Monitoring support system, monitoring support method, and program

Abstract

This system provides a monitoring support system, monitoring support method, and program that utilizes information from the daily conversations and terminal usage of elderly people to understand their condition, and provides information to family members and others in response to changes in their condition. [Solution] The monitoring support system includes: a data acquisition unit that acquires dialogue data entered between the elderly person and a dialogue agent and operation logs on the elderly person's terminal used by the elderly person; an analysis unit that calculates a state score representing the elderly person's state based on the dialogue data and the operation logs; a determination unit that detects a change in the elderly person's state based on the difference between the state score calculated by the analysis unit and a reference value based on state scores previously calculated for the elderly person, and determines a notification level corresponding to the change in state; and a notification unit that sends a notification with content corresponding to the notification level determined by the determination unit to a family terminal associated with the elderly person.

Description

[[Technical Field]]

[0001] The present invention relates to elderly monitoring technology, and particularly relates to a monitoring support system, a monitoring support method, and a program that estimate the condition and changes in the condition of an elderly person using dialogue data and operation logs, and notify the result to the family. [[Background Art]]

[0002] Conventionally, methods using camera monitoring, sensor devices such as human motion sensors, or wearable devices have been generally used for monitoring elderly people. However, these methods have problems including concerns about privacy infringement caused by imaging, physical burden from wearing the device, and issues related to installation and maintenance costs.

[0003] In addition, simple safety confirmation methods such as safety notification via push buttons or scheduled check-ins make it difficult to grasp subtle changes in the condition of elderly people, such as increased feelings of loneliness, signs of cognitive function decline, or disruption of life rhythms. For this reason, there is a demand for technology that continuously grasps the condition of an elderly person by using information that can be naturally acquired in the elderly person's daily life.

[0004] One example of information that can be naturally acquired in daily life is dialogue data between a user and an information processing device. In recent years, along with the development of dialogue technology using artificial intelligence, technology for estimating a user's emotion, mood, intention, etc. from dialogue data has been proposed. For example, Patent Document 1 is known as a technology for estimating emotion or mood from dialogue, and Patent Document 2 is known as a technology for continuously tracking emotion or intention from past dialogue data. [[Prior Art Documents]] [[Patent Documents]]

[0005] [[Patent Document 1]] Japanese Unexamined Patent Publication No. 2023-175380 [[Patent Document 2]] Japanese Unexamined Patent Publication No. 2025-170730 [[Summary of the Invention]] [Problems that the invention aims to solve]

[0006] However, while the above-mentioned conventional technology can estimate emotions or intentions from dialogue data, it has not adequately considered how to continuously monitor changes in the elderly person's condition using these estimation results, and how to detect abnormalities or signs of change based on the normal state, which differs for each elderly person. Furthermore, a mechanism for adaptively controlling the content or level of notifications sent to family members, etc., according to the degree of detected changes in condition has not been adequately implemented.

[0007] This invention has been made in view of the above problems, and aims to provide a monitoring support system, a monitoring support method, and a program that can grasp the condition of an elderly person by using their daily conversation information and terminal usage information, and provide information to family members, etc., in accordance with changes in that condition. [Means for solving the problem]

[0008] A monitoring support system according to one aspect of the present invention includes: a data acquisition unit that acquires dialogue data entered between the elderly person and a dialogue agent and operation logs on the elderly person's terminal used by the elderly person; an analysis unit that calculates a state score representing the elderly person's state based on the dialogue data and the operation logs; a determination unit that detects a change in the elderly person's state based on the difference between the state score calculated by the analysis unit and a reference value based on state scores previously calculated for the elderly person, and determines a notification level corresponding to the change in state; and a notification unit that sends a notification with content corresponding to the notification level determined by the determination unit to a family terminal associated with the elderly person. According to this aspect of the monitoring support system, the elderly person's state can be grasped by using the elderly person's daily dialogue data and information on terminal usage, and information corresponding to changes in that state can be provided to the family, thus supporting the monitoring of the elderly person without using additional sensors, surveillance cameras, or wearable devices.

[0009] Furthermore, a monitoring support method according to one aspect of the present invention includes the following steps: an acquisition step in which a computer acquires dialogue data entered between the elderly person and a dialogue agent and an operation log on the elderly person's terminal used by the elderly person; a calculation step in which a computer calculates a state score representing the elderly person's state based on the dialogue data and the operation log; a determination step in which a computer detects a change in the elderly person's state based on the difference between the state score calculated in the calculation step and a reference value based on state scores previously calculated for the elderly person, and determines a notification level corresponding to the change in state; and a notification step in which a notification corresponding to the notification level determined in the determination step is sent to a family terminal associated with the elderly person. This aspect of the monitoring support method can achieve the same effects as the above-described monitoring support system.

[0010] Furthermore, a program according to one aspect of the present invention is a program that causes a computer to execute each step of the above-described monitoring support method. According to this aspect of the program, the same effects as the above-described monitoring support system and monitoring support method can be achieved. [Effects of the Invention]

[0011] According to the present invention, it is possible to understand the condition of elderly people by utilizing their daily conversation information and terminal usage information, and to provide information to family members or others in response to changes in their condition. Furthermore, it is possible to support the monitoring of elderly people without using additional sensors, surveillance cameras, or wearable devices. [Brief explanation of the drawing]

[0012] [Figure 1] This is a block diagram showing the overall configuration of a monitoring support system according to one embodiment of the present invention. [Figure 2] This figure shows an example of the hardware configuration of the cloud server according to this embodiment. [Figure 3] This is a block diagram showing an example of the functional configuration of a terminal for the elderly according to this embodiment. [Figure 4] This is a block diagram showing an example of the functional configuration of the cloud server according to this embodiment. [Figure 5] This flowchart shows the flow of dialogue processing and dialogue data processing according to this embodiment. [Figure 6] This flowchart shows the flow of the state analysis process according to this embodiment. [Figure 7] This flowchart shows the flow of the state change determination process and notification process according to this embodiment. [Figure 8] This flowchart shows the flow of the reference value update process and the personalization model update process according to this embodiment. [Figure 9] This flowchart shows the flow of disaster monitoring processing according to this embodiment. [Figure 10] This flowchart shows the flow of emergency notification processing and peace notification processing according to this embodiment. [Modes for carrying out the invention]

[0013] One embodiment of the present invention will be described below with reference to the drawings. In each drawing, the same components are denoted by the same reference numerals, and redundant explanations are omitted.

[0014] Figure 1 is a block diagram showing the overall configuration of a monitoring support system 1 according to one embodiment of the present invention. As shown in Figure 1, the monitoring support system 1 comprises an elderly terminal 10 used by the elderly, a cloud server 20, and a family terminal 30 used by family members. The elderly terminal 10, the cloud server 20, and the family terminal 30 are connected to each other so as to be able to communicate with each other via a communication network NW such as the Internet.

[0015] The elderly terminal 10 is constituted by a smartphone, a tablet terminal, a smart speaker, a personal computer, or any other information processing terminal. The elderly terminal 10 acquires information related to dialogues conducted between the elderly user and a dialogue agent and information related to terminal usage, and transmits said information to the cloud server 20. The family terminal 30 is constituted by a smartphone, a tablet terminal, a personal computer, or any other information processing terminal, and receives and displays monitoring notifications transmitted from the cloud server 20.

[0016] The cloud server 20 is a server device that executes various types of information processing in the monitoring support system 1. The cloud server 20 accumulates information acquired from the elderly terminal 10, and executes analysis related to the state of the elderly user and notification control based on said information.

[0017] Furthermore, the cloud server 20 is communicably connected to an external device 40 that provides weather information, earthquake information, and other disaster information, and to an operation management platform 50 that operates and manages the monitoring support service. The external device 40 is constituted by a weather information providing server, a disaster information distribution server, or any other information providing system. The operation management platform 50 manages user information, notification setting information, and other management information.

[0018] Note that data transmitted and received between the elderly terminal 10 and the cloud server 20 is preferably encrypted to ensure communication security and protect user privacy.

[0019] Figure 2 is a diagram showing an example of the hardware configuration of the cloud server 20 according to the present embodiment. The cloud server 20 includes at least a Central Processing Unit (CPU) 21, a storage device 22, a communication Interface (I / F) 23, an input device 24, and an output device 25.

[0020] The CPU 21 functions as a control unit that controls the entire cloud server 20. By reading and executing programs and various data stored in the storage device 22, it performs tasks such as acquiring dialogue data and operation logs transmitted from the elderly terminal 10, analyzing the elderly person's condition, determining changes in condition, and notifying the family terminal 30.

[0021] The storage device 22 comprises volatile memory and non-volatile memory. The storage device 22 stores the operating system, the monitoring support program, and various data. The various data includes dialogue data, operation logs, status scores, reference values, notification history, user information, and other information necessary for providing the monitoring support service.

[0022] Communication I / F23 communicates data with the elderly terminal 10, family terminal 30, external device 40, and operation management platform 50 via wired or wireless communication lines. Communication I / F23 can connect to the Internet, mobile communication networks, local area networks (LANs), and other communication networks.

[0023] The input device 24 includes a keyboard, mouse, touch panel, etc., and accepts various setting information from the administrator. The output device 25 includes a display, speaker, printer, etc., and outputs usage status, notification history, system management information, and other information.

[0024] The various functional units in the cloud server 20 are realized by the execution of programs stored in the storage device 22 by the CPU 21, based on these hardware configurations. These programs may be provided stored on optical discs, magnetic discs, semiconductor memory, or other non-temporary recording media, or they may be distributed via a communication network NW.

[0025] The elderly terminal 10 and the family terminal 30 can also be configured as information processing terminals equipped with a CPU, storage device, communication interface, input device, output device, etc. The detailed hardware configuration is the same as that of the cloud server 20, so a detailed explanation is omitted.

[0026] Figure 3 is a block diagram showing an example of the functional configuration of the elderly terminal 10 according to this embodiment. As shown in Figure 3, the elderly terminal 10 includes a dialogue interface 101, an operation log recording unit 102, an operation reception unit 103, and a location information acquisition unit 104. Each of these functions may be realized by the processor of the elderly terminal 10 executing a program stored in a storage device.

[0027] The dialogue interface 101 is a function that mediates dialogue between the elderly person and the dialogue agent, and supports at least one of voice input, voice output, text input, and text display. For example, the dialogue interface 101 acquires the voice spoken by the elderly person and sends it to the cloud server 20, and presents the response received from the cloud server 20 to the elderly person as voice or text.

[0028] The dialogue agent provided via the dialogue interface 101 generates different dialogue content depending on the current time. Specifically, if the dialogue agent determines that it is the first dialogue of the day, it determines the time of day based on the current time and generates dialogue content appropriate to that time of day.

[0029] For example, during the period from 0:00 to 11:00, the system generates dialogue content that includes morning greetings, confirmation of sleep status, confirmation of breakfast, presentation of the day's weather forecast, and confirmation or prompting for input of the day's itinerary. Furthermore, during the period from 12:00 to 17:00, the system generates dialogue content that includes midday greetings, confirmation of lunch, confirmation of exercise, presentation of the weather forecast, and confirmation of the itinerary.

[0030] Furthermore, during the period from 6 PM to 8 PM, the system generates dialogue content that includes evening greetings, confirmation of dinner, presentation of the weather forecast, and confirmation of the itinerary. Additionally, during the period from 8 PM to midnight, the system generates dialogue content that includes evening greetings, a review of the day, confirmation of medication, confirmation of the next day's itinerary, and encouragement to go to bed.

[0031] If the elderly terminal 10 is not connected to the communication network NW, the conversational agent may determine the time of day using time information based on the internal clock of the elderly terminal 10.

[0032] The operation log recording unit 102 records operation logs that show the usage status of the elderly terminal 10. The operation logs include application launch history, number of operations, operation time, usage time, screen transition history, response time, and inactivity time. The operation logs recorded by the operation log recording unit 102 are sent to the cloud server 20 at predetermined timings and used for status analysis described later.

[0033] The operation reception unit 103 is a function that accepts various operations from elderly people, and includes an SOS button and a safety button. The SOS button is an input means for elderly people to operate in an emergency, and if a predetermined operation such as continuous pressing for a predetermined time or longer is detected, the transmission of an emergency notification is initiated. The safety button is an input means for elderly people to actively notify others of their safety or well-being, and if an operation on this button is detected, a safety notification is sent to the family terminal 30. Predetermined operations may include long-pressing the SOS button, tapping in a predetermined pattern, voice commands, and other operations.

[0034] The location information acquisition unit 104 acquires the location information of the elderly terminal 10 using GPS (Global Positioning System), mobile phone base station information, wireless LAN access point information, and other positioning means. The acquired location information is used for checking the safety of the elderly in the event of a disaster, sending emergency notifications, or understanding the elderly person's movements.

[0035] The elderly person terminal 10 is equipped with an application for realizing the monitoring support service according to this embodiment. The elderly person can engage in daily conversations with a conversational agent via voice or text through this application. In addition, one or more family terminals 30 that will be used as notification recipients are pre-associated and registered with the application. The registration information may include the names of family members, contact information, messaging service account information, notification priority, and notification conditions. As a result, the cloud server 20 can send notifications to the appropriate family terminal 30 when it detects a change in the elderly person's condition, the occurrence of a disaster, or an emergency operation.

[0036] Figure 4 is a block diagram showing an example of the functional configuration of the cloud server 20 according to this embodiment. As shown in Figure 4, the cloud server 20 includes a data acquisition unit 201, an analysis unit 202, a determination unit 203, a learning unit 204, a notification unit 205, and a storage unit 206. Each of these units is realized by the CPU 21 executing a program stored in the storage device 22.

[0037] The data acquisition unit 201 receives dialogue data and operation logs transmitted from the elderly terminal 10. The dialogue data includes voice data, text data, dialogue date and time information, and response history exchanged between the elderly person and the dialogue agent. The operation log includes application launch history, number of operations, usage time, screen transition history, response time, inactivity time, and other information indicating terminal usage status. The data acquisition unit 201 may also acquire location information transmitted from the elderly terminal 10 and disaster information acquired from the external device 40.

[0038] The analysis unit 202 extracts features from dialogue data and operation logs to perform a state analysis of elderly individuals. The extracted features include the number of occurrences of words expressing negative emotions, the number of occurrences of words related to health, emotion scores, the number of dialogues on the day, average response time, activity time, number of application launches, inactivity time, and indicators representing the regularity of daily rhythms. The analysis unit 202 includes the dialogue data processing unit 2021, the activity analysis unit 2022, the psychological analysis unit 2023, the health analysis unit 2024, the integration unit 2025, and the language feature calculation unit 2026.

[0039] If the dialogue data is audio data, the dialogue data processing unit 2021 performs preprocessing on the dialogue data, including conversion to text data by speech recognition, noise reduction, correction of typographical errors, and standardization of inconsistencies in notation. Subsequently, the dialogue data processing unit 2021 performs natural language analysis, including morphological analysis, syntactic analysis, sentiment analysis, and intent analysis, on the preprocessed dialogue data to extract entities related to health-related information, lifestyle information, medication / hospital visit information, and psychological / emotional information. The extracted information is generated as structured data, such as JSON format data, stored in the storage unit 206, and also provided for various analysis processes by the analysis unit 202.

[0040] The Activity Analysis Unit 2022 estimates the daily rhythm of elderly individuals based on operation logs and uses the estimated daily rhythm to calculate an activity score. The Activity Analysis Unit 2022 also calculates the activity score based on activity time, application launch counts, and inactivity time.

[0041] Psychological Analysis Department 2023 estimates the emotional state of elderly individuals based on the results of natural language analysis and calculates an emotional score representing that emotional state. Furthermore, Psychological Analysis Department 2023 calculates a psychological score based on the number of occurrences of words expressing negative emotions, the emotional score, and the change from the previous day. Health Analysis Department 2024 calculates a health score based on the number of occurrences of health-related words, sleep patterns, medication status, and utterances regarding physical condition.

[0042] The activity score calculated by the Activity Analysis Department 2022, the psychological score calculated by the Psychological Analysis Department 2023, and the health score calculated by the Health Analysis Department 2024 are each calculated as values ​​between 0 and 100. In one embodiment, each score is calculated using the following formula, and if the calculated value is less than 0, it is corrected to 0, and if it is greater than 100, it is corrected to 100. Psychological score = 100 - (2 × number of occurrences of words expressing negative emotions) + (10 × emotion score) - (5 × negative change from the previous day) Activity score = 50 + (0.5 × activity time [minutes]) + (2 × number of activations) - (0.3 × inactivity time [hours]) Health score = 70 - (3 × number of occurrences of health-related words) - (deductions based on sleep problems) The above coefficients are merely examples and may be changed or adjusted for each elderly person.

[0043] The linguistic feature calculation unit 2026 calculates linguistic features from dialogue data and uses them to calculate a psychological score or a health score. The linguistic features used include an index representing vocabulary diversity (type-token ratio), which is the number of different words in an utterance divided by the total number of words; the frequency of demonstrative pronouns; the frequency of self-negating words; semantic coherence, which represents the degree of logical relationship between sentences; and changes in the usage of dialects or idioms. For example, if the index representing vocabulary diversity decreases or the frequency of demonstrative pronouns exceeds a predetermined threshold, the psychological score or health score may be lowered as an indication of word retrieval impairment. Similarly, if the frequency of self-negating words increases, the psychological score or health score may be lowered as an indication of depressive tendencies.

[0044] The language feature calculation unit 2026 may set a vocabulary standard for each elderly person using dialogue data from a predetermined period, for example, two weeks, from the start of use. If an index representing vocabulary diversity falls below a predetermined percentage, for example, 20%, relative to the vocabulary standard, the psychological score or health score may be reduced. This allows for the absorption of individual differences such as educational level and enables evaluation based on the elderly person's own past condition.

[0045] The integration unit 2025 integrates the activity score, psychological score, and health score to calculate an overall status score. In one embodiment, the overall status score is calculated by multiplying the activity score, psychological score, and health score by weighting coefficients w1, w2, and w3 set for each elderly person, and then adding them together. Overall State Score = w1 × Activity Score + w2 × Psychological Score + w3 × Health Score The weight coefficients w1 to w3 may be optimized for each elderly person by the learning unit 204 described later. The analysis unit 202 may also calculate the day-to-day change rate, the weekly average change rate, and the abnormal trend flag as state changes. The overall state score is one form of the state score, and hereafter the overall state score will also be simply referred to as the state score.

[0046] The determination unit 203 compares the current state score with a reference value generated or updated by the learning unit 204 (described later) and calculates the difference as the change amount. The change amount may include the difference between the current state score and the reference value, and the rate of change between the current state score and the past average state score. Based on the change amount, the determination unit 203 detects a change in the elderly person's state and, depending on which of the multiple ranges set in stages the change amount belongs to, determines one notification level from among the multiple notification levels associated with each of those ranges.

[0047] For example, the determination unit 203 may determine that the state is normal if the amount of change is less than 10% of the reference value, and may not issue an immediate notification. Alternatively, as an example of determining the notification level, the unit may determine that the notification level is 1 (minor) if the amount of change is 10% or more but less than 20% of the reference value, notification level 2 (caution) if it is 20% or more but less than 40%, notification level 3 (warning) if it is 40% or more but less than 60%, and notification level 4 (urgent) if it is 60% or more.

[0048] Furthermore, the determination unit 203 may determine the notification level by considering not only the change in the status score, but also whether or not the dialogue data contains phrases indicating an anomaly, whether or not an SOS operation has been detected, or whether or not no response has been given for a long period of time. For example, if the change is between 40% and 60% of the baseline value, or if phrases indicating an anomaly are detected, notification level 3 may be determined, and if the change is 60% or more of the baseline value, if an SOS operation is detected, or if there has been no response for a long period of time, notification level 4 may be determined.

[0049] The notification unit 205 sends a notification to one or more family terminals 30 associated with the elderly person, according to the notification level determined by the determination unit 203. The notification unit 205 may change at least one of the number of recipients and the type of notification means, according to the notification level. The notification means can include messaging services, push notifications, email, and telephone. For example, the notification unit 205 may send a notification to a messaging service account associated with the family terminal 30, and a messaging service such as LINE (registered trademark) may be used as a specific example.

[0050] The notification content may include an overview of the elderly person's condition, details of any changes in their condition, suggested actions, and matters to be confirmed. For example, at notification level 1, only a periodic report may be included; at notification level 2, a notification including a mild warning may be sent; at notification level 3, an immediate notification may be sent; and at notification level 4, an emergency notification may be sent to multiple family terminals 30 or pre-registered emergency contacts.

[0051] Furthermore, the notification unit 205 may generate and send periodic reports to the family terminal 30. In one embodiment, the periodic reports are generated in different ways depending on the number of conversations and analysis results for the day. For example, a green report may be generated if the number of conversations for the day exceeds 50 or if the health status is determined to be good; a yellow report may be generated if the number of conversations for the day is between 10 and 50 or if no abnormalities are detected; and a red report may be generated if the number of conversations for the day is 10 or less or if abnormalities are detected.

[0052] The periodic report may not simply list the activity history in chronological order, but rather generate content that converts the analysis results obtained from dialogue data and operation logs into expressions and stages representing the physical and mental state of the elderly person. The content of notifications sent by the notification unit 205, the recipient, the time of transmission, and the confirmation status are recorded in the storage unit 206 as notification history.

[0053] The learning unit 204 continuously stores dialogue data, operation logs, behavioral history, and external environment data in the storage unit 206, and sequentially generates or updates standard values ​​for each elderly person based on this stored data. The initial value of the standard value may be set as the average value of the condition score calculated over a predetermined period from the start of use, for example, two weeks.

[0054] In one embodiment, the updated reference value is calculated as the sum of the value obtained by multiplying the reference value before the update by a first weight and the value obtained by multiplying the status score for the day by a second weight smaller than the first weight. For example, if the first weight is 0.9 and the second weight is 0.1, the updated reference value may be calculated by the following formula. Updated baseline value = 0.9 × previous baseline value + 0.1 × current day's status score This allows for the gradual reflection of recent changes in the condition of elderly individuals while maintaining long-term trends in their condition.

[0055] The learning unit 204 may exclude status scores from updating the reference value on days when the determination unit 203 detects a status change above a predetermined notification level, for example, notification level 3 or higher. This prevents the influence of temporary outliers or sudden events from being reflected in the reference value. In addition, when updating the reference value, the learning unit 204 may automatically adjust the tolerance range for the reference value by considering the day of the week, time of day, season, and external environmental factors.

[0056] Furthermore, the learning unit 204 may update the personalized model, including the weight coefficients w1 to w3, using the accumulated data. For example, the learning unit 204 may acquire whether or not the family has confirmed or responded to the notification as feedback information, and update the weight coefficients w1 to w3 at a predetermined learning rate in a direction that reduces false notifications. This makes it possible to achieve state determination that is adapted to the lifestyle habits and behavioral characteristics of each elderly person.

[0057] In this embodiment, each process may be executed according to a predetermined schedule. For example, the reception of dialogue data and operation logs by the data acquisition unit 201 may be performed as needed, the calculation of the status score by the analysis unit 202 and the notification by the notification unit 205 may be performed at a predetermined time each day (e.g., 8pm), and the updating of the reference value by the learning unit 204 may be performed at a predetermined time each day (e.g., 11pm). In addition, the memory unit 206 may store profile information, reference values, daily status scores, dialogue history, behavior history, and notification history associated with each elderly person.

[0058] Next, the dialogue processing and dialogue data processing according to this embodiment will be described. Figure 5 is a flowchart showing the flow of dialogue processing and dialogue data processing according to this embodiment. First, the cloud server 20 obtains the current time (step S101). The current time may be obtained based on time information held by the cloud server 20, or it may be obtained based on time information transmitted from the elderly terminal 10.

[0059] Next, the cloud server 20 refers to the dialogue history stored in the memory unit 206 and determines whether or not it is the first dialogue of the day (step S102). For example, if there is no dialogue history for the day, the cloud server 20 determines that it is the first dialogue of the day.

[0060] If it is determined that this is the first interaction of the day (Step S102: Yes), the cloud server 20 determines the time of day based on the current time (Step S103). The time of day is classified into, for example, morning, noon, evening, and night.

[0061] Next, the cloud server 20 uses a dialogue agent to generate dialogue content corresponding to the determined time period and sends it to the elderly terminal 10 (step S104). For example, in the morning, it generates dialogue content including morning greetings, confirmation of sleep status, confirmation of breakfast, presentation of the day's weather forecast, and confirmation or prompting input of the day's itinerary. In the midday period, it generates dialogue content including midday greetings, confirmation of lunch, confirmation of exercise, presentation of the weather forecast, and confirmation of the itinerary. Furthermore, in the evening, it generates dialogue content including evening greetings, confirmation of dinner, presentation of the weather forecast, and confirmation of the itinerary, and in the nighttime, it generates dialogue content including evening greetings, a review of the day, confirmation of medication, confirmation of the next day's itinerary, and prompting to go to bed.

[0062] On the other hand, if it is determined that this is not the first conversation of the day (Step S102: No), the cloud server 20 uses the conversation agent to perform normal conversation processing (Step S105). In normal conversation processing, responses are given in accordance with the input of the elderly person, small talk is conducted, health checks are made, living conditions are checked, and other conversations take place.

[0063] Subsequently, the data acquisition unit 201 of the cloud server 20 acquires dialogue data and operation logs from the elderly terminal 10 (step S106). The dialogue data includes voice data or text data, and the operation logs include application launch history, number of operations, usage time, response time, and inactivity time.

[0064] Next, the dialogue data processing unit 2021 performs preprocessing on the acquired dialogue data (step S107). Preprocessing includes conversion to text data by speech recognition, noise reduction, correction of typographical errors and omissions, and standardization of inconsistencies in notation.

[0065] Next, the dialogue data processing unit 2021 performs natural language analysis on the preprocessed dialogue data (step S108). Natural language analysis includes morphological analysis, syntactic analysis, sentiment analysis, and intent analysis.

[0066] Next, the dialogue data processing unit 2021 extracts entities such as health-related information, lifestyle information, medication / hospital visit information, and psychological / emotional information based on the results of natural language processing, and generates structured data containing said entities (step S109).

[0067] Subsequently, the cloud server 20 stores the generated structured data in the storage unit 206 and links to the state analysis processing by the analysis unit 202 (step S110). This makes it possible to use the content of the elderly person's daily conversations and terminal usage status to calculate a state score representing the elderly person's condition and to determine changes in that condition.

[0068] Through the above process, the cloud server 20 can naturally collect information about elderly people within the flow of everyday conversations, and structure and store that information. Therefore, it becomes possible to continuously monitor the condition of elderly people without imposing any special operational burden on them.

[0069] Next, the state analysis process according to this embodiment will be described. Figure 6 is a flowchart showing the flow of the state analysis process according to this embodiment.

[0070] First, the analysis unit 202 extracts features based on the dialogue data and operation logs acquired from the elderly terminal 10 (step S201). The extracted features include the number of occurrences of words expressing negative emotions, the number of occurrences of words related to health, the emotion score, the number of dialogues on the day, the average response time, the activity time, the number of times the application was launched, the inactivity time, and an index representing the regularity of the daily rhythm.

[0071] Next, the activity analysis unit 2022 calculates an activity score based on the extracted features and operation logs (step S202). For example, the activity analysis unit 2022 calculates the activity score using the elderly person's activity time, the number of times the application was launched, and the inactivity time. Alternatively, the activity analysis unit 2022 may estimate the elderly person's daily rhythm from the operation logs and use the estimation results to calculate the activity score.

[0072] Next, the Psychological Analysis Unit 2023 calculates a psychological score based on the results of natural language analysis and the extracted features (step S203). For example, the Psychological Analysis Unit 2023 calculates a psychological score using the number of occurrences of words representing negative emotions, the emotion score, and the change from the previous day.

[0073] Next, the health analysis unit 2024 calculates a health score based on the extracted features (step S204). For example, the health analysis unit 2024 calculates a health score using the number of occurrences of health-related words, sleep status, medication status, and spoken content related to physical condition.

[0074] Next, the language feature calculation unit 2026 calculates language features from the dialogue data and reflects the calculated language features in the psychological score or health score (step S205). For example, indicators representing vocabulary diversity, the frequency of demonstrative pronouns, the frequency of self-negating words, semantic consistency, and changes in the usage patterns of dialects or idioms may be calculated, and the psychological score or health score may be adjusted according to the results.

[0075] Next, the integration unit 2025 integrates the activity score, psychological score, and health score to calculate an overall status score (step S206). In one embodiment, the integration unit 2025 calculates the overall status score by multiplying each score by weighting coefficients w1, w2, and w3 set for each elderly person and adding them together.

[0076] Next, the analysis unit 202 calculates the state change based on the calculated overall state score (step S207). The state change includes the day-to-day change rate, the weekly average change rate, and an abnormal trend flag. The analysis unit 202 outputs this information to the determination unit 203 for use in the subsequent notification level determination process.

[0077] Through the above processing, the analysis unit 202 can comprehensively evaluate the activity status, psychological state, and health status from the dialogue data and operation logs, and calculate an overall status score that represents the elderly person's condition. This makes it possible to generate basic information for early detection of changes in the elderly person's condition.

[0078] Next, the state change determination process and notification process according to this embodiment will be described. Figure 7 is a flowchart showing the flow of the state change determination process and notification process according to this embodiment.

[0079] First, the determination unit 203 compares the current day's status score with a reference value generated or updated by the learning unit 204 at a predetermined time, for example, 8pm every day, and calculates the amount of change based on the difference (step S301). The amount of change may include the difference between the current day's status score and the reference value, the rate of change between the current day's status score and the past average status score, etc.

[0080] Next, the determination unit 203 determines whether the calculated change is 10% or more of the reference value (step S302).

[0081] If the change is determined to be less than 10% of the standard value (Step S302: No), the determination unit 203 determines that the elderly person's condition is normal and does not perform immediate notification (Step S303). In this case, the process may be terminated, or the process may proceed to reflect the day's analysis results in the periodic report.

[0082] On the other hand, if it is determined that the amount of change is 10% or more of the reference value (step S302: Yes), the determination unit 203 determines the range to which the amount of change belongs (step S304).

[0083] Next, the determination unit 203 determines the notification level according to the determined range (step S305). For example, if the amount of change is 10% or more but less than 20% of the reference value, notification level 1 may be determined; if it is 20% or more but less than 40%, notification level 2 may be determined; if it is 40% or more but less than 60%, notification level 3 may be determined; and if it is 60% or more, notification level 4 may be determined.

[0084] Next, the notification unit 205 determines the recipients and notification methods based on the determined notification level (step S306). For example, the number of recipients may be increased as the notification level increases, or a more urgent notification method may be selected from messaging services, push notifications, email, and telephone calls.

[0085] Next, the notification unit 205 sends a notification to the determined recipient (step S307). For example, the notification unit 205 may send a notification to the messaging service account associated with the family terminal 30, and specifically, it may send the notification using LINE® or the like. The notification content may include an overview of the elderly person's condition, details of any changes in their condition, recommended actions, and points to confirm.

[0086] Subsequently, the notification unit 205 records the content of the transmitted notification, the recipient, the transmission time, and the confirmation status, etc., as a notification history in the storage unit 206 (step S308).

[0087] Through the above process, the determination unit 203 can quantitatively determine the change in the elderly person's condition relative to the standard value set for each elderly person, and determine a notification level according to the degree of change. The notification unit 205 can then select an appropriate notification destination and notification method according to the notification level and send the notification. As a result, it becomes possible to quickly and appropriately notify family members of changes in the elderly person's condition.

[0088] Next, the reference value update process and the personalized model update process according to this embodiment will be described. Figure 8 is a flowchart showing the flow of the reference value update process and the personalized model update process according to this embodiment.

[0089] First, the learning unit 204 acquires the status score for the day at a predetermined time, for example, at 11:00 PM every day (step S401). The status score for the day may be the overall status score calculated by the analysis unit 202.

[0090] Next, the learning unit 204 determines whether the current day is an abnormal day with a predetermined notification level or higher (step S402). For example, if the determination unit 203 determines that the notification level is 3 or higher, the current day may be determined to be an abnormal day.

[0091] If the day is determined to be an abnormal day (Step S402: Yes), the learning unit 204 excludes the status score for that day from updating the reference value (Step S403). This prevents the influence of temporary abnormal values ​​or sudden events from being reflected in the reference value.

[0092] On the other hand, if it is determined that the day is not an abnormal day (Step S402: No), the learning unit 204 updates the reference value based on the reference value before the update and the status score for the day (Step S404). For example, the learning unit 204 may calculate the updated reference value according to the following formula. Updated baseline value = 0.9 × previous baseline value + 0.1 × current day's status score

[0093] Next, the learning unit 204 adjusts the tolerance range for the reference value based on the day of the week, time of day, season, and external environmental factors (step S405). For example, the tolerance range may be expanded during seasons when activity levels tend to decrease or during bad weather, and narrowed during normal times.

[0094] Next, the learning unit 204 updates the weight coefficients w1 to w3 based on the accumulated dialogue data, operation logs, notification history, and feedback information (step S406). For example, the presence or absence of confirmation or response to notifications by family members may be used as learning data, and the weight coefficients w1 to w3 may be adjusted in a direction that reduces false notifications.

[0095] Through the above processing, the learning unit 204 can continuously update reference values ​​that reflect the trends in the condition changes of each elderly person, and can also construct a personalized model adapted to the lifestyle and behavioral characteristics of each elderly person. As a result, it becomes possible to improve the accuracy of condition change judgment and to realize appropriate monitoring support while suppressing unnecessary notifications.

[0096] Next, the disaster monitoring process according to this embodiment will be described. Figure 9 is a flowchart showing the flow of the disaster monitoring process according to this embodiment.

[0097] First, the cloud server 20 acquires the location information of the elderly terminal 10 obtained by the location information acquisition unit 104 (step S501). The location information may be location information based on GPS information, mobile phone base station information, or wireless LAN access point information, etc.

[0098] Next, the cloud server 20 acquires disaster information from the external device 40 (step S502). The disaster information may include earthquake information, tsunami information, typhoon information, heavy rain information, flood information, landslide information, and other disaster prevention information.

[0099] In one embodiment, the cloud server 20 can determine disaster risk based on the location information of the elderly person's terminal 10 and disaster information obtained from an external device 40. For example, if an earthquake or tsunami of magnitude 3 or higher occurs at the elderly person's location, or if a warning for a storm, typhoon, or blizzard is issued on that day, the cloud server 20 may determine that a disaster risk exists. Information regarding earthquakes may be obtained in real time, while information regarding other natural disasters may be obtained within a predetermined time (e.g., 3 hours).

[0100] Next, the cloud server 20 determines whether or not there is a disaster risk at the location of the elderly terminal 10 based on the acquired location information and disaster information (step S503). For example, this determination may be made by comparing the location information of the elderly terminal 10 with the warning zone, evacuation order zone, or evacuation advisory zone included in the disaster information.

[0101] If it is determined that there is no disaster risk at the location (Step S503: No), the cloud server 20 terminates processing.

[0102] On the other hand, if it is determined that there is a disaster risk at the location (Step S503: Yes), the cloud server 20 sends a disaster notification to the elderly terminal 10 (Step S504). The disaster notification may include information such as the type of disaster, the expected level of danger, evacuation information, a request for safety confirmation, and information about evacuation locations.

[0103] Next, the cloud server 20 sends a disaster notification to the family terminal 30 (step S505). The notification may include the type of disaster, the location of the elderly person's terminal 10, the status of the elderly person's safety confirmation, and the results of the safety confirmation.

[0104] Next, the cloud server 20 records the content of the sent notification, the recipient, the time of transmission, and the confirmation status as a notification history in the storage unit 206 (step S506).

[0105] Through the above process, the cloud server 20 can determine the disaster risk by combining the location information of the elderly terminal 10 with disaster information obtained from the external device 40, and send appropriate disaster notifications to the elderly and their families. As a result, it becomes possible to quickly confirm the safety of the elderly and provide evacuation support in the event of a disaster.

[0106] Next, the emergency notification process and peace notification process according to this embodiment will be described. Figure 10 is a flowchart showing the flow of the emergency notification process and peace notification process according to this embodiment.

[0107] First, the cloud server 20 determines whether the SOS button has been pressed for a predetermined time or longer, based on the operation information transmitted from the operation reception unit 103 of the elderly terminal 10 (step S601). In one embodiment, the predetermined time may be 30 seconds.

[0108] If it is determined that the SOS button has been pressed for 30 seconds or more (Step S601: Yes), the cloud server 20 generates an emergency notification and sends it to the family terminal 30 (Step S602). The emergency notification may include the location information of the elderly terminal 10, the notification time, and information regarding emergency contact. In addition, the location information may be obtained by the location information acquisition unit 104.

[0109] On the other hand, if it is determined that the SOS button has not been pressed for more than 30 seconds (step S601: No), the cloud server 20 determines whether or not the Ping An button has been operated (step S603).

[0110] If it is determined that the safety button has been operated (step S603: Yes), the cloud server 20 generates a safety notification and sends it to the family terminal 30 (step S604). The safety notification may include the notification time, information indicating the safety of the elderly person, and a brief message if necessary.

[0111] On the other hand, if it is determined that the Ping An button has not been operated (step S603: No), the cloud server 20 does not perform the notification process and terminates the process.

[0112] Through the above process, the cloud server 20 can send emergency notifications in response to the elderly person pressing the SOS button, and also send safety notifications in response to the safety button being pressed. As a result, the elderly person can quickly request assistance in emergencies and easily notify their family of their safety or well-being on a daily basis.

[0113] In this embodiment, at least some of the functions of the cloud server 20 may be executed on the elderly terminal 10. Furthermore, the calculation formulas for each score, the thresholds, and the weight coefficients are illustrative examples and may be modified as appropriate without departing from the spirit of the present invention. Additionally, the dialogue agent may be implemented using a rule-based method, a machine learning model method, or a combination thereof.

[0114] The present invention may be implemented as the monitoring support system 1 described above, as a monitoring support method, or as a program for causing a computer to execute the method, or as a non-temporary computer-readable recording medium on which the program is recorded.

[0115] The technologies disclosed herein are not limited to the embodiments, examples, and modifications described above. The technologies disclosed herein can be implemented in various configurations without departing from the spirit of the invention. The technical features of the embodiments, examples, and modifications described above that correspond to the technical features of each form described in the summary of the invention may be substituted and combined in order to solve some or all of the above-described problems or to achieve some or all of the above-described effects. Furthermore, technical features not described as essential in this specification may be deleted as necessary. [Explanation of Symbols]

[0116] 1…Monitoring and support system 10…Elderly terminals 101…Interactive Interface 102... Operation log recording unit 103...Operation reception desk 104...Location information acquisition unit 20…Cloud Server 21…CPU 22…Storage device 23…Communication I / F 24…Input devices 25…Output device 201...Data Acquisition Unit 202…Analysis Department 2021…Dialogue Data Processing Unit 2022…Activity Analysis Department 2023…Psychological Analysis Department 2024…Health Analysis Department 2025…Integration Department 2026…Language Feature Calculation Unit 203...Judgment section 204…Learning Department 205…Notification department 206...Storage section 30...Family terminal 40...External device 50…Operational Management Platform NW...Communication Network

Claims

1. A data acquisition unit acquires dialogue data from an elderly person's terminal used by an elderly person, which includes dialogue data showing the content of a conversation between the elderly person and a dialogue agent, the content of the conversation input by the elderly person in the conversation via voice or text, and the content of the dialogue agent's response to the elderly person, as well as an operation log from the elderly person's terminal. An analysis unit that calculates a state score representing the state of the elderly person based on the dialogue data and the operation log, the analysis unit that estimates the emotional state of the elderly person by natural language processing of the dialogue data, uses the estimated emotional state in calculating the state score, and estimates the elderly person's daily rhythm based on the operation log, and uses the estimated daily rhythm in calculating the state score, A determination unit detects a change in the elderly person's condition based on the difference between the condition score calculated by the analysis unit and a reference value based on condition scores previously calculated for the elderly person, and determines a notification level corresponding to the change in condition. A notification unit sends a notification with content corresponding to the notification level determined by the determination unit to a family terminal associated with the elderly person, A monitoring and support system equipped with these features.

2. In the monitoring support system described in claim 1, The analysis unit calculates an activity score representing the elderly person's activity level, a psychological score representing their psychological state, and a health score representing their health state. It then integrates the activity score, psychological score, and health score to calculate an overall status score, and uses this overall status score as the status score. A monitoring and support system.

3. In the monitoring support system described in claim 2, The analysis unit calculates the overall status score by multiplying each of the activity score, the psychological score, and the health score by a weighting coefficient set for each elderly person and adding them together. A monitoring and support system.

4. In the monitoring support system described in claim 3, The learning unit further comprises a unit that calculates the reference value based on the history of the condition score previously calculated for the elderly person, and updates the reference value sequentially. A monitoring and support system.

5. In the monitoring support system described in claim 4, The learning unit excludes the state score for days on which the determination unit has detected a state change of a predetermined notification level or higher from updating the reference value. A monitoring and support system.

6. In the monitoring support system described in claim 5, The learning unit calculates the updated reference value as the sum of the value obtained by multiplying the reference value before the update by a first weight and the value obtained by multiplying the state score for the day by a second weight smaller than the first weight. A monitoring and support system.

7. In the monitoring support system described in claim 6, The determination unit determines one of the notification levels associated with each of the multiple ranges, depending on which of the multiple ranges set in stages the difference belongs to. A monitoring and support system.

8. In the monitoring support system described in claim 7, The notification unit transmits the notification, varying at least one of the number of recipients and the type of notification means according to the notification level determined by the determination unit. A monitoring and support system.

9. In the monitoring support system according to claim 8, The notification unit sends the notification to the messaging service account associated with the family terminal. A monitoring and support system.

10. In the monitoring support system described in claim 2, The analysis unit calculates at least one of the following linguistic features from the dialogue data: an index representing vocabulary diversity in the elderly person's speech, the frequency of demonstrative pronouns, and the frequency of self-negating words. This linguistic feature is then used to calculate at least one of the psychological score and the health score. A monitoring and support system.

11. In the monitoring support system according to claim 10, When the analysis unit calculates an index representing vocabulary diversity as a linguistic feature, it sets a vocabulary standard based on the index representing vocabulary diversity previously calculated for the elderly person, and when the index representing vocabulary diversity falls by more than a predetermined percentage relative to the vocabulary standard, it lowers at least one of the psychological score and the health score. A monitoring and support system.

12. In the monitoring support system according to any one of claims 1 to 11, The data acquisition unit further acquires the location information of the elderly terminal and disaster information acquired from an external device. The determination unit determines whether there is a disaster risk at the location of the elderly person based on the location information and the disaster information, When the notification unit determines that there is a risk of disaster, it sends a disaster-related notification to the elderly terminal and the family terminal. A monitoring and support system.

13. In the monitoring support system according to any one of claims 1 to 11, The notification unit, when it detects a predetermined operation on the elderly terminal, sends an emergency notification to the family terminal, including the location information of the elderly terminal. A monitoring and support system.

14. Computers An acquisition step of acquiring dialogue data from an elderly person's terminal used by an elderly person, which includes dialogue data showing the content of a conversation between the elderly person and a conversational agent, the content of the conversation input by the elderly person in the conversation via voice or text, and the content of the conversational agent's response to the elderly person, and an operation log from the elderly person's terminal. A calculation step for calculating a state score representing the state of the elderly person based on the dialogue data and the operation log, comprising: estimating the emotional state of the elderly person by natural language processing of the dialogue data, using the estimated emotional state in calculating the state score, and estimating the elderly person's daily rhythm based on the operation log, and using the estimated daily rhythm in calculating the state score; A determination step which detects a change in the elderly person's condition based on the difference between the condition score calculated in the calculation step and a reference value based on condition scores previously calculated for the elderly person, and determines a notification level corresponding to the change in condition, A notification step which sends a notification with content corresponding to the notification level determined in the determination step to the family terminal associated with the elderly person, A monitoring and support method that implements this.

15. A program for causing a computer to execute the monitoring support method described in claim 14.

Images