Apparatus and method for monitoring resident safety

The integration of quantum deep learning for resident safety monitoring improves detection accuracy by analyzing lifestyle and environmental data, addressing sensor malfunctions, and enabling rapid response to potential dangers.

KR103046204B1Active Publication Date: 2026-11-10S & D
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
KR1020260030987
Authority / Receiving Office
KR · KR
Patent Type
Patents
Current Assignee / Owner
Filing Date
2026-02-19
Publication Date
2026-11-10
Estimated Expiration
2046-02-19

AI Technical Summary

Technical Problem

Conventional resident safety monitoring technologies struggle to accurately distinguish between normal daily changes and dangerous situations due to reliance on simple threshold judgments or rule-based analysis, failing to consider individual lifestyle habits and environmental differences, leading to false positives and undetected results, especially in cases of decreased consciousness or sudden health issues.

Method used

A resident safety monitoring device and method that integrates quantum deep learning to analyze lifestyle, climate, and personal activity data, using reference patterns defined by resident and environmental information to detect anomalies, and includes a data verification model to handle sensor malfunctions, ensuring accurate detection and personalized monitoring.

Benefits of technology

Enhances detection accuracy by understanding resident living patterns in a three-dimensional manner, reduces false alarms from sensor defects, and supports rapid response to potential dangers through personalized monitoring and timely notifications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 112026020371241-PAT00001_ABST
    Figure 112026020371241-PAT00001_ABST
Patent Text Reader

Abstract

The embodiments disclosed in this specification relate to a resident safety monitoring device and method using artificial intelligence. According to one embodiment, the resident safety monitoring device includes a program for performing resident safety monitoring, a memory in which reference patterns for environmental conditions defined by combining resident information including the age and health status of a resident and external environment information are stored, and one or more processors, and operates by executing the program stored in the memory; collects resident-specific data including household energy consumption data, external environment data, and personal activity data; inputs the collected resident-specific data into an analysis model learned based on the reference patterns to obtain a predicted behavior pattern; and performs rule-based anomaly detection based on the actual behavior pattern obtained by processing the collected resident-specific data and the predicted behavior pattern to determine an anomaly stage.
Need to check novelty before this filing date? Find Prior Art

Description

Technology Field

[0001] The embodiments disclosed in this specification relate to an artificial intelligence-based resident safety monitoring device and method, and more specifically, to a device and method for monitoring resident safety by analyzing living data based on quantum artificial intelligence. Background Technology

[0002] With the aging of the population and the increase in single-person households, so-called "lonely deaths"—deaths occurring after living alone for extended periods without contact with the outside world—are emerging as a social issue. In particular, for the elderly, those with chronic diseases, and socially vulnerable groups, it is difficult to immediately recognize or report to the outside world even if abnormalities in daily life or health deterioration occur, which limits the appropriate response when an accident occurs.

[0003] In this regard, conventional methods have involved making periodic calls to check on residents or conducting on-site inspections, or using simple manual safety devices such as emergency call buttons. However, since these methods rely on the voluntary response or operation of residents, their practical preventive effect is limited in situations such as decreased consciousness, sudden health issues, or prolonged inactivity.

[0004] Recently, attempts have been made to estimate the condition of residents by collecting various lifestyle data to compensate for these limitations. In this regard, Korean Patent Publication No. 10-2021-0123046 discloses an elderly care device using a thermal imaging algorithm that can manage the condition of elderly people living alone in advance using thermal imaging sensors, etc.

[0005] However, conventional technologies often rely primarily on simple threshold judgments or rule-based analysis of individual sensor data, which has the problem of failing to adequately reflect individual lifestyle habits or environmental differences. Consequently, it is difficult to accurately distinguish between normal daily changes and dangerous situations, and there is a possibility of false positives or undetected results. Furthermore, limitations have been pointed out regarding the failure to consider multiple heterogeneous data integrally or the lack of real-time updates to status assessment results.

[0006] Therefore, there is an increasing need for resident safety monitoring technology that can comprehensively analyze various lifestyle, environmental, and personal data to more precisely assess the safety status of residents and detect and respond to abnormal signs at an early stage.

[0007] Meanwhile, the aforementioned background technology is technical information that the inventor possessed for the derivation of the present invention or acquired during the process of deriving the present invention, and it cannot be considered as prior art disclosed to the general public prior to the filing of the present invention. The problem to be solved

[0008] The embodiments disclosed in this specification aim to provide a resident safety monitoring device and method that collects lifestyle data, climate data, and personal activity data, and detects abnormal signs of resident safety based on the collected data.

[0009] The embodiments disclosed in this specification are intended to present a resident safety monitoring device and method that detect abnormal signs of resident safety using quantum deep learning.

[0010] The embodiments disclosed in this specification aim to provide a resident safety monitoring device and method capable of identifying correlations between living data, detecting sensor malfunctions when some of the living data are missing or delayed, and reducing the impact on the sensor data when a sensor malfunction is detected. means of solving the problem

[0011] According to one embodiment, as a technical means for achieving the aforementioned technical task, a resident safety monitoring device is disclosed that includes a memory in which reference patterns for environmental conditions defined by combining resident information including the age and health status of a resident and external environment information are stored, and one or more processors, and operates by executing a program stored in the memory; collects resident-specific data including energy consumption data by household, external environment data, and personal activity data; inputs the collected resident-specific data into an analysis model learned based on the reference patterns to obtain a predicted behavior pattern; and performs rule-based anomaly detection based on the actual behavior pattern obtained by processing the collected resident-specific data and the predicted behavior pattern to determine an anomaly stage.

[0012] According to another embodiment, a resident safety monitoring method is disclosed, comprising the steps of: collecting resident-specific data including household energy consumption data, external environment data, and personal activity data, which is performed by a resident safety monitoring device; obtaining a predicted behavior pattern by inputting the collected resident-specific data into an analysis model learned based on a reference pattern for environmental conditions defined by combining resident information including the age and health status of the resident and external environment information; and determining an abnormal sign stage by performing rule-based abnormal sign detection based on the actual behavior pattern obtained by processing the predicted behavior pattern and the collected resident-specific data.

[0013] According to another embodiment, a recording medium is disclosed in which a program for performing a resident safety monitoring method is stored and which is performed by a resident safety monitoring device. The resident safety monitoring method includes the steps of: collecting resident-specific data including household energy consumption data, external environment data, and personal activity data; obtaining a predicted behavior pattern by inputting the collected resident-specific data into an analysis model learned based on a reference pattern for environmental conditions defined by combining resident information including the age and health status of the resident and external environment information; and determining an abnormal sign stage by performing rule-based anomaly detection based on the actual behavior pattern obtained by processing the predicted behavior pattern and the collected resident-specific data.

[0014] According to another embodiment, a computer program stored on a computer-readable recording medium is disclosed to perform the resident safety monitoring method, which is performed by a resident safety monitoring device. The resident safety monitoring method comprises the steps of: collecting resident-specific data including household energy consumption data, external environment data, and personal activity data; obtaining a predicted behavior pattern by inputting the collected resident-specific data into an analysis model learned based on a reference pattern for environmental conditions defined by combining resident information including the age and health status of the resident and external environment information; and determining an abnormality stage by performing rule-based anomaly detection based on the actual behavior pattern obtained by processing the predicted behavior pattern and the collected resident-specific data. Effects of the invention

[0015] According to any one of the aforementioned means for solving the problem, beyond simply verifying energy consumption, by integrating and analyzing in-house sensor data and activity data from external agencies in a time-series manner, it is possible to understand the actual living patterns of residents in a more three-dimensional way and obtain precise predictive behavioral patterns that reflect the residents' lifestyle habits, thereby improving detection accuracy.

[0016] In addition, according to any one of the aforementioned means for solving the problem, even if data fixation or omission occurs where the data value of specific sensor data remains fixed at a specific value without significant fluctuation over time, false alarms caused by sensor defects can be effectively blocked by cross-verifying the actual activity of the resident through data from other sensors that have a high dependency on the sensor where the data fixation or omission was detected.

[0017] In addition, according to any one of the aforementioned means for solving the problem, by introducing a quantum deep learning model, non-linear and complex correlations between various sensors that are difficult to capture with existing classical algorithms can be processed quickly and accurately. In particular, by identifying subtle anomaly patterns hidden within large-scale time-series data using the principles of quantum superposition and entanglement, the performance of anomaly detection can be improved.

[0018] In addition, according to any one of the aforementioned means for solving the problem, the condition of a resident can be managed in more detail through a logic that systematically transitions from the interest stage to the danger stage of the four abnormal sign stages by combining rules of numerical deviation and duration with the inference results of an AI-based analysis model, and by immediately sending notifications to guardians and control centers when the danger stage is confirmed, it is possible to support rapid rescue activities within the golden time.

[0019] In addition, according to any one of the aforementioned means of solving the problem, by dynamically determining a standard pattern for each resident based on the similarity of environmental conditions, it is possible to provide a personalized monitoring service that comprehensively considers the resident's age, health status, and seasonal factors. This can prevent malfunctions caused by the application of uniform standards and have the effect of establishing a safety net that is universally applicable to households with diverse living environments and individual lifestyle characteristics.

[0020] The effects obtainable from the disclosed embodiments are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art to which the disclosed embodiments belong from the description below. Brief explanation of the drawing

[0021] FIG. 1 is a configuration diagram illustrating a safety monitoring system according to one embodiment. FIG. 2 is a block diagram illustrating the configuration of a resident safety monitoring device according to one embodiment. Figures 3 and 4 are drawings illustrating the resident safety monitoring process. FIG. 5 is a flowchart illustrating a resident safety monitoring method according to one embodiment. Specific details for implementing the invention

[0022] Various embodiments are described in detail below with reference to the attached drawings. The embodiments described below may be implemented in various different forms. In order to explain the features of the embodiments more clearly, detailed descriptions of matters widely known to those skilled in the art to which the following embodiments belong have been omitted. Additionally, parts of the drawings unrelated to the description of the embodiments have been omitted, and similar parts throughout the specification have been given similar reference numerals.

[0023] Throughout the specification, when a configuration is described as being "connected" to another configuration, this includes not only cases where they are "directly connected," but also cases where they are "connected with another configuration in between." Furthermore, when a configuration is described as "including" another configuration, this means that, unless specifically stated otherwise, it does not exclude other configurations but may include additional configurations.

[0024] Additionally, terms including ordinal numbers such as 'first' or 'second' used herein may be used to describe various components, but said components should not be limited by said terms. Such terms are used solely for the purpose of distinguishing one component from other components.

[0025] First, we will define the terms used in this specification.

[0026] 'Behavioral patterns' refer to quantitative data clusters generated by combining household energy consumption data and individual activity data collected over a specified period in a time-series manner. Behavioral patterns may take the form of multidimensional vector data in which feature quantities corresponding to multiple data items for identifying the activity status of residents are arranged along a time axis.

[0027] 'Predicted behavior patterns' refer to standardized time-series data output from an analysis model, which can mean behavior patterns inferred for specific environmental conditions based on the analysis model's learning history.

[0028] A 'reference pattern' refers to a behavioral pattern generated based on clustering resident-specific data according to the similarity of environmental conditions, representing the standard behavioral pattern of a resident under specific environmental conditions. The reference pattern can be utilized as a reference dataset defining a normal range to determine whether the aforementioned predicted behavioral pattern and the actual behavioral pattern described later are abnormal.

[0029] 'Actual behavior patterns' refer to resident-specific data collected in real-time that has been processed into time-series data, and may mean behavior patterns having the same data dimensions and feature configurations to enable comparison with the aforementioned predicted behavior patterns.

[0030] "Resident" refers to a person who is subject to safety monitoring and performs daily activities while residing in or staying for a certain period in a household, which is a residential space. In this specification, the description is based on the assumption that one resident resides and lives in one household, and unless otherwise noted, data collected per household is considered to be data collected for one resident.

[0031] The embodiments will be described in detail below with reference to the attached drawings.

[0032] FIG. 1 is a configuration diagram illustrating a resident safety monitoring system according to one embodiment, and FIG. 2 is a block diagram illustrating a resident safety monitoring device according to one embodiment. The resident safety monitoring system (1) may include a resident safety monitoring device (100), an external server (30), a metering device connected to a household (20) that is a residence, a sensor provided within the household (20) to collect the resident's activities, and a guardian's terminal (40) or control center (not shown) to receive notifications in case of abnormal signs. In this case, the sensor may be a non-contact sensor.

[0033] A resident safety monitoring device (100, hereinafter referred to as the "monitoring device") according to one embodiment collects data for each resident or household and monitors whether a problem regarding resident safety has occurred based on the collected data. Although the monitoring device (100) in FIG. 1 is illustrated as being implemented as a server, it is not limited thereto and may be implemented as an electronic terminal or as a server-client system.

[0034] The electronic terminal can be implemented as a computer, portable terminal, or wearable device capable of connecting to a remote server via a network or connecting to other electronic terminals and servers. Here, the computer includes, for example, a notebook, desktop, or laptop equipped with a web browser, and the portable terminal can include, for example, any type of handheld-based wireless communication device that ensures portability and mobility, such as PCS (Personal Communication System), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), GSM (Global System for Mobile communications), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet), smartphone, and Mobile WiMAX (Mobile Worldwide Interoperability for Microwave Access). In addition, wearable devices are types of information processing devices that can be worn directly on the human body, such as watches, glasses, accessories, clothing, and shoes, and can connect to a server at a remote location or to another terminal via a network, either directly or through another information processing device.

[0035] And the server can be implemented as a computing device capable of communicating with electronic terminals and networks, or as a cloud computing server.

[0036] The monitoring device (100) can detect abnormal signs using an analysis model that learns the behavioral patterns of residents based on collected data for each resident. In addition, by performing real-time warning notifications and response procedures according to the results of abnormal sign detection, it can support early response to resident safety risks and contribute to ensuring resident safety.

[0037] The monitoring device (100) can collect data about the resident (i.e., data per resident) through a metering device and sensor connected to the household (20) that is the residence, and an external server (30). The monitoring device (100) can collect energy consumption data, such as electricity, gas, water, and heating usage data, through the metering device connected to the household (20) that is the residence. In addition, the monitoring device (100) can collect internal environmental data, such as the illuminance, temperature, and humidity of the household (20), as well as door entry / exit information and electronic product usage information included in personal activity data, through a non-contact IoT sensor installed within the household (20).

[0038] The external server (30) is a server where a public platform for providing environmental data such as weather or an operating system of a medical institution such as a hospital or pharmacy operates, and the monitoring device (100) can collect external environmental data such as weather, temperature, and season, personal activity data including records of visits to medical institutions such as pharmacies and hospitals, and social worker visit history from the external server (30). In addition, if the resident lives in a multi-unit dwelling, the external server (30) may be a server where a multi-unit dwelling management system operates, and the monitoring device (100) can collect door entry / exit information, parking system usage information, and weekly or monthly management fee payment history from the external server (30). Furthermore, the monitoring device (100) can collect personal information of the resident including age, gender, family information, etc. Data containing sensitive information, such as records of visits to medical institutions, can be collected by pseudonymizing or encrypting.

[0039] The monitoring device (100) can detect abnormal signs by using an artificial intelligence-based analysis model that learns the behavioral patterns of residents based on collected resident-specific data. Specifically, the control unit (120) can cluster energy consumption data by household, environmental data inside and outside the household, and personal activity data collected through the communication unit (130) according to environmental conditions.

[0040] Here, environmental conditions may include external environmental information such as temperature, weather, and season, as well as time and day information, and multidimensional parameters including the age and health status of the resident. The control unit (120) can define multiple reference patterns by grouping behavioral patterns within similar contexts according to these environmental conditions.

[0041] Furthermore, the monitoring device (100) can determine the abnormal sign stage as one of four grades—normal, interest, caution, or danger—by analyzing the deviation between the data prediction value calculated through an artificial intelligence-based analysis model and the real-time collected data using a rule-based algorithm. Additionally, the monitoring device (100) can ensure the reliability of detection by detecting sensor malfunctions in real-time during the data preprocessing stage, excluding error data, or interpolating data values ​​determined to be error data based on a reference pattern, and can provide an optimized notification to the guardian's terminal (40) or control center when an abnormal sign is detected.

[0042] The monitoring device (100) can input the resident's energy consumption data and personal activity data, as well as the resident's medical history, into an analysis model so that the artificial intelligence model can learn the resident's standard pattern reflecting the resident's health condition, and then use the learned artificial intelligence model to detect abnormal signs.

[0043] Meanwhile, when the monitoring device (100) is implemented as a single electronic terminal, the monitoring device (100) may include a mmWave radar and a lightweight analysis model, and may be configured to monitor the behavior of residents and detect abnormal signs using the same, and to enable immediate response when a dangerous situation occurs.

[0044] For example, the monitoring device (100) can collect the movement and activity amount of a resident within the household, which is included in the personal activity data, through an IoT sensor that includes a location calculation function provided in the household (20). The IoT sensor may include an ultrasonic sensor, and the location of the resident can be calculated using the ultrasonic sensor. The data collected through the IoT sensor can be transmitted to the monitoring device (100) via Wi-Fi or BLE communication.

[0045] The monitoring device (100) can detect abnormal signs and determine whether there is a dangerous situation based on collected data per resident, including data collected through IoT sensors, using a lightweight analysis model, and if it is determined to be a dangerous situation, it can proceed with notification or response procedures.

[0046] Hereinafter, based on FIGS. 2 to 4, a resident safety monitoring device according to one embodiment will be described in detail.

[0047] Referring to FIG. 2, a resident safety monitoring device (100) according to one embodiment may include a memory (110), a control unit (120), a communication unit (130), and an input / output unit (140).

[0048] The memory (110) can be configured using various types of memory such as DRAM, SSD, etc., and the memory (110) may include multiple distributed memory storage units, and the control unit (120) may separate and store the collected data by resident in multiple distributed memory storage units.

[0049] A program for resident safety monitoring and data for this purpose may be installed and stored in the memory (110). For example, a resident safety monitoring method, an artificial intelligence-based analysis model for resident safety monitoring that learns resident behavior patterns to detect abnormal signs, a data verification model, etc., may be installed and stored in the memory (110) in the form of a program.

[0050] Additionally, the memory (110) may store a threshold for determining the abnormal sign stage, measures to be taken when an abnormal sign is detected, and collected data per resident.

[0051] The control unit (120) is configured to include at least one processor, such as a CPU or GPU, and can perform the resident safety monitoring method presented below by executing a program stored in memory (110). For example, the control unit (120) can implement an artificial neural network by executing a program stored in memory (110) by a processor, and can perform anomaly detection based on data collected using the artificial neural network.

[0052] Additionally, the control unit (120) can control other components included in the resident safety monitoring device (100). For example, the control unit (120) may store the history of abnormal sign detection in memory (110), provide the history of abnormal sign detection for a predetermined period through the input / output unit (140), or, if the current abnormal sign level is at the 'danger' level as a result of abnormal sign detection, transmit the fact of abnormal sign detection to relevant agencies such as guardians, control centers, and emergency rescue teams through the communication unit (130). The process of the control unit (120) performing resident safety monitoring will be described in detail below with reference to other drawings.

[0053] Meanwhile, the communication unit (130) can perform wired or wireless communication with other devices or networks. For example, the communication unit (130) can collect energy consumption data such as water, electricity, and heating gas at a predetermined interval, or collect weather data from a public data portal at a predetermined interval.

[0054] To this end, the communication unit (130) may include a communication module that supports at least one of various wired and wireless communication methods, and the communication module may be implemented in the form of a chipset. The wireless communication supported by the communication unit (130) may be, for example, WiFi (Wireless Fidelity), Wi-Fi Direct, Bluetooth, UWB (Ultra-Wide Band), or NFC (Near Field Communication).

[0055] The input / output unit (140) may include a display panel or wearable display device for displaying resident safety monitoring results, and additionally, an output device such as a speaker. The input / output unit (140) may also include various types of input devices (e.g., keyboard, touchscreen, camera, etc.) for receiving input such as a resident safety monitoring request from an administrator.

[0056] Hereinafter, a method for monitoring resident safety performed by a resident safety monitoring device according to one embodiment is described in detail by the control unit (120) executing a program stored in memory (110). Unless otherwise specifically stated, the processes described below are performed by the control unit (120) executing a program stored in memory (110).

[0057] FIG. 3 is a diagram illustrating the operation of resident safety monitoring according to one embodiment. Referring to FIG. 3, the control unit (120) may include a preprocessing module (310) and an analysis module (320).

[0058] The control unit (120) can collect data by resident at a predetermined period through the communication unit (130) and store it in memory (110). The control unit (120) can collect data according to a preset schedule and can store the collected data by resident along with the time of collection or the time of recording. Here, the collection period may be weekly, monthly, daily, hourly, etc., and for example, the collection period may be 1 hour.

[0059] The control unit (120) can collect energy consumption data, including heating, water, gas, and electricity usage data for each household, through a metering device connected to the household (20). The collected data can be stored in memory (110). The energy consumption data may include at least one of a recording time or a collection time, and the control unit (120) can calculate the real-time usage, monthly usage, and average usage of heating, water, gas, and electricity for each household based on the collected energy consumption data and store them in memory (110).

[0060] Additionally, the control unit (120) can collect internal environmental data, including illumination, temperature, and humidity within the household, as well as door entry / exit information and electronic product usage information included in personal activity data, through an IoT sensor provided within the household (20). Additionally, the control unit (120) can collect movement and activity levels within the household included in personal activity data through an IoT sensor provided within the household (20) that includes a location information collection function. For example, the IoT sensor may be attached to the front door, and the control unit (120) can verify the entry and exit of the resident through the sensor attached to the front door.

[0061] The control unit (120) can execute a preprocessing module (310) to perform preprocessing for analyzing collected data by resident. For example, the control unit (120) can check whether energy consumption data, external environment data, internal environment data, and personal activity data included in the data by resident are collected normally, and based on the results of the check, perform missing value processing, noise removal, and data consistency verification, and then process the data by resident into behavioral patterns and environmental conditions of a time-series data structure based on the recording time or collection time. The preprocessed data by resident is used as input data for detecting abnormal signs and can be utilized as reference data for classifying abnormal sign stages into four grades: normal, interest, caution, and danger.

[0062] Meanwhile, the control unit (120) can verify the reliability of normal data collection, stable storage, and real-time anomaly detection functions through unit tests.

[0063] First, the control unit (120) tests the collection of data for each household. Specifically, the control unit (120) can verify whether the usage data value included in the energy consumption data for each household within the complex is a valid value other than zero, generated at a promised time interval, and stored primarily. In addition, by repeatedly collecting the stored data according to a set cycle and verifying whether it is properly linked with the database, the normal operation of the data collection and storage process can be confirmed. Furthermore, regarding the data that is properly linked, it can verify whether there are any omissions or errors in each data by performing data loading verification or querying.

[0064] Meanwhile, the control unit (120) is included in the preprocessing module (310) and can determine whether the energy consumption data is collected normally by using a deep learning-based data verification model (311) that learns the correlation between energy consumption data collected from sensors included in the remote metering device.

[0065] The control unit (120) can generate a correlation map that quantifies the interdependence between sensors for measuring each data item included in the data for each resident, based on an attention mechanism derived from the internal operation process of the learned data verification model (311). The correlation map defines weights for other items that are likely to change in conjunction with a specific data item when a change in that data item occurs, and can be updated in real time according to the learning results of the data verification model (311).

[0066] The control unit (120) can perform a cross-check to determine whether the sensor that generated the data is malfunctioning when a data stuck state, in which there is no omission, transmission delay, or change in value, is detected in the collected data per resident. Here, a data stuck state means a state in which the numerical value of the data remains fixed at a specific value without significant change despite the passage of time.

[0067] Specifically, the control unit (120) analyzes whether there is activity data for a second sensor that has a high dependency on the correlation map with the first sensor where the occurrence of the problem data was detected, when problem data is detected in which the value of the problem data included in the data for each resident does not change or data has not been collected. Activity data refers to a significant value indicating a physical change of the object measured by the sensor.

[0068] If it is confirmed that activity data for the second sensor exists, the control unit (120) may determine that an error has occurred due to a technical defect in the first sensor. In contrast, if no activity data is detected from the second sensor, the control unit (120) may determine the current state as a normal static data collection state.

[0069] If it is determined that an error has occurred due to a technical defect, the control unit (120) can obtain a replacement value that can replace the problem data based on other sensor data collected normally. Alternatively, the control unit (120) can lower the weight in the calculation process to reduce the contribution of the error data to the final abnormal sign detection result.

[0070] Through this stepwise verification and correction configuration, the control unit (120) can stably maintain the reliability of detecting abnormal signs throughout the system even when some sensors fail or the communication environment is poor.

[0071] Meanwhile, the control unit (120) can determine whether energy consumption data is collected normally by using a data verification model (311) implemented with a quantum deep learning model including a Quantum Neural Network (QNN) or a Variational Quantum Circuit (VQC). The control unit (120) performs a Quantum Feature Map process that converts the collected energy consumption data in the form of a time series into a quantum state, through which classical data can be encoded into a quantum state in a high-dimensional Hilbert space. The data verification model (311) implemented with a quantum deep learning model maps each sensor data to a quantum bit (Qubit) and learns the non-linear correlation between sensors using a quantum entanglement gate. In particular, the control unit (120) can derive a quantum correlation value between different energy consumption data of electricity, gas, water, and heating by analyzing the expectation value measured at the output terminal of the quantum circuit.

[0072] Meanwhile, when problem data is detected, the control unit (120) can determine whether normal data is collected by performing cross-verification based on the entanglement strength formed within the quantum circuit.

[0073] When the correlation value between the first sensor, which detects the occurrence of problem data, and the second sensor, which is in a quantum entanglement state, is greater than or equal to a threshold, the control unit (120) may refer to the activity data of the second sensor. At this time, if significant activity data is detected by the second sensor but no change in the quantum state of the first sensor is observed, the control unit (120) may finally determine the first sensor to be in a technically defective state. When utilizing quantum deep learning, it is possible to analyze subtle phase changes or complex multivariate causal relationships that are difficult for general artificial intelligence models to capture through quantum interference patterns, thereby providing the advantage of improved precision in defect detection.

[0074] Additionally, regarding the problem data, the control unit (120) can operate a quantum generation model that takes the quantum states of the remaining sensors in a normal state as input. The quantum generation model can restore the inferred quantum state of the defective sensor based on information from other sensors in an entangled relationship, and then decode it back into classical numerical data to replace the problem data. As described above, the data flow of the entire system can be maintained without interruption even in the event of a physical failure of the device performing data collection, and the reliability of the replacement data can be maximized through high-dimensional correlation analysis.

[0075] Meanwhile, the control unit (120) verifies whether the data stored in the memory (110) is normal. At this time, the data is stored in a table designed considering the nature and characteristics of the data, and the normality of data storage and access is verified through querying using a query language. In addition, the distributed storage and processing functions of external data can be verified by checking whether the stored data is properly visualized using a dashboard program.

[0076] Additionally, the control unit (120) may provide an API capable of performing additional linkage for the collection, storage, and processing of energy consumption data, and the collected data per resident may be preprocessed to be suitable for a big data collection, storage, processing, and analysis environment. Additionally, the collected data per resident may be used for the purposes of cumulative trend analysis, identification of anomaly occurrence times, identification of time intervals, verification of data consistency, verification of data integrity, and additional data refinement.

[0077] The control unit (120) can execute the analysis module (320) to analyze data by resident, identify the behavioral patterns of residents by household, and determine the standard patterns of residents by household. At this time, the control unit (120) can use Federation Query analysis technology to integrate and query data distributed in different memory storages using common SQL, thereby efficiently integrating and analyzing physically distributed data as if it existed in a single database.

[0078] The control unit (120) can determine multiple standard patterns by clustering past resident-specific data stored in memory (110) based on multidimensional variables. At this time, the control unit (120) can derive standard patterns that systematize standard lifestyles that a resident may exhibit under a specific environmental context by grouping household-specific energy consumption data and personal activity data collected through the communication unit (130) according to similarities in environmental conditions. Here, environmental conditions can be defined as a combination of multidimensional parameters including temperature and weather information provided from an external server (30) and time and day of the week information confirmed through the communication unit (130), as well as the resident's age, health status, and seasonal information. The control unit (120) can define multiple standard patterns by first grouping data based on the resident's age group and the presence or absence of underlying diseases, and then subdividing and clustering the data within each group according to seasonal factors and weather conditions.

[0079] For example, the control unit (120) can set environmental conditions that combine elements of a cold weather condition and a night time period for a group of residents who are elderly and have a specific underlying disease, and process the dataset generated under said conditions into a single cluster to generate a dedicated standard pattern. This standard pattern may mean a universal behavioral pattern model that a resident with a specific age and health condition may exhibit under a specific environment.

[0080] Furthermore, the control unit (120) can immediately monitor the resident's medical history or health status information, which is one of the components of the environmental conditions, when it is updated in real time through an external server (30). If a significant change in the health status is detected, the control unit (120) can dynamically reconstruct the existing clustering results or update the reference pattern by reflecting the changed information as a new environmental condition parameter. The reference pattern established through this dynamic re-clustering process can be input into an artificial intelligence-based analysis model and used as a statistical basis to determine whether there is a deviation in real-time data, thereby maintaining anomaly detection performance optimized for the resident's current physical condition.

[0081] The control unit (120) can identify an optimal standard pattern corresponding to the current situation based on environmental information included in the data collected in real time for each resident and the latest health status information of the resident. In addition, the control unit (120) can derive a predicted behavioral pattern for the environmental conditions at a specific point in time for the resident using an analysis model (321) included in the analysis module (320).

[0082] Here, the analysis model (321) can be implemented as an algorithm including at least one of a linear regression model, a recurrent neural network (RNN), a long short-term memory (LSTM), and a transformer, and can be learned based on clustered reference patterns.

[0083] The control unit (120) defines a normal range based on the maximum value, minimum value, and average value of the data accumulated within the identified reference pattern, and can check whether the actual behavior pattern exists within the normal range defined by the reference pattern (e.g., the interval between the minimum value and the maximum value). If the actual behavior pattern does not exist within the normal range defined by the reference pattern (e.g., the interval between the minimum value and the maximum value), the abnormal sign level can be set to the 'caution' level.

[0084] The control unit (120) can determine the resident's condition as an abnormal sign stage of 4 levels (grades) by comparing actual behavior patterns and predicted behavior patterns. In this process, the analysis model (321) is configured to directly calculate the reliability of each pattern by checking the dispersion of data for each standard pattern. The control unit (120) dynamically optimizes the accuracy of the final prediction and analysis by assigning a low weight to standard patterns with low reliability calculated through the analysis model and assigning a high weight to standard patterns with high reliability.

[0085] Specifically, the control unit (120) can determine a normal stage when both the actual behavior pattern based on real-time data and the predicted behavior pattern based on the analysis model exist within the normal range defined by the reference pattern, which may mean that the activity pattern of the resident matches the standard scenario for each environmental condition.

[0086] Meanwhile, the control unit (120) may determine the abnormal sign stage as a level of interest if both the actual behavior pattern and the predicted behavior pattern deviate from the normal range and the numerical deviation between the actual behavior pattern and the predicted behavior pattern exceeds a first deviation threshold. If the state of the level of interest persists beyond a first time threshold, the abnormal sign stage may be determined as a level of caution, which may mean a state in which there is a high probability that an abnormal physical or environmental situation of the resident has occurred. The control unit (120) may finally determine the level of danger if the state of the level of caution persists beyond a second time threshold or if a state of data fixation of the actual behavior pattern occurs. When the level of danger is confirmed, an emergency rescue notification may be sent to the guardian terminal (40) or the control center via the communication unit (130).

[0087] Since the control unit (12) excludes problem data caused by sensor defects or communication failures in advance through the aforementioned data verification model (311), it can identify the loss of variability appearing in the actual behavior pattern as a data fixation state caused by a physical emergency of the resident. That is, the control unit (120) can determine that a case where the rate of change of the actual behavior pattern remains below a preset lower limit is a top-priority risk situation. If the control unit determines that the aforementioned data fixation state has occurred in the actual behavior pattern, it can identify this as a more direct and critical rescue signal than the predicted deviation by the analysis model (321). Accordingly, the control unit (120) can immediately determine the abnormal sign stage as a risk stage without waiting for a separate period of time, and can maximize the response speed by transmitting an emergency rescue notification to the guardian terminal or control center.

[0088] Meanwhile, the control unit (120) can detect abnormal signs using an analysis model (321) implemented with a quantum deep learning model including a Quantum Neural Network (QNN) or a Variational Quantum Circuit (VQC).

[0089] Quantum deep learning-based analysis models can effectively handle non-linear correlations contained in high-dimensional time-series datasets. To solve the problem of computational complexity that occurs when existing AI-based analysis models extract features from multivariate data, quantum deep learning-based analysis models can utilize the principles of superposition and entanglement of quantum states.

[0090] The control unit (120) encodes the collected energy consumption data and personal activity data into the input state of the quantum circuit. Subsequently, a quantum deep learning-based analysis model measures the entanglement strength between each data through learned quantum gate operations, thereby determining the similarity between a reference pattern under specific environmental conditions and real-time data in a high-dimensional Hilbert space.

[0091] This quantum deep learning-based analysis method enables the early detection of subtle data deviations that are difficult to capture by classical linear regression or general neural network models, and consequently provides the advantage of maximizing the reliability of anomaly detection for the prevention of solitary deaths.

[0092] Additionally, the control unit (120) may detect abnormal signs in a resident by using the analysis model (321) as the main analysis model and additionally considering analysis results derived from one or more sub-analysis models dedicated to performing specific logic. The main analysis model is organically linked with the sub-analysis models, so that the technical judgment indicators provided by each sub-analysis model can be utilized as factors for integrated analysis.

[0093] FIG. 4 is a diagram illustrating a resident safety monitoring process of another embodiment. Since the preprocessing module (410) and data verification model (411) of FIG. 4 perform the same functions as the preprocessing module (310) and data verification model (311) of FIG. 3, respectively, a detailed description thereof will be omitted.

[0094] Referring to FIG. 4, the analysis module (420) may include a main analysis model (421), a first sub-analysis model (422), and a second sub-analysis model (423). Although FIG. 4 describes the analysis module (420) as including two sub-analysis models, according to the embodiment, the analysis module (420) may include either the first sub-analysis model (422) and the second sub-analysis model (423) and the main analysis model (421).

[0095] The main analysis model (421) can be configured to perform a function corresponding to the analysis model (321) of FIG. 3 described above. Specifically, the main analysis model (421) can identify an optimal reference pattern corresponding to the current situation from memory (110) based on environmental information collected in real time and the latest health status information of the resident.

[0096] The main analysis model (421) checks whether the actual behavior pattern is within the normal range defined by the reference pattern, and if it is within the normal range, it can determine whether there is a primary abnormal sign by comparing the predicted behavior pattern obtained from the learned main analysis model (421) with the actual behavior pattern. Furthermore, the main analysis model (421) can determine the final abnormal sign stage by utilizing the judgment indicators received from the sub-analysis models as factors for integrated analysis. The main analysis model (421) is organically linked with the sub-analysis models and can reflect the technical judgment indicators provided by each sub-analysis model as weights for integrated analysis.

[0097] The first sub-analysis model (422) can precisely analyze the active operation ability of a resident in response to rapid changes in the external environment. The first sub-analysis model (422) can measure the time interval between the point when the external temperature reaches a specific threshold and the point when a significant change in actual energy consumption data occurs, and calculate this as the response delay time.

[0098] The first sub-analysis model (422) can compare the calculated response delay time with standard delay time data within a reference pattern managed by the main analysis model (421). The reference pattern can be identified by the main analysis model based on a cluster of historical data containing external temperature fluctuation ranges, temperature zones, and time zone information similar to the present.

[0099] The first sub-analysis model (422) can determine whether the currently measured response delay time exceeds the normal range defined in the reference pattern. In particular, if the operation signal of the heating and cooling device is not detected within a preset time despite a sudden change in temperature, making it impossible to calculate the response delay time itself, the first sub-analysis model (422) can identify this as an abnormal sign due to a physical emergency of the resident or loss of ability to operate the device and transmit it to the main analysis model (421).

[0100] The second sub-analysis model (423) can identify significant activities of residents based on resident-specific data and define the flow of behavioral patterns corresponding to those activities. Here, individual activities may refer to the results of identifying significant activities of residents identified from various types of sensor data, including electricity, water, and gas usage and door opening / closing signals.

[0101] The second sub-analysis model (423) can learn the unique lifestyle routines of each resident to generate a standard sequence and include it in a standard pattern. For example, a series of activities such as using the bathroom after waking up, using kitchen power, and opening and closing the front door can be structured into a single visualized scenario that includes the order of occurrence and the time elapsed between them.

[0102] The second sub-analysis model (423) can construct a sequence of actions currently in progress in real time based on resident-specific data collected in real time. Additionally, specifically, the second sub-analysis model (423) can determine time-series consistency by comparing the sequence of actions generated in real time with a standard sequence stored in a reference pattern. At this time, the second sub-analysis model (423) can identify phenomena where the order of events within the sequence is mixed up or the time interval between specific sensor detections deviates from the standard deviation of past data and becomes abnormally long, even if the individual measurement values ​​of each sensor exist within a normal range.

[0103] Furthermore, the second sub-analysis model (423) can calculate an entropy value indicating the degree of disorder within the resident's living pattern. The second sub-analysis model (423) can calculate the degree of deviation of the real-time behavior pattern from the learned standard routine as an information entropy value based on a probability distribution, and can identify cases where the entropy value exceeds a threshold and increases rapidly as abnormal signs due to the breakdown of the living rhythm and provide this to the main analysis model (421). The second sub-analysis model (423) can provide such sequence analysis results and entropy fluctuation values ​​to the main analysis model (421).

[0104] The main analysis model (421) can determine the resident's condition in four stages of abnormal signs by integrating analysis data received from at least one of the first sub-analysis model (422) and the second sub-analysis model (423) into the comparison results of the predicted behavior pattern and the actual behavior pattern.

[0105] The main analysis model (421) can determine the resident's condition by integrating behavioral reactivity analysis data received from the first sub-analysis model (422) with the comparison results of the predicted behavior pattern and the actual behavior pattern. Specifically, the main analysis model (421) can determine the abnormal sign stage as a normal stage if the actual behavior pattern and the predicted behavior pattern exist within a normal range and the reaction delay time identified by the first sub-analysis model (422) is within the set attention delay time. On the other hand, the main analysis model (421) can determine the abnormal sign stage as a caution stage if, even if the actual behavior pattern and the predicted behavior pattern exist within a normal range, the reaction delay time identified by the first sub-analysis model (422) exceeds the attention delay time. This may be intended to derive advanced detection results that consider the resident's actual operational state in response to the external environment context, going beyond simple numerical deviation analysis.

[0106] Alternatively, the main analysis model (421) may determine the state of the resident by integrating entropy analysis data received from the second sub-analysis model (423) with the comparison results of the predicted behavior pattern and the actual behavior pattern. The main analysis model (421) may determine the state of interest if it receives an analysis result from the second sub-analysis model (423) indicating that the disorder of the living pattern has increased, within the normal range defined by the reference pattern, where the actual behavior pattern and the predicted behavior pattern are within the reference pattern. Conversely, even if the main analysis model (421) detects a possibility of anomaly because some values ​​of the actual behavior pattern have temporarily deviated, if the entropy change of the entire behavior pattern analyzed by the second sub-analysis model (423) is stable, it may be considered a one-time behavior to prevent false positives.

[0107] For example, if the flow of the actual behavior pattern deviates significantly from the usual routine and the entropy change increases above a threshold, the main analysis model may determine it as a stage of interest. Conversely, even if the main analysis model detects a potential anomaly, if the entropy change of the entire behavior pattern analyzed by the second sub-analysis model (423) is stable, the main analysis model (421) may consider this as temporary overuse to prevent false detection. In other words, even if the main analysis model (421) detects a potential anomaly because some values ​​of the real-time pattern temporarily deviate from the threshold range of the reference pattern, if the entropy change of the entire behavior sequence analyzed by the second sub-analysis model (423) is stable and the flow of the sequence maintains the usual routine, the main analysis model (421) may consider this as simple temporary overuse or one-time behavior to prevent false detection.

[0108] As another example, the main analysis model (421) can calculate the final anomaly level by combining the analysis data of the first sub-analysis model (422) and the second sub-analysis model (423) with the comparison results of the predicted behavior pattern and the actual behavior pattern. The main analysis model (421) can determine the final anomaly level by integrating its own pattern comparison results and the results of each sub-analysis model (422, 423) in a weighted sum or ensemble-based majority method. If the weighted sum method is followed, the main analysis model (421) can assign dynamic weights based on environmental conditions and time context to the analysis data received from each sub-analysis model (422, 423).

[0109] The main analysis model (421) can adjust the contribution of each sub-analysis model according to the situation in which a specific environmental event is detected. For example, in a cold wave situation, a high weight can be assigned to the behavioral reactivity analysis result of the first sub-analysis model (422). On the other hand, in normal situations, a high weight can be assigned to the entropy analysis result of the second sub-analysis model (423) to precisely monitor whether the resident's lifestyle pattern is maintained.

[0110] For example, even if there is no numerical deviation in the actual data, if entropy rises and the reaction delay time begins to increase simultaneously, this can be determined as a stage of interest, and continuous monitoring can be strengthened. If a combination of numerical deviation, a surge in entropy, or a state of inability to respond occurs, this can be immediately upgraded to a stage of caution or danger. Furthermore, the main analysis model (421) can determine the state of the aforementioned caution stage as a danger stage if it persists beyond a second time threshold or if a data fixation state occurs in the actual behavior pattern. Here, a data fixation state in the actual behavior pattern may refer to a state in which the numerical values ​​constituting the actual behavior pattern remain fixed at a specific value without significant fluctuation despite the passage of time. In particular, if the collected actual behavior pattern remains in a static state despite the predicted behavior pattern derived by the main analysis model (421) indicating an increase in activity, this can be identified as a top-priority danger situation.

[0111] When the danger level is finally determined, the control unit (120) can send an emergency rescue notification to a guardian terminal (40) or a control center via the communication unit (130). At this time, the control unit (120) can support rapid rescue activities by transmitting the resident's last activity location, real-time health status information, and environmental condition data at that time together.

[0112] Meanwhile, the control unit (120) may be configured to automatically update the reference pattern at a predetermined interval (e.g., 24-hour interval) to reflect the latest lifestyle of the resident. During this periodic update process, the analysis model may also be retrained based on a newly collected learning dataset, and in particular, for a quantum-based analysis model, parameter tuning and quantum circuit optimization may be performed in parallel.

[0113] The control unit (120) can perform a process of optimizing rotation angle parameters assigned to each quantum gate within the variational quantum circuit (VQC). This process can be carried out in a hybrid manner incorporating classical optimization algorithms, and parameters can be updated in a direction that minimizes the objective function based on the latest trends of personal activity data and energy consumption data. Through this, the analysis model can precisely reflect subtle changes in the resident's lifestyle patterns in the probability distribution of the quantum state.

[0114] In addition, the control unit (120) can dynamically reconfigure the structure of the quantum circuit according to the dimension or complexity of the collected data. For example, in an environment where the correlation between data is simple, the depth of the quantum gate can be reduced to increase computational efficiency, and during periods of high activity, the density of the entanglement gate can be increased to enhance analysis precision. Such circuit optimization can provide the effect of increasing the real-time responsiveness of the analysis model while minimizing the possibility of errors caused by computational noise.

[0115] Additionally, the control unit (120) can manage the analysis model by version as the analysis model update progresses, and can perform performance verification including sensitivity and accuracy for each version of the analysis model. In order to secure conditions for optimal performance manifestation, the control unit (120) can accumulate, record, and manage the improvement history of the algorithm and the performance verification results in the memory (110).

[0116] For an analysis model that has been updated, the control unit (120) reviews and applies the applicability to an actual system based on predefined performance indicators, and through this systematic management, the analysis model can flexibly respond to changes in personal activity data and energy consumption data and maintain a high level of reliability in detecting abnormal signs.

[0117] Meanwhile, the results of abnormal sign detection can be classified into multiple abnormal sign stages, such as normal, interest, caution, and danger stages, depending on the severity. When the abnormal sign stage reaches a danger level, the control unit (120) can respond early to dangerous situations such as solitary death by sending an emergency notification to the guardian terminal (40) or the control center.

[0118] Meanwhile, the control unit (120) can provide standardized maintenance and management services for data for each resident, and to this end, it can provide a resident interface that visualizes real-time energy usage for each residence. At this time, the resident interface can be designed intuitively by considering the various environments and behavioral patterns of the residents, and can be implemented so that anyone can easily use it by prioritizing accessibility and ease of use. In addition, the resident interface can be systematically organized into a modular structure to facilitate future functional expansion and maintenance.

[0119] For example, the resident interface can be implemented to allow for the identification of risk, caution, and normal statuses for each household at a glance through changes in color and shape. This enables managers to intuitively perceive complex analysis results without the need for separate interpretation.

[0120] Additionally, the resident interface may include a schedule management module for data processing and operation, a module responsible for data collection, transformation, and processing, and a development module for data analysis and model development. These components can be linked through an integrated dashboard, and the control unit (120) can visually provide the overall system status and processing status through the integrated dashboard.

[0121] In addition, the control unit (120) may include a data integration visualization screen in the resident interface, and may display an apartment building image in conjunction with status information at the floor and unit level, and may be able to grasp the risk status of the entire apartment and specific areas at a glance.

[0122] The administrator screen can be structured around data visualization and utilizes color-based graphs and tables to enable easy identification of anomalies in individual households. This allows administrators to efficiently monitor the status of multiple households and respond quickly.

[0123] The control unit (120) performs real-time resident safety monitoring and can automatically generate daily reports, weekly reports, and monthly reports based on the real-time monitoring results. The generated reports may be provided along with the results of an abnormal sign determination.

[0124] When an abnormal sign is detected at the 'danger' level, the control unit (120) can keep and manage the data necessary to automatically send an SMS notification to the recipient up to date, and can send real-time monitoring results and generated reports to the person in charge via email. For example, a daily report can be sent once a day, a weekly report once a week, and a monthly report once a month.

[0125] Such a resident interface and system configuration can improve the usability and effectiveness of the system by effectively conveying anomaly detection results and supporting the judgment and response of residents and managers.

[0126] In the embodiments above, the term 'part' refers to a software or hardware component such as a field programmable gate array (FPGA) or an ASIC, and the 'part' performs certain roles. However, the 'part' is not limited to software or hardware. The 'part' may be configured to reside in an addressable storage medium or may be configured to run one or more processors. Thus, as an example, the 'part' includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables.

[0127] The functions provided within the components and 'parts' can be combined into fewer components and 'parts' or separated from additional components and 'parts'.

[0128] In addition, the components and '~parts' may be implemented to play one or more CPUs within the device or secure multimedia card.

[0129] FIG. 5 is a flowchart illustrating a resident safety monitoring method according to one embodiment. The resident safety monitoring method illustrated in FIG. 5 includes steps processed chronologically in the resident safety monitoring device (100) illustrated in FIG. 1 to 4. Therefore, even if details are omitted below, the details described above regarding the resident safety monitoring device (100) illustrated in FIG. 1 to 4 can also be used in the resident safety monitoring method according to the embodiment illustrated in FIG. 5.

[0130] Referring to FIG. 5, the monitoring device (100) can collect data per resident, including energy consumption data per household, external environment data, and personal activity data (S510). Specifically, the monitoring device (100) can collect energy consumption data per household, including at least one of electricity, gas, water, and heating usage, by linking with a remote metering device provided in the household, and can collect at least one of internal environment information, door entry / exit information, and home appliance usage information included in personal activity data by analyzing signals obtained from sensors provided in the household. In addition, the monitoring device (100) can collect the external environment data, including weather and temperature information, and the history of use of nursing facilities included in the personal activity data from an external server.

[0131] Next, the monitoring device (100) can obtain a predicted behavior pattern by inputting collected data into an analysis model learned based on a standard pattern for environmental conditions defined by combining resident information including the age and health status of the resident and external environment information, and can determine an abnormal sign stage by performing rule-based abnormal sign detection based on the actual behavior pattern obtained by processing the predicted behavior pattern and the collected data (S520).

[0132] At this time, the monitoring device (100) may determine the normal stage when the numerical deviation is less than the first deviation threshold, and determine the abnormal sign stage as the interest stage when the numerical deviation is greater than or equal to the first deviation threshold. Additionally, the monitoring device (100) may determine the abnormal sign stage as the caution stage when the state of the interest stage persists beyond the first time threshold.

[0133] The monitoring device (100) may determine the abnormal sign stage as a danger stage if the state of the warning stage persists beyond a second time threshold, or if a data fixation state occurs for at least some of the data included in the actual behavioral pattern, not the problem data caused by a technical defect. In the case of the danger stage, the monitoring device (100) may send a notification to a guardian terminal or a control center.

[0134] According to an embodiment, the monitoring device (100) may further execute one or more sub-analysis models that analyze the collected resident-specific data using an algorithm different from the analysis model, and may finally determine the abnormal sign stage by integrating and analyzing the analysis data received from the sub-analysis models along with the predicted behavioral pattern and the actual behavioral pattern.

[0135] The resident safety monitoring method described through FIG. 5 may also be implemented in the form of a computer-readable medium that stores commands and data executable by a computer. In this case, the commands and data may be stored in the form of program code, and when executed by a processor, they may generate a specific program module to perform a specific operation. Furthermore, the computer-readable medium may be any available medium accessible by a computer and includes both volatile and non-volatile media, as well as removable and non-removable media. Additionally, the computer-readable medium may be a computer recording medium, which may include both volatile and non-volatile, removable and non-removable media implemented by any method or technique for storing information such as computer-readable commands, data structures, program modules, or other data. For example, the computer recording medium may be a magnetic storage medium such as an HDD and an SSD, an optical recording medium such as a CD, DVD, or Blu-ray disc, or a memory contained in a server accessible via a network.

[0136] Additionally, the resident safety monitoring method described through FIG. 5 may be implemented as a computer program (or computer program product) containing instructions executable by a computer. The computer program includes programmable machine instructions processed by a processor and may be implemented in a high-level programming language, an object-oriented programming language, assembly language, or machine language. Additionally, the computer program may be recorded on a tangible computer-readable recording medium (e.g., memory, hard disk, magnetic / optical medium, or SSD (Solid-State Drive), etc.).

[0137] Accordingly, the resident safety monitoring method described through FIG. 5 can be implemented by executing a computer program as described above by a computing device. The computing device may include at least some of a processor, memory, a storage device, a high-speed interface connected to the memory and a high-speed expansion port, and a low-speed interface connected to a low-speed bus and a storage device. Each of these components is connected to one another using various buses and may be mounted on a common motherboard or mounted in other suitable ways.

[0138] Here, the processor can process instructions within the computing device, such as instructions stored in memory or storage devices to display graph information for providing a Graphic User Interface (GUI) on external input / output devices, such as a display connected to a high-speed interface. In another embodiment, a plurality of processors and / or a plurality of buses may be appropriately utilized with a plurality of memories and memory types. Additionally, the processor may be implemented as a chipset comprising chips including a plurality of independent analog and / or digital processors.

[0139] In addition, memory stores information within a computing device. For example, memory may consist of volatile memory units or a set thereof. As another example, memory may consist of non-volatile memory units or a set thereof. Furthermore, memory may be other forms of computer-readable media, such as magnetic or optical discs.

[0140] And memory can provide a large amount of storage space to a computing device. Memory may be a computer-readable medium or a configuration containing such a medium, and may include, for example, devices or other configurations within a Storage Area Network (SAN), and may be a floppy disk device, a hard disk device, an optical disk device, or a tape device, flash memory, or other similar semiconductor memory device or device array.

[0141] The embodiments described above are for illustrative purposes only, and those skilled in the art will understand that the embodiments described above can be easily modified into other specific forms without altering the technical concept or essential features of the embodiments described above. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0142] The scope of protection sought through this specification is defined by the claims set forth below rather than by the detailed description above, and should be interpreted to include all modifications or variations derived from the meaning and scope of the claims and the concept of equivalents. Explanation of the symbols

[0143] 100: Resident safety monitoring device 110: Memory 120: Control unit 130: Communication unit 140: Input / Output unit

Claims

Claim 1 In an artificial intelligence-based resident safety monitoring device, a memory storing a program for performing resident safety monitoring and standard patterns for environmental conditions defined by combining resident information, including the age and health status of the resident, and external environment information;The system includes one or more processors and operates by executing a program stored in memory, collects resident-specific data including household energy consumption data, external environment data, and personal activity data, obtains predicted behavior patterns by inputting the collected resident-specific data into an analysis model learned based on the reference patterns, and includes a control unit that determines an anomaly stage by performing rule-based anomaly detection based on the actual behavior patterns obtained by processing the collected resident-specific data into a time-series data structure based on the recording time or collection time and the predicted behavior patterns, wherein the reference patterns are behavior patterns generated based on clustering resident-specific data according to the similarity of environmental conditions, and the behavior patterns have the form of multidimensional vector data in which feature quantities corresponding to multiple data items included in household energy consumption data and personal activity data collected over a predetermined period are arranged along a time axis, and the control unit executes a data verification model that learns the correlation between each data item included in the resident-specific data, and generates a correlation map that quantifies the interdependence between sensors for measuring each data item by the data verification model, wherein the correlation map has weights for other data items that are highly likely to change in conjunction with the change of a specific data item A device defined and generated based on an attention mechanism derived from the internal computation process of the data verification model, and detecting whether the first sensor is malfunctioning by checking the existence of activity data of a second sensor having a dependency on the correlation map with a first sensor measuring the problem data when problem data is detected among the data included in the resident-specific data where the value does not change or has not been collected. Claim 2 A device according to claim 1, wherein the control unit determines that an error has occurred due to a technical defect of the first sensor when the existence of a second sensor having a dependency on the correlation map with the first sensor measuring the problem data is confirmed and activity data of the second sensor exists, and when it is determined that an error has occurred due to a technical defect of the first sensor, the control unit obtains a replacement value to replace the problem data or lowers the weight to reduce the contribution of the problem data to the anomaly detection result. Claim 3 In paragraph 2, the apparatus wherein at least one of the data verification model and the analysis model is a quantum deep learning model. Claim 4 In claim 1, the control unit determines a plurality of standard patterns by clustering resident-specific data collected over a predetermined period based on the similarity of the environmental conditions. Claim 5 A device according to claim 1, wherein the control unit collects energy consumption data for each household, including at least one of electricity, gas, water, and heating usage, by linking with a remote metering device provided in the household, and collects at least one of internal environment information, door entry / exit information, and home appliance usage information included in the personal activity data by analyzing a signal obtained from a sensor provided in the household, and collects external environment data including weather and temperature information from an external server and medical institution usage history included in the personal activity data. Claim 6 A device according to claim 1, wherein the control unit determines the abnormal sign stage according to a preset rule based on the numerical deviation between the actual behavior pattern and the predicted behavior pattern and the duration of the abnormal sign, wherein the normal stage is determined when the numerical deviation is less than a first deviation threshold, the abnormal sign stage is determined as an interest stage when the numerical deviation is greater than or equal to the first deviation threshold, the abnormal sign stage is determined as a caution stage when the state of the interest stage persists exceeding a first time threshold, the abnormal sign stage is determined as a danger stage when the state of the caution stage persists exceeding a second time threshold or a state of data fixation of the actual behavior pattern occurs, and in the case of the danger stage, transmits a notification to a guardian terminal or a control center. Claim 7 A device according to claim 1, wherein the control unit further executes one or more sub-analysis models that analyze the collected resident-specific data using an algorithm different from the analysis model, and integrates and analyzes the analysis data received from the sub-analysis models together with the predicted behavior pattern and the actual behavior pattern to finally determine the abnormal sign stage. Claim 8 A method for monitoring resident safety performed by a resident safety monitoring device, comprising the step of collecting resident-specific data including household-specific energy consumption data, external environment data, and personal activity data; The method comprises the step of obtaining predicted behavioral patterns by inputting collected resident-specific data into an analysis model learned based on reference patterns for environmental conditions defined by combining resident information, including the age and health status of residents, and external environmental information, and determining an anomaly stage by performing rule-based anomaly detection based on actual behavioral patterns obtained by processing the predicted behavioral patterns and the collected resident-specific data into a time-series data structure based on the recording time or collection time, wherein the reference patterns are behavioral patterns generated based on clustering resident-specific data according to the similarity of environmental conditions, and the behavioral patterns have the form of multidimensional vector data in which feature quantities corresponding to multiple data items included in household energy consumption data and personal activity data collected over a predetermined period are arranged along a time axis, and the method further comprises executing a data validation model learned for the correlation between each data item included in the resident-specific data to generate a correlation map that quantifies the interdependence between sensors for measuring each data item by the data validation model, wherein the correlation map defines weights for other data items that are highly likely to change in conjunction with the change of a specific data item, and based on an attention mechanism derived from the internal computation process of the data validation model A method comprising further including a step of generating, wherein the step of determining the abnormal sign step includes, when problem data that has not changed or has not been collected among the data included in the resident-specific data is detected, checking whether there is activity data of a second sensor having a dependency on the correlation map with a first sensor measuring the problem data, thereby detecting whether the first sensor is malfunctioning. Claim 9 In claim 8, the step of detecting whether the first sensor is malfunctioning comprises: a step of determining that an error has occurred due to a technical defect of the first sensor if the existence of a second sensor having a dependency on the correlation map with the first sensor measuring the problem data is confirmed and activity data of the second sensor exists; and a step of obtaining a replacement value to replace the problem data or lowering the weight to reduce the contribution of the problem data to the anomaly detection result if it is determined that an error has occurred due to a technical defect of the first sensor. Claim 10 In claim 8, the step of determining an abnormal sign stage comprises: determining a normal stage when the numerical deviation between the actual behavior pattern and the predicted behavior pattern is less than a first deviation threshold, and determining the abnormal sign stage as an interest stage when the numerical deviation is greater than or equal to the first deviation threshold; determining the abnormal sign stage as a caution stage when the state of the interest stage persists exceeding a first time threshold; and determining the abnormal sign stage as a danger stage when the state of the caution stage persists exceeding a second time threshold or a state of data fixation of the actual behavior pattern occurs, and transmitting a notification to a guardian terminal or control center in the case of the danger stage. Claim 11 A computer program that is performed by a resident safety monitoring device and is stored on a computer-readable recording medium to perform the method described in paragraph 8. Claim 12 A computer-readable recording medium having a computer program that performs the method described in paragraph 8.

Citation Information

Patent Citations

  • Integrated Monitoring System and Method for Emergency based on Big data and Artificial Intelligence for Resident in House

    KR1020220066712A

  • Apparatus and method for predicting risk for using collected information of non-contact type sensing device and energy consumption information

    KR1020240098825A

  • Method for determining abnormal behavior of occupants using non-contact sensor data based context awareness model and energy segmentation model

    KR1020240151035A

  • Artificial intelligence-based emergency response system for the elderly

    KR102160463B1