI models.

[0011] "Device" refers to an electronic device used by a user, such as a mobile phone, tablet, or laptop.

[0012] "Training" is the process of teaching an AI model patterns and rules using collected data.

[0013] An "improved AI model" is a machine learning model in which multiple local models are integrated through federated learning, improving its overall performance and accuracy.

[0014] "Assistance" refers to the act of using an improved AI model to predict the user's daily choices and actions, and assisting them in their daily lives by making suggestions and automatically implementing them.

[0015] "Data collection" is the process of capturing and recording information about user activity through sensors, log files, etc. [Brief explanation of the drawings]

[0016] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11]FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0017] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0018] First, the terms used in the following description will be explained.

[0019] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0020] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

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

[0022] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0023] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0024] [First embodiment]

[0025] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0026] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0027] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. 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. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0028] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0029] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0030] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including 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.

[0031] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0032] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0033] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0034] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0035] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0036] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0037] The present invention is a system for supporting the daily activities of a user, and specific embodiments thereof will be described below.

[0038] Behavioral data collection

[0039] Device: Collects real-time activity information as users operate their devices, including the amount of time they spend using specific applications, their location, web browsing history, and click patterns, allowing for a detailed understanding of user behavior.

[0040] Data Preprocessing

[0041] Terminal: The collected raw data often contains noise and missing values. Therefore, the collected data is cleansed, missing values ​​are imputed, and the format is standardized. For example, an algorithm is used to impute missing parts of time series data.

[0042] Training a local AI model

[0043] On-device: The pre-processed data is used to train a local AI model on the device, which can learn your behavioral patterns and preferences and make personalized predictions, such as what you're likely to view next based on your past web browsing history.

[0044] Sending a local model to the server

[0045] On the device: Once trained, the local model is encrypted and sent to a server for security purposes. This process occurs periodically and has built-in mechanisms to protect user privacy.

[0046] Federated learning

[0047] Server: Aggregates the received local models and performs federated learning. This federated learning integrates the behavioral patterns of multiple users to generate an AI model with overall higher accuracy. This process is iterative, and the model improves with each new data point. For example, a powerful model reflecting overall trends can be created based on a user's news article browsing patterns.

[0048] Deploying improved AI models

[0049] Server: The improved AI model generated through federated learning is re-encrypted and distributed to each user's device, enabling the model to be provided in a form optimized for each device.

[0050] User support

[0051] Device: Supports the user's daily activities based on the deployed improved AI model. Specifically, it predicts the user's behavioral patterns and makes appropriate suggestions based on them. For example, it can provide the optimal outfit and weather forecast based on the user's morning commute time. It can also suggest restaurants at lunchtime based on the user's past dining history.

[0052] Specific examples

[0053] When choosing what to wear

[0054] Device: When the user wakes up in the morning, the device will suggest the best outfit for the day based on weather forecast data and past clothing selection data. For example, if rain is forecast, suggestions will be made based on the clothes the user has liked to wear on rainy days in the past.

[0055] When choosing a meal

[0056] On your device: As lunchtime approaches, your device will suggest suitable restaurants based on your dining history and current location. If you tend to frequent certain cuisines or restaurants, suggestions will be based on that. For example, if you frequently ate Japanese food on past Mondays, Japanese restaurants will be prioritized.

[0057] In this way, by implementing the present invention, it is possible to automatically support the user's daily choices and actions, saving time and effort, allowing the user to focus on important decisions and improving the quality of life.

[0058] The processing flow will be explained below.

[0059] Step 1:

[0060] Device: Collects behavioral data when users operate devices. Specifically, it stores application usage history, location information, web browsing history, click patterns, and other data in real time logs.

[0061] Step 2:

[0062] Terminal: Preprocessing the collected raw data. Specifically, noise is removed, missing data is filled, and data formats are standardized. For example, incomplete location data is filled with the nearest known location.

[0063] Step 3:

[0064] On-device: The preprocessed data is used to train a local AI model, which uses machine learning algorithms to locally build a model that learns user behavior patterns and uses the training dataset to improve the model's accuracy.

[0065] Step 4:

[0066] Terminal: The terminal transmits the trained local AI model to the server. Specifically, the model is encrypted for security purposes and uploaded to the server using a secure communication channel. The transmission is usually done at regular intervals (e.g., once a day).

[0067] Step 5:

[0068] Server: Aggregates local AI models received from multiple users. Specifically, it integrates the parameters of each local model to generate a single integrated model. In this process, it uses a federated learning algorithm to efficiently integrate the information from each model.

[0069] Step 6:

[0070] Server: Performs federated learning to generate an AI model with higher overall accuracy. Specifically, it retrains the aggregated model and creates an improved model that reflects the data of all users.

[0071] Step 7:

[0072] Server: The improved AI model is then sent back to the device. Specifically, the improved model is encrypted and sent to each user's device via a secure communication channel.

[0073] Step 8:

[0074] Device: The deployed and improved AI model is used to assist users in their daily activities. Specifically, the model predicts users' behavioral patterns and makes real-time suggestions, such as suggesting the best outfit to wear based on the time of day in the morning or recommending restaurants suitable for lunchtime. It also automates tasks that users frequently perform.

[0075] Example 1

[0076] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0077] Conventional systems have difficulty effectively collecting and analyzing data on users' daily activities, and are unable to provide optimized support for individual users. There are also issues with preprocessing the collected data and with learning methods to improve the accuracy of AI models. There is a particular need to generate highly accurate AI models while protecting user privacy. Therefore, there is a need to build a system that can efficiently and effectively support users' daily activities.

[0078] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0079] In this invention, the server includes means for collecting motion data, means for preprocessing the collected motion data, means for training a local AI model using the preprocessed data, means for transmitting the trained local AI model to the server, means for aggregating multiple local AI models and performing federated learning, means for distributing an improved AI model to a user's computer, and means for supporting the user's daily activities based on the distributed improved AI model. This makes it possible to efficiently collect, preprocess, and analyze data related to the user's daily activities, and to provide highly accurate, optimized support for each user.

[0080] "Behavioral Data" refers to information about a user's behavior, including application usage history, location information, click patterns, and internet browsing history.

[0081] "Preprocessing" refers to the process of removing noise, filling in missing values, standardizing formats, and otherwise preparing collected raw data in an analyzable format.

[0082] A "local AI model" is an artificial intelligence model that is trained on an individual device to learn user behavior patterns and preferences and make individually customized predictions.

[0083] "Federated learning" is a distributed machine learning technique that aggregates local AI models sent from multiple devices on a server to generate a single, highly accurate global AI model.

[0084] "Encryption" is a technique for converting data or models into a secure format so that third parties cannot access them.

[0085] "Computer" refers to all devices used by users, including smartphones, tablets, and personal computers.

[0086] "Supporting users' daily activities" means automating various choices and actions in daily life based on an improved AI model, and providing appropriate suggestions and support.

[0087] The present invention relates to a system for supporting daily activities of a user, and specific embodiments thereof will be described below.

[0088] Behavioral data collection

[0089] Devices: Every time a user interacts with a device, the device collects activity information in real time. This data includes application usage history, location information, click patterns, and internet browsing history. For example, a smartphone may obtain a user's current location through GPS and record the URLs of websites visited during web browsing.

[0090] Data Preprocessing

[0091] Terminal: The collected raw data may contain noise and missing values, so it is cleansed and standardized into a consistent format. For example, a time series imputation algorithm is used to impute missing data. This process allows for more accurate data analysis.

[0092] Training a local AI model

[0093] On-device: The pre-processed data is used to train a local AI model on the device. This model uses machine learning algorithms (e.g., random forest or LSTM) to learn user behavior patterns and preferences, allowing it to predict what information and actions users are likely to view next.

[0094] Sending a local model to the server

[0095] On the device: Once the local model is trained, it is encrypted for added security and sent to the server using the AES algorithm via the HTTPS protocol.

[0096] Federated learning

[0097] Server: The server receives the local AI models sent from multiple devices and performs federated learning based on them. This process integrates the knowledge of each model to generate a highly accurate global AI model. For example, a federated learning algorithm can be used.

[0098] Deploying improved AI models

[0099] Server: The improved AI model is re-encrypted and delivered to each user's device via a secure channel. This delivery model is provided in a form optimized for each device.

[0100] User support

[0101] Device: Based on the deployed improved AI model, the device assists users in their daily activities, for example, suggesting the best outfit to wear based on their morning commute time, or suggesting restaurants to choose from at lunchtime based on their past dining history.

[0102] Specific examples

[0103] When choosing what to wear

[0104] Device: When the user wakes up in the morning, the device retrieves weather forecast data and suggests the best outfit for the day based on past clothing choices. For example, if rain is forecast, suggestions will be made based on the clothes the user has liked to wear on rainy days in the past.

[0105] When choosing a meal

[0106] Device: When lunchtime approaches, the device will acquire the user's current location information and refer to their past dining history to suggest suitable restaurants. For example, if the user chose Japanese food frequently on the previous Monday, Japanese restaurants will be prioritized.

[0107] Prompt Sentence Examples

[0108] "Search a database for what kind of clothes the user has chosen for what weather in the past, and suggest clothes that suit tomorrow's weather."

[0109] This system efficiently collects and analyzes data and generates highly accurate AI models to effectively support users' daily activities.

[0110] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0111] Step 1:

[0112] Behavioral data collection

[0113] Device: Collects real-time activity information every time you interact with your device, including application launch times, location, click patterns, and internet browsing history.

[0114] Input: User device operation information.

[0115] Output: The raw data collected.

[0116] Specific behavior: For example, when a user opens a smartphone browser and visits a specific website, the device records the URL and the time spent browsing.

[0117] Step 2:

[0118] Data Preprocessing

[0119] Terminal: The raw data collected is denoised, imputed, and standardised into a consistent format. Specifically, missing data is imputed using a time series imputation algorithm.

[0120] Input: The raw data collected.

[0121] Output: Cleansed preprocessed data.

[0122] Specific behavior: If there are any jumps in the location data, the device will delete the data and convert it into consistent data.

[0123] Step 3:

[0124] Training a local AI model

[0125] On the device: A local AI model is trained based on the preprocessed data, specifically using machine learning algorithms (e.g., random forest or LSTM) to learn user behavior patterns.

[0126] Input: Preprocessed data.

[0127] Output: A trained local AI model.

[0128] What it does: Uses website browsing history data to generate a model that predicts which website you're likely to visit next.

[0129] Step 4:

[0130] Sending a local model to the server

[0131] On the device: The trained local model is encrypted and sent to the server over a secure channel, specifically using the AES algorithm and the HTTPS protocol.

[0132] Input: A trained local AI model.

[0133] Output: The encrypted local AI model sent to the server.

[0134] What happens: The device encrypts the local AI model and sends an HTTPS request to the server.

[0135] Step 5:

[0136] Federated learning

[0137] Server: Aggregates local models sent from multiple devices and performs federated learning. Specifically, it uses a federated learning algorithm.

[0138] Input: Multiple local AI models sent from each device.

[0139] Output: An improved global AI model.

[0140] Specific operation: The server integrates the weights of each local model to improve the overall model accuracy.

[0141] Step 6:

[0142] Deploying improved AI models

[0143] Server: The improved AI model is encrypted and delivered to each user's device using the AES algorithm via the HTTPS protocol.

[0144] Input: Improved global AI model.

[0145] Output: An improved AI model delivered to the user's device.

[0146] What it does: The server encrypts the global AI model and sends an HTTPS request to each user.

[0147] Step 7:

[0148] User support

[0149] The device: Based on the deployed improved AI model, it assists the user in their daily activities, such as suggesting outfits based on the daily weather forecast and past clothing data, or suggesting nearby restaurants when it's lunchtime.

[0150] Input: Improved AI model, current conditions (weather forecast, location, etc.).

[0151] Output: Specific suggestions to the user.

[0152] Specific behavior: When the user wakes up in the morning, the system sends a message saying, "It's forecast to rain today. We recommend the jacket you wore on the last rainy day." When lunchtime approaches, the system notifies the user, "We recommend a nearby Japanese restaurant."

[0153] (Application example 1)

[0154] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0155] Currently, technologies for autonomous vehicles that can analyze users' driving behavior and traffic conditions in real time and provide optimal route and parking information are not fully established. As a result, traffic congestion and a lack of parking spaces result in time loss, stress for users, and inefficient driving. Another issue is that information is provided uniformly without taking into account each user's unique driving patterns, resulting in a lack of optimal support. To solve these problems, it is necessary to build an advanced driving assistance system that collects and analyzes user-specific behavioral data and uses further improved AI models.

[0156] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0157] In this invention, the server includes means for collecting user behavioral data, means for preprocessing the collected behavioral data, means for training a local AI model using the preprocessed data, means for transmitting the trained local AI model to the server, means for aggregating multiple local AI models and performing federated learning, means for distributing an improved AI model to the user's device, means for predicting driving behavior based on the deployed improved AI model and providing optimal driving route and parking information, and means for generating prompt sentences to present an optimal driving plan using the generative AI model. This makes it possible to provide optimal driving assistance based on the user's unique driving patterns and real-time traffic conditions.

[0158] "Behavioral Data" refers to various data generated when you operate your device, including your app usage history, location information, click patterns, web browsing history, and driving data.

[0159] "Preprocessing" refers to the process of cleansing the collected behavioral data, filling in missing values, and standardizing the format.

[0160] A "local AI model" is an artificial intelligence model that is trained on a device using pre-processed data to learn user behavior patterns and preferences.

[0161] "Federated learning" is a learning method that aggregates multiple local AI models to generate an overall more accurate AI model.

[0162] "Deployment" is the process of placing an improved AI model on a user's device and actually running it.

[0163] "Driving behavior" refers to specific behavioral patterns and driver preferences when driving a vehicle, including speed, acceleration, and choice of driving route.

[0164] "Driving route" refers to the optimal route to a destination, which is selected based on real-time traffic conditions and the driver's past driving patterns.

[0165] "Parking lot information" refers to information about the location and availability of parking lots necessary for users to park near their destination.

[0166] A "generative AI model" is an artificial intelligence model that is generated through federated learning and individual training and is used to predict user behavior patterns.

[0167] A "prompt" is an instruction used by a generative AI model to suggest an optimal driving plan.

[0168] This invention is a driving assistance system for autonomous vehicles that collects, preprocesses, and analyzes user behavior data and provides optimal driving routes and parking information to improve driving efficiency. Specifically, the system has the following steps and configuration:

[0169] Behavioral data collection

[0170] The server collects user behavior data through sensing devices and APIs. This data includes app usage history, location information, click patterns, web browsing history, driving data (time, speed, weather, traffic conditions, etc.) For example, if a user uses a vehicle during their morning commute, data such as the route, driving time, and speed are collected.

[0171] Data Preprocessing

[0172] The collected data is first cleansed, missing values ​​are imputed, and the format is standardized. The server then performs preprocessing, applying algorithms to impute missing parts of time series data, for example. This process uses Python and the data processing library Pandas.

[0173] Training a local AI model

[0174] The preprocessed data is used on the user's device to train a local AI model, which is built using major AI libraries such as TENSORFLOW® and Keras. For example, based on driving data from the user's commute, the model learns the optimal driving route and suggests it for the next commute.

[0175] Sending a local model to the server

[0176] Once trained, the local AI model is encrypted and sent to the server, using an encryption algorithm to ensure user privacy.

[0177] Federated learning

[0178] The server aggregates the received local models and performs federated learning, which generates a more accurate AI model. This process uses a distributed learning algorithm on the server side.

[0179] Deploying improved AI models

[0180] The improved AI model is then re-encrypted and delivered to each user's device, with the model provided in a form optimized for that device.

[0181] User support

[0182] The improved AI model deployed on the device supports the user's daily activities. For example, when driving in the morning, the device will suggest the optimal driving route and parking lot based on the weather forecast and past driving data. It also obtains traffic conditions in real time and optimizes the driving route based on that information.

[0183] Specific examples

[0184] For example, if a user commutes to work in an autonomous vehicle, the "Smart Driving Assistant" app will suggest the optimal driving route based on the weather forecast, traffic conditions, and past driving data before the user starts driving in the morning. If necessary, the app will also notify the user in real time of the availability of suitable parking spaces. Specific examples of prompts are as follows:

[0185] Example prompt:

[0186] "Build an AI model that uses user behavior pattern data (time, location information, speed, weather, traffic conditions) to suggest the next destination and the optimal driving route. Material 1: Driving data from October 2023. Material 2: Current traffic condition data. Output: Suggest the optimal route and next destination to the user."

[0187] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0188] Step 1:

[0189] The server collects user behavioral data (app usage history, location information, click patterns, web browsing history, driving data) through sensing devices and APIs.

[0190] Input: Various user behavior data

[0191] Output: Raw data collected

[0192] Specific operation: The server calls the API and obtains data from the user's smartphone and vehicle sensors.

[0193] Step 2:

[0194] The server preprocesses the collected behavioral data, specifically cleansing the data, filling in missing values, and standardizing the format.

[0195] Input: Raw data collected

[0196] Output: Cleansed data

[0197] Specific operation: The server uses Python and Pandas to create a data frame, impute missing values, and unify the format of each data.

[0198] Step 3:

[0199] The device uses the preprocessed data to train a local AI model.

[0200] Input: Preprocessed data

[0201] Output: A trained local AI model

[0202] Specific operation: The device uses TensorFlow and Keras to build a local AI model and trains the model using data as input.

[0203] Step 4:

[0204] The device encrypts the trained local AI model and sends it to the server.

[0205] Input: A trained local AI model

[0206] Output: Encrypted local AI model

[0207] Specific operation: The terminal encrypts the model using an encryption algorithm and sends it to the server over the network.

[0208] Step 5:

[0209] The server aggregates multiple local AI models and performs federated learning.

[0210] Input: Multiple encrypted local AI models

[0211] Output: An improved AI model generated through federated learning.

[0212] What it does: The server decrypts the encrypted model training data and aggregates the model using a distributed learning algorithm.

[0213] Step 6:

[0214] The server re-encrypts the improved AI model and distributes it to each user's device.

[0215] Input: Improved AI model

[0216] Output: An encrypted, improved AI model

[0217] What it does: The server encrypts the model and delivers it in a format optimized for the user's specific device.

[0218] Step 7:

[0219] The device will then use the deployed improved AI model to assist the user in their daily activities, for example by providing optimal driving routes and parking information.

[0220] Input: Improved AI model, current user data (location, traffic, weather, etc.)

[0221] Output: Driving route and parking information suggestions

[0222] Specific operation: The device uses the generative AI model to process real-time data, generate prompts, and notify the user of optimal suggestions.

[0223] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0224] The present invention combines an emotion engine with a system for supporting the daily activities of users, and specific embodiments thereof are described below.

[0225] Behavioral data collection

[0226] Device: Collects real-time behavioral data as users interact with their devices, including application usage, location, web browsing, and click patterns. The emotion engine also collects emotional data by analyzing users' voice, facial expressions, and text inputs.

[0227] Data Preprocessing

[0228] Terminal: Collected behavioral and emotional data is preprocessed. For behavioral data, noise is removed, missing values ​​are filled, and data formats are standardized. For emotional data, the voice, facial expression, and text input data analyzed by the emotion engine are processed. For example, emotional information extracted from voice data is standardized into text format.

[0229] Training a local AI model

[0230] On-device: The preprocessed behavioral and emotional data is used to train a local AI model on the device. This model can learn the user's behavioral patterns and emotional changes and make personalized predictions. For example, it can learn how a user behaves in a particular emotional state.

[0231] Sending a local model to the server

[0232] On-device: Once trained, the local AI model is encrypted and sent to a server for security purposes. This process occurs periodically and has built-in mechanisms to protect user privacy.

[0233] Federated learning

[0234] Server: Aggregates local AI models received from multiple users and performs federated learning. This integrates the behavioral patterns and emotional information of multiple users to generate an AI model with higher overall accuracy. For example, it comprehensively learns how a user behaves in a specific emotional state.

[0235] Deploying improved AI models

[0236] Server: The improved AI model generated through federated learning is then distributed to the devices again. The distributed model is provided in a form optimized for each device.

[0237] User support

[0238] Device: Based on the deployed and improved AI model, the device supports the user's daily activities. Specifically, it makes real-time suggestions based on behavioral patterns and emotional information. For example, if the user is feeling stressed, it will suggest relaxation activities or provide content to help them change their mood.

[0239] Specific examples

[0240] When choosing what to wear

[0241] Device: When the user wakes up in the morning, the device will suggest the best outfit for the day based on weather forecast data, past clothing choices, and the user's current emotional state. For example, if the user is feeling stressed, the device will suggest comfortable clothing.

[0242] When choosing a meal

[0243] Device: When lunchtime approaches, the device will suggest suitable restaurants based on the user's past dining history, current location, and emotional state. If the user tends to frequent certain cuisines or restaurants, suggestions will be based on that. For example, if the user is tired, the device will prioritize restaurants that serve nutritious meals.

[0244] In this way, the implementation of the present invention can automatically support users' daily choices and actions, saving them time and effort, and by taking into account the user's emotional state, more personalized suggestions can be made, improving their quality of life.

[0245] The processing flow will be explained below.

[0246] Step 1:

[0247] Device: Collects behavioral data when users operate the device. Specifically, it stores real-time logs of users' application usage history, location information, web browsing history, click patterns, etc. Furthermore, it utilizes an emotion engine to collect emotional data from users' voice, facial expressions, and text input.

[0248] Step 2:

[0249] Terminal: Preprocessing the collected raw data and emotion data. Preprocessing of behavioral data involves removing noise, filling in missing data, and standardizing data formats. Preprocessing of emotion data involves standardizing the emotion information extracted from audio data into text format. It also standardizes the emotion labels extracted from facial images.

[0250] Step 3:

[0251] Device: Trains a local AI model using preprocessed behavioral and emotional data. Specifically, it creates a local machine learning model to learn user behavior patterns and emotional changes, and then uses the training dataset to improve the model's accuracy.

[0252] Step 4:

[0253] Device: The device transmits the trained local AI model to the server. First, the model is encrypted and uploaded to the server using a secure communication protocol. This transmission is done periodically to protect the user's privacy.

[0254] Step 5:

[0255] Server: Aggregates local AI models received from multiple users. Specifically, it integrates the parameters of each local model to generate a single integrated model. This process uses a federated learning algorithm to efficiently integrate the information from each model.

[0256] Step 6:

[0257] Server: Performs federated learning to generate a more accurate AI model overall. Retrains the aggregated model to create an improved model that reflects the data and sentiment information of all users.

[0258] Step 7:

[0259] Server: The improved AI model is then sent back to the device. The improved model is then encrypted and sent to each user's device via a secure communication protocol. At this time, the model is adjusted to be optimized for each device.

[0260] Step 8:

[0261] Device: The deployed and improved AI model is used to assist the user in their daily activities. The model makes real-time suggestions based on the user's behavioral patterns and emotional information. For example, if the user is feeling stressed, it will suggest relaxation activities. If the user is tired, it will recommend restaurants with nutritious meals.

[0262] Example 2

[0263] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0264] Currently, most behavioral support systems make suggestions based solely on the user's behavioral data, lacking detailed support that takes into account the user's emotional state. Furthermore, while secure management and use of collected data is required from the perspective of privacy protection, systems that address this need are limited. Furthermore, methods for efficiently aggregating individually trained local AI models and improving accuracy through federated learning are also insufficient. To address these issues, the present invention provides a federated learning system that combines behavioral data and emotional data.

[0265] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0266] In this invention, the server includes means for collecting user behavioral data and emotional data, means for preprocessing the collected behavioral data and emotional data, means for training a local AI model using the preprocessed data, means for encrypting the trained local AI model and transmitting it to the server, means for aggregating multiple local AI models and performing federated learning, means for delivering an improved AI model to the user's device, and means for supporting the user's daily behavior and emotional state based on the delivered improved AI model. This enables personalized suggestions based on the user's behavior and emotions, and makes it possible to improve the accuracy of the model while safely managing data.

[0267] "Behavioral data" refers to data such as application usage history, location information, click patterns, and web browsing history generated when a user operates a device.

[0268] "Emotion data" refers to data related to emotions generated from the user's voice, facial expressions, and text input.

[0269] "Preprocessing" refers to processing of collected data to remove noise, fill in missing data, and standardize the data format.

[0270] A "local AI model" is an artificial intelligence model that is trained on a user's device and makes individually customized predictions.

[0271] "Encryption" is the process of transforming data using a specific algorithm to keep the information confidential.

[0272] A "server" is a computer system that aggregates multiple local AI models and provides the computational resources to perform federated learning.

[0273] "Federated learning" is a learning method that integrates knowledge gained from multiple local AI models to generate an AI model with high overall accuracy.

[0274] An "improved AI model" refers to an AI model whose accuracy has been improved through federated learning.

[0275] "Deployment" is the process of placing software or AI models in a specific location or device and making them operational.

[0276] "Support" refers to providing appropriate suggestions and advice based on the user's daily behavior and emotional state.

[0277] The present invention combines an emotion engine with a system for supporting users' daily activities, and is mainly composed of the following steps.

[0278] Behavioral data collection

[0279] Device: When a user operates a device, behavioral data such as application usage history, location information, web browsing history, and click patterns are collected in real time. An emotion engine is also used to simultaneously collect emotional data from voice, facial expressions, and text input. The emotion engine uses voice recognition software and facial expression recognition software. For example, Google® Cloud Speech-to-Text can be used for voice recognition, and Microsoft® Azure® Face API can be used for facial expression recognition.

[0280] Data Preprocessing

[0281] Device: Noise is removed from the collected behavioral and emotional data, missing data is filled in, and the data format is standardized. For emotional data, information extracted from voice and facial expression data is standardized and unified in text format. For example, voice data is converted to text using Google Cloud Speech-to-Text, and sentiment analysis is performed using Azure Text Analytics.

[0282] Training a local AI model

[0283] On-device: The preprocessed behavioral and emotional data is used to train a local AI model. TensorFlow is used for training, and the model learns the relationship between a user's behavioral patterns and emotions. For example, the model can predict how a user will behave in a given emotional state.

[0284] Sending a local model to the server

[0285] Terminal: After completing training, the local AI model is encrypted and sent to the server. The encryption is performed using AES (Advanced Encryption Standard) technology. Specifically, the model data is encrypted using Python's cryptography library.

[0286] Federated learning

[0287] Server: Aggregates local AI models sent from multiple users and performs federated learning. This integrates the behavioral patterns and emotional data of multiple users to generate a more accurate AI model. TensorFlow Federated is used for federated learning.

[0288] Deploying improved AI models

[0289] Server: The improved model generated through federated learning is distributed to each device. The distribution is encrypted and ensures secure reception by each device.

[0290] User support

[0291] Device: Based on an improved AI model, the device supports the user's daily activities and emotional state. Specifically, it makes real-time suggestions based on behavioral and emotional data. For example, if the user is feeling stressed, it will suggest relaxation activities. It also combines weather forecast data, past behavioral data, and emotional state to make specific suggestions.

[0292] Specific examples

[0293] When choosing what to wear

[0294] Device: When the user wakes up, the system suggests the most appropriate outfit based on weather forecast data, past clothing choices, and the user's current emotional state. For example, if the user is feeling stressed, it suggests clothing that prioritizes comfort. The weather forecast API uses OpenWeatherMap.

[0295] When choosing a meal

[0296] Device: When lunchtime approaches, the system suggests suitable restaurants based on the user's past meal history, current location, and emotional information. For example, if the user is tired, it will prioritize restaurants that serve nutritious meals. The location service uses the Google Maps API.

[0297] Prompt Sentence Examples

[0298] Example input: Design a system that makes lunch suggestions based on the user's eating history, location, and emotional state. Explain how the system works.

[0299] In this way, by implementing the invention, personalized suggestions based on the user's behavior and emotions become possible, and highly accurate support can be provided while safely managing data.

[0300] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0301] Step 1:

[0302] Collecting behavioral and emotional data

[0303] Terminal: When a user operates a device, behavioral data such as application usage history, location information, web browsing history, and click patterns are collected in real time. An emotion engine is also used to collect emotional data from voice, facial expressions, and text input. The input is sensor data and user operation data, and the output is the collected raw data. For example, voice data is obtained from a microphone, and location information is obtained from a GPS.

[0304] Step 2:

[0305] Data Preprocessing

[0306] Terminal: The collected behavioral and emotional data is subjected to noise removal, missing data completion, and data format unification. Voice and facial expression data is standardized and unified into text format. The input is the collected raw data, and the output is preprocessed clean data. Specific operations include using Python's pandas library to complete missing data and unify the time format. In addition, the voice data is converted to text using Google Cloud Speech-to-Text, and emotions are analyzed using Azure Text Analytics.

[0307] Step 3:

[0308] Training a local AI model

[0309] Terminal: A local AI model is trained using preprocessed behavioral data and emotion data. The input is the preprocessed dataset, and the output is the trained local AI model. TensorFlow is used for training, and a model is built that can predict what behavior will be taken in a specific emotional state. Specifically, a neural network is built using TensorFlow's Keras API, and the dataset is input to train the model.

[0310] Step 4:

[0311] Encrypting and sending the local model

[0312] Terminal: The local AI model that has completed training is encrypted and sent to the server. The input is the trained local AI model, and the output is the encrypted and sent model. The encryption is performed using AES (Advanced Encryption Standard) technology. Specifically, the model data is encrypted using Python's cryptography library, and uploaded to the server via the HTTPS protocol using the requests library.

[0313] Step 5:

[0314] Server-based federated learning

[0315] Server: Aggregates local AI models sent by multiple users and performs federated learning. The input is an encrypted local model, and the output is an AI model improved through federated learning. TensorFlow Federated is used for federated learning. Specifically, it runs a federated learning algorithm and combines the parameters of multiple local models to generate a new model.

[0316] Step 6:

[0317] Encoding and delivering the improved model

[0318] Server: Encodes and encrypts the improved model generated by federated learning for distribution to each device. The input is the improved model, and the output is the encrypted and encoded model. Specifically, the improved model data is encoded in Base64 and then encrypted with AES.

[0319] Step 7:

[0320] Receive and deploy the improved model

[0321] Terminal: Decrypts and decodes the received improved model and deploys it locally. The input is the encrypted and encoded model, and the output is the decoded and deployed model. Specifically, the encryption is performed using Python's cryptography library, and the model is loaded using Model.load().

[0322] Step 8:

[0323] Supporting users' daily activities and emotional state

[0324] Device: Based on an improved AI model, the device supports the user's behavior and emotional state. The input is the user's current emotional state and behavioral data, and the output is real-time support suggestions. Specifically, the device uses the model's predictive capabilities to generate appropriate suggestions based on the user's current emotional state and behavioral history, and displays them as notifications.

[0325] (Application example 2)

[0326] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0327] Conventional autonomous driving systems do not take into account the emotional state of the user, which results in the inability to reduce the user's stress and discomfort. Therefore, real-time driving mode adjustment and behavior suggestions based on emotional data are necessary.

[0328] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting behavioral data and emotional data of the user, means for preprocessing the collected behavioral data and emotional data, and means for training a local AI model using the preprocessed data and emotional data. This makes it possible to automatically adjust the driving mode based on the emotional state of the user and to suggest appropriate actions.

[0329] "Behavioral Data" is data that records a user's everyday behavior, such as their application usage history, location information, click patterns, and web browsing history.

[0330] "Emotion data" refers to data that includes emotional information extracted from a user's voice, facial expression, and text input.

[0331] "Preprocessing" refers to processes such as removing noise from collected data, filling in missing values, and standardizing data formats.

[0332] A "local AI model" is an artificial intelligence model that learns user behavioral patterns and emotional changes and makes individually customized predictions.

[0333] A "server" is a centralized computer system that aggregates multiple local AI models and performs federated learning.

[0334] "Federated learning" is a machine learning technique that integrates local AI models received from multiple devices to generate a more accurate AI model.

[0335] An "improved AI model" is an artificial intelligence model that is generated through federated learning and has overall higher accuracy.

[0336] "Driving mode adjustment" refers to changing the driving mode of an autonomous vehicle based on the emotional state of the user.

[0337] "Behavioral Suggestion" refers to suggesting specific activities or content based on a user's current behavioral patterns and emotional state.

[0338] The present invention is a system for adjusting driving modes and suggesting actions in response to the user's daily behavior and emotional state in an autonomous vehicle. This system is realized through cooperation between a terminal and a server.

[0339] Collecting behavioral and emotional data

[0340] Device: Through sensors, cameras, and microphones installed inside the autonomous vehicle in which the user is riding, the device collects user behavioral data (application usage history, location information, click patterns, web browsing history) and emotional data (voice, facial expressions, text input) in real time.

[0341] Data Preprocessing

[0342] Terminal: The collected behavioral and emotional data undergoes noise removal, missing values ​​are filled in, and the data format is standardized. In particular, for emotional data, processing such as converting voice data into text format is performed.

[0343] Training a local AI model

[0344] On-device: Using pre-processed behavioral and emotional data, a local AI model is trained. This model is customized for each user and learns what behaviors and requests are predicted in specific emotional states.

[0345] Server submission and federated learning of local models

[0346] Device: Once trained, the local AI model is encrypted and sent to the server, where it aggregates the local AI models sent by multiple users and performs federated learning, resulting in an improved AI model with greater overall accuracy.

[0347] Deploying improved AI models

[0348] Server: The improved AI model generated through federated learning is then distributed to the devices again. This model is provided in a form optimized for each device.

[0349] User support

[0350] Device: The deployed improved AI model can adjust the driving mode and suggest appropriate actions in real time based on the user's behavioral patterns and emotional information. For example, if the user is feeling stressed, the driving mode can be set to "Relaxation Mode" and play relaxing music. If the user is having fun, the driving mode can be changed to "Dynamic Mode" and suggest a guided city tour.

[0351] Examples of concrete examples and prompts

[0352] Example 1: When the user is feeling stressed, the vehicle switches to "relaxation mode" and suggests "playing relaxing music."

[0353] Example 2: If the user is happy, the system switches to "dynamic mode" and suggests a "guided city tour."

[0354] An example of a prompt sentence to input to a generative AI model is:

[0355] "Generate appropriate driving modes and behavior suggestions based on the user's emotional data."

[0356] In this way, the present invention enables an autonomous vehicle to adjust its driving mode and suggest actions in accordance with the user's emotional state, providing a safer and more comfortable driving experience.

[0357] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0358] Step 1:

[0359] The device collects user behavioral data (application usage history, location information, click patterns, web browsing history) and emotional data (voice, facial expressions, text input). The input at this stage is raw data obtained from sensors, cameras, and microphones, and the collected raw data is obtained as the output.

[0360] Step 2:

[0361] The collected behavioral and emotional data is preprocessed on the device. The input raw data undergoes processes such as noise removal, missing value completion, and data format standardization before being output as preprocessed data. Specific operations include converting voice data into text format.

[0362] Step 3:

[0363] The device uses the preprocessed behavioral and emotional data to train a local AI model. The input to this process is the preprocessed data, which is then used for learning and analysis. The output is a local AI model that is customized for each user.

[0364] Step 4:

[0365] The device encrypts the trained local AI model and sends it to the server. The input of this process is the trained local AI model, and the output is the encrypted model data. Specifically, a data encryption algorithm is used.

[0366] Step 5:

[0367] The server aggregates multiple local AI models and performs federated learning. The input to this process is the local AI models received from multiple devices, and the output is an improved AI model with higher overall accuracy. Specifically, learning is performed using a federated learning algorithm that integrates multiple models.

[0368] Step 6:

[0369] The server distributes the improved AI model to each device. The input to this process is the improved AI model generated by federated learning, and the output is the model data that is redistributed to each device. Specific operations use a secure data transfer protocol.

[0370] Step 7:

[0371] The device adjusts the driving mode based on the user's behavioral patterns and emotional information using the deployed improved AI model. The input for this process is the improved AI model and real-time collected behavioral and emotional data, and the output is an adjusted driving mode and suggested actions. Specifically, if the emotional state is "stressed," the device switches to "relaxed mode," and if the emotional state is "happy," the device switches to "dynamic mode."

[0372] Step 8:

[0373] The device displays and executes suggested actions to the user. The input of this process is the result of adjusting the driving mode based on the improved AI model, and the output is specific suggested actions to the user (e.g., playing relaxing music, starting a guided city tour). Specific actions include activating a trigger to perform the selected activity.

[0374] The specific processing unit 290 transmits 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 audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user 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 audio data.

[0375] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0376] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0377] [Second embodiment]

[0378] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0379] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0380] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. 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. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0381] 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, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0382] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0383] Camera 42 is a small digital camera equipped with an optical system including 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, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0384] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0385] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0386] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0387] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0388] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0389] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0390] The present invention is a system for supporting the daily activities of a user, and specific embodiments thereof will be described below.

[0391] Behavioral data collection

[0392] Device: Collects real-time activity information as users operate their devices, including the amount of time they spend using specific applications, their location, web browsing history, and click patterns, allowing for a detailed understanding of user behavior.

[0393] Data Preprocessing

[0394] Terminal: The collected raw data often contains noise and missing values. Therefore, the collected data is cleansed, missing values ​​are imputed, and the format is standardized. For example, an algorithm is used to impute missing parts of time series data.

[0395] Training a local AI model

[0396] On-device: The pre-processed data is used to train a local AI model on the device, which can learn your behavioral patterns and preferences and make personalized predictions, such as what you're likely to view next based on your past web browsing history.

[0397] Sending a local model to the server

[0398] On the device: Once trained, the local model is encrypted and sent to a server for security purposes. This process occurs periodically and has built-in mechanisms to protect user privacy.

[0399] Federated learning

[0400] Server: Aggregates the received local models and performs federated learning. This federated learning integrates the behavioral patterns of multiple users to generate an AI model with overall higher accuracy. This process is iterative, and the model improves with each new data point. For example, a powerful model reflecting overall trends can be created based on a user's news article browsing patterns.

[0401] Deploying improved AI models

[0402] Server: The improved AI model generated through federated learning is re-encrypted and distributed to each user's device, enabling the model to be provided in a form optimized for each device.

[0403] User support

[0404] Device: Supports the user's daily activities based on the deployed improved AI model. Specifically, it predicts the user's behavioral patterns and makes appropriate suggestions based on them. For example, it can provide the optimal outfit and weather forecast based on the user's morning commute time. It can also suggest restaurants at lunchtime based on the user's past dining history.

[0405] Specific examples

[0406] When choosing what to wear

[0407] Device: When the user wakes up in the morning, the device will suggest the best outfit for the day based on weather forecast data and past clothing selection data. For example, if rain is forecast, suggestions will be made based on the clothes the user has liked to wear on rainy days in the past.

[0408] When choosing a meal

[0409] On your device: As lunchtime approaches, your device will suggest suitable restaurants based on your dining history and current location. If you tend to frequent certain cuisines or restaurants, suggestions will be based on that. For example, if you frequently ate Japanese food on past Mondays, Japanese restaurants will be prioritized.

[0410] In this way, by implementing the present invention, it is possible to automatically support the user's daily choices and actions, saving time and effort, allowing the user to focus on important decisions and improving the quality of life.

[0411] The processing flow will be explained below.

[0412] Step 1:

[0413] Device: Collects behavioral data when users operate devices. Specifically, it stores application usage history, location information, web browsing history, click patterns, and other data in real time logs.

[0414] Step 2:

[0415] Terminal: Preprocessing the collected raw data. Specifically, noise is removed, missing data is filled, and data formats are standardized. For example, incomplete location data is filled with the nearest known location.

[0416] Step 3:

[0417] On-device: The preprocessed data is used to train a local AI model, which uses machine learning algorithms to locally build a model that learns user behavior patterns and uses the training dataset to improve the model's accuracy.

[0418] Step 4:

[0419] Terminal: The terminal transmits the trained local AI model to the server. Specifically, the model is encrypted for security purposes and uploaded to the server using a secure communication channel. The transmission is usually done at regular intervals (e.g., once a day).

[0420] Step 5:

[0421] Server: Aggregates local AI models received from multiple users. Specifically, it integrates the parameters of each local model to generate a single integrated model. In this process, it uses a federated learning algorithm to efficiently integrate the information from each model.

[0422] Step 6:

[0423] Server: Performs federated learning to generate an AI model with higher overall accuracy. Specifically, it retrains the aggregated model and creates an improved model that reflects the data of all users.

[0424] Step 7:

[0425] Server: The improved AI model is then sent back to the device. Specifically, the improved model is encrypted and sent to each user's device via a secure communication channel.

[0426] Step 8:

[0427] Device: The deployed and improved AI model is used to assist users in their daily activities. Specifically, the model predicts users' behavioral patterns and makes real-time suggestions, such as suggesting the best outfit to wear based on the time of day in the morning or recommending restaurants suitable for lunchtime. It also automates tasks that users frequently perform.

[0428] Example 1

[0429] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0430] Conventional systems have difficulty effectively collecting and analyzing data on users' daily activities, and are unable to provide optimized support for individual users. There are also issues with preprocessing the collected data and with learning methods to improve the accuracy of AI models. There is a particular need to generate highly accurate AI models while protecting user privacy. Therefore, there is a need to build a system that can efficiently and effectively support users' daily activities.

[0431] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0432] In this invention, the server includes means for collecting motion data, means for preprocessing the collected motion data, means for training a local AI model using the preprocessed data, means for transmitting the trained local AI model to the server, means for aggregating multiple local AI models and performing federated learning, means for distributing an improved AI model to a user's computer, and means for supporting the user's daily activities based on the distributed improved AI model. This makes it possible to efficiently collect, preprocess, and analyze data related to the user's daily activities, and to provide highly accurate, optimized support for each user.

[0433] "Behavioral Data" refers to information about a user's behavior, including application usage history, location information, click patterns, and internet browsing history.

[0434] "Preprocessing" refers to the process of removing noise, filling in missing values, standardizing formats, and otherwise preparing collected raw data in an analyzable format.

[0435] A "local AI model" is an artificial intelligence model that is trained on an individual device to learn user behavior patterns and preferences and make individually customized predictions.

[0436] "Federated learning" is a distributed machine learning technique that aggregates local AI models sent from multiple devices on a server to generate a single, highly accurate global AI model.

[0437] "Encryption" is a technique for converting data or models into a secure format so that third parties cannot access them.

[0438] "Computer" refers to all devices used by users, including smartphones, tablets, and personal computers.

[0439] "Supporting users' daily activities" means automating various choices and actions in daily life based on an improved AI model, and providing appropriate suggestions and support.

[0440] The present invention relates to a system for supporting daily activities of a user, and specific embodiments thereof will be described below.

[0441] Behavioral data collection

[0442] Devices: Every time a user interacts with a device, the device collects activity information in real time. This data includes application usage history, location information, click patterns, and internet browsing history. For example, a smartphone may obtain a user's current location through GPS and record the URLs of websites visited during web browsing.

[0443] Data Preprocessing

[0444] Terminal: The collected raw data may contain noise and missing values, so it is cleansed and standardized into a consistent format. For example, a time series imputation algorithm is used to impute missing data. This process allows for more accurate data analysis.

[0445] Training a local AI model

[0446] On-device: The pre-processed data is used to train a local AI model on the device. This model uses machine learning algorithms (e.g., random forest or LSTM) to learn user behavior patterns and preferences, allowing it to predict what information and actions users are likely to view next.

[0447] Sending a local model to the server

[0448] On the device: Once the local model is trained, it is encrypted for added security and sent to the server using the AES algorithm via the HTTPS protocol.

[0449] Federated learning

[0450] Server: The server receives the local AI models sent from multiple devices and performs federated learning based on them. This process integrates the knowledge of each model to generate a highly accurate global AI model. For example, a federated learning algorithm can be used.

[0451] Deploying improved AI models

[0452] Server: The improved AI model is re-encrypted and delivered to each user's device via a secure channel. This delivery model is provided in a form optimized for each device.

[0453] User support

[0454] Device: Based on the deployed improved AI model, the device assists users in their daily activities, for example, suggesting the best outfit to wear based on their morning commute time, or suggesting restaurants to choose from at lunchtime based on their past dining history.

[0455] Specific examples

[0456] When choosing what to wear

[0457] Device: When the user wakes up in the morning, the device retrieves weather forecast data and suggests the best outfit for the day based on past clothing choices. For example, if rain is forecast, suggestions will be made based on the clothes the user has liked to wear on rainy days in the past.

[0458] When choosing a meal

[0459] Device: When lunchtime approaches, the device will acquire the user's current location information and refer to their past dining history to suggest suitable restaurants. For example, if the user chose Japanese food frequently on the previous Monday, Japanese restaurants will be prioritized.

[0460] Prompt Sentence Examples

[0461] "Search a database for what kind of clothes the user has chosen for what weather in the past, and suggest clothes that suit tomorrow's weather."

[0462] This system efficiently collects and analyzes data and generates highly accurate AI models to effectively support users' daily activities.

[0463] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0464] Step 1:

[0465] Behavioral data collection

[0466] Device: Collects real-time activity information every time you interact with your device, including application launch times, location, click patterns, and internet browsing history.

[0467] Input: User device operation information.

[0468] Output: The raw data collected.

[0469] Specific behavior: For example, when a user opens a smartphone browser and visits a specific website, the device records the URL and the time spent browsing.

[0470] Step 2:

[0471] Data Preprocessing

[0472] Terminal: The raw data collected is denoised, imputed, and standardised into a consistent format. Specifically, missing data is imputed using a time series imputation algorithm.

[0473] Input: The raw data collected.

[0474] Output: Cleansed preprocessed data.

[0475] Specific behavior: If there are any jumps in the location data, the device will delete the data and convert it into consistent data.

[0476] Step 3:

[0477] Training a local AI model

[0478] On the device: A local AI model is trained based on the preprocessed data, specifically using machine learning algorithms (e.g., random forest or LSTM) to learn user behavior patterns.

[0479] Input: Preprocessed data.

[0480] Output: A trained local AI model.

[0481] What it does: Uses website browsing history data to generate a model that predicts which website you're likely to visit next.

[0482] Step 4:

[0483] Sending a local model to the server

[0484] On the device: The trained local model is encrypted and sent to the server over a secure channel, specifically using the AES algorithm and the HTTPS protocol.

[0485] Input: A trained local AI model.

[0486] Output: The encrypted local AI model sent to the server.

[0487] What happens: The device encrypts the local AI model and sends an HTTPS request to the server.

[0488] Step 5:

[0489] Federated learning

[0490] Server: Aggregates local models sent from multiple devices and performs federated learning. Specifically, it uses a federated learning algorithm.

[0491] Input: Multiple local AI models sent from each device.

[0492] Output: An improved global AI model.

[0493] Specific operation: The server integrates the weights of each local model to improve the overall model accuracy.

[0494] Step 6:

[0495] Deploying improved AI models

[0496] Server: The improved AI model is encrypted and delivered to each user's device using the AES algorithm via the HTTPS protocol.

[0497] Input: Improved global AI model.

[0498] Output: An improved AI model delivered to the user's device.

[0499] What it does: The server encrypts the global AI model and sends an HTTPS request to each user.

[0500] Step 7:

[0501] User support

[0502] The device: Based on the deployed improved AI model, it assists the user in their daily activities, such as suggesting outfits based on the daily weather forecast and past clothing data, or suggesting nearby restaurants when it's lunchtime.

[0503] Input: Improved AI model, current conditions (weather forecast, location, etc.).

[0504] Output: Specific suggestions to the user.

[0505] Specific behavior: When the user wakes up in the morning, the system sends a message saying, "It's forecast to rain today. We recommend the jacket you wore on the last rainy day." When lunchtime approaches, the system notifies the user, "We recommend a nearby Japanese restaurant."

[0506] (Application example 1)

[0507] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0508] Currently, technologies for autonomous vehicles that can analyze users' driving behavior and traffic conditions in real time and provide optimal route and parking information are not fully established. As a result, traffic congestion and a lack of parking spaces result in time loss, stress for users, and inefficient driving. Another issue is that information is provided uniformly without taking into account each user's unique driving patterns, resulting in a lack of optimal support. To solve these problems, it is necessary to build an advanced driving assistance system that collects and analyzes user-specific behavioral data and uses further improved AI models.

[0509] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0510] In this invention, the server includes means for collecting user behavioral data, means for preprocessing the collected behavioral data, means for training a local AI model using the preprocessed data, means for transmitting the trained local AI model to the server, means for aggregating multiple local AI models and performing federated learning, means for distributing an improved AI model to the user's device, means for predicting driving behavior based on the deployed improved AI model and providing optimal driving route and parking information, and means for generating prompt sentences to present an optimal driving plan using the generative AI model. This makes it possible to provide optimal driving assistance based on the user's unique driving patterns and real-time traffic conditions.

[0511] "Behavioral Data" refers to various data generated when you operate your device, including your app usage history, location information, click patterns, web browsing history, and driving data.

[0512] "Preprocessing" refers to the process of cleansing the collected behavioral data, filling in missing values, and standardizing the format.

[0513] A "local AI model" is an artificial intelligence model that is trained on a device using pre-processed data to learn user behavior patterns and preferences.

[0514] "Federated learning" is a learning method that aggregates multiple local AI models to generate an overall more accurate AI model.

[0515] "Deployment" is the process of placing an improved AI model on a user's device and actually running it.

[0516] "Driving behavior" refers to specific behavioral patterns and driver preferences when driving a vehicle, including speed, acceleration, and choice of driving route.

[0517] "Driving route" refers to the optimal route to a destination, which is selected based on real-time traffic conditions and the driver's past driving patterns.

[0518] "Parking lot information" refers to information about the location and availability of parking lots necessary for users to park near their destination.

[0519] A "generative AI model" is an artificial intelligence model that is generated through federated learning and individual training and is used to predict user behavior patterns.

[0520] A "prompt" is an instruction used by a generative AI model to suggest an optimal driving plan.

[0521] This invention is a driving assistance system for autonomous vehicles that collects, preprocesses, and analyzes user behavior data and provides optimal driving routes and parking information to improve driving efficiency. Specifically, the system has the following steps and configuration:

[0522] Behavioral data collection

[0523] The server collects user behavior data through sensing devices and APIs. This data includes app usage history, location information, click patterns, web browsing history, driving data (time, speed, weather, traffic conditions, etc.) For example, if a user uses a vehicle during their morning commute, data such as the route, driving time, and speed are collected.

[0524] Data Preprocessing

[0525] The collected data is first cleansed, missing values ​​are imputed, and the format is standardized. The server then performs preprocessing, applying algorithms to impute missing parts of time series data, for example. This process uses Python and the data processing library Pandas.

[0526] Training a local AI model

[0527] The preprocessed data is used on the user's device to train a local AI model, built using major AI libraries such as TensorFlow and Keras. For example, based on driving data from the user's commute, the model can learn the optimal driving route and suggest it for the next commute.

[0528] Sending a local model to the server

[0529] Once trained, the local AI model is encrypted and sent to the server, using an encryption algorithm to ensure user privacy.

[0530] Federated learning

[0531] The server aggregates the received local models and performs federated learning, which generates a more accurate AI model. This process uses a distributed learning algorithm on the server side.

[0532] Deploying improved AI models

[0533] The improved AI model is then re-encrypted and delivered to each user's device, with the model provided in a form optimized for that device.

[0534] User support

[0535] The improved AI model deployed on the device supports the user's daily activities. For example, when driving in the morning, the device will suggest the optimal driving route and parking lot based on the weather forecast and past driving data. It also obtains traffic conditions in real time and optimizes the driving route based on that information.

[0536] Specific examples

[0537] For example, if a user commutes to work in an autonomous vehicle, the "Smart Driving Assistant" app will suggest the optimal driving route based on the weather forecast, traffic conditions, and past driving data before the user starts driving in the morning. If necessary, the app will also notify the user in real time of the availability of suitable parking spaces. Specific examples of prompts are as follows:

[0538] Example prompt:

[0539] "Build an AI model that uses user behavior pattern data (time, location information, speed, weather, traffic conditions) to suggest the next destination and the optimal driving route. Material 1: Driving data from October 2023. Material 2: Current traffic condition data. Output: Suggest the optimal route and next destination to the user."

[0540] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0541] Step 1:

[0542] The server collects user behavioral data (app usage history, location information, click patterns, web browsing history, driving data) through sensing devices and APIs.

[0543] Input: Various user behavior data

[0544] Output: Raw data collected

[0545] Specific operation: The server calls the API and obtains data from the user's smartphone and vehicle sensors.

[0546] Step 2:

[0547] The server preprocesses the collected behavioral data, specifically cleansing the data, filling in missing values, and standardizing the format.

[0548] Input: Raw data collected

[0549] Output: Cleansed data

[0550] Specific operation: The server uses Python and Pandas to create a data frame, impute missing values, and unify the format of each data.

[0551] Step 3:

[0552] The device uses the preprocessed data to train a local AI model.

[0553] Input: Preprocessed data

[0554] Output: A trained local AI model

[0555] Specific operation: The device uses TensorFlow and Keras to build a local AI model and trains the model using data as input.

[0556] Step 4:

[0557] The device encrypts the trained local AI model and sends it to the server.

[0558] Input: A trained local AI model

[0559] Output: Encrypted local AI model

[0560] Specific operation: The terminal encrypts the model using an encryption algorithm and sends it to the server over the network.

[0561] Step 5:

[0562] The server aggregates multiple local AI models and performs federated learning.

[0563] Input: Multiple encrypted local AI models

[0564] Output: An improved AI model generated through federated learning.

[0565] What it does: The server decrypts the encrypted model training data and aggregates the model using a distributed learning algorithm.

[0566] Step 6:

[0567] The server re-encrypts the improved AI model and distributes it to each user's device.

[0568] Input: Improved AI model

[0569] Output: An encrypted, improved AI model

[0570] What it does: The server encrypts the model and delivers it in a format optimized for the user's specific device.

[0571] Step 7:

[0572] The device will then use the deployed improved AI model to assist the user in their daily activities, for example by providing optimal driving routes and parking information.

[0573] Input: Improved AI model, current user data (location, traffic, weather, etc.)

[0574] Output: Driving route and parking information suggestions

[0575] Specific operation: The device uses the generative AI model to process real-time data, generate prompts, and notify the user of optimal suggestions.

[0576] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0577] The present invention combines an emotion engine with a system for supporting the daily activities of users, and specific embodiments thereof are described below.

[0578] Behavioral data collection

[0579] Device: Collects real-time behavioral data as users interact with their devices, including application usage, location, web browsing, and click patterns. The emotion engine also collects emotional data by analyzing users' voice, facial expressions, and text inputs.

[0580] Data Preprocessing

[0581] Terminal: Collected behavioral and emotional data is preprocessed. For behavioral data, noise is removed, missing values ​​are filled, and data formats are standardized. For emotional data, the voice, facial expression, and text input data analyzed by the emotion engine are processed. For example, emotional information extracted from voice data is standardized into text format.

[0582] Training a local AI model

[0583] On-device: The preprocessed behavioral and emotional data is used to train a local AI model on the device. This model can learn the user's behavioral patterns and emotional changes and make personalized predictions. For example, it can learn how a user behaves in a particular emotional state.

[0584] Sending a local model to the server

[0585] On-device: Once trained, the local AI model is encrypted and sent to a server for security purposes. This process occurs periodically and has built-in mechanisms to protect user privacy.

[0586] Federated learning

[0587] Server: Aggregates local AI models received from multiple users and performs federated learning. This integrates the behavioral patterns and emotional information of multiple users to generate an AI model with higher overall accuracy. For example, it comprehensively learns how a user behaves in a specific emotional state.

[0588] Deploying improved AI models

[0589] Server: The improved AI model generated through federated learning is then distributed to the devices again. The distributed model is provided in a form optimized for each device.

[0590] User support

[0591] Device: Based on the deployed and improved AI model, the device supports the user's daily activities. Specifically, it makes real-time suggestions based on behavioral patterns and emotional information. For example, if the user is feeling stressed, it will suggest relaxation activities or provide content to help them change their mood.

[0592] Specific examples

[0593] When choosing what to wear

[0594] Device: When the user wakes up in the morning, the device will suggest the best outfit for the day based on weather forecast data, past clothing choices, and the user's current emotional state. For example, if the user is feeling stressed, the device will suggest comfortable clothing.

[0595] When choosing a meal

[0596] Device: When lunchtime approaches, the device will suggest suitable restaurants based on the user's past dining history, current location, and emotional state. If the user tends to frequent certain cuisines or restaurants, suggestions will be based on that. For example, if the user is tired, the device will prioritize restaurants that serve nutritious meals.

[0597] In this way, the implementation of the present invention can automatically support users' daily choices and actions, saving them time and effort, and by taking into account the user's emotional state, more personalized suggestions can be made, improving their quality of life.

[0598] The processing flow will be explained below.

[0599] Step 1:

[0600] Device: Collects behavioral data when users operate the device. Specifically, it stores real-time logs of users' application usage history, location information, web browsing history, click patterns, etc. Furthermore, it utilizes an emotion engine to collect emotional data from users' voice, facial expressions, and text input.

[0601] Step 2:

[0602] Terminal: Preprocessing the collected raw data and emotion data. Preprocessing of behavioral data involves removing noise, filling in missing data, and standardizing data formats. Preprocessing of emotion data involves standardizing the emotion information extracted from audio data into text format. It also standardizes the emotion labels extracted from facial images.

[0603] Step 3:

[0604] Device: Trains a local AI model using preprocessed behavioral and emotional data. Specifically, it creates a local machine learning model to learn user behavior patterns and emotional changes, and then uses the training dataset to improve the model's accuracy.

[0605] Step 4:

[0606] Device: The device transmits the trained local AI model to the server. First, the model is encrypted and uploaded to the server using a secure communication protocol. This transmission is done periodically to protect the user's privacy.

[0607] Step 5:

[0608] Server: Aggregates local AI models received from multiple users. Specifically, it integrates the parameters of each local model to generate a single integrated model. This process uses a federated learning algorithm to efficiently integrate the information from each model.

[0609] Step 6:

[0610] Server: Performs federated learning to generate a more accurate AI model overall. Retrains the aggregated model to create an improved model that reflects the data and sentiment information of all users.

[0611] Step 7:

[0612] Server: The improved AI model is then sent back to the device. The improved model is then encrypted and sent to each user's device via a secure communication protocol. At this time, the model is adjusted to be optimized for each device.

[0613] Step 8:

[0614] Device: The deployed and improved AI model is used to assist the user in their daily activities. The model makes real-time suggestions based on the user's behavioral patterns and emotional information. For example, if the user is feeling stressed, it will suggest relaxation activities. If the user is tired, it will recommend restaurants with nutritious meals.

[0615] Example 2

[0616] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0617] Currently, most behavioral support systems make suggestions based solely on the user's behavioral data, lacking detailed support that takes into account the user's emotional state. Furthermore, while secure management and use of collected data is required from the perspective of privacy protection, systems that address this need are limited. Furthermore, methods for efficiently aggregating individually trained local AI models and improving accuracy through federated learning are also insufficient. To address these issues, the present invention provides a federated learning system that combines behavioral data and emotional data.

[0618] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0619] In this invention, the server includes means for collecting user behavioral data and emotional data, means for preprocessing the collected behavioral data and emotional data, means for training a local AI model using the preprocessed data, means for encrypting the trained local AI model and transmitting it to the server, means for aggregating multiple local AI models and performing federated learning, means for delivering an improved AI model to the user's device, and means for supporting the user's daily behavior and emotional state based on the delivered improved AI model. This enables personalized suggestions based on the user's behavior and emotions, and makes it possible to improve the accuracy of the model while safely managing data.

[0620] "Behavioral data" refers to data such as application usage history, location information, click patterns, and web browsing history generated when a user operates a device.

[0621] "Emotion data" refers to data related to emotions generated from the user's voice, facial expressions, and text input.

[0622] "Preprocessing" refers to processing of collected data to remove noise, fill in missing data, and standardize the data format.

[0623] A "local AI model" is an artificial intelligence model that is trained on a user's device and makes individually customized predictions.

[0624] "Encryption" is the process of transforming data using a specific algorithm to keep the information confidential.

[0625] A "server" is a computer system that aggregates multiple local AI models and provides the computational resources to perform federated learning.

[0626] "Federated learning" is a learning method that integrates knowledge gained from multiple local AI models to generate an AI model with high overall accuracy.

[0627] An "improved AI model" refers to an AI model whose accuracy has been improved through federated learning.

[0628] "Deployment" is the process of placing software or AI models in a specific location or device and making them operational.

[0629] "Support" refers to providing appropriate suggestions and advice based on the user's daily behavior and emotional state.

[0630] The present invention combines an emotion engine with a system for supporting users' daily activities, and is mainly composed of the following steps.

[0631] Behavioral data collection

[0632] Device: When a user operates a device, behavioral data such as application usage history, location information, web browsing history, and click patterns are collected in real time. An emotion engine is also used to simultaneously collect emotional data from voice, facial expressions, and text input. The emotion engine uses voice recognition software and facial expression recognition software. For example, Google Cloud Speech-to-Text can be used for voice recognition, and Microsoft Azure Face API can be used for facial expression recognition.

[0633] Data Preprocessing

[0634] Device: Noise is removed from the collected behavioral and emotional data, missing data is filled in, and the data format is standardized. For emotional data, information extracted from voice and facial expression data is standardized and unified in text format. For example, voice data is converted to text using Google Cloud Speech-to-Text, and sentiment analysis is performed using Azure Text Analytics.

[0635] Training a local AI model

[0636] On-device: The preprocessed behavioral and emotional data is used to train a local AI model. TensorFlow is used for training, and the model learns the relationship between a user's behavioral patterns and emotions. For example, the model can predict how a user will behave in a given emotional state.

[0637] Sending a local model to the server

[0638] Terminal: After completing training, the local AI model is encrypted and sent to the server. The encryption is performed using AES (Advanced Encryption Standard) technology. Specifically, the model data is encrypted using Python's cryptography library.

[0639] Federated learning

[0640] Server: Aggregates local AI models sent from multiple users and performs federated learning. This integrates the behavioral patterns and emotional data of multiple users to generate a more accurate AI model. TensorFlow Federated is used for federated learning.

[0641] Deploying improved AI models

[0642] Server: The improved model generated through federated learning is distributed to each device. The distribution is encrypted and ensures secure reception by each device.

[0643] User support

[0644] Device: Based on an improved AI model, the device supports the user's daily activities and emotional state. Specifically, it makes real-time suggestions based on behavioral and emotional data. For example, if the user is feeling stressed, it will suggest relaxation activities. It also combines weather forecast data, past behavioral data, and emotional state to make specific suggestions.

[0645] Specific examples

[0646] When choosing what to wear

[0647] Device: When the user wakes up, the system suggests the most appropriate outfit based on weather forecast data, past clothing choices, and the user's current emotional state. For example, if the user is feeling stressed, it suggests clothing that prioritizes comfort. The weather forecast API uses OpenWeatherMap.

[0648] When choosing a meal

[0649] Device: When lunchtime approaches, the system suggests suitable restaurants based on the user's past meal history, current location, and emotional information. For example, if the user is tired, it will prioritize restaurants that serve nutritious meals. The location service uses the Google Maps API.

[0650] Prompt Sentence Examples

[0651] Example input: Design a system that makes lunch suggestions based on the user's eating history, location, and emotional state. Explain how the system works.

[0652] In this way, by implementing the invention, personalized suggestions based on the user's behavior and emotions become possible, and highly accurate support can be provided while safely managing data.

[0653] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0654] Step 1:

[0655] Collecting behavioral and emotional data

[0656] Terminal: When a user operates a device, behavioral data such as application usage history, location information, web browsing history, and click patterns are collected in real time. An emotion engine is also used to collect emotional data from voice, facial expressions, and text input. The input is sensor data and user operation data, and the output is the collected raw data. For example, voice data is obtained from a microphone, and location information is obtained from a GPS.

[0657] Step 2:

[0658] Data Preprocessing

[0659] Terminal: The collected behavioral and emotional data is subjected to noise removal, missing data completion, and data format unification. Voice and facial expression data is standardized and unified into text format. The input is the collected raw data, and the output is preprocessed clean data. Specific operations include using Python's pandas library to complete missing data and unify the time format. In addition, the voice data is converted to text using Google Cloud Speech-to-Text, and emotions are analyzed using Azure Text Analytics.

[0660] Step 3:

[0661] Training a local AI model

[0662] Terminal: A local AI model is trained using preprocessed behavioral data and emotion data. The input is the preprocessed dataset, and the output is the trained local AI model. TensorFlow is used for training, and a model is built that can predict what behavior will be taken in a specific emotional state. Specifically, a neural network is built using TensorFlow's Keras API, and the dataset is input to train the model.

[0663] Step 4:

[0664] Encrypting and sending the local model

[0665] Terminal: The local AI model that has completed training is encrypted and sent to the server. The input is the trained local AI model, and the output is the encrypted and sent model. The encryption is performed using AES (Advanced Encryption Standard) technology. Specifically, the model data is encrypted using Python's cryptography library, and uploaded to the server via the HTTPS protocol using the requests library.

[0666] Step 5:

[0667] Server-based federated learning

[0668] Server: Aggregates local AI models sent by multiple users and performs federated learning. The input is an encrypted local model, and the output is an AI model improved through federated learning. TensorFlow Federated is used for federated learning. Specifically, it runs a federated learning algorithm and combines the parameters of multiple local models to generate a new model.

[0669] Step 6:

[0670] Encoding and delivering the improved model

[0671] Server: Encodes and encrypts the improved model generated by federated learning for distribution to each device. The input is the improved model, and the output is the encrypted and encoded model. Specifically, the improved model data is encoded in Base64 and then encrypted with AES.

[0672] Step 7:

[0673] Receive and deploy the improved model

[0674] Terminal: Decrypts and decodes the received improved model and deploys it locally. The input is the encrypted and encoded model, and the output is the decoded and deployed model. Specifically, the encryption is performed using Python's cryptography library, and the model is loaded using Model.load().

[0675] Step 8:

[0676] Supporting users' daily activities and emotional state

[0677] Device: Based on an improved AI model, the device supports the user's behavior and emotional state. The input is the user's current emotional state and behavioral data, and the output is real-time support suggestions. Specifically, the device uses the model's predictive capabilities to generate appropriate suggestions based on the user's current emotional state and behavioral history, and displays them as notifications.

[0678] (Application example 2)

[0679] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0680] Conventional autonomous driving systems do not take into account the emotional state of the user, which results in the inability to reduce the user's stress and discomfort. Therefore, real-time driving mode adjustment and behavior suggestions based on emotional data are necessary.

[0681] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting behavioral data and emotional data of the user, means for preprocessing the collected behavioral data and emotional data, and means for training a local AI model using the preprocessed data and emotional data. This makes it possible to automatically adjust the driving mode based on the emotional state of the user and to suggest appropriate actions.

[0682] "Behavioral Data" is data that records a user's everyday behavior, such as their application usage history, location information, click patterns, and web browsing history.

[0683] "Emotion data" refers to data that includes emotional information extracted from a user's voice, facial expression, and text input.

[0684] "Preprocessing" refers to processes such as removing noise from collected data, filling in missing values, and standardizing data formats.

[0685] A "local AI model" is an artificial intelligence model that learns user behavioral patterns and emotional changes and makes individually customized predictions.

[0686] A "server" is a centralized computer system that aggregates multiple local AI models and performs federated learning.

[0687] "Federated learning" is a machine learning technique that integrates local AI models received from multiple devices to generate a more accurate AI model.

[0688] An "improved AI model" is an artificial intelligence model that is generated through federated learning and has overall higher accuracy.

[0689] "Driving mode adjustment" refers to changing the driving mode of an autonomous vehicle based on the emotional state of the user.

[0690] "Behavioral Suggestion" refers to suggesting specific activities or content based on a user's current behavioral patterns and emotional state.

[0691] The present invention is a system for adjusting driving modes and suggesting actions in response to the user's daily behavior and emotional state in an autonomous vehicle. This system is realized through cooperation between a terminal and a server.

[0692] Collecting behavioral and emotional data

[0693] Device: Through sensors, cameras, and microphones installed inside the autonomous vehicle in which the user is riding, the device collects user behavioral data (application usage history, location information, click patterns, web browsing history) and emotional data (voice, facial expressions, text input) in real time.

[0694] Data Preprocessing

[0695] Terminal: The collected behavioral and emotional data undergoes noise removal, missing values ​​are filled in, and the data format is standardized. In particular, for emotional data, processing such as converting voice data into text format is performed.

[0696] Training a local AI model

[0697] On-device: Using pre-processed behavioral and emotional data, a local AI model is trained. This model is customized for each user and learns what behaviors and requests are predicted in specific emotional states.

[0698] Server submission and federated learning of local models

[0699] Device: Once trained, the local AI model is encrypted and sent to the server, where it aggregates the local AI models sent by multiple users and performs federated learning, resulting in an improved AI model with greater overall accuracy.

[0700] Deploying improved AI models

[0701] Server: The improved AI model generated through federated learning is then distributed to the devices again. This model is provided in a form optimized for each device.

[0702] User support

[0703] Device: The deployed improved AI model can adjust the driving mode and suggest appropriate actions in real time based on the user's behavioral patterns and emotional information. For example, if the user is feeling stressed, the driving mode can be set to "Relaxation Mode" and play relaxing music. If the user is having fun, the driving mode can be changed to "Dynamic Mode" and suggest a guided city tour.

[0704] Examples of concrete examples and prompts

[0705] Example 1: When the user is feeling stressed, the vehicle switches to "relaxation mode" and suggests "playing relaxing music."

[0706] Example 2: If the user is happy, the system switches to "dynamic mode" and suggests a "guided city tour."

[0707] An example of a prompt sentence to input to a generative AI model is:

[0708] "Generate appropriate driving modes and behavior suggestions based on the user's emotional data."

[0709] In this way, the present invention enables an autonomous vehicle to adjust its driving mode and suggest actions in accordance with the user's emotional state, providing a safer and more comfortable driving experience.

[0710] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0711] Step 1:

[0712] The device collects user behavioral data (application usage history, location information, click patterns, web browsing history) and emotional data (voice, facial expressions, text input). The input at this stage is raw data obtained from sensors, cameras, and microphones, and the collected raw data is obtained as the output.

[0713] Step 2:

[0714] The collected behavioral and emotional data is preprocessed on the device. The input raw data undergoes processes such as noise removal, missing value completion, and data format standardization before being output as preprocessed data. Specific operations include converting voice data into text format.

[0715] Step 3:

[0716] The device uses the preprocessed behavioral and emotional data to train a local AI model. The input to this process is the preprocessed data, which is then used for learning and analysis. The output is a local AI model that is customized for each user.

[0717] Step 4:

[0718] The device encrypts the trained local AI model and sends it to the server. The input of this process is the trained local AI model, and the output is the encrypted model data. Specifically, a data encryption algorithm is used.

[0719] Step 5:

[0720] The server aggregates multiple local AI models and performs federated learning. The input to this process is the local AI models received from multiple devices, and the output is an improved AI model with higher overall accuracy. Specifically, learning is performed using a federated learning algorithm that integrates multiple models.

[0721] Step 6:

[0722] The server distributes the improved AI model to each device. The input to this process is the improved AI model generated by federated learning, and the output is the model data that is redistributed to each device. Specific operations use a secure data transfer protocol.

[0723] Step 7:

[0724] The device adjusts the driving mode based on the user's behavioral patterns and emotional information using the deployed improved AI model. The input for this process is the improved AI model and real-time collected behavioral and emotional data, and the output is an adjusted driving mode and suggested actions. Specifically, if the emotional state is "stressed," the device switches to "relaxed mode," and if the emotional state is "happy," the device switches to "dynamic mode."

[0725] Step 8:

[0726] The device displays and executes suggested actions to the user. The input of this process is the result of adjusting the driving mode based on the improved AI model, and the output is specific suggested actions to the user (e.g., playing relaxing music, starting a guided city tour). Specific actions include activating a trigger to perform the selected activity.

[0727] The specific processing unit 290 transmits 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 audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating 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 audio data.

[0728] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0729] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0730] [Third embodiment]

[0731] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0732] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0733] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. 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. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0734] The headset type 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, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0735] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0736] Camera 42 is a small digital camera equipped with an optical system including 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, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0737] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0738] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0739] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0740] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0741] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0742] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0743] The present invention is a system for supporting the daily activities of a user, and specific embodiments thereof will be described below.

[0744] Behavioral data collection

[0745] Device: Collects real-time activity information as users operate their devices, including the amount of time they spend using specific applications, their location, web browsing history, and click patterns, allowing for a detailed understanding of user behavior.

[0746] Data Preprocessing

[0747] Terminal: The collected raw data often contains noise and missing values. Therefore, the collected data is cleansed, missing values ​​are imputed, and the format is standardized. For example, an algorithm is used to impute missing parts of time series data.

[0748] Training a local AI model

[0749] On-device: The pre-processed data is used to train a local AI model on the device, which can learn your behavioral patterns and preferences and make personalized predictions, such as what you're likely to view next based on your past web browsing history.

[0750] Sending a local model to the server

[0751] On the device: Once trained, the local model is encrypted and sent to a server for security purposes. This process occurs periodically and has built-in mechanisms to protect user privacy.

[0752] Federated learning

[0753] Server: Aggregates the received local models and performs federated learning. This federated learning integrates the behavioral patterns of multiple users to generate an AI model with overall higher accuracy. This process is iterative, and the model improves with each new data point. For example, a powerful model reflecting overall trends can be created based on a user's news article browsing patterns.

[0754] Deploying improved AI models

[0755] Server: The improved AI model generated through federated learning is re-encrypted and distributed to each user's device, enabling the model to be provided in a form optimized for each device.

[0756] User support

[0757] Device: Supports the user's daily activities based on the deployed improved AI model. Specifically, it predicts the user's behavioral patterns and makes appropriate suggestions based on them. For example, it can provide the optimal outfit and weather forecast based on the user's morning commute time. It can also suggest restaurants at lunchtime based on the user's past dining history.

[0758] Specific examples

[0759] When choosing what to wear

[0760] Device: When the user wakes up in the morning, the device will suggest the best outfit for the day based on weather forecast data and past clothing selection data. For example, if rain is forecast, suggestions will be made based on the clothes the user has liked to wear on rainy days in the past.

[0761] When choosing a meal

[0762] On your device: As lunchtime approaches, your device will suggest suitable restaurants based on your dining history and current location. If you tend to frequent certain cuisines or restaurants, suggestions will be based on that. For example, if you frequently ate Japanese food on past Mondays, Japanese restaurants will be prioritized.

[0763] In this way, by implementing the present invention, it is possible to automatically support the user's daily choices and actions, saving time and effort, allowing the user to focus on important decisions and improving the quality of life.

[0764] The processing flow will be explained below.

[0765] Step 1:

[0766] Device: Collects behavioral data when users operate devices. Specifically, it stores application usage history, location information, web browsing history, click patterns, and other data in real time logs.

[0767] Step 2:

[0768] Terminal: Preprocessing the collected raw data. Specifically, noise is removed, missing data is filled, and data formats are standardized. For example, incomplete location data is filled with the nearest known location.

[0769] Step 3:

[0770] On-device: The preprocessed data is used to train a local AI model, which uses machine learning algorithms to locally build a model that learns user behavior patterns and uses the training dataset to improve the model's accuracy.

[0771] Step 4:

[0772] Terminal: The terminal transmits the trained local AI model to the server. Specifically, the model is encrypted for security purposes and uploaded to the server using a secure communication channel. The transmission is usually done at regular intervals (e.g., once a day).

[0773] Step 5:

[0774] Server: Aggregates local AI models received from multiple users. Specifically, it integrates the parameters of each local model to generate a single integrated model. In this process, it uses a federated learning algorithm to efficiently integrate the information from each model.

[0775] Step 6:

[0776] Server: Performs federated learning to generate an AI model with higher overall accuracy. Specifically, it retrains the aggregated model and creates an improved model that reflects the data of all users.

[0777] Step 7:

[0778] Server: The improved AI model is then sent back to the device. Specifically, the improved model is encrypted and sent to each user's device via a secure communication channel.

[0779] Step 8:

[0780] Device: The deployed and improved AI model is used to assist users in their daily activities. Specifically, the model predicts users' behavioral patterns and makes real-time suggestions, such as suggesting the best outfit to wear based on the time of day in the morning or recommending restaurants suitable for lunchtime. It also automates tasks that users frequently perform.

[0781] Example 1

[0782] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0783] Conventional systems have difficulty effectively collecting and analyzing data on users' daily activities, and are unable to provide optimized support for individual users. There are also issues with preprocessing the collected data and with learning methods to improve the accuracy of AI models. There is a particular need to generate highly accurate AI models while protecting user privacy. Therefore, there is a need to build a system that can efficiently and effectively support users' daily activities.

[0784] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0785] In this invention, the server includes means for collecting motion data, means for preprocessing the collected motion data, means for training a local AI model using the preprocessed data, means for transmitting the trained local AI model to the server, means for aggregating multiple local AI models and performing federated learning, means for distributing an improved AI model to a user's computer, and means for supporting the user's daily activities based on the distributed improved AI model. This makes it possible to efficiently collect, preprocess, and analyze data related to the user's daily activities, and to provide highly accurate, optimized support for each user.

[0786] "Behavioral Data" refers to information about a user's behavior, including application usage history, location information, click patterns, and internet browsing history.

[0787] "Preprocessing" refers to the process of removing noise, filling in missing values, standardizing formats, and otherwise preparing collected raw data in an analyzable format.

[0788] A "local AI model" is an artificial intelligence model that is trained on an individual device to learn user behavior patterns and preferences and make individually customized predictions.

[0789] "Federated learning" is a distributed machine learning technique that aggregates local AI models sent from multiple devices on a server to generate a single, highly accurate global AI model.

[0790] "Encryption" is a technique for converting data or models into a secure format so that third parties cannot access them.

[0791] "Computer" refers to all devices used by users, including smartphones, tablets, and personal computers.

[0792] "Supporting users' daily activities" means automating various choices and actions in daily life based on an improved AI model, and providing appropriate suggestions and support.

[0793] The present invention relates to a system for supporting daily activities of a user, and specific embodiments thereof will be described below.

[0794] Behavioral data collection

[0795] Devices: Every time a user interacts with a device, the device collects activity information in real time. This data includes application usage history, location information, click patterns, and internet browsing history. For example, a smartphone may obtain a user's current location through GPS and record the URLs of websites visited during web browsing.

[0796] Data Preprocessing

[0797] Terminal: The collected raw data may contain noise and missing values, so it is cleansed and standardized into a consistent format. For example, a time series imputation algorithm is used to impute missing data. This process allows for more accurate data analysis.

[0798] Training a local AI model

[0799] On-device: The pre-processed data is used to train a local AI model on the device. This model uses machine learning algorithms (e.g., random forest or LSTM) to learn user behavior patterns and preferences, allowing it to predict what information and actions users are likely to view next.

[0800] Sending a local model to the server

[0801] On the device: Once the local model is trained, it is encrypted for added security and sent to the server using the AES algorithm via the HTTPS protocol.

[0802] Federated learning

[0803] Server: The server receives the local AI models sent from multiple devices and performs federated learning based on them. This process integrates the knowledge of each model to generate a highly accurate global AI model. For example, a federated learning algorithm can be used.

[0804] Deploying improved AI models

[0805] Server: The improved AI model is re-encrypted and delivered to each user's device via a secure channel. This delivery model is provided in a form optimized for each device.

[0806] User support

[0807] Device: Based on the deployed improved AI model, the device assists users in their daily activities, for example, suggesting the best outfit to wear based on their morning commute time, or suggesting restaurants to choose from at lunchtime based on their past dining history.

[0808] Specific examples

[0809] When choosing what to wear

[0810] Device: When the user wakes up in the morning, the device retrieves weather forecast data and suggests the best outfit for the day based on past clothing choices. For example, if rain is forecast, suggestions will be made based on the clothes the user has liked to wear on rainy days in the past.

[0811] When choosing a meal

[0812] Device: When lunchtime approaches, the device will acquire the user's current location information and refer to their past dining history to suggest suitable restaurants. For example, if the user chose Japanese food frequently on the previous Monday, Japanese restaurants will be prioritized.

[0813] Prompt Sentence Examples

[0814] "Search a database for what kind of clothes the user has chosen for what weather in the past, and suggest clothes that suit tomorrow's weather."

[0815] This system efficiently collects and analyzes data and generates highly accurate AI models to effectively support users' daily activities.

[0816] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0817] Step 1:

[0818] Behavioral data collection

[0819] Device: Collects real-time activity information every time you interact with your device, including application launch times, location, click patterns, and internet browsing history.

[0820] Input: User device operation information.

[0821] Output: The raw data collected.

[0822] Specific behavior: For example, when a user opens a smartphone browser and visits a specific website, the device records the URL and the time spent browsing.

[0823] Step 2:

[0824] Data Preprocessing

[0825] Terminal: The raw data collected is denoised, imputed, and standardised into a consistent format. Specifically, missing data is imputed using a time series imputation algorithm.

[0826] Input: The raw data collected.

[0827] Output: Cleansed preprocessed data.

[0828] Specific behavior: If there are any jumps in the location data, the device will delete the data and convert it into consistent data.

[0829] Step 3:

[0830] Training a local AI model

[0831] On the device: A local AI model is trained based on the preprocessed data, specifically using machine learning algorithms (e.g., random forest or LSTM) to learn user behavior patterns.

[0832] Input: Preprocessed data.

[0833] Output: A trained local AI model.

[0834] What it does: Uses website browsing history data to generate a model that predicts which website you're likely to visit next.

[0835] Step 4:

[0836] Sending a local model to the server

[0837] On the device: The trained local model is encrypted and sent to the server over a secure channel, specifically using the AES algorithm and the HTTPS protocol.

[0838] Input: A trained local AI model.

[0839] Output: The encrypted local AI model sent to the server.

[0840] What happens: The device encrypts the local AI model and sends an HTTPS request to the server.

[0841] Step 5:

[0842] Federated learning

[0843] Server: Aggregates local models sent from multiple devices and performs federated learning. Specifically, it uses a federated learning algorithm.

[0844] Input: Multiple local AI models sent from each device.

[0845] Output: An improved global AI model.

[0846] Specific operation: The server integrates the weights of each local model to improve the overall model accuracy.

[0847] Step 6:

[0848] Deploying improved AI models

[0849] Server: The improved AI model is encrypted and delivered to each user's device using the AES algorithm via the HTTPS protocol.

[0850] Input: Improved global AI model.

[0851] Output: An improved AI model delivered to the user's device.

[0852] What it does: The server encrypts the global AI model and sends an HTTPS request to each user.

[0853] Step 7:

[0854] User support

[0855] The device: Based on the deployed improved AI model, it assists the user in their daily activities, such as suggesting outfits based on the daily weather forecast and past clothing data, or suggesting nearby restaurants when it's lunchtime.

[0856] Input: Improved AI model, current conditions (weather forecast, location, etc.).

[0857] Output: Specific suggestions to the user.

[0858] Specific behavior: When the user wakes up in the morning, the system sends a message saying, "It's forecast to rain today. We recommend the jacket you wore on the last rainy day." When lunchtime approaches, the system notifies the user, "We recommend a nearby Japanese restaurant."

[0859] (Application example 1)

[0860] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0861] Currently, technologies for autonomous vehicles that can analyze users' driving behavior and traffic conditions in real time and provide optimal route and parking information are not fully established. As a result, traffic congestion and a lack of parking spaces result in time loss, stress for users, and inefficient driving. Another issue is that information is provided uniformly without taking into account each user's unique driving patterns, resulting in a lack of optimal support. To solve these problems, it is necessary to build an advanced driving assistance system that collects and analyzes user-specific behavioral data and uses further improved AI models.

[0862] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0863] In this invention, the server includes means for collecting user behavioral data, means for preprocessing the collected behavioral data, means for training a local AI model using the preprocessed data, means for transmitting the trained local AI model to the server, means for aggregating multiple local AI models and performing federated learning, means for distributing an improved AI model to the user's device, means for predicting driving behavior based on the deployed improved AI model and providing optimal driving route and parking information, and means for generating prompt sentences to present an optimal driving plan using the generative AI model. This makes it possible to provide optimal driving assistance based on the user's unique driving patterns and real-time traffic conditions.

[0864] "Behavioral Data" refers to various data generated when you operate your device, including your app usage history, location information, click patterns, web browsing history, and driving data.

[0865] "Preprocessing" refers to the process of cleansing the collected behavioral data, filling in missing values, and standardizing the format.

[0866] A "local AI model" is an artificial intelligence model that is trained on a device using pre-processed data to learn user behavior patterns and preferences.

[0867] "Federated learning" is a learning method that aggregates multiple local AI models to generate an overall more accurate AI model.

[0868] "Deployment" is the process of placing an improved AI model on a user's device and actually running it.

[0869] "Driving behavior" refers to specific behavioral patterns and driver preferences when driving a vehicle, including speed, acceleration, and choice of driving route.

[0870] "Driving route" refers to the optimal route to a destination, which is selected based on real-time traffic conditions and the driver's past driving patterns.

[0871] "Parking lot information" refers to information about the location and availability of parking lots necessary for users to park near their destination.

[0872] A "generative AI model" is an artificial intelligence model that is generated through federated learning and individual training and is used to predict user behavior patterns.

[0873] A "prompt" is an instruction used by a generative AI model to suggest an optimal driving plan.

[0874] This invention is a driving assistance system for autonomous vehicles that collects, preprocesses, and analyzes user behavior data and provides optimal driving routes and parking information to improve driving efficiency. Specifically, the system has the following steps and configuration:

[0875] Behavioral data collection

[0876] The server collects user behavior data through sensing devices and APIs. This data includes app usage history, location information, click patterns, web browsing history, driving data (time, speed, weather, traffic conditions, etc.) For example, if a user uses a vehicle during their morning commute, data such as the route, driving time, and speed are collected.

[0877] Data Preprocessing

[0878] The collected data is first cleansed, missing values ​​are imputed, and the format is standardized. The server then performs preprocessing, applying algorithms to impute missing parts of time series data, for example. This process uses Python and the data processing library Pandas.

[0879] Training a local AI model

[0880] The preprocessed data is used on the user's device to train a local AI model, built using major AI libraries such as TensorFlow and Keras. For example, based on driving data from the user's commute, the model can learn the optimal driving route and suggest it for the next commute.

[0881] Sending a local model to the server

[0882] Once trained, the local AI model is encrypted and sent to the server, using an encryption algorithm to ensure user privacy.

[0883] Federated learning

[0884] The server aggregates the received local models and performs federated learning, which generates a more accurate AI model. This process uses a distributed learning algorithm on the server side.

[0885] Deploying improved AI models

[0886] The improved AI model is then re-encrypted and delivered to each user's device, with the model provided in a form optimized for that device.

[0887] User support

[0888] The improved AI model deployed on the device supports the user's daily activities. For example, when driving in the morning, the device will suggest the optimal driving route and parking lot based on the weather forecast and past driving data. It also obtains traffic conditions in real time and optimizes the driving route based on that information.

[0889] Specific examples

[0890] For example, if a user commutes to work in an autonomous vehicle, the "Smart Driving Assistant" app will suggest the optimal driving route based on the weather forecast, traffic conditions, and past driving data before the user starts driving in the morning. If necessary, the app will also notify the user in real time of the availability of suitable parking spaces. Specific examples of prompts are as follows:

[0891] Example prompt:

[0892] "Build an AI model that uses user behavior pattern data (time, location information, speed, weather, traffic conditions) to suggest the next destination and the optimal driving route. Material 1: Driving data from October 2023. Material 2: Current traffic condition data. Output: Suggest the optimal route and next destination to the user."

[0893] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0894] Step 1:

[0895] The server collects user behavioral data (app usage history, location information, click patterns, web browsing history, driving data) through sensing devices and APIs.

[0896] Input: Various user behavior data

[0897] Output: Raw data collected

[0898] Specific operation: The server calls the API and obtains data from the user's smartphone and vehicle sensors.

[0899] Step 2:

[0900] The server preprocesses the collected behavioral data, specifically cleansing the data, filling in missing values, and standardizing the format.

[0901] Input: Raw data collected

[0902] Output: Cleansed data

[0903] Specific operation: The server uses Python and Pandas to create a data frame, impute missing values, and unify the format of each data.

[0904] Step 3:

[0905] The device uses the preprocessed data to train a local AI model.

[0906] Input: Preprocessed data

[0907] Output: A trained local AI model

[0908] Specific operation: The device uses TensorFlow and Keras to build a local AI model and trains the model using data as input.

[0909] Step 4:

[0910] The device encrypts the trained local AI model and sends it to the server.

[0911] Input: A trained local AI model

[0912] Output: Encrypted local AI model

[0913] Specific operation: The terminal encrypts the model using an encryption algorithm and sends it to the server over the network.

[0914] Step 5:

[0915] The server aggregates multiple local AI models and performs federated learning.

[0916] Input: Multiple encrypted local AI models

[0917] Output: An improved AI model generated through federated learning.

[0918] What it does: The server decrypts the encrypted model training data and aggregates the model using a distributed learning algorithm.

[0919] Step 6:

[0920] The server re-encrypts the improved AI model and distributes it to each user's device.

[0921] Input: Improved AI model

[0922] Output: An encrypted, improved AI model

[0923] What it does: The server encrypts the model and delivers it in a format optimized for the user's specific device.

[0924] Step 7:

[0925] The device will then use the deployed improved AI model to assist the user in their daily activities, for example by providing optimal driving routes and parking information.

[0926] Input: Improved AI model, current user data (location, traffic, weather, etc.)

[0927] Output: Driving route and parking information suggestions

[0928] Specific operation: The device uses the generative AI model to process real-time data, generate prompts, and notify the user of optimal suggestions.

[0929] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0930] The present invention combines an emotion engine with a system for supporting the daily activities of users, and specific embodiments thereof are described below.

[0931] Behavioral data collection

[0932] Device: Collects real-time behavioral data as users interact with their devices, including application usage, location, web browsing, and click patterns. The emotion engine also collects emotional data by analyzing users' voice, facial expressions, and text inputs.

[0933] Data Preprocessing

[0934] Terminal: Collected behavioral and emotional data is preprocessed. For behavioral data, noise is removed, missing values ​​are filled, and data formats are standardized. For emotional data, the voice, facial expression, and text input data analyzed by the emotion engine are processed. For example, emotional information extracted from voice data is standardized into text format.

[0935] Training a local AI model

[0936] On-device: The preprocessed behavioral and emotional data is used to train a local AI model on the device. This model can learn the user's behavioral patterns and emotional changes and make personalized predictions. For example, it can learn how a user behaves in a particular emotional state.

[0937] Sending a local model to the server

[0938] On-device: Once trained, the local AI model is encrypted and sent to a server for security purposes. This process occurs periodically and has built-in mechanisms to protect user privacy.

[0939] Federated learning

[0940] Server: Aggregates local AI models received from multiple users and performs federated learning. This integrates the behavioral patterns and emotional information of multiple users to generate an AI model with higher overall accuracy. For example, it comprehensively learns how a user behaves in a specific emotional state.

[0941] Deploying improved AI models

[0942] Server: The improved AI model generated through federated learning is then distributed to the devices again. The distributed model is provided in a form optimized for each device.

[0943] User support

[0944] Device: Based on the deployed and improved AI model, the device supports the user's daily activities. Specifically, it makes real-time suggestions based on behavioral patterns and emotional information. For example, if the user is feeling stressed, it will suggest relaxation activities or provide content to help them change their mood.

[0945] Specific examples

[0946] When choosing what to wear

[0947] Device: When the user wakes up in the morning, the device will suggest the best outfit for the day based on weather forecast data, past clothing choices, and the user's current emotional state. For example, if the user is feeling stressed, the device will suggest comfortable clothing.

[0948] When choosing a meal

[0949] Device: When lunchtime approaches, the device will suggest suitable restaurants based on the user's past dining history, current location, and emotional state. If the user tends to frequent certain cuisines or restaurants, suggestions will be based on that. For example, if the user is tired, the device will prioritize restaurants that serve nutritious meals.

[0950] In this way, the implementation of the present invention can automatically support users' daily choices and actions, saving them time and effort, and by taking into account the user's emotional state, more personalized suggestions can be made, improving their quality of life.

[0951] The processing flow will be explained below.

[0952] Step 1:

[0953] Device: Collects behavioral data when users operate the device. Specifically, it stores real-time logs of users' application usage history, location information, web browsing history, click patterns, etc. Furthermore, it utilizes an emotion engine to collect emotional data from users' voice, facial expressions, and text input.

[0954] Step 2:

[0955] Terminal: Preprocessing the collected raw data and emotion data. Preprocessing of behavioral data involves removing noise, filling in missing data, and standardizing data formats. Preprocessing of emotion data involves standardizing the emotion information extracted from audio data into text format. It also standardizes the emotion labels extracted from facial images.

[0956] Step 3:

[0957] Device: Trains a local AI model using preprocessed behavioral and emotional data. Specifically, it creates a local machine learning model to learn user behavior patterns and emotional changes, and then uses the training dataset to improve the model's accuracy.

[0958] Step 4:

[0959] Device: The device transmits the trained local AI model to the server. First, the model is encrypted and uploaded to the server using a secure communication protocol. This transmission is done periodically to protect the user's privacy.

[0960] Step 5:

[0961] Server: Aggregates local AI models received from multiple users. Specifically, it integrates the parameters of each local model to generate a single integrated model. This process uses a federated learning algorithm to efficiently integrate the information from each model.

[0962] Step 6:

[0963] Server: Performs federated learning to generate a more accurate AI model overall. Retrains the aggregated model to create an improved model that reflects the data and sentiment information of all users.

[0964] Step 7:

[0965] Server: The improved AI model is then sent back to the device. The improved model is then encrypted and sent to each user's device via a secure communication protocol. At this time, the model is adjusted to be optimized for each device.

[0966] Step 8:

[0967] Device: The deployed and improved AI model is used to assist the user in their daily activities. The model makes real-time suggestions based on the user's behavioral patterns and emotional information. For example, if the user is feeling stressed, it will suggest relaxation activities. If the user is tired, it will recommend restaurants with nutritious meals.

[0968] Example 2

[0969] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0970] Currently, most behavioral support systems make suggestions based solely on the user's behavioral data, lacking detailed support that takes into account the user's emotional state. Furthermore, while secure management and use of collected data is required from the perspective of privacy protection, systems that address this need are limited. Furthermore, methods for efficiently aggregating individually trained local AI models and improving accuracy through federated learning are also insufficient. To address these issues, the present invention provides a federated learning system that combines behavioral data and emotional data.

[0971] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0972] In this invention, the server includes means for collecting user behavioral data and emotional data, means for preprocessing the collected behavioral data and emotional data, means for training a local AI model using the preprocessed data, means for encrypting the trained local AI model and transmitting it to the server, means for aggregating multiple local AI models and performing federated learning, means for delivering an improved AI model to the user's device, and means for supporting the user's daily behavior and emotional state based on the delivered improved AI model. This enables personalized suggestions based on the user's behavior and emotions, and makes it possible to improve the accuracy of the model while safely managing data.

[0973] "Behavioral data" refers to data such as application usage history, location information, click patterns, and web browsing history generated when a user operates a device.

[0974] "Emotion data" refers to data related to emotions generated from the user's voice, facial expressions, and text input.

[0975] "Preprocessing" refers to processing of collected data to remove noise, fill in missing data, and standardize the data format.

[0976] A "local AI model" is an artificial intelligence model that is trained on a user's device and makes individually customized predictions.

[0977] "Encryption" is the process of transforming data using a specific algorithm to keep the information confidential.

[0978] A "server" is a computer system that aggregates multiple local AI models and provides the computational resources to perform federated learning.

[0979] "Federated learning" is a learning method that integrates knowledge gained from multiple local AI models to generate an AI model with high overall accuracy.

[0980] An "improved AI model" refers to an AI model whose accuracy has been improved through federated learning.

[0981] "Deployment" is the process of placing software or AI models in a specific location or device and making them operational.

[0982] "Support" refers to providing appropriate suggestions and advice based on the user's daily behavior and emotional state.

[0983] The present invention combines an emotion engine with a system for supporting users' daily activities, and is mainly composed of the following steps.

[0984] Behavioral data collection

[0985] Device: When a user operates a device, behavioral data such as application usage history, location information, web browsing history, and click patterns are collected in real time. An emotion engine is also used to simultaneously collect emotional data from voice, facial expressions, and text input. The emotion engine uses voice recognition software and facial expression recognition software. For example, Google Cloud Speech-to-Text can be used for voice recognition, and Microsoft Azure Face API can be used for facial expression recognition.

[0986] Data Preprocessing

[0987] Device: Noise is removed from the collected behavioral and emotional data, missing data is filled in, and the data format is standardized. For emotional data, information extracted from voice and facial expression data is standardized and unified in text format. For example, voice data is converted to text using Google Cloud Speech-to-Text, and sentiment analysis is performed using Azure Text Analytics.

[0988] Training a local AI model

[0989] On-device: The preprocessed behavioral and emotional data is used to train a local AI model. TensorFlow is used for training, and the model learns the relationship between a user's behavioral patterns and emotions. For example, the model can predict how a user will behave in a given emotional state.

[0990] Sending a local model to the server

[0991] Terminal: After completing training, the local AI model is encrypted and sent to the server. The encryption is performed using AES (Advanced Encryption Standard) technology. Specifically, the model data is encrypted using Python's cryptography library.

[0992] Federated learning

[0993] Server: Aggregates local AI models sent from multiple users and performs federated learning. This integrates the behavioral patterns and emotional data of multiple users to generate a more accurate AI model. TensorFlow Federated is used for federated learning.

[0994] Deploying improved AI models

[0995] Server: The improved model generated through federated learning is distributed to each device. The distribution is encrypted and ensures secure reception by each device.

[0996] User support

[0997] Device: Based on an improved AI model, the device supports the user's daily activities and emotional state. Specifically, it makes real-time suggestions based on behavioral and emotional data. For example, if the user is feeling stressed, it will suggest relaxation activities. It also combines weather forecast data, past behavioral data, and emotional state to make specific suggestions.

[0998] Specific examples

[0999] When choosing what to wear

[1000] Device: When the user wakes up, the system suggests the most appropriate outfit based on weather forecast data, past clothing choices, and the user's current emotional state. For example, if the user is feeling stressed, it suggests clothing that prioritizes comfort. The weather forecast API uses OpenWeatherMap.

[1001] When choosing a meal

[1002] Device: When lunchtime approaches, the system suggests suitable restaurants based on the user's past meal history, current location, and emotional information. For example, if the user is tired, it will prioritize restaurants that serve nutritious meals. The location service uses the Google Maps API.

[1003] Prompt Sentence Examples

[1004] Example input: Design a system that makes lunch suggestions based on the user's eating history, location, and emotional state. Explain how the system works.

[1005] In this way, by implementing the invention, personalized suggestions based on the user's behavior and emotions become possible, and highly accurate support can be provided while safely managing data.

[1006] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1007] Step 1:

[1008] Collecting behavioral and emotional data

[1009] Terminal: When a user operates a device, behavioral data such as application usage history, location information, web browsing history, and click patterns are collected in real time. An emotion engine is also used to collect emotional data from voice, facial expressions, and text input. The input is sensor data and user operation data, and the output is the collected raw data. For example, voice data is obtained from a microphone, and location information is obtained from a GPS.

[1010] Step 2:

[1011] Data Preprocessing

[1012] Terminal: The collected behavioral and emotional data is subjected to noise removal, missing data completion, and data format unification. Voice and facial expression data is standardized and unified into text format. The input is the collected raw data, and the output is preprocessed clean data. Specific operations include using Python's pandas library to complete missing data and unify the time format. In addition, the voice data is converted to text using Google Cloud Speech-to-Text, and emotions are analyzed using Azure Text Analytics.

[1013] Step 3:

[1014] Training a local AI model

[1015] Terminal: A local AI model is trained using preprocessed behavioral data and emotion data. The input is the preprocessed dataset, and the output is the trained local AI model. TensorFlow is used for training, and a model is built that can predict what behavior will be taken in a specific emotional state. Specifically, a neural network is built using TensorFlow's Keras API, and the dataset is input to train the model.

[1016] Step 4:

[1017] Encrypting and sending the local model

[1018] Terminal: The local AI model that has completed training is encrypted and sent to the server. The input is the trained local AI model, and the output is the encrypted and sent model. The encryption is performed using AES (Advanced Encryption Standard) technology. Specifically, the model data is encrypted using Python's cryptography library, and uploaded to the server via the HTTPS protocol using the requests library.

[1019] Step 5:

[1020] Server-based federated learning

[1021] Server: Aggregates local AI models sent by multiple users and performs federated learning. The input is an encrypted local model, and the output is an AI model improved through federated learning. TensorFlow Federated is used for federated learning. Specifically, it runs a federated learning algorithm and combines the parameters of multiple local models to generate a new model.

[1022] Step 6:

[1023] Encoding and delivering the improved model

[1024] Server: Encodes and encrypts the improved model generated by federated learning for distribution to each device. The input is the improved model, and the output is the encrypted and encoded model. Specifically, the improved model data is encoded in Base64 and then encrypted with AES.

[1025] Step 7:

[1026] Receive and deploy the improved model

[1027] Terminal: Decrypts and decodes the received improved model and deploys it locally. The input is the encrypted and encoded model, and the output is the decoded and deployed model. Specifically, the encryption is performed using Python's cryptography library, and the model is loaded using Model.load().

[1028] Step 8:

[1029] Supporting users' daily activities and emotional state

[1030] Device: Based on an improved AI model, the device supports the user's behavior and emotional state. The input is the user's current emotional state and behavioral data, and the output is real-time support suggestions. Specifically, the device uses the model's predictive capabilities to generate appropriate suggestions based on the user's current emotional state and behavioral history, and displays them as notifications.

[1031] (Application example 2)

[1032] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1033] Conventional autonomous driving systems do not take into account the emotional state of the user, which results in the inability to reduce the user's stress and discomfort. Therefore, real-time driving mode adjustment and behavior suggestions based on emotional data are necessary.

[1034] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting behavioral data and emotional data of the user, means for preprocessing the collected behavioral data and emotional data, and means for training a local AI model using the preprocessed data and emotional data. This makes it possible to automatically adjust the driving mode based on the emotional state of the user and to suggest appropriate actions.

[1035] "Behavioral Data" is data that records a user's everyday behavior, such as their application usage history, location information, click patterns, and web browsing history.

[1036] "Emotion data" refers to data that includes emotional information extracted from a user's voice, facial expression, and text input.

[1037] "Preprocessing" refers to processes such as removing noise from collected data, filling in missing values, and standardizing data formats.

[1038] A "local AI model" is an artificial intelligence model that learns user behavioral patterns and emotional changes and makes individually customized predictions.

[1039] A "server" is a centralized computer system that aggregates multiple local AI models and performs federated learning.

[1040] "Federated learning" is a machine learning technique that integrates local AI models received from multiple devices to generate a more accurate AI model.

[1041] An "improved AI model" is an artificial intelligence model that is generated through federated learning and has overall higher accuracy.

[1042] "Driving mode adjustment" refers to changing the driving mode of an autonomous vehicle based on the emotional state of the user.

[1043] "Behavioral Suggestion" refers to suggesting specific activities or content based on a user's current behavioral patterns and emotional state.

[1044] The present invention is a system for adjusting driving modes and suggesting actions in response to the user's daily behavior and emotional state in an autonomous vehicle. This system is realized through cooperation between a terminal and a server.

[1045] Collecting behavioral and emotional data

[1046] Device: Through sensors, cameras, and microphones installed inside the autonomous vehicle in which the user is riding, the device collects user behavioral data (application usage history, location information, click patterns, web browsing history) and emotional data (voice, facial expressions, text input) in real time.

[1047] Data Preprocessing

[1048] Terminal: The collected behavioral and emotional data undergoes noise removal, missing values ​​are filled in, and the data format is standardized. In particular, for emotional data, processing such as converting voice data into text format is performed.

[1049] Training a local AI model

[1050] On-device: Using pre-processed behavioral and emotional data, a local AI model is trained. This model is customized for each user and learns what behaviors and requests are predicted in specific emotional states.

[1051] Server submission and federated learning of local models

[1052] Device: Once trained, the local AI model is encrypted and sent to the server, where it aggregates the local AI models sent by multiple users and performs federated learning, resulting in an improved AI model with greater overall accuracy.

[1053] Deploying improved AI models

[1054] Server: The improved AI model generated through federated learning is then distributed to the devices again. This model is provided in a form optimized for each device.

[1055] User support

[1056] Device: The deployed improved AI model can adjust the driving mode and suggest appropriate actions in real time based on the user's behavioral patterns and emotional information. For example, if the user is feeling stressed, the driving mode can be set to "Relaxation Mode" and play relaxing music. If the user is having fun, the driving mode can be changed to "Dynamic Mode" and suggest a guided city tour.

[1057] Examples of concrete examples and prompts

[1058] Example 1: When the user is feeling stressed, the vehicle switches to "relaxation mode" and suggests "playing relaxing music."

[1059] Example 2: If the user is happy, the system switches to "dynamic mode" and suggests a "guided city tour."

[1060] An example of a prompt sentence to input to a generative AI model is:

[1061] "Generate appropriate driving modes and behavior suggestions based on the user's emotional data."

[1062] In this way, the present invention enables an autonomous vehicle to adjust its driving mode and suggest actions in accordance with the user's emotional state, providing a safer and more comfortable driving experience.

[1063] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1064] Step 1:

[1065] The device collects user behavioral data (application usage history, location information, click patterns, web browsing history) and emotional data (voice, facial expressions, text input). The input at this stage is raw data obtained from sensors, cameras, and microphones, and the collected raw data is obtained as the output.

[1066] Step 2:

[1067] The collected behavioral and emotional data is preprocessed on the device. The input raw data undergoes processes such as noise removal, missing value completion, and data format standardization before being output as preprocessed data. Specific operations include converting voice data into text format.

[1068] Step 3:

[1069] The device uses the preprocessed behavioral and emotional data to train a local AI model. The input to this process is the preprocessed data, which is then used for learning and analysis. The output is a local AI model that is customized for each user.

[1070] Step 4:

[1071] The device encrypts the trained local AI model and sends it to the server. The input of this process is the trained local AI model, and the output is the encrypted model data. Specifically, a data encryption algorithm is used.

[1072] Step 5:

[1073] The server aggregates multiple local AI models and performs federated learning. The input to this process is the local AI models received from multiple devices, and the output is an improved AI model with higher overall accuracy. Specifically, learning is performed using a federated learning algorithm that integrates multiple models.

[1074] Step 6:

[1075] The server distributes the improved AI model to each device. The input to this process is the improved AI model generated by federated learning, and the output is the model data that is redistributed to each device. Specific operations use a secure data transfer protocol.

[1076] Step 7:

[1077] The device adjusts the driving mode based on the user's behavioral patterns and emotional information using the deployed improved AI model. The input for this process is the improved AI model and real-time collected behavioral and emotional data, and the output is an adjusted driving mode and suggested actions. Specifically, if the emotional state is "stressed," the device switches to "relaxed mode," and if the emotional state is "happy," the device switches to "dynamic mode."

[1078] Step 8:

[1079] The device displays and executes suggested actions to the user. The input of this process is the result of adjusting the driving mode based on the improved AI model, and the output is specific suggested actions to the user (e.g., playing relaxing music, starting a guided city tour). Specific actions include activating a trigger to perform the selected activity.

[1080] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating 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 audio data.

[1081] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1082] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1083] [Fourth embodiment]

[1084] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1085] 7, a 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.

[1086] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. 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. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1087] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1088] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1089] Camera 42 is a small digital camera equipped with an optical system including 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, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1090] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1091] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1092] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1093] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1094] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1095] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1096] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1097] The present invention is a system for supporting the daily activities of a user, and specific embodiments thereof will be described below.

[1098] Behavioral data collection

[1099] Device: Collects real-time activity information as users operate their devices, including the amount of time they spend using specific applications, their location, web browsing history, and click patterns, allowing for a detailed understanding of user behavior.

[1100] Data Preprocessing

[1101] Terminal: The collected raw data often contains noise and missing values. Therefore, the collected data is cleansed, missing values ​​are imputed, and the format is standardized. For example, an algorithm is used to impute missing parts of time series data.

[1102] Training a local AI model

[1103] On-device: The pre-processed data is used to train a local AI model on the device, which can learn your behavioral patterns and preferences and make personalized predictions, such as what you're likely to view next based on your past web browsing history.

[1104] Sending a local model to the server

[1105] On the device: Once trained, the local model is encrypted and sent to a server for security purposes. This process occurs periodically and has built-in mechanisms to protect user privacy.

[1106] Federated learning

[1107] Server: Aggregates the received local models and performs federated learning. This federated learning integrates the behavioral patterns of multiple users to generate an AI model with overall higher accuracy. This process is iterative, and the model improves with each new data point. For example, a powerful model reflecting overall trends can be created based on a user's news article browsing patterns.

[1108] Deploying improved AI models

[1109] Server: The improved AI model generated through federated learning is re-encrypted and distributed to each user's device, enabling the model to be provided in a form optimized for each device.

[1110] User support

[1111] Device: Supports the user's daily activities based on the deployed improved AI model. Specifically, it predicts the user's behavioral patterns and makes appropriate suggestions based on them. For example, it can provide the optimal outfit and weather forecast based on the user's morning commute time. It can also suggest restaurants at lunchtime based on the user's past dining history.

[1112] Specific examples

[1113] When choosing what to wear

[1114] Device: When the user wakes up in the morning, the device will suggest the best outfit for the day based on weather forecast data and past clothing selection data. For example, if rain is forecast, suggestions will be made based on the clothes the user has liked to wear on rainy days in the past.

[1115] When choosing a meal

[1116] On your device: As lunchtime approaches, your device will suggest suitable restaurants based on your dining history and current location. If you tend to frequent certain cuisines or restaurants, suggestions will be based on that. For example, if you frequently ate Japanese food on past Mondays, Japanese restaurants will be prioritized.

[1117] In this way, by implementing the present invention, it is possible to automatically support the user's daily choices and actions, saving time and effort, allowing the user to focus on important decisions and improving the quality of life.

[1118] The processing flow will be explained below.

[1119] Step 1:

[1120] Device: Collects behavioral data when users operate devices. Specifically, it stores application usage history, location information, web browsing history, click patterns, and other data in real time logs.

[1121] Step 2:

[1122] Terminal: Preprocessing the collected raw data. Specifically, noise is removed, missing data is filled, and data formats are standardized. For example, incomplete location data is filled with the nearest known location.

[1123] Step 3:

[1124] On-device: The preprocessed data is used to train a local AI model, which uses machine learning algorithms to locally build a model that learns user behavior patterns and uses the training dataset to improve the model's accuracy.

[1125] Step 4:

[1126] Terminal: The terminal transmits the trained local AI model to the server. Specifically, the model is encrypted for security purposes and uploaded to the server using a secure communication channel. The transmission is usually done at regular intervals (e.g., once a day).

[1127] Step 5:

[1128] Server: Aggregates local AI models received from multiple users. Specifically, it integrates the parameters of each local model to generate a single integrated model. In this process, it uses a federated learning algorithm to efficiently integrate the information from each model.

[1129] Step 6:

[1130] Server: Performs federated learning to generate an AI model with higher overall accuracy. Specifically, it retrains the aggregated model and creates an improved model that reflects the data of all users.

[1131] Step 7:

[1132] Server: The improved AI model is then sent back to the device. Specifically, the improved model is encrypted and sent to each user's device via a secure communication channel.

[1133] Step 8:

[1134] Device: The deployed and improved AI model is used to assist users in their daily activities. Specifically, the model predicts users' behavioral patterns and makes real-time suggestions, such as suggesting the best outfit to wear based on the time of day in the morning or recommending restaurants suitable for lunchtime. It also automates tasks that users frequently perform.

[1135] Example 1

[1136] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1137] Conventional systems have difficulty effectively collecting and analyzing data on users' daily activities, and are unable to provide optimized support for individual users. There are also issues with preprocessing the collected data and with learning methods to improve the accuracy of AI models. There is a particular need to generate highly accurate AI models while protecting user privacy. Therefore, there is a need to build a system that can efficiently and effectively support users' daily activities.

[1138] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1139] In this invention, the server includes means for collecting motion data, means for preprocessing the collected motion data, means for training a local AI model using the preprocessed data, means for transmitting the trained local AI model to the server, means for aggregating multiple local AI models and performing federated learning, means for distributing an improved AI model to a user's computer, and means for supporting the user's daily activities based on the distributed improved AI model. This makes it possible to efficiently collect, preprocess, and analyze data related to the user's daily activities, and to provide highly accurate, optimized support for each user.

[1140] "Behavioral Data" refers to information about a user's behavior, including application usage history, location information, click patterns, and internet browsing history.

[1141] "Preprocessing" refers to the process of removing noise, filling in missing values, standardizing formats, and otherwise preparing collected raw data in an analyzable format.

[1142] A "local AI model" is an artificial intelligence model that is trained on an individual device to learn user behavior patterns and preferences and make individually customized predictions.

[1143] "Federated learning" is a distributed machine learning technique that aggregates local AI models sent from multiple devices on a server to generate a single, highly accurate global AI model.

[1144] "Encryption" is a technique for converting data or models into a secure format so that third parties cannot access them.

[1145] "Computer" refers to all devices used by users, including smartphones, tablets, and personal computers.

[1146] "Supporting users' daily activities" means automating various choices and actions in daily life based on an improved AI model, and providing appropriate suggestions and support.

[1147] The present invention relates to a system for supporting daily activities of a user, and specific embodiments thereof will be described below.

[1148] Behavioral data collection

[1149] Devices: Every time a user interacts with a device, the device collects activity information in real time. This data includes application usage history, location information, click patterns, and internet browsing history. For example, a smartphone may obtain a user's current location through GPS and record the URLs of websites visited during web browsing.

[1150] Data Preprocessing

[1151] Terminal: The collected raw data may contain noise and missing values, so it is cleansed and standardized into a consistent format. For example, a time series imputation algorithm is used to impute missing data. This process allows for more accurate data analysis.

[1152] Training a local AI model

[1153] On-device: The pre-processed data is used to train a local AI model on the device. This model uses machine learning algorithms (e.g., random forest or LSTM) to learn user behavior patterns and preferences, allowing it to predict what information and actions users are likely to view next.

[1154] Sending a local model to the server

[1155] On the device: Once the local model is trained, it is encrypted for added security and sent to the server using the AES algorithm via the HTTPS protocol.

[1156] Federated learning

[1157] Server: The server receives the local AI models sent from multiple devices and performs federated learning based on them. This process integrates the knowledge of each model to generate a highly accurate global AI model. For example, a federated learning algorithm can be used.

[1158] Deploying improved AI models

[1159] Server: The improved AI model is re-encrypted and delivered to each user's device via a secure channel. This delivery model is provided in a form optimized for each device.

[1160] User support

[1161] Device: Based on the deployed improved AI model, the device assists users in their daily activities, for example, suggesting the best outfit to wear based on their morning commute time, or suggesting restaurants to choose from at lunchtime based on their past dining history.

[1162] Specific examples

[1163] When choosing what to wear

[1164] Device: When the user wakes up in the morning, the device retrieves weather forecast data and suggests the best outfit for the day based on past clothing choices. For example, if rain is forecast, suggestions will be made based on the clothes the user has liked to wear on rainy days in the past.

[1165] When choosing a meal

[1166] Device: When lunchtime approaches, the device will acquire the user's current location information and refer to their past dining history to suggest suitable restaurants. For example, if the user chose Japanese food frequently on the previous Monday, Japanese restaurants will be prioritized.

[1167] Prompt Sentence Examples

[1168] "Search a database for what kind of clothes the user has chosen for what weather in the past, and suggest clothes that suit tomorrow's weather."

[1169] This system efficiently collects and analyzes data and generates highly accurate AI models to effectively support users' daily activities.

[1170] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1171] Step 1:

[1172] Behavioral data collection

[1173] Device: Collects real-time activity information every time you interact with your device, including application launch times, location, click patterns, and internet browsing history.

[1174] Input: User device operation information.

[1175] Output: The raw data collected.

[1176] Specific behavior: For example, when a user opens a smartphone browser and visits a specific website, the device records the URL and the time spent browsing.

[1177] Step 2:

[1178] Data Preprocessing

[1179] Terminal: The raw data collected is denoised, imputed, and standardised into a consistent format. Specifically, missing data is imputed using a time series imputation algorithm.

[1180] Input: The raw data collected.

[1181] Output: Cleansed preprocessed data.

[1182] Specific behavior: If there are any jumps in the location data, the device will delete the data and convert it into consistent data.

[1183] Step 3:

[1184] Training a local AI model

[1185] On the device: A local AI model is trained based on the preprocessed data, specifically using machine learning algorithms (e.g., random forest or LSTM) to learn user behavior patterns.

[1186] Input: Preprocessed data.

[1187] Output: A trained local AI model.

[1188] What it does: Uses website browsing history data to generate a model that predicts which website you're likely to visit next.

[1189] Step 4:

[1190] Sending a local model to the server

[1191] On the device: The trained local model is encrypted and sent to the server over a secure channel, specifically using the AES algorithm and the HTTPS protocol.

[1192] Input: A trained local AI model.

[1193] Output: The encrypted local AI model sent to the server.

[1194] What happens: The device encrypts the local AI model and sends an HTTPS request to the server.

[1195] Step 5:

[1196] Federated learning

[1197] Server: Aggregates local models sent from multiple devices and performs federated learning. Specifically, it uses a federated learning algorithm.

[1198] Input: Multiple local AI models sent from each device.

[1199] Output: An improved global AI model.

[1200] Specific operation: The server integrates the weights of each local model to improve the overall model accuracy.

[1201] Step 6:

[1202] Deploying improved AI models

[1203] Server: The improved AI model is encrypted and delivered to each user's device using the AES algorithm via the HTTPS protocol.

[1204] Input: Improved global AI model.

[1205] Output: An improved AI model delivered to the user's device.

[1206] What it does: The server encrypts the global AI model and sends an HTTPS request to each user.

[1207] Step 7:

[1208] User support

[1209] The device: Based on the deployed improved AI model, it assists the user in their daily activities, such as suggesting outfits based on the daily weather forecast and past clothing data, or suggesting nearby restaurants when it's lunchtime.

[1210] Input: Improved AI model, current conditions (weather forecast, location, etc.).

[1211] Output: Specific suggestions to the user.

[1212] Specific behavior: When the user wakes up in the morning, the system sends a message saying, "It's forecast to rain today. We recommend the jacket you wore on the last rainy day." When lunchtime approaches, the system notifies the user, "We recommend a nearby Japanese restaurant."

[1213] (Application example 1)

[1214] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1215] Currently, technologies for autonomous vehicles that can analyze users' driving behavior and traffic conditions in real time and provide optimal route and parking information are not fully established. As a result, traffic congestion and a lack of parking spaces result in time loss, stress for users, and inefficient driving. Another issue is that information is provided uniformly without taking into account each user's unique driving patterns, resulting in a lack of optimal support. To solve these problems, it is necessary to build an advanced driving assistance system that collects and analyzes user-specific behavioral data and uses further improved AI models.

[1216] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1217] In this invention, the server includes means for collecting user behavioral data, means for preprocessing the collected behavioral data, means for training a local AI model using the preprocessed data, means for transmitting the trained local AI model to the server, means for aggregating multiple local AI models and performing federated learning, means for distributing an improved AI model to the user's device, means for predicting driving behavior based on the deployed improved AI model and providing optimal driving route and parking information, and means for generating prompt sentences to present an optimal driving plan using the generative AI model. This makes it possible to provide optimal driving assistance based on the user's unique driving patterns and real-time traffic conditions.

[1218] "Behavioral Data" refers to various data generated when you operate your device, including your app usage history, location information, click patterns, web browsing history, and driving data.

[1219] "Preprocessing" refers to the process of cleansing the collected behavioral data, filling in missing values, and standardizing the format.

[1220] A "local AI model" is an artificial intelligence model that is trained on a device using pre-processed data to learn user behavior patterns and preferences.

[1221] "Federated learning" is a learning method that aggregates multiple local AI models to generate an overall more accurate AI model.

[1222] "Deployment" is the process of placing an improved AI model on a user's device and actually running it.

[1223] "Driving behavior" refers to specific behavioral patterns and driver preferences when driving a vehicle, including speed, acceleration, and choice of driving route.

[1224] "Driving route" refers to the optimal route to a destination, which is selected based on real-time traffic conditions and the driver's past driving patterns.

[1225] "Parking lot information" refers to information about the location and availability of parking lots necessary for users to park near their destination.

[1226] A "generative AI model" is an artificial intelligence model that is generated through federated learning and individual training and is used to predict user behavior patterns.

[1227] A "prompt" is an instruction used by a generative AI model to suggest an optimal driving plan.

[1228] This invention is a driving assistance system for autonomous vehicles that collects, preprocesses, and analyzes user behavior data and provides optimal driving routes and parking information to improve driving efficiency. Specifically, the system has the following steps and configuration:

[1229] Behavioral data collection

[1230] The server collects user behavior data through sensing devices and APIs. This data includes app usage history, location information, click patterns, web browsing history, driving data (time, speed, weather, traffic conditions, etc.) For example, if a user uses a vehicle during their morning commute, data such as the route, driving time, and speed are collected.

[1231] Data Preprocessing

[1232] The collected data is first cleansed, missing values ​​are imputed, and the format is standardized. The server then performs preprocessing, applying algorithms to impute missing parts of time series data, for example. This process uses Python and the data processing library Pandas.

[1233] Training a local AI model

[1234] The preprocessed data is used on the user's device to train a local AI model, built using major AI libraries such as TensorFlow and Keras. For example, based on driving data from the user's commute, the model can learn the optimal driving route and suggest it for the next commute.

[1235] Sending a local model to the server

[1236] Once trained, the local AI model is encrypted and sent to the server, using an encryption algorithm to ensure user privacy.

[1237] Federated learning

[1238] The server aggregates the received local models and performs federated learning, which generates a more accurate AI model. This process uses a distributed learning algorithm on the server side.

[1239] Deploying improved AI models

[1240] The improved AI model is then re-encrypted and delivered to each user's device, with the model provided in a form optimized for that device.

[1241] User support

[1242] The improved AI model deployed on the device supports the user's daily activities. For example, when driving in the morning, the device will suggest the optimal driving route and parking lot based on the weather forecast and past driving data. It also obtains traffic conditions in real time and optimizes the driving route based on that information.

[1243] Specific examples

[1244] For example, if a user commutes to work in an autonomous vehicle, the "Smart Driving Assistant" app will suggest the optimal driving route based on the weather forecast, traffic conditions, and past driving data before the user starts driving in the morning. If necessary, the app will also notify the user in real time of the availability of suitable parking spaces. Specific examples of prompts are as follows:

[1245] Example prompt:

[1246] "Build an AI model that uses user behavior pattern data (time, location information, speed, weather, traffic conditions) to suggest the next destination and the optimal driving route. Material 1: Driving data from October 2023. Material 2: Current traffic condition data. Output: Suggest the optimal route and next destination to the user."

[1247] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1248] Step 1:

[1249] The server collects user behavioral data (app usage history, location information, click patterns, web browsing history, driving data) through sensing devices and APIs.

[1250] Input: Various user behavior data

[1251] Output: Raw data collected

[1252] Specific operation: The server calls the API and obtains data from the user's smartphone and vehicle sensors.

[1253] Step 2:

[1254] The server preprocesses the collected behavioral data, specifically cleansing the data, filling in missing values, and standardizing the format.

[1255] Input: Raw data collected

[1256] Output: Cleansed data

[1257] Specific operation: The server uses Python and Pandas to create a data frame, impute missing values, and unify the format of each data.

[1258] Step 3:

[1259] The device uses the preprocessed data to train a local AI model.

[1260] Input: Preprocessed data

[1261] Output: A trained local AI model

[1262] Specific operation: The device uses TensorFlow and Keras to build a local AI model and trains the model using data as input.

[1263] Step 4:

[1264] The device encrypts the trained local AI model and sends it to the server.

[1265] Input: A trained local AI model

[1266] Output: Encrypted local AI model

[1267] Specific operation: The terminal encrypts the model using an encryption algorithm and sends it to the server over the network.

[1268] Step 5:

[1269] The server aggregates multiple local AI models and performs federated learning.

[1270] Input: Multiple encrypted local AI models

[1271] Output: An improved AI model generated through federated learning.

[1272] What it does: The server decrypts the encrypted model training data and aggregates the model using a distributed learning algorithm.

[1273] Step 6:

[1274] The server re-encrypts the improved AI model and distributes it to each user's device.

[1275] Input: Improved AI model

[1276] Output: An encrypted, improved AI model

[1277] What it does: The server encrypts the model and delivers it in a format optimized for the user's specific device.

[1278] Step 7:

[1279] The device will then use the deployed improved AI model to assist the user in their daily activities, for example by providing optimal driving routes and parking information.

[1280] Input: Improved AI model, current user data (location, traffic, weather, etc.)

[1281] Output: Driving route and parking information suggestions

[1282] Specific operation: The device uses the generative AI model to process real-time data, generate prompts, and notify the user of optimal suggestions.

[1283] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1284] The present invention combines an emotion engine with a system for supporting the daily activities of users, and specific embodiments thereof are described below.

[1285] Behavioral data collection

[1286] Device: Collects real-time behavioral data as users interact with their devices, including application usage, location, web browsing, and click patterns. The emotion engine also collects emotional data by analyzing users' voice, facial expressions, and text inputs.

[1287] Data Preprocessing

[1288] Terminal: Collected behavioral and emotional data is preprocessed. For behavioral data, noise is removed, missing values ​​are filled, and data formats are standardized. For emotional data, the voice, facial expression, and text input data analyzed by the emotion engine are processed. For example, emotional information extracted from voice data is standardized into text format.

[1289] Training a local AI model

[1290] On-device: The preprocessed behavioral and emotional data is used to train a local AI model on the device. This model can learn the user's behavioral patterns and emotional changes and make personalized predictions. For example, it can learn how a user behaves in a particular emotional state.

[1291] Sending a local model to the server

[1292] On-device: Once trained, the local AI model is encrypted and sent to a server for security purposes. This process occurs periodically and has built-in mechanisms to protect user privacy.

[1293] Federated learning

[1294] Server: Aggregates local AI models received from multiple users and performs federated learning. This integrates the behavioral patterns and emotional information of multiple users to generate an AI model with higher overall accuracy. For example, it comprehensively learns how a user behaves in a specific emotional state.

[1295] Deploying improved AI models

[1296] Server: The improved AI model generated through federated learning is then distributed to the devices again. The distributed model is provided in a form optimized for each device.

[1297] User support

[1298] Device: Based on the deployed and improved AI model, the device supports the user's daily activities. Specifically, it makes real-time suggestions based on behavioral patterns and emotional information. For example, if the user is feeling stressed, it will suggest relaxation activities or provide content to help them change their mood.

[1299] Specific examples

[1300] When choosing what to wear

[1301] Device: When the user wakes up in the morning, the device will suggest the best outfit for the day based on weather forecast data, past clothing choices, and the user's current emotional state. For example, if the user is feeling stressed, the device will suggest comfortable clothing.

[1302] When choosing a meal

[1303] Device: When lunchtime approaches, the device will suggest suitable restaurants based on the user's past dining history, current location, and emotional state. If the user tends to frequent certain cuisines or restaurants, suggestions will be based on that. For example, if the user is tired, the device will prioritize restaurants that serve nutritious meals.

[1304] In this way, the implementation of the present invention can automatically support users' daily choices and actions, saving them time and effort, and by taking into account the user's emotional state, more personalized suggestions can be made, improving their quality of life.

[1305] The processing flow will be explained below.

[1306] Step 1:

[1307] Device: Collects behavioral data when users operate the device. Specifically, it stores real-time logs of users' application usage history, location information, web browsing history, click patterns, etc. Furthermore, it utilizes an emotion engine to collect emotional data from users' voice, facial expressions, and text input.

[1308] Step 2:

[1309] Terminal: Preprocessing the collected raw data and emotion data. Preprocessing of behavioral data involves removing noise, filling in missing data, and standardizing data formats. Preprocessing of emotion data involves standardizing the emotion information extracted from audio data into text format. It also standardizes the emotion labels extracted from facial images.

[1310] Step 3:

[1311] Device: Trains a local AI model using preprocessed behavioral and emotional data. Specifically, it creates a local machine learning model to learn user behavior patterns and emotional changes, and then uses the training dataset to improve the model's accuracy.

[1312] Step 4:

[1313] Device: The device transmits the trained local AI model to the server. First, the model is encrypted and uploaded to the server using a secure communication protocol. This transmission is done periodically to protect the user's privacy.

[1314] Step 5:

[1315] Server: Aggregates local AI models received from multiple users. Specifically, it integrates the parameters of each local model to generate a single integrated model. This process uses a federated learning algorithm to efficiently integrate the information from each model.

[1316] Step 6:

[1317] Server: Performs federated learning to generate a more accurate AI model overall. Retrains the aggregated model to create an improved model that reflects the data and sentiment information of all users.

[1318] Step 7:

[1319] Server: The improved AI model is then sent back to the device. The improved model is then encrypted and sent to each user's device via a secure communication protocol. At this time, the model is adjusted to be optimized for each device.

[1320] Step 8:

[1321] Device: The deployed and improved AI model is used to assist the user in their daily activities. The model makes real-time suggestions based on the user's behavioral patterns and emotional information. For example, if the user is feeling stressed, it will suggest relaxation activities. If the user is tired, it will recommend restaurants with nutritious meals.

[1322] Example 2

[1323] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1324] Currently, most behavioral support systems make suggestions based solely on the user's behavioral data, lacking detailed support that takes into account the user's emotional state. Furthermore, while secure management and use of collected data is required from the perspective of privacy protection, systems that address this need are limited. Furthermore, methods for efficiently aggregating individually trained local AI models and improving accuracy through federated learning are also insufficient. To address these issues, the present invention provides a federated learning system that combines behavioral data and emotional data.

[1325] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1326] In this invention, the server includes means for collecting user behavioral data and emotional data, means for preprocessing the collected behavioral data and emotional data, means for training a local AI model using the preprocessed data, means for encrypting the trained local AI model and transmitting it to the server, means for aggregating multiple local AI models and performing federated learning, means for delivering an improved AI model to the user's device, and means for supporting the user's daily behavior and emotional state based on the delivered improved AI model. This enables personalized suggestions based on the user's behavior and emotions, and makes it possible to improve the accuracy of the model while safely managing data.

[1327] "Behavioral data" refers to data such as application usage history, location information, click patterns, and web browsing history generated when a user operates a device.

[1328] "Emotion data" refers to data related to emotions generated from the user's voice, facial expressions, and text input.

[1329] "Preprocessing" refers to processing of collected data to remove noise, fill in missing data, and standardize the data format.

[1330] A "local AI model" is an artificial intelligence model that is trained on a user's device and makes individually customized predictions.

[1331] "Encryption" is the process of transforming data using a specific algorithm to keep the information confidential.

[1332] A "server" is a computer system that aggregates multiple local AI models and provides the computational resources to perform federated learning.

[1333] "Federated learning" is a learning method that integrates knowledge gained from multiple local AI models to generate an AI model with high overall accuracy.

[1334] An "improved AI model" refers to an AI model whose accuracy has been improved through federated learning.

[1335] "Deployment" is the process of placing software or AI models in a specific location or device and making them operational.

[1336] "Support" refers to providing appropriate suggestions and advice based on the user's daily behavior and emotional state.

[1337] The present invention combines an emotion engine with a system for supporting users' daily activities, and is mainly composed of the following steps.

[1338] Behavioral data collection

[1339] Device: When a user operates a device, behavioral data such as application usage history, location information, web browsing history, and click patterns are collected in real time. An emotion engine is also used to simultaneously collect emotional data from voice, facial expressions, and text input. The emotion engine uses voice recognition software and facial expression recognition software. For example, Google Cloud Speech-to-Text can be used for voice recognition, and Microsoft Azure Face API can be used for facial expression recognition.

[1340] Data Preprocessing

[1341] Device: Noise is removed from the collected behavioral and emotional data, missing data is filled in, and the data format is standardized. For emotional data, information extracted from voice and facial expression data is standardized and unified in text format. For example, voice data is converted to text using Google Cloud Speech-to-Text, and sentiment analysis is performed using Azure Text Analytics.

[1342] Training a local AI model

[1343] On-device: The preprocessed behavioral and emotional data is used to train a local AI model. TensorFlow is used for training, and the model learns the relationship between a user's behavioral patterns and emotions. For example, the model can predict how a user will behave in a given emotional state.

[1344] Sending a local model to the server

[1345] Terminal: After completing training, the local AI model is encrypted and sent to the server. The encryption is performed using AES (Advanced Encryption Standard) technology. Specifically, the model data is encrypted using Python's cryptography library.

[1346] Federated learning

[1347] Server: Aggregates local AI models sent from multiple users and performs federated learning. This integrates the behavioral patterns and emotional data of multiple users to generate a more accurate AI model. TensorFlow Federated is used for federated learning.

[1348] Deploying improved AI models

[1349] Server: The improved model generated through federated learning is distributed to each device. The distribution is encrypted and ensures secure reception by each device.

[1350] User support

[1351] Device: Based on an improved AI model, the device supports the user's daily activities and emotional state. Specifically, it makes real-time suggestions based on behavioral and emotional data. For example, if the user is feeling stressed, it will suggest relaxation activities. It also combines weather forecast data, past behavioral data, and emotional state to make specific suggestions.

[1352] Specific examples

[1353] When choosing what to wear

[1354] Device: When the user wakes up, the system suggests the most appropriate outfit based on weather forecast data, past clothing choices, and the user's current emotional state. For example, if the user is feeling stressed, it suggests clothing that prioritizes comfort. The weather forecast API uses OpenWeatherMap.

[1355] When choosing a meal

[1356] Device: When lunchtime approaches, the system suggests suitable restaurants based on the user's past meal history, current location, and emotional information. For example, if the user is tired, it will prioritize restaurants that serve nutritious meals. The location service uses the Google Maps API.

[1357] Prompt Sentence Examples

[1358] Example input: Design a system that makes lunch suggestions based on the user's eating history, location, and emotional state. Explain how the system works.

[1359] In this way, by implementing the invention, personalized suggestions based on the user's behavior and emotions become possible, and highly accurate support can be provided while safely managing data.

[1360] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1361] Step 1:

[1362] Collecting behavioral and emotional data

[1363] Terminal: When a user operates a device, behavioral data such as application usage history, location information, web browsing history, and click patterns are collected in real time. An emotion engine is also used to collect emotional data from voice, facial expressions, and text input. The input is sensor data and user operation data, and the output is the collected raw data. For example, voice data is obtained from a microphone, and location information is obtained from a GPS.

[1364] Step 2:

[1365] Data Preprocessing

[1366] Terminal: The collected behavioral and emotional data is subjected to noise removal, missing data completion, and data format unification. Voice and facial expression data is standardized and unified into text format. The input is the collected raw data, and the output is preprocessed clean data. Specific operations include using Python's pandas library to complete missing data and unify the time format. In addition, the voice data is converted to text using Google Cloud Speech-to-Text, and emotions are analyzed using Azure Text Analytics.

[1367] Step 3:

[1368] Training a local AI model

[1369] Terminal: A local AI model is trained using preprocessed behavioral data and emotion data. The input is the preprocessed dataset, and the output is the trained local AI model. TensorFlow is used for training, and a model is built that can predict what behavior will be taken in a specific emotional state. Specifically, a neural network is built using TensorFlow's Keras API, and the dataset is input to train the model.

[1370] Step 4:

[1371] Encrypting and sending the local model

[1372] Terminal: The local AI model that has completed training is encrypted and sent to the server. The input is the trained local AI model, and the output is the encrypted and sent model. The encryption is performed using AES (Advanced Encryption Standard) technology. Specifically, the model data is encrypted using Python's cryptography library, and uploaded to the server via the HTTPS protocol using the requests library.

[1373] Step 5:

[1374] Server-based federated learning

[1375] Server: Aggregates local AI models sent by multiple users and performs federated learning. The input is an encrypted local model, and the output is an AI model improved through federated learning. TensorFlow Federated is used for federated learning. Specifically, it runs a federated learning algorithm and combines the parameters of multiple local models to generate a new model.

[1376] Step 6:

[1377] Encoding and delivering the improved model

[1378] Server: Encodes and encrypts the improved model generated by federated learning for distribution to each device. The input is the improved model, and the output is the encrypted and encoded model. Specifically, the improved model data is encoded in Base64 and then encrypted with AES.

[1379] Step 7:

[1380] Receive and deploy the improved model

[1381] Terminal: Decrypts and decodes the received improved model and deploys it locally. The input is the encrypted and encoded model, and the output is the decoded and deployed model. Specifically, the encryption is performed using Python's cryptography library, and the model is loaded using Model.load().

[1382] Step 8:

[1383] Supporting users' daily activities and emotional state

[1384] Device: Based on an improved AI model, the device supports the user's behavior and emotional state. The input is the user's current emotional state and behavioral data, and the output is real-time support suggestions. Specifically, the device uses the model's predictive capabilities to generate appropriate suggestions based on the user's current emotional state and behavioral history, and displays them as notifications.

[1385] (Application example 2)

[1386] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1387] Conventional autonomous driving systems do not take into account the emotional state of the user, which results in the inability to reduce the user's stress and discomfort. Therefore, real-time driving mode adjustment and behavior suggestions based on emotional data are necessary.

[1388] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting behavioral data and emotional data of the user, means for preprocessing the collected behavioral data and emotional data, and means for training a local AI model using the preprocessed data and emotional data. This makes it possible to automatically adjust the driving mode based on the emotional state of the user and to suggest appropriate actions.

[1389] "Behavioral Data" is data that records a user's everyday behavior, such as their application usage history, location information, click patterns, and web browsing history.

[1390] "Emotion data" refers to data that includes emotional information extracted from a user's voice, facial expression, and text input.

[1391] "Preprocessing" refers to processes such as removing noise from collected data, filling in missing values, and standardizing data formats.

[1392] A "local AI model" is an artificial intelligence model that learns user behavioral patterns and emotional changes and makes individually customized predictions.

[1393] A "server" is a centralized computer system that aggregates multiple local AI models and performs federated learning.

[1394] "Federated learning" is a machine learning technique that integrates local AI models received from multiple devices to generate a more accurate AI model.

[1395] An "improved AI model" is an artificial intelligence model that is generated through federated learning and has overall higher accuracy.

[1396] "Driving mode adjustment" refers to changing the driving mode of an autonomous vehicle based on the emotional state of the user.

[1397] "Behavioral Suggestion" refers to suggesting specific activities or content based on a user's current behavioral patterns and emotional state.

[1398] The present invention is a system for adjusting driving modes and suggesting actions in response to the user's daily behavior and emotional state in an autonomous vehicle. This system is realized through cooperation between a terminal and a server.

[1399] Collecting behavioral and emotional data

[1400] Device: Through sensors, cameras, and microphones installed inside the autonomous vehicle in which the user is riding, the device collects user behavioral data (application usage history, location information, click patterns, web browsing history) and emotional data (voice, facial expressions, text input) in real time.

[1401] Data Preprocessing

[1402] Terminal: The collected behavioral and emotional data undergoes noise removal, missing values ​​are filled in, and the data format is standardized. In particular, for emotional data, processing such as converting voice data into text format is performed.

[1403] Training a local AI model

[1404] On-device: Using pre-processed behavioral and emotional data, a local AI model is trained. This model is customized for each user and learns what behaviors and requests are predicted in specific emotional states.

[1405] Server submission and federated learning of local models

[1406] Device: Once trained, the local AI model is encrypted and sent to the server, where it aggregates the local AI models sent by multiple users and performs federated learning, resulting in an improved AI model with greater overall accuracy.

[1407] Deploying improved AI models

[1408] Server: The improved AI model generated through federated learning is then distributed to the devices again. This model is provided in a form optimized for each device.

[1409] User support

[1410] Device: The deployed improved AI model can adjust the driving mode and suggest appropriate actions in real time based on the user's behavioral patterns and emotional information. For example, if the user is feeling stressed, the driving mode can be set to "Relaxation Mode" and play relaxing music. If the user is having fun, the driving mode can be changed to "Dynamic Mode" and suggest a guided city tour.

[1411] Examples of concrete examples and prompts

[1412] Example 1: When the user is feeling stressed, the vehicle switches to "relaxation mode" and suggests "playing relaxing music."

[1413] Example 2: If the user is happy, the system switches to "dynamic mode" and suggests a "guided city tour."

[1414] An example of a prompt sentence to input to a generative AI model is:

[1415] "Generate appropriate driving modes and behavior suggestions based on the user's emotional data."

[1416] In this way, the present invention enables an autonomous vehicle to adjust its driving mode and suggest actions in accordance with the user's emotional state, providing a safer and more comfortable driving experience.

[1417] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1418] Step 1:

[1419] The device collects user behavioral data (application usage history, location information, click patterns, web browsing history) and emotional data (voice, facial expressions, text input). The input at this stage is raw data obtained from sensors, cameras, and microphones, and the collected raw data is obtained as the output.

[1420] Step 2:

[1421] The collected behavioral and emotional data is preprocessed on the device. The input raw data undergoes processes such as noise removal, missing value completion, and data format standardization before being output as preprocessed data. Specific operations include converting voice data into text format.

[1422] Step 3:

[1423] The device uses the preprocessed behavioral and emotional data to train a local AI model. The input to this process is the preprocessed data, which is then used for learning and analysis. The output is a local AI model that is customized for each user.

[1424] Step 4:

[1425] The device encrypts the trained local AI model and sends it to the server. The input of this process is the trained local AI model, and the output is the encrypted model data. Specifically, a data encryption algorithm is used.

[1426] Step 5:

[1427] The server aggregates multiple local AI models and performs federated learning. The input to this process is the local AI models received from multiple devices, and the output is an improved AI model with higher overall accuracy. Specifically, learning is performed using a federated learning algorithm that integrates multiple models.

[1428] Step 6:

[1429] The server distributes the improved AI model to each device. The input to this process is the improved AI model generated by federated learning, and the output is the model data that is redistributed to each device. Specific operations use a secure data transfer protocol.

[1430] Step 7:

[1431] The device adjusts the driving mode based on the user's behavioral patterns and emotional information using the deployed improved AI model. The input for this process is the improved AI model and real-time collected behavioral and emotional data, and the output is an adjusted driving mode and suggested actions. Specifically, if the emotional state is "stressed," the device switches to "relaxed mode," and if the emotional state is "happy," the device switches to "dynamic mode."

[1432] Step 8:

[1433] The device displays and executes suggested actions to the user. The input of this process is the result of adjusting the driving mode based on the improved AI model, and the output is specific suggested actions to the user (e.g., playing relaxing music, starting a guided city tour). Specific actions include activating a trigger to perform the selected activity.

[1434] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating 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.

[1435] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1436] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1437] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1438] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1439] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1440] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1441] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1442] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1443] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1444] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1445] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1446] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may 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 executes the specific processing in accordance with the specific processing program 56.

[1447] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1448] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1449] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1450] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1451] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1452] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1453] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1454] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1455] The following is further disclosed regarding the above embodiment.

[1456] (Claim 1)

[1457] A means of collecting user behavior data;

[1458] means for preprocessing the collected behavioral data;

[1459] means for training a local AI model using the preprocessed data;

[1460] a means for transmitting the trained local AI model to a server;

[1461] A means of aggregating multiple local AI models and conducting federated learning;

[1462] A means for delivering the improved AI model to the user's device; and

[1463] A system that includes a means to assist users in their daily activities based on the delivered improved AI model.

[1464] (Claim 2)

[1465] 2. The system of claim 1, wherein the behavioral data includes a user's app usage history, location information, click patterns, and web browsing history.

[1466] (Claim 3)

[1467] The system of claim 1, wherein the preprocessed data undergoes processing including filling in missing data and standardizing format.

[1468] "Example 1"

[1469] (Claim 1)

[1470] A means for collecting user behavior data;

[1471] means for pre-processing the collected motion data;

[1472] means for training a local AI model on the device using the preprocessed data;

[1473] means for transmitting the trained local AI model to a server;

[1474] A means of aggregating multiple local AI models and conducting federated learning;

[1475] a means for delivering the improved AI model to a user's computer;

[1476] A system that includes a means of assisting users in their daily activities based on the delivered improved AI model.

[1477] (Claim 2)

[1478] 10. The system of claim 1, wherein the behavioral data includes a user's application usage history, location information, click patterns, and internet browsing history.

[1479] (Claim 3)

[1480] The system of claim 1, wherein the preprocessed data undergoes processing including filling in missing data and standardizing format.

[1481] "Application Example 1"

[1482] (Claim 1)

[1483] A means of collecting user behavior data;

[1484] means for preprocessing the collected behavioral data;

[1485] means for training a local AI model using the preprocessed data;

[1486] a means for transmitting the trained local AI model to a server;

[1487] A means of aggregating multiple local AI models and conducting federated learning;

[1488] A means for delivering the improved AI model to the user's device; and

[1489] A means to predict driving behavior based on the deployed improved AI model and provide optimal driving routes and parking information;

[1490] A system including a means for generating prompt sentences to present an optimal driving plan using a generative AI model.

[1491] (Claim 2)

[1492] 2. The system of claim 1, wherein the behavioral data includes a user's app usage history, location information, click patterns, web browsing history, and driving data.

[1493] (Claim 3)

[1494] The system of claim 1, wherein the preprocessed data undergoes processing including filling in missing data and standardizing format.

[1495] "Example 2: Combining Emotion Engines"

[1496] (Claim 1)

[1497] a means for collecting user behavioral and emotional data;

[1498] means for preprocessing the collected behavioral and emotional data;

[1499] means for training a local AI model using the preprocessed data;

[1500] A means to encrypt the trained local AI model and send it to the server;

[1501] A means of aggregating multiple local AI models and conducting federated learning;

[1502] A means for delivering the improved AI model to the user's device;

[1503] A system that includes a means to support the user's daily activities and emotional state based on the delivered improved AI model.

[1504] (Claim 2)

[1505] 10. The system of claim 1, wherein the behavioral data includes a user's application usage history, location information, click patterns, and web browsing history.

[1506] (Claim 3)

[1507] 2. The system of claim 1, wherein the preprocessed data includes filling in missing data and unifying data formats to standardize emotion data.

[1508] "Application example 2 when combining emotion engines"

[1509] (Claim 1)

[1510] a means for collecting user behavioral and emotional data;

[1511] means for preprocessing the collected behavioral and emotional data;

[1512] means for training a local AI model using the preprocessed data and the emotion data;

[1513] a means for transmitting the trained local AI model to a server;

[1514] A means of aggregating multiple local AI models and conducting federated learning;

[1515] A means for delivering the improved AI model to the user's device; and

[1516] The system includes a means for adjusting driving modes and suggested actions in response to the user's emotions based on the delivered improved AI model.

[1517] (Claim 2)

[1518] 2. The system of claim 1, wherein the behavioral data includes a user's application usage history, location information, click patterns, and web browsing history, and the emotional data includes voice, facial expressions, and text input.

[1519] (Claim 3)

[1520] 2. The system of claim 1, wherein the preprocessed data includes filling in missing data and standardizing formats, and the preprocessed emotion data includes processing to standardize emotion information extracted from audio data into text format. [Explanation of symbols]

[1521] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. A means of collecting user behavior data; means for preprocessing the collected behavioral data; means for training a local AI model using the preprocessed data; a means for transmitting the trained local AI model to a server; A means of aggregating multiple local AI models and conducting federated learning; A means for delivering the improved AI model to the user's device; and A system that includes a means to assist users in their daily activities based on the delivered improved AI model.

2. The system of claim 1 , wherein the behavioral data includes a user's app usage history, location information, click patterns, and web browsing history.

3. The system of claim 1, wherein the preprocessed data undergoes processing including missing data completion and format unification.

Images