Abnormality detecting system for solitary living housing / elderly housing and program

Abstract

To enable another person to recognize a potential or actual abnormality in a solitary person or an elderly person requiring particular monitoring, based on information related to water supply and other lifelines.SOLUTION: In an abnormality notification system 1 that notifies a degree of abnormality of a resident based on usage information, which includes information including water usage amount and water usage time, the system comprises: an evaluation model generation unit 304 that uses a history of past usage information in the housing as training data, and generates an evaluation model using machine learning with input being usage information since the latest predetermined time and elapsed time since the last water usage, and with output being an evaluation of a degree of abnormality of the resident in the housing; a usage information input unit 301 that receives water usage information from an NCU 205; an abnormality evaluation unit 305 that evaluates the degree of abnormality of the resident in the housing using the evaluation model based on the input usage information; and a notification unit 306 that notifies a terminal device previously associated with the resident.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to an anomaly detection system and program for single-person and elderly housing, and in particular to a technology that uses machine learning to accurately grasp the degree of deviation from the target resident's water usage patterns. [Background technology]

[0002] Conventionally, there is known a technology to check the safety of elderly people living alone by monitoring the usage of so-called lifelines such as water, gas, and electricity using smart meters, etc. For example, if the water meter does not work for more than a day even though the elderly person is supposed to be at home, there is a concern that they may be unable to move around inside the house and may have an accident. In such cases, the applicant has developed a technology such as that described in Reference 3, which allows for email forwarding to pre-registered relatives and backup support from security companies.

[0003] However, the conventional technology has the following problems. Naturally, each resident's lifestyle patterns are different. For example, some people use a lot of water in the morning and evening for meals, toilets, and baths, and barely use water in between, while others use small amounts of water frequently during the day. Therefore, detecting anomalies based on the time elapsed since the last use has low accuracy and is not very useful. Similarly, setting a threshold for screening purposes and detecting anomalies based on the time elapsed since the final threshold was exceeded also has the problem of not providing particularly high accuracy. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2011-053859 [Patent Document 2] Patent Publication No. 2013-027598 [Patent Document 3] Patent Publication No. 2021-068193 Summary of the Invention [Problem to be solved by the invention]

[0005] The present invention has been made in consideration of the above, and aims to provide technology that enables others to recognize possible abnormalities in people living alone or the elderly, who particularly require overt and potential monitoring, based on information on water and other lifelines. [Means for solving the problem]

[0006] The abnormality notification system described in claim 1 is an abnormality notification system that notifies the degree of abnormality in residents of a single-person home or a home for elderly people based on usage information that includes water usage amount and the time of water usage, and is characterized by comprising: an evaluation model generation means that uses the history of past usage information in the home as training data, inputs usage information since the most recent fixed time and the time elapsed since the last water usage, and generates an evaluation model through machine learning that outputs an evaluation of the degree of abnormality in the resident of the home; a usage information input means that inputs water usage information from an NCU connected to the water meter of the home; an abnormality evaluation means that uses the evaluation model generated by the evaluation model generation means to evaluate the degree of abnormality in the resident of the home based on the usage information input by the usage information input means; and a notification means that notifies a terminal device previously associated with the resident of the degree of abnormality evaluated by the abnormality evaluation means.

[0007] In other words, the invention of claim 1 makes it possible to evaluate the degree of abnormality in a resident using an evaluation model that has been machine-learned for each individual residence based on information such as the amount of water used, the time of use, and the time elapsed since the last use.

[0008] Residents of elderly housing naturally include elderly couples and elderly parents and children. Abnormality is defined in a broad sense as something that is not normal, and includes, for example, water usage exceeding the expected amount, as well as water usage being below the expected amount or not at all. The degree of abnormality is not particularly limited, but could be, for example, the degree of deviation from the expected water usage pattern (or group of patterns) of the resident. Specifically, in the summer, an elderly woman living alone may normally use a fixed amount of water (here, using a flush toilet), but if this fixed amount of water usage becomes significantly longer, there is concern that she may be experiencing early-stage dehydration. Because individual learning models are constructed using machine learning, individual usage patterns and individual actions can be separated and analyzed, and abnormalities can be detected and notified even if the person is not aware of them. For example, if a person regularly uses a small amount of water after meals (here, brushing their teeth), but then gradually this frequency decreases, it could be inferred that this is an early stage of dementia. Furthermore, if a person continues to use approximately 200 liters of water on a regular basis but then stops using the water, this can be used to screen for the possibility that they are not bathing, which means they are not feeling well.If, although this is done regularly, they start to use more than 400 liters of water from a certain point on, this can also be used to screen for the possibility that they are leaving the water running, which means they may have dementia. In addition, the degree of abnormality can be evaluated using multiple levels, such as levels 1 to 5, and can also be combined with information from different perspectives, such as urgency levels of "high," "low," and "none." Usage information does not consist only of the amount of water used and the time of water use, but the training data can also use information related to water use, such as the day of the week, weather, temperature, humidity, existing facilities, holidays, as well as online information and cloud information as appropriate. For example, "after the nearest fixed time" may be 96 hours before the date and time of evaluation. The NCU can be attached to the water meter (smart meter) or integrated into it. It is also possible to use the LPWA (Low Power Wide Area) wireless communication method (LTE lines can be used). Terminal devices broadly include smartphones, mobile phones, PCs, and even landline phones depending on the specifications. Examples of terminal devices associated with a resident include terminal devices of the resident's relatives, a terminal device of a neighborhood association, and a terminal device of a security company. The notification by the notification means can be by sending an email, a voice message, an alarm sound, etc. The notification destination is not limited to one, but can be multiple.

[0009] The anomaly notification system of claim 2 is the anomaly detection system of claim 1, characterized in that the usage information also includes gas usage amount, gas usage time, and electricity usage trends, the evaluation model generation means adds the elapsed time since the last gas usage to the input and generates an evaluation model by machine learning, the output of which is an evaluation of the degree of anomaly for the resident of the house, and the usage information input means inputs usage information from an NCU that is also connected to the gas meter and electricity meter of the house.

[0010] That is, the invention according to claim 2 effectively uses the NCU to connect smart meters related to lifelines, thereby realizing more accurate detection of abnormalities in a complex manner.

[0011] The anomaly detection system described in claim 3 is characterized in that, in the anomaly detection system described in claim 1 or 2, it is equipped with a notification control means that controls the notification means to expand or change the notification destinations based on the degree of anomaly evaluated by the anomaly evaluation means.

[0012] In other words, the invention of claim 3 can notify the appropriate notification recipient of an abnormality by classifying and categorizing the abnormality in advance.

[0013] The anomaly detection system of claim 4 is the anomaly detection system of claim 1 or 2, further comprising a basic information input means for inputting basic information characterizing a large number of residents who are system users, such as the resident's age, sex, height, weight, medical history, breakfast start time, lunch start time, dinner start time, bedtime, wake-up time, sleep duration, and other basic information; and a correlated person extraction means for extracting highly correlated residents based on the basic information input by the basic information input means, and the evaluation model generation means generates an evaluation model using the usage information of highly correlated residents extracted by the correlated person extraction means as training data.

[0014] In other words, the invention of claim 4 can generate a model by utilizing usage information from a large number of similar residents, thereby contributing to the realization of more accurate anomaly detection.

[0015] Highly correlated means similar, and similarity can be characterized from various perspectives.

[0016] The anomaly detection program described in claim 5 is an anomaly detection program for operating the anomaly detection system described in any one of claims 1 to 4, characterized in that it causes a computer that constructs the system to function as each of the means defined in the claims.

[0017] That is, the invention of claim 5 makes it possible to evaluate the degree of abnormality in a resident using a machine-learned evaluation model based on information such as water usage, time of usage, and time elapsed since the last usage. Furthermore, by effectively using the NCU to connect smart meters related to lifelines, it is possible to realize more accurate abnormality detection in a complex manner. By pre-classifying and categorizing abnormalities, it is possible to notify the appropriate notification recipients of abnormalities. Furthermore, it is possible to generate a model using usage information from many similar residents, contributing to the realization of more accurate abnormality detection.

[0018] The specific program is as follows: <Anomaly detection program corresponding to claim 1> A program for constructing an abnormality notification system that notifies the degree of abnormality of residents of single-person homes or elderly housing based on usage information that includes water usage amount and water usage time, Computer, an evaluation model generation means for generating an evaluation model by machine learning using a history of past usage information in the residence as training data, inputting usage information since a fixed time and the time elapsed since the last water usage, and outputting an evaluation of the degree of abnormality of the resident of the residence; a usage information input means for inputting water usage information from an NCU connected to a water meter of the residence; an abnormality evaluation means for evaluating the degree of abnormality of the resident of the house based on the usage information input by the usage information input means, using the evaluation model generated by the evaluation model generation means; and a notification means for notifying a terminal device associated in advance with the resident of the degree of abnormality evaluated by the abnormality evaluation means; An anomaly detection program characterized by functioning as <Anomaly detection program corresponding to claim 2> The usage information includes the amount of gas used, the time of gas use, and the trend of electricity usage. The evaluation model generation means is configured to generate an evaluation model by machine learning, the evaluation model having an input of the time elapsed since the last gas usage and an output of the evaluation of the degree of abnormality of the resident of the house. The usage information input means is configured to input usage information from an NCU connected to the gas meter and the electricity meter of the house. The anomaly detection program is characterized by causing the anomaly detection program to function. <Anomaly detection program corresponding to claim 3> Computers, and more a notification control means for controlling the notification means to expand or change the notification destinations based on the degree of abnormality evaluated by the abnormality evaluation means; The anomaly detection program as described above, characterized in that it functions as <Anomaly detection program corresponding to claim 4> Computers, moreover, A basic information input means for inputting basic information that characterizes a large number of residents who are system users, such as the age, sex, height, weight, medical history, breakfast start time, lunch start time, dinner start time, bedtime, wake-up time, sleep time, and other basic information; a correlated person extraction means for extracting residents with high correlation based on the basic information input by the basic information input means; It functions as The anomaly detection system is characterized in that the evaluation model generation means is configured to generate an evaluation model using the usage information of highly correlated residents extracted by the correlated person extraction means as training data. [Effects of the Invention]

[0019] According to the present invention, it is possible to provide a technology that enables others to recognize possible abnormalities in people living alone or elderly people who particularly require overt and potential monitoring, based on information on water and other lifelines. [Brief explanation of the drawings]

[0020] [Figure 1] FIG. 1 is an explanatory diagram showing an example of the configuration of an anomaly detection system. [Figure 2] 2 is a diagram illustrating an example of the hardware configuration of a terminal device 100 and a meter device 200. FIG. [Figure 3] 1 is an explanatory diagram showing an example of the functional configuration of an anomaly detection system 1. FIG. [Figure 4] This is an example of the notification content sent to a security company (email content sent to personal computer KP). DETAILED DESCRIPTION OF THE INVENTION

[0021] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. Here, an example will be described in which a meter reading contracting company provides an ancillary service of detecting abnormalities in residents of single-person homes and elderly housing. Below, an anomaly detection system will be described in which the anomaly detection program of the present invention is installed on a personal computer owned by the contracting company. Furthermore, single-person homes (whose residents are not limited to elderly people) and elderly housing will be referred to simply as homes as appropriate.

[0022] <System configuration example> FIG. 1 is an explanatory diagram showing an example of the configuration of an anomaly detection system. The anomaly detection system 1 is configured by a terminal device 100 provided at a meter reading contracting company CM, a meter device 200 provided at a residence J, and a network N. House J also receives water, gas, and electricity supplies from a waterworks bureau W, a gas company G, and an electric company D, respectively. There are many houses J (J1, J2, J3, ...), but in this embodiment, house J1 will be described as a representative house J. When the meter reading contracting company CM detects an abnormality at house J, it notifies pre-registered contacts, and in this case, the notification destinations are assumed to be the mobile phone M of the child of the resident of house J, a personal computer CP at the local community center, and a personal computer KP at the security company.

[0023] <Configuration of terminal device> The terminal device 100 can be a general-purpose personal computer, so its external configuration will be omitted and only the hardware configuration will be described. FIG. 2 is a diagram illustrating an example of the hardware configuration of the terminal device 100 and the meter device 200. As shown in FIG. The terminal device 100 has, as its hardware configuration, a CPU 101, a ROM 102, a RAM 103, a hard disk (HD) 104, a graphics board 105, an LCD monitor 106, a keyboard (K / B) 107, a network interface 108, and a mouse 109.

[0024] The CPU 101, together with the OS, controls the entire terminal device 100, evaluates the degree of abnormality in each residence J based on the usage history of water, electricity, and gas, and also performs processing to notify the associated mobile phone M and personal computers CP and KP. In addition, the CPU 101 also controls the temporary storage of work data stored on the hard disk 104 in the RAM 103 .

[0025] The ROM 102 stores a boot program and the like. Depending on the mode of use, the ROM 102 may also store a control program for the terminal device 100. The RAM 103 is used as a work area for the CPU 101. Specifically, it temporarily stores the contents of data read from the hard disk 104, the contents of programs, and the like.

[0026] The graphics board 105 sends an image signal to be output to the liquid crystal monitor 106. The graphics board 105 includes a GPU and an image output interface (image output I / F), and outputs an image processed by the GPU to the liquid crystal monitor 106.

[0027] The network interface 108 connects the terminal device 100 to the network N via Wi-Fi or a 4G line.

[0028] The hard disk 104 is made up of an application section 110 and a data storage section 130 .

[0029] The application unit 110 includes an OS 111 that controls the entire terminal device 100, and an evaluation notification program 112 that evaluates the degree of abnormality and issues a predetermined notification. The evaluation notification program 112 includes an input storage program 121 that inputs the ID and usage information from each NCU 205 (described later) and stores it in the data storage unit 130, a model generation program 122 that uses machine learning to generate an evaluation model for each residence J that evaluates the degree of abnormality in the use of water, gas, and electricity at that residence J using the history of past usage information as training data, an evaluation program 123 that evaluates the degree of abnormality for each residence J via the evaluation model based on usage information for a fixed period in the immediate vicinity, and a notification program 124 that notifies the mobile phone M and / or personal computer CP and / or personal computer KP of the degree of abnormality. The usage information is composed of water usage information, gas usage information, and electricity usage information, with the water usage information consisting of the amount of water used and the time of use, the gas usage information consisting of the amount of gas used and the time of gas use, and the electricity usage information consisting of the amount of electricity used at each time. The input storage program 121 also inputs basic information that characterizes the occupants of each house J and stores it in the data storage unit 130. Here, the basic information refers to additional information that is used when constructing an evaluation model to increase the accuracy of anomaly evaluations using the evaluation model, and examples of this information include the occupants' age, sex, height, weight, medical history, breakfast start time, lunch start time, dinner start time, bedtime, wake-up time, and sleep duration. Furthermore, when generating an evaluation model, the model generation program 122 uses, in addition to the basic information, the usage information of other residents who have similar basic information, thereby generating an evaluation model with higher accuracy. The functional configurations described below are realized by the OS 111 or the evaluation notification program 112 alone or in combination, and in some cases in cooperation with the data storage unit 130 .

[0030] The data storage unit 130 has a usage information storage unit 131 that stores usage information for each residence J, and a basic information storage unit 132 that similarly stores basic information for each residence J. The data storage unit 130 also has an evaluation model storage unit 133 that stores a machine-learned evaluation model for each residence J. Note that the evaluation model is re-learned from time to time (as usage information is accumulated over time) and updated to a more reliable model. For each house J, contact information in case of an emergency and other related information are also stored in the data storage unit 130.

[0031] <Configuration of meter device> The meter device 200 is made up of a water meter 201, a gas meter 202, and an electricity meter 203, each configured as a smart meter, and an NCU (Network Control Unit) 205 connected to these meters. Here, an example configuration is shown in which the water meter 201 and NCU 205 are housed in a meter box MB, and the gas meter 202 and electricity meter are separately connected to the NCU 205.

[0032] <System Functional Configuration> Next, a description will be given of the functional configuration of the anomaly detection system 1. FIG. The anomaly detection system 1 has, as its functional configuration, a usage information input unit 301, a basic information input unit 302, a correlated person extraction unit 303, an evaluation model generation unit 304, an anomaly evaluation unit 305, a notification unit 306, and a notification control unit 307.

[0033] The usage information input unit 301 sequentially inputs usage information (water usage information, gas usage information, electricity usage information) from the NCU 205 connected to the smart meter of each residence J. The functions of the usage information input unit 301 can be realized by, for example, the network interface 108, the RAM 103, the input storage program 121, the usage information storage unit 131, and the like.

[0034] The basic information input unit 302 inputs basic information about a large number of residents who are users of the anomaly detection system 1. The basic information input unit 302 can realize its functions by, for example, the network interface 108, the input storage program 121, the basic information storage unit 132, and the like.

[0035] The correlated person extraction unit 303 extracts residents with high correlation based on the basic information input by the basic information input unit 302. By weighting the information during learning, it becomes possible to build a more accurate individual evaluation model for each dwelling unit. The correlated individual extraction unit 303 can realize its functions using, for example, the basic information storage unit 132, the RAM 103, the CPU 101, the model generation program 122, and the like.

[0036] The evaluation model generation unit 304 uses the history of past usage information for each residence J as training data, takes the usage information for the most recent 96 hours and the time elapsed since the last water and gas usage as input, and generates an evaluation model by machine learning that outputs an evaluation of the degree of abnormality in the resident of the residence. At this time, the usage information of highly correlated residents extracted by the correlated person extraction unit 303 is also used as training data as appropriate to generate the evaluation model. The assessment model takes into consideration a combination of water usage information, gas usage information, and electricity usage information, and if electricity is being used as usual, but water or gas has not been used for an unusual amount of time since the last time it was used, it can be determined that something unusual is occurring within the dwelling (the output can increase over time to levels such as 20%, 40%, 70%, etc.). On the other hand, if electricity usage is also lower than usual, it can be determined that the resident is out, such as staying overnight, so the probability of an abnormality occurring is initially set to 20%, but the probability only increases gradually thereafter. The history of residents with high correlations can also be used as reference. For example, if people have similar medical histories, their lifestyle patterns may be similar, and it is possible to build a learning model by taking into account such abnormal data patterns of others. The evaluation model generation unit 304 can realize its functions by, for example, the CPU 101, RAM 103, model generation program 122, evaluation model storage unit 133, basic information storage unit 132, usage information storage unit 131, and the like.

[0037] The abnormality evaluation unit 305 uses the evaluation model generated by the evaluation model generation unit 304 to evaluate the degree of abnormality of the resident of the house based on the usage information input by the usage information input unit 301. This degree of abnormality is most objectively expressed as a percentage or numerical value as described above. The evaluation may also include adding the nature of the abnormality. For example, an evaluation may be made to infer that the person is unable to move within the dwelling, that the person has forgotten to take medication, that the faucet has been left open and there is concern about flooding, or that the laundry has piled up and the quality of life has declined. For example, the evaluation may be "high urgency: physical," "medium importance: medication," "medium importance: environment," or "low urgency: quality of life." The function of the anomaly evaluation unit 305 can be realized by, for example, the evaluation program 123, the evaluation model storage unit 133, the usage information storage unit 131, the network interface 108, and the like.

[0038] The notification unit 306 notifies the terminal device associated with the resident in advance of the degree of abnormality evaluated by the abnormality evaluation unit 305. The notification control unit 307 controls the notification unit 306 to expand or change the notification recipients based on the degree of abnormality evaluated by the abnormality evaluation unit 305. For example, if the evaluation is "high urgency: physical," notifications are sent by email or other means to the resident's children (their mobile phones M), the community center (the personal computer CP attached to it), and the security company (the personal computer KP attached to it). If the evaluation is "low urgency: quality of life," notifications are sent only to the children.

[0039] The functions of the notification unit 306 can be realized by, for example, the notification program 124, the data storage unit 130, the OS 111, the network interface 108, and the like. The notification control unit 307 can realize its functions by, for example, the notification program 124 and the OS 111.

[0040] An example of the notification content (email content sent to the personal computer KP) addressed to the security company is shown in FIG. Depending on the specifications, the meter reading contracting company CM may be security company K (or conversely, security company K may be contracted to read the meters). Note that in the above example, the meter reading work of the meter reading contracting company CM is not explained, but it goes without saying that the meter reading work can be (is) performed using the information individually input from the smart meters. Furthermore, although the system configuration has been described as being primarily comprised of a personal computer, this does not preclude a configuration in which the system is constructed by utilizing partial cloud computing as appropriate. [Industrial Applicability]

[0041] The above examples are intended for homes for single-person dwellings or homes for elderly people, but they can also be used in ordinary homes, and can detect and notify not only water leaks and electrical leaks, but also things like gas fan heaters or air conditioners being left on when the resident is away, or forgetting to stop filling the bathroom with hot water. [Explanation of symbols]

[0042] 1. Anomaly detection system 100 Terminal Device 101 CPU 103 RAM 104 Hard Disk 105 Graphics Board 108 Network Interface 110 Application Section 111 OS 112 Evaluation Notification Program 121 Input Storage Program 122 Model Generation Program 123 Evaluation Program 124 Notification Program 130 Data storage unit 131 Usage information storage unit 132 Basic information storage section 133 Evaluation model storage section 200 Meter Device 201 Water meter 202 Gas meter 203 Electricity Meter 205 NCU (Network Control Unit) 301 Usage information input section 302 Basic Information Input Section 303 Correlation Extraction Unit 304 Evaluation Model Generation Unit 305 Anomaly Evaluation Department 306 Notification Department 307 Notification control section CM Meter Reading Contractor D Electric company G Gas Company W Water Bureau J Housing K: Security company, KP: Security company <に備わるパーソナルコンピュータC: Local community center, CP: Personal computer at Community Center C M mobile phone N Network

Claims

1. An abnormality notification system that notifies the degree of abnormality of residents of a single-person residence or elderly residence based on usage information that is information including water usage amount and water usage time, an evaluation model generation means for generating an evaluation model by machine learning using a history of past usage information for the residence as training data, inputting usage information from a fixed time onward and the time elapsed since the last water usage, and outputting an evaluation of the degree of abnormality of the resident of the residence; a usage information input means for inputting water usage information from an NCU connected to a water meter of the residence; an abnormality evaluation means for evaluating the degree of abnormality of the resident of the house based on the usage information input by the usage information input means, using the evaluation model generated by the evaluation model generation means; a notification means for notifying a terminal device associated with the resident in advance of the degree of abnormality evaluated by the abnormality evaluation means; An anomaly detection system comprising:

2. The usage information includes the amount of gas used, the time of gas use, and the trend of electricity usage. The evaluation model generation means generates an evaluation model by machine learning, which adds the elapsed time since the last gas usage as an input and outputs an evaluation of the degree of abnormality of the resident of the house; 2. The anomaly detection system according to claim 1, wherein the usage information input means inputs usage information from an NCU that is also connected to a gas meter and an electricity meter of the house.

3. The anomaly detection system according to claim 1 or 2, further comprising a notification control means for controlling the notification means to expand or change the notification destinations based on the degree of anomaly evaluated by the anomaly evaluation means.

4. a basic information input means for inputting basic information that characterizes a large number of residents who are system users, such as the age, sex, height, weight, medical history, breakfast start time, lunch start time, dinner start time, bedtime, wake-up time, sleep time, and other basic information; a correlated person extraction means for extracting residents who have a high correlation based on the basic information input by the basic information input means; Equipped with The anomaly detection system according to claim 1 or 2, characterized in that the evaluation model generation means generates the evaluation model using the usage information of highly correlated residents extracted by the correlated person extraction means as training data.

5. An anomaly detection program for operating the anomaly detection system according to any one of claims 1 to 4, An anomaly detection program that causes a computer that constitutes the system to function as each of the means defined in the claims.

Images