Wandering alert device, wandering alert method, wandering alert program, artificial intelligence learning device for wandering alert device, artificial intelligence learning method for wandering alert method, and artificial intelligence learning program

Abstract

To provide a wandering alert device that notifies only protected persons who may wander. [Solution] The wandering alert device disclosed herein includes an input data receiving unit, an estimation unit, a wandering possibility determination unit, a sensor signal receiving unit, and an alert output unit. The input data receiving unit receives sleep data and basic physical data related to the protected person. The estimation unit inputs the received data into a trained artificial intelligence, causing the artificial intelligence to estimate the protected person's activities of daily living data. The wandering possibility determination unit determines whether or not the protected person is likely to wander based on the data estimated by the artificial intelligence. The sensor signal receiving unit receives a signal from a bed exit sensor that detects the protected person getting out of bed. The alert output unit outputs a wandering alert based on the received signal when the protected person continues to get out of bed for a period of time that exceeds a standard bed exit time, provided that the wandering possibility determination unit has determined that there is a possibility of wandering.

Description

[Technical Field]

[0001] The present invention relates to a wandering alert device, a wandering alert method, a wandering alert program, which detects wandering of a protected person and issues an alert, as well as an artificial intelligence learning device for the wandering alert device, an artificial intelligence learning method for the wandering alert method, and an artificial intelligence learning program. [Background technology]

[0002] There is known a technology that monitors the sleep state of a protected person such as a patient or care recipient using a sensor such as a sleep sensor, and if getting out of bed is detected, issues a "bed leaving alert" to the guardian (see Patent Document 1). There is also known a technology that determines that the person has "wandered" and issues a "wandering alert" if they do not return to a lying position within a predetermined time after getting out of bed is detected (see Patent Document 2).

[0003] However, among protected persons, there are some who are likely to wander and for whom a wandering alert should be issued if they continue to get out of bed for more than a predetermined time, while there are also some who are not likely to wander and do not require a wandering alert. Therefore, issuing a wandering alert uniformly based on the time they get out of bed has the problem of inaccuracy of the alert. If the possibility of wandering were known for each protected person, it would be possible to select in advance only those protected persons who are likely to wander as targets for the wandering alert, but there is a problem in that it is not easy to determine whether or not there is a possibility of wandering. Incidentally, Patent Document 3 by the same applicant discloses a technology for estimating daily living activity data of an occupational therapy subject using artificial intelligence based on the subject's sleep data and basic physical data. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2020-129413 [Patent Document 2] Japanese Patent Application Publication No. 09-327442 [Patent Document 3] Patent No. 6994262 Summary of the Invention [Problem to be solved by the invention]

[0005] The present invention has been made in consideration of the above problems, and aims to provide a wandering alert device, a wandering alert method, and a wandering alert program that target wandering alert notifications only to protected persons who are likely to wander.A further aim of the present invention is to provide an artificial intelligence learning device for the wandering alert device, an artificial intelligence learning method for the wandering alert method, and an artificial intelligence learning program. [Means for solving the problem]

[0006] To achieve the above object, a first aspect of the present invention is a wandering alert device comprising an input data receiving unit, an estimation unit, a wandering possibility determination unit, a sensor signal receiving unit, and an alert output unit. The input data receiving unit receives input data including sleep data relating to the sleep of the protected person and basic data relating to the body of the protected person. The estimation unit inputs the input data received by the input data receiving unit into a trained artificial intelligence, causing the artificial intelligence to calculate estimated data of the activities of daily living data relating to the protected person's activities of daily living. The wandering possibility determination unit determines whether or not the protected person is likely to wander based on the estimated data calculated by the artificial intelligence. The sensor signal receiving unit receives a signal from a bed exit sensor that detects the protected person leaving bed. The alert output unit outputs a wandering alert when the time the protected person continues to get out of bed exceeds a standard bed getting-out time based on the signal received by the sensor signal receiving unit, on the condition that the wandering possibility determination unit has determined that the protected person is likely to wander. The activities of daily living data include data on motor function and cognitive function. The sleep data includes data expressed in numerical values ​​based on data measured by a sleep sensor. The basic data includes data expressed in numerical values. The activities of daily living data includes data evaluated in a numerical scale.

[0007] According to this configuration, the trained artificial intelligence estimates the protected person's activities of daily living (ADL) related to their motor and cognitive functions based on the objective sleep data and basic data of the protected person. Using this ADL data, the wandering possibility determination unit can objectively determine whether the protected person is likely to wander. The alert output unit does not uniformly output a wandering alert when it detects that the protected person's continuous bed time has exceeded the reference bed time, but rather outputs a wandering alert only if the protected person has been determined to be "likely to wander" by the wandering possibility determination unit. This realizes a wandering alert notification only for protected persons who are likely to wander. Furthermore, because the input data and estimated data include numerical data or data that is graded, it is easy for the AI ​​to calculate the estimated data. The trained artificial intelligence may be part of the wandering alert device or may be an external device, such as one located on an external cloud server.

[0008] A second aspect of the present invention is the wandering alert device of the first aspect, further comprising a determination output unit and an alert necessity receiving unit. The determination output unit outputs that the wandering possibility determination unit has determined that there is a possibility of wandering for the protected person. The alert necessity receiving unit receives an input of whether a wandering alert is necessary for the protected person who is likely to wander, output by the determination output unit. Then, the alert output unit outputs the wandering alert with the further condition that there has been an input of a wandering alert necessity for the protected person.

[0009] According to this configuration, when the alert output unit detects that the protected person's continuous time out of bed has exceeded the standard time out of bed, it outputs a wandering alert on the condition that the wandering possibility determination unit has determined that there is a "possibility of wandering" and that the guardian or other person has input that a wandering alert is required, thereby achieving more accurate wandering alert notifications targeted only at protected persons who are likely to wander.

[0010] A third aspect of the present invention is a wandering alert device according to the first or second aspect, wherein the wandering possibility determination unit determines that there is a possibility of wandering when the estimated data regarding the motor function of the protected person exceeds the standard motor function level and the estimated data regarding the cognitive function does not reach the standard cognitive function level.

[0011] With this configuration, a protected person whose motor function is high enough to allow them to wander, but whose cognitive function is low enough to lead to wandering, can be appropriately determined to be at risk of wandering.

[0012] A fourth aspect of the present invention is a wandering alert device according to the first or second aspect, wherein the wandering possibility determination unit has another estimation unit that inputs the estimated daily living activity data into another trained artificial intelligence to estimate whether or not there is a possibility of wandering, and the result of the estimation by the other artificial intelligence is used as the result of determining whether or not there is a possibility of wandering.

[0013] According to this configuration, the trained separate AI is made to estimate the possibility of wandering from the estimated daily living activity data of the protected person, so it is possible to easily determine whether or not there is a possibility of wandering for the protected person. Note that the trained separate AI may be part of the wandering alert device of this configuration, or may be an external device, such as one placed on an external cloud server.

[0014] A fifth aspect of the present invention is the wandering alert device according to the fourth aspect, wherein the other estimation unit also causes the other artificial intelligence to estimate the standard bed-out time, and the alert output unit compares the standard bed-out time estimated by the other artificial intelligence with the time the protected person continues to stay out of bed.

[0015] According to this configuration, a separate trained artificial intelligence is used to estimate the standard bed exit time for each protected person, so that wandering alert notifications targeted only at protected persons who are likely to wander can be issued with greater accuracy, taking into account the standard bed exit time which differs for each protected person.

[0016] A sixth aspect of the present invention provides a wandering alert device comprising an input data receiving unit, an estimation unit, a sensor signal receiving unit, and an alert output unit. The input data receiving unit receives input data including sleep data related to the sleep of the protected person and basic data related to the protected person's body. The estimation unit inputs the input data received by the input data receiving unit into a trained artificial intelligence (AI) to estimate whether the protected person is likely to wander. The sensor signal receiving unit receives a signal from a bed exit sensor that detects the protected person getting out of bed. The alert output unit outputs a wandering alert when the AI ​​has estimated, based on the signal received by the sensor signal receiving unit, that the protected person's continuous bed exit time exceeds a reference bed exit time, provided that the AI ​​has estimated that the protected person is likely to wander. The sleep data includes numerical data based on data measured by the sleep sensor. The basic data also includes numerical data.

[0017] According to this configuration, the trained artificial intelligence estimates whether the protected person is likely to wander based on the sleep data and basic data, which are objective data of the protected person. When the alert output unit detects that the protected person's continuous bed time has exceeded the standard bed time, it does not output a wandering alert uniformly. Instead, it outputs a wandering alert only if the protected person is estimated by the artificial intelligence to be "likely to wander." This realizes a wandering alert notification targeting only protected persons who are likely to wander. Furthermore, because the input data includes data expressed in numerical form, it is easy for the artificial intelligence to calculate the estimated data. The trained artificial intelligence may be part of the wandering alert device of this configuration, or it may be an external device, such as one located on an external cloud server.

[0018] A seventh aspect of the present invention is the wandering alert device according to the sixth aspect, further comprising a determination output unit and an alert necessity receiving unit. The determination output unit outputs that the artificial intelligence has estimated that the protected person may have wandered. The alert necessity receiving unit receives input as to whether a wandering alert is necessary for the protected person who is likely to have wandered, as output by the determination output unit. The alert output unit outputs the wandering alert with the further condition that a wandering alert is necessary for the protected person.

[0019] According to this configuration, when the alert output unit detects that the protected person's continuous time out of bed exceeds the standard time out of bed, it outputs a wandering alert on the condition that the artificial intelligence has estimated that there is a "possibility of wandering" and that the guardian or other person has input that a wandering alert is required, thereby achieving more accurate wandering alert notifications targeted only at protected persons who are likely to wander.

[0020] According to an eighth aspect of the present invention, in the wandering alert device according to the sixth or seventh aspect, the estimation unit also causes the artificial intelligence to estimate the standard bed-out time. Also, the alert output unit compares the standard bed-out time estimated by the artificial intelligence with the time the protected person continues to be out of bed.

[0021] With this configuration, the trained artificial intelligence is able to estimate the standard bed-exit time for each protected person, so that wandering alert notifications targeted only at protected persons who are likely to wander can be achieved with greater accuracy, taking into account the standard bed-exit time which differs for each protected person.

[0022] A ninth aspect of the present invention is a wandering alert device according to any one of the first to fifth aspects, wherein the data on activities of daily living is data on items related to activities of daily living, including data evaluated on multiple levels for at least some of the evaluation items defined in the Functional Independence Assessment Method.

[0023] According to this configuration, the estimated data includes data suitable for calculation of the estimated data, so that the calculation of the estimated data can be performed with higher accuracy. Note that in this configuration, "at least a portion of the evaluation items defined in the Functional Independence Assessment Method" is sufficient as long as it is "data that can be evaluated on a multiple-level scale" and is not limited to seven levels.

[0024] A tenth aspect of the present invention is a wandering alert device according to any one of the first to ninth aspects, wherein the sleep data is data on sleep-related items including at least some of breathing, pulse, sleep time, sleep rhythm, number of times turning over in bed, number of body movements, number of times going to the toilet, and toilet time.

[0025] According to this configuration, the sleep data includes data suitable for calculating the estimated data, so that the estimated data can be calculated with higher accuracy.

[0026] An eleventh aspect of the present invention is a wandering alert device according to any one of the first to tenth aspects, wherein the basic data is data on basic items including at least some of age, sex, height, weight, medical history, level of care required, and BMI (Body Mass Index).

[0027] According to this configuration, the basic data includes data suitable for calculating the estimated data, so that the calculation of the estimated data can be performed with higher accuracy.

[0028] A twelfth aspect of the present invention is the wandering alert device according to any one of the first to eleventh aspects, wherein the sleep sensor also serves as the bed exit sensor.

[0029] According to this configuration, the external configuration for using the wandering alert device is further simplified.

[0030] A thirteenth aspect of the present invention is a wandering alert method, comprising an input data reception process, an estimation process, a wandering possibility determination process, a sensor signal reception process, and an alert output process. In the input data reception process, a computer receives input data including sleep data related to the sleep of the protected person and basic data related to the body of the protected person. In the estimation process, the computer inputs the input data received in the input data reception process into a trained artificial intelligence, causing the artificial intelligence to calculate estimated data of the activities of daily living data related to the protected person's activities of daily living. In the wandering possibility determination process, the computer determines whether or not the protected person is likely to wander based on the estimated data calculated by the artificial intelligence. In the sensor signal reception process, the computer receives a signal from a bed exit sensor that detects the protected person leaving bed. The alert output process outputs a wandering alert when the computer determines, based on the signal received by the sensor signal reception process, that the protected person continues to get out of bed for a period exceeding a standard bed getting-out time, provided that the wandering possibility determination process determines that the protected person is likely to wander.The activities of daily living data include data on motor function and cognitive function.The sleep data includes data expressed in numerical values ​​based on data measured by a sleep sensor.The basic data includes data expressed in numerical values.The activities of daily living data includes data evaluated in a numerical scale.

[0031] The method according to this configuration corresponds to the wandering alert method realized by the wandering alert device according to the first aspect. Note that the trained artificial intelligence may be part of the computer that executes the method according to this configuration, or may be an external device, such as one placed on an external cloud server.

[0032] A fourteenth aspect of the present invention is the wandering alert method according to the thirteenth aspect, further comprising a determination output process and an alert necessity receiving process. In the determination output process, the computer outputs that the wandering possibility determination process has determined that there is a possibility of wandering for the protected person. In the alert necessity receiving process, the computer receives an input of whether a wandering alert is necessary for the protected person who is likely to wander, output by the determination output process. Then, the alert output process outputs the wandering alert with the further condition that there has been an input of a wandering alert necessity for the protected person.

[0033] The method with this configuration corresponds to the wandering alert method realized by the wandering alert device according to the second aspect.

[0034] A fifteenth aspect of the present invention is a wandering alert method according to the thirteenth or fourteenth aspect, in which the wandering possibility determination process determines that there is a possibility of wandering when the estimated data regarding the motor function of the protected person exceeds the standard motor function level and the estimated data regarding the cognitive function does not reach the standard cognitive function level.

[0035] The method with this configuration corresponds to the wandering alert method realized by the wandering alert device according to the third aspect.

[0036] A 16th aspect of the present invention is a wandering alert method according to the 13th or 14th aspect, wherein the wandering possibility determination process includes another estimation process that inputs the estimated daily living activity data into another trained artificial intelligence to estimate whether or not there is a possibility of wandering, and the result of the estimation by the other artificial intelligence is used as the result of determining whether or not there is a possibility of wandering.

[0037] The method with this configuration corresponds to the wandering alert method realized by the wandering alert device according to the fourth aspect.

[0038] A seventeenth aspect of the present invention is the wandering alert method according to the sixteenth aspect, wherein the separate estimation process also causes the separate artificial intelligence to estimate the reference time to get out of bed, and the alert output process compares the reference time to get out of bed estimated by the separate artificial intelligence with the time the protected person continues to get out of bed.

[0039] The method with this configuration corresponds to the wandering alert method realized by the wandering alert device according to the fifth aspect.

[0040] An eighteenth aspect of the present invention provides a wandering alert method, comprising an input data reception process, an estimation process, a sensor signal reception process, and an alert output process. In the input data reception process, a computer receives input data including sleep data related to the sleep of the protected person and basic data related to the protected person's body. In the estimation process, the computer inputs the input data received by the input data reception process into a trained artificial intelligence (AI) to estimate whether the protected person is likely to wander. In the sensor signal reception process, the computer receives a signal from a bed exit sensor that detects the protected person getting out of bed. In the alert output process, the computer outputs a wandering alert if the AI ​​has estimated, based on the signal received by the sensor signal reception process, that the protected person's continuous bed exit time exceeds a reference bed exit time. The sleep data includes numerical data based on data measured by a sleep sensor. The basic data includes data expressed in numerical values.

[0041] The method with this configuration corresponds to the wandering alert method realized by the wandering alert device according to the sixth aspect.

[0042] A 19th aspect of the present invention is the wandering alert method according to the 18th aspect, further comprising a determination output process and an alert necessity receiving process. In the determination output process, the computer outputs that the artificial intelligence has estimated that the protected person may have wandered. In the alert necessity receiving process, the computer receives input of whether a wandering alert is necessary for the protected person who is likely to have wandered, as output by the determination output process. Then, the alert output process outputs the wandering alert with the further condition that a wandering alert is necessary for the protected person.

[0043] The method with this configuration corresponds to the wandering alert method realized by the wandering alert device according to the seventh aspect.

[0044] A twentieth aspect of the present invention is the wandering alert method according to the eighteenth or nineteenth aspect, wherein the estimation process also causes the artificial intelligence to estimate the standard time to get out of bed. The alert output process compares the standard time to get out of bed estimated by the artificial intelligence with the time the protected person continues to get out of bed.

[0045] The method with this configuration corresponds to the wandering alert method realized by the wandering alert device according to the eighth aspect.

[0046] A twenty-first aspect of the present invention is a wandering alert program which, when read by a computer, causes the computer to execute the wandering alert method according to any one of the thirteenth to twentieth aspects.

[0047] According to the program having this configuration, the wandering alert method according to any one of the thirteenth to twentieth aspects is realized by a computer.

[0048] A 22nd aspect of the present invention provides an artificial intelligence learning device for a wandering alert device, which trains the other artificial intelligence used in the wandering alert device according to the fourth aspect, and includes an additional input data receiving unit, a teacher data receiving unit, and a learning unit. The additional input data receiving unit receives, as additional input data, data on activities of daily living, including data on the motor function and cognitive function of the protected person. The teacher data receiving unit receives teacher data, which corresponds to the additional input data and includes data indicating whether the protected person is likely to wander or has wandered in the past. The learning unit inputs the additional input data received by the additional input data receiving unit and the teacher data received by the teacher data receiving unit into the other artificial intelligence, thereby training the other artificial intelligence to estimate the teacher data from the other input data. The activity of daily living data includes data evaluated in a numerical scale.

[0049] According to this configuration, the separate artificial intelligence used by the wandering alert device according to the fourth aspect is constructed by learning. Note that the separate artificial intelligence may be part of the artificial intelligence learning device according to this configuration, or may be an external device, such as one placed on an external cloud server.

[0050] A 23rd aspect of the present invention is an artificial intelligence learning device according to the 22nd aspect, wherein the teacher data also includes actual data on the standard bed leaving time for the protected person, which is used by the wandering alert device according to the 5th aspect.

[0051] According to this configuration, the different artificial intelligence used by the wandering alert device according to the fifth aspect is constructed by learning.

[0052] A 24th aspect of the present invention provides an artificial intelligence learning device for a wandering alert device, which trains the artificial intelligence used in the wandering alert device according to the 6th or 7th aspect. The device includes an input data receiving unit, a teacher data receiving unit, and a teacher data receiving unit. The input data receiving unit receives input data including sleep data related to the sleep of the protected person and basic data related to the protected person's body. The teacher data receiving unit receives teacher data corresponding to the input data, including data indicating whether the protected person is likely to wander or has previously wandered. The learning unit inputs the input data received by the input data receiving unit and the teacher data received by the teacher data receiving unit into the artificial intelligence, thereby training the artificial intelligence to estimate the teacher data from the input data. The sleep data includes numerical data based on data measured by a sleep sensor. The basic data also includes numerical data.

[0053] According to this configuration, the artificial intelligence used by the wandering alert device according to the sixth or seventh aspect is constructed by learning. Note that the artificial intelligence may be part of the artificial intelligence learning device according to this configuration, or may be an external device, such as one placed on an external cloud server.

[0054] A 25th aspect of the present invention is an artificial intelligence learning device according to the 24th aspect, wherein the teacher data also includes actual data on the standard bed-out time for the protected person, which is used by the wandering alert device according to the 8th aspect.

[0055] According to this configuration, the artificial intelligence used by the wandering alert device according to the eighth aspect is constructed by learning.

[0056] A 26th aspect of the present invention provides an artificial intelligence learning method for the wandering alert method, which trains the other artificial intelligence used in the wandering alert method according to the 16th aspect, and includes a separate input data reception process, a teacher data reception process, and a learning process. In the separate input data reception process, a computer receives, as separate input data, data on activities of daily living, including data on motor function and cognitive function. In the teacher data reception process, the computer receives teacher data, which corresponds to the separate input data, and includes data indicating whether the protected person is likely to wander or has wandered in the past. In the learning process, the computer inputs the other input data received in the separate input data reception process and the teacher data received in the teacher data reception process into the other artificial intelligence, thereby training the other artificial intelligence to estimate the teacher data from the other input data. The activity of daily living data includes data evaluated in a numerical scale.

[0057] According to this configuration, the separate artificial intelligence used in the wandering alert method according to the sixteenth aspect is constructed by learning. Note that the separate artificial intelligence may be part of the computer that executes the method of this configuration, or may be an external device, such as one placed on an external cloud server.

[0058] A 27th aspect of the present invention is an artificial intelligence learning method according to the 26th aspect, wherein the teacher data also includes actual data on the standard bed exit time for the protected person, which is used in the wandering alert method according to the 17th aspect.

[0059] According to this configuration, the separate artificial intelligence used in the wandering alert method according to the seventeenth aspect is constructed by learning.

[0060] A 28th aspect of the present invention provides an artificial intelligence learning method for a wandering alert method, which trains the artificial intelligence used in the wandering alert method according to the 18th or 19th aspect. The method includes an input data reception process, a teacher data reception process, and a learning process. In the input data reception process, a computer receives input data including sleep data related to the sleep of the protected person and basic data related to the protected person's body. In the teacher data reception process, the computer receives teacher data corresponding to the input data, including data indicating whether the protected person is likely to wander or has previously wandered. In the learning process, the computer inputs the input data received by the input data reception process and the teacher data received by the teacher data reception process into the artificial intelligence, thereby training the artificial intelligence to estimate the teacher data from the input data. The sleep data includes numerical data based on data measured by a sleep sensor. The basic data also includes numerical data.

[0061] According to this configuration, the artificial intelligence used in the wandering alert method according to the 18th or 19th aspect is constructed by learning. Note that the artificial intelligence may be part of a computer that executes the method of this configuration, or may be an external device, such as one placed on an external cloud server.

[0062] A 29th aspect of the present invention is an artificial intelligence learning method according to the 28th aspect, wherein the training data also includes actual data on the standard bed-out time for the protected person, which is used in the wandering alert method according to the 20th aspect.

[0063] According to this configuration, the artificial intelligence used in the wandering alert method according to the twentieth aspect is constructed by learning.

[0064] A 30th aspect of the present invention is an artificial intelligence learning program which, when read by a computer, causes the computer to execute the artificial intelligence learning method according to any one of the 26th to 29th aspects.

[0065] According to the program with this configuration, the artificial intelligence learning method according to any one of the 26th to 29th aspects is realized by a computer. [Effects of the Invention]

[0066] As described above, the present invention provides a wandering alert device, a wandering alert method, and a wandering alert program that send a wandering alert only to protected persons who are likely to wander.The present invention also provides an artificial intelligence learning device for the wandering alert device, an artificial intelligence learning method for the wandering alert method, and an artificial intelligence learning program. [Brief explanation of the drawings]

[0067] [Figure 1] FIG. 1 is a schematic explanatory diagram illustrating an example of an outline of data processing by a wandering alert device according to an embodiment of the present invention. [Figure 2] 1 is a diagram illustrating the configuration of a wandering alert system including a wandering alert device according to an embodiment of the present invention; [Figure 3] 3 is a block diagram illustrating the configuration of the wandering alert device of FIG. 2. FIG. [Figure 4] 4 is a flowchart illustrating the process flow of a wandering alert method implemented by the wandering alert device illustrated in FIG. 3. [Figure 5] FIG. 4 is a table showing an example of input data and output data of the artificial intelligence used by the wandering alert device shown in FIG. 3. [Figure 6] FIG. 4 is a schematic diagram illustrating the conceptual configuration of artificial intelligence used by the wandering alert device illustrated in FIG. 3. [Figure 7] FIG. 10 is a block diagram illustrating the configuration of a wandering alert device according to another embodiment of the present invention. [Figure 8]8 is a flowchart illustrating the process flow of a wandering alert method implemented by the wandering alert device illustrated in FIG. 7. [Figure 9] FIG. 10 is a block diagram illustrating the configuration of a wandering alert device according to yet another embodiment of the present invention. [Figure 10] 10 is a flowchart illustrating the process flow of a wandering alert method implemented by the wandering alert device illustrated in FIG. 9. [Figure 11] 10 is a flowchart illustrating the flow of processing of an artificial intelligence learning method implemented by the wandering alert device illustrated in FIGS. 3, 7, and 9. [Figure 12] 1 is a screen diagram showing an example of an image displayed on the screen of a computer 10. FIG. DETAILED DESCRIPTION OF THE INVENTION

[0068] (Outline of the embodiment) 1 is a schematic explanatory diagram illustrating an example of an outline of data processing by a wandering alert device according to an embodiment of the present invention. The wandering alert device generally operates as follows.

[0069] Based on sleep data and basic data of a protected person such as a patient or care recipient, artificial intelligence (AI) estimates daily living activity data related to motor and cognitive functions. Sleep data is acquired by a sleep sensor and includes, for example, heart rate and respiratory rate. Basic data includes, for example, age, gender, and BMI. Based on the estimated daily living activity data, a rule-based or AI-based determination is made as to whether the protected person is a target for a wandering alert.

[0070] When a sleep sensor detects that a protected person has gotten out of bed, and the time spent getting out of bed exceeds the standard time, a wandering alert is output, provided that the protected person has been determined to be an alert target. The determined alert target is presented to the guardian who is providing care and nursing, and after confirmation, the alert target can be finally determined (processing indicated by the dotted line in the figure). If artificial intelligence is used to determine the alert target, it is also possible to estimate the appropriate standard time to get out of bed for each protected person (processing indicated by the dotted line in the figure).

[0071] Furthermore, it is also possible for a single AI system to directly determine whether or not a person is eligible for an alert based on sleep data and basic data (the process enclosed by the thin two-dot chain line in the diagram). In this case, it is also possible to estimate the appropriate standard time to get out of bed for each protected person (the process enclosed by the dotted line in the diagram).

[0072] (First embodiment) 2 is a diagram illustrating the configuration of a wandering alert system including a wandering alert device according to one embodiment of the present invention. In addition to wandering alert device 101, this wandering alert system 100 includes sleep sensor 1, network 5, and servers 7 and 9. Sleep sensor 1, network 5, and servers 7 and 9 are devices connected to wandering alert device 101 and cooperate with it.

[0073] The wandering alert device 101 is a device that contributes to appropriately protecting wandering protected persons 11 by outputting a wandering alert based on sleep data and basic data of the protected person 11, such as a patient or care recipient, who is the target of the wandering alert. In the illustrated example, the wandering alert device 101 is incorporated into a computer 10. That is, by installing and activating a specific application in the computer 10, a processing device (processor) such as a central processing unit (CPU) of the computer 10 functions as the wandering alert device 101.

[0074] In the illustrated example, it is assumed that the protected persons 11 are many care recipients living in nursing facilities such as nursing homes, and that the computer 10 is managed by the nursing facility. Also, in the illustrated example, it is assumed that the computer 10 is a personal computer (PC), but it may also be a small portable terminal such as a smartphone. Alternatively, the functions of the computer 10 may be shared between a PC and a smartphone.

[0075] The sleep sensor 1 is a sensor that automatically acquires sleep data of the protected person 11 and has a communication function to transmit the acquired data wirelessly or otherwise. The sleep data is data related to the sleep of the subject 11, such as the sleeping time, the number of times the subject turns over in sleep, breathing while sleeping, pulse rate, etc. The sleep sensor 1 also functions as a bed exit sensor that detects when the protected person 11 leaves bed. In the illustrated example, the sleep sensor 1 is a mat-shaped sensor that is placed under the bedding on which the protected person 11 lies. Sleep sensors 1 of this type are already commercially available and well known. Each item of sleep data will be described later.

[0076] The sleep sensor 1 is connected to, for example, a Wi-Fi router 3. The Wi-Fi router 3 can connect the sleep sensors 1 used by many protected persons 11 to the wandering alert device 101 via a LAN (local area network; for example, wireless LAN) within the facility, or to an external network 5. In the latter case, the wandering alert device 101 can receive the sleep data transmitted by the sleep sensor 1 via the network 5. In the illustrated example, the network 5 is the Internet. If the sleep sensor 1 has the functionality of the Wi-Fi router 3 built in, for example, it can also be connected directly to the network 5.

[0077] The server 7 is owned by a facility such as a hospital and is connected to the network 5, and stores basic data such as the medical records of the protected person 11. The server 7 may also be owned by an external provider and used by a facility such as a hospital. The wandering alert device 101 can obtain basic data such as the age and medical history of the protected person 11 by accessing the server 7. Information leakage can be prevented by requiring the input of, for example, an identification code (ID) and password in communication between the wandering alert device 101 and the server 7. The basic data may be input manually by the administrator or operator of the wandering alert device 101, rather than through communication with the server 7. Each item of basic data will be described later.

[0078] The server 9 is connected to the network 5 and has constructed an artificial intelligence that can be used through the network 5. The wandering alert device 101 uses the artificial intelligence to estimate daily living activity data based on the sleep data and basic data of the protected person 11. The daily living activity data is data on items related to daily living activities such as eating, using the toilet, defecation, transferring to the toilet, moving and walking, understanding, problem solving, and memory. Each item of daily living data will be described later.

[0079] The artificial intelligence may be built into the computer 10 as part of the wandering alert device 101, or may be built into the computer 10 separately from the wandering alert device 101 so that it is accessible by the wandering alert device 101, or may be artificial intelligence external to the computer 10, such as artificial intelligence provided by the server 9. Note that, even in communications between the wandering alert device 101 and the server 9, information leakage can be prevented by requiring the input of, for example, an identification code (ID) or password.

[0080] Fig. 3 is a block diagram illustrating the configuration of the wandering alert device 101. Fig. 4 is a flowchart illustrating the processing flow of the wandering alert method implemented by the wandering alert device 101. The wandering alert device 101 has an interface 13, an input data receiving unit 15, a teacher data receiving unit 17, an estimation unit 19, a learning unit 21, an artificial intelligence 23, a wandering possibility determination unit 41, a sensor signal receiving unit 43, and an alert output unit 45. The wandering alert device 101 may further have a determination output unit 47 and an alert necessity receiving unit 49.

[0081] The interface 13 is a device part that enables communication between the wandering alert device 101 itself and external devices according to a predetermined protocol for each external device. Communication between the wandering alert device 101 and the sleep sensor 1, Wi-Fi router 3, servers 7 and 9, input device 27 such as a keyboard, output device 29 such as a printer or display, and storage medium 31 such as a USB memory or CD-ROM is performed through the interface 13.

[0082] The input data receiving unit 15 receives input data including sleep data of the protected person 11 and basic data of the protected person 11 (step S1). The estimation unit 19 inputs the input data received by the input data receiving unit 15 to the artificial intelligence 23, causing the artificial intelligence 23 to calculate estimated data on the daily living activity data of the protected person 11 (step S3). The daily living activity data includes data on motor function and data on cognitive function. If the artificial intelligence 23 has already been trained, it outputs estimated data with high accuracy on the daily living activity data.

[0083] The wandering possibility determination unit 41 determines whether or not there is a possibility that the protected person 11 has wandered, based on the estimated data calculated by the artificial intelligence 23 (step S5). The processing of step S5 is performed, for example, through steps S21, S23, and S25. In step S21, it is determined whether or not the data on motor function, among the data on activities of daily living estimated by the artificial intelligence 23, exceeds the reference motor function level and the data on cognitive function does not reach the reference cognitive function level. When the data on motor function includes multiple items, for example, a simple average value of those items or a weighted average value taking into account the influence of each item is compared with the reference motor function level. The same applies to the data on cognitive function. If the result of the determination is "No," it is determined that there is no possibility of wandering (step S23), and if it is "Yes," it is determined that there is a possibility of wandering (step S25).

[0084] By appropriately setting the reference motor function level and reference cognitive function level, it is possible to appropriately determine that a protected person 11 whose motor function is high enough to allow them to wander and whose cognitive function is low enough to lead them to wander is likely to wander. The reference motor function level and reference cognitive function level are input, for example, by the input device 27, and are stored in a memory (not shown) built into the wandering alert device 101.

[0085] If there is new input data (Yes in step S11), the process returns to step S1. If there is no new input data (No in step S11), the process proceeds to step S13. In step S13, the sensor signal receiving unit 43 receives a signal from a bed exit sensor that detects the protected person 11 getting out of bed. In the illustrated example, the sleep sensor 1 also serves as the bed exit sensor. On the other hand, it is also possible to use a bed exit sensor separate from the sleep sensor 1 to detect getting out of bed.

[0086] In step S15, if the time that the protected person 11 continues to get out of bed exceeds the reference bed getting out time 46 based on the signal received by the sensor signal receiving unit 43 (Yes in step S41), the alert output unit 45 outputs a wandering alert (step S45) on the condition that the wandering possibility determination unit 41 has determined that there is a possibility that the protected person 11 has wandered (Yes in step S43). The reference bed getting out time 46 is input by, for example, the input device 27, and is held in a memory (not shown) built into the wandering alert device 101.

[0087] The wandering alert output by the alert output unit 45 is transmitted to, for example, the output device 29 via the interface 13. This allows the guardian who is caring for or nursing the protected person 11 in the facility to recognize that the protected person 11 has begun to wander or is highly suspected of doing so, and makes it possible to appropriately protect the protected person 11.

[0088] The alert output unit 45 can also incorporate the guardian's judgment as to whether or not a wandering alert is required for the protected person 11. The judgment output unit 47 and the alert necessity receiving unit 49 are provided for this purpose. That is, the judgment output unit 47 outputs that the wandering possibility judgment unit 41 has judged that the protected person 11 is "possibly wandering" (step S7). The alert necessity receiving unit 49 receives an input from the guardian as to whether or not a wandering alert is required for the protected person 11 who is likely to wander and has been output by the judgment output unit 47 (step S9). Then, the alert output unit 45 outputs a wandering alert when the time that the protected person 11 continues to get out of bed exceeds the standard bed getting out time 46, with the further condition that the wandering possibility determination unit 41 has determined that the protected person 11 is likely to wander and that a "wandering alert required" has been input for the protected person 11 (steps S41, 43, 45). As a result, a wandering alert can be issued with higher accuracy only to the protected person 11 who is likely to wander.

[0089] If the process should not be ended (No in step S17), the process returns to step S11. If there is no new input data (No in step S11), the sensor signal receiving unit 43 receives a signal from the bed exit sensor again (step S13). If there is new input data (Yes in step S11), the input data accepting unit 15 accepts the new input data (step S1). If the process should be ended (Yes in step S17), the process ends.

[0090] By undergoing machine learning, the artificial intelligence 23 is able to output highly accurate estimation data. The wandering alert device 101 has a teacher data receiving unit 17 and a learning unit 21, which makes it possible for the wandering alert device 101 to train the artificial intelligence 23 by itself, without using an external artificial intelligence learning device. In other words, the wandering alert device 101 also has a built-in artificial intelligence learning device that trains the artificial intelligence 23 through machine learning.

[0091] When wandering alert device 101 performs machine learning, input data receiving unit 15 receives input data including sleep data and basic data, and teacher data receiving unit 17 receives teacher data, which is correct daily living activity data corresponding to the input data. Learning unit 21 inputs the input data received by input data receiving unit 15 and the teacher data received by teacher data receiving unit 17 to artificial intelligence 23, thereby training artificial intelligence 23 to estimate the teacher data from the input data. By inputting a large number of pairs of input data and teacher data that are associated with each other into wandering alert device 101, learning of artificial intelligence 23 progresses and the accuracy of estimation improves.

[0092] In the past, sleep data and basic data collected for various protected persons 11, as well as data on activities of daily living obtained by actual measurements corresponding to these data, were associated with each other and recorded, for example, in storage medium 31. By doing so, input data receiving unit 15 and teacher data receiving unit 17 can sequentially read out a large amount of data required for learning from storage medium 31, and learning unit 21 can repeatedly learn artificial intelligence 23 for each piece of read data. In this way, wandering alert device 101 can switch between two operating modes: an estimation mode in which estimated data is calculated and output using artificial intelligence 23, and a learning mode in which artificial intelligence 23 undergoes machine learning. The switching of operating modes can be instructed, for example, by input device 27.

[0093] In the example of FIG. 3, the artificial intelligence 23 is incorporated into the computer 10 as part of the wandering alert device 101. Alternatively, an artificial intelligence built on an external server 9, etc., may be used, as illustrated by the dotted line in FIG. 3. In this case, the estimation unit 19 and the learning unit 21 operate the external artificial intelligence via the network 5, etc. The estimation unit 19 inputs the input data received by the input data receiving unit 15 to the external trained artificial intelligence, causing it to output estimated data. The estimated data is sent to the wandering possibility determination unit 41, for example, via the interface 13. Furthermore, the learning unit 21 inputs the input data received by the input data receiving unit 15 and the teacher data received by the teacher data receiving unit 17 to the external artificial intelligence, thereby training the external artificial intelligence to estimate the teacher data from the input data. When an external artificial intelligence is used, the artificial intelligence 23, which is part of the wandering alert device 101, is unnecessary.

[0094] As mentioned above, the computer 10 incorporating the wandering alert device 101 may be a small portable terminal (not shown) such as a smartphone, or some of the functions may be shared by the small portable terminal. For example, the function of executing a learning mode that trains the artificial intelligence 23 through machine learning may be shared by a PC, and the functions of the wandering alert device 101 other than the learning function may be shared by the small portable terminal. In this case, the guardian can understand the status of the protected person 11 by the wandering alert displayed on the screen of the small portable terminal he or she carries or simultaneously output as audio. When the functions of the computer 10 are shared by the small portable terminal, the hardware and processing load of the small portable terminal can be further reduced by using an external artificial intelligence built in the server 9, etc., instead of the artificial intelligence 23.

[0095] FIG. 5 is a table showing examples of input data and output data of the artificial intelligence 23 used by the wandering alert device 101. FIG. 5(a) shows examples of sleep data and environmental data during sleep, FIG. 5(b) shows basic data, and FIG. 5(c) shows data on activities of daily living. The sleep data, environmental data, and basic data are provided as input data to the artificial intelligence 23. The data on activities of daily living is output data from the artificial intelligence 23. Below, an example is given of how each piece of data can be expressed for handling by the wandering alert device 101. It is obvious that this is merely an example, and other ways of expressing it are also possible.

[0096] Among the sleep data (see Figure 5(a)), sleep time and toilet time are expressed in hours (h), such as 6.5. Sleep rhythms are expressed as a time series of changes in sleep and wakefulness, and are expressed as a data string indicating whether a person is asleep or awake at 15-minute intervals from the time they lie down until, say, nine hours later (awake, wakeful, sleep, sleep, sleep, sleep, wakeful, wakeful, sleep, ...). This also allows us to determine the time it takes to fall asleep after lying down, which is an indicator of how well or poorly a person falls asleep. "Sleep" and "wakefulness" are represented by pre-assigned codes, such as "1" and "0." The number of times people turn over in bed, the number of body movements, and the number of times they toilet are represented by natural numbers, such as 1, 2, and 3. The number of body movements refers to the number of movements performed while lying down, excluding turning over in bed, such as moving their legs or reaching out from under the covers. The number of times they toilet refers to the number of times they leave bed to use the toilet during their sleep time. Respiration and pulse are expressed as the number of times per minute.

[0097] The environmental data, room temperature, humidity, and illuminance, are expressed as numerical values ​​based on the units of temperature, humidity, and illuminance, respectively. The sleep data and environmental data described above are acquired by sleep sensor 1. Alternatively, the sleep data may be generated by an application on computer 10 (for example, by estimation unit 19). That is, data such as lying down, sleeping, waking, turning over, and body movements are generated by sleep sensor 1 itself or an application from pressure changes, heart rate, respiratory rate, and the like sensed by sleep sensor 1. Even if it is not possible to identify which part of the body has moved, it is possible to detect that the body movement is not turning over.

[0098] Of the subjective evaluation data included in the sleep data, sleepiness is represented by one of the options "slept well," "can't say," or "didn't sleep well," and each option is represented by a pre-assigned code, such as the numbers "1," "2," or "3." Alternatively, each option may be represented by a code corresponding to selection or non-selection, such as a code "1" or "0." Fatigue is represented by, for example, "feeling refreshed," "normal," or "tired." These subjective evaluation data are input, for example, by a guardian in charge of nursing or care, who asks the protected person 11 about the condition and operates the input device 27.

[0099] The bedding in the environmental data is photographic data of the bedding. The photographic data is obtained, for example, by a guardian taking a photograph using a camera. The image data obtained by photographing is input to the wandering alert device 101, for example, via the memory 31. The image data is expressed by a set of pixel values.

[0100] The basic data (see FIG. 5(b)) is obtained, for example, from the hospital server 7. Alternatively, it may be entered into the computer 10, for example, manually. Of the basic data, age, height, and weight are represented by numerical values ​​based on those units. Gender is represented by a code corresponding to male or female, for example, a numerical value such as "0" or "1". Medical history is represented by a code assigned in advance to various disease names, for example, a numerical value such as "0", "1", "2", etc. Alternatively, each disease name may be represented by a code corresponding to "absent" or "present", for example, a numerical value such as "0", "1". The level of care required indicates the level of care required, and is represented by a numerical value on an eight-level scale, for example.

[0101] The example activities of daily living data (see Figure 5(c)) includes 18 items based on the Functional Independence Measure (FIM), a known effective tool for occupational therapy evaluation. The extent to which the protected person 11 can perform each activity independently is assessed on a seven-point scale (1 to 7). The assessment consists of cognitive and motor items. The cognitive items include five items: understanding, expression, social interaction, problem solving, and memory. The motor items include the remaining 13 items: eating, grooming, wiping, dressing, toileting, bladder management, bowel management, transfers (activities of getting up and down), and mobility. Each item is expressed as a numerical value corresponding to a score. The activities of daily living data also include "fall risk." Fall risk assesses the likelihood of falling and is expressed as a two-level value, for example, "0" or "1," corresponding to "high probability" and "low probability," respectively, or as a number of more detailed values, such as "0," "1," "2," and "3." Each item of the FIM may also be expressed using a number other than seven levels, for example, three levels.

[0102] For activity of daily living data expressed as discontinuous numerical values ​​in seven levels, for example, the wandering possibility determination unit 41 converts the estimated value calculated by the artificial intelligence 23 to the nearest numerical value among the seven levels, for example by rounding it off. Since the converted data is based on the estimated data of activity of daily living data calculated by the artificial intelligence 23, it is still estimated data of activity of daily living data.

[0103] One of the present inventors, with many years of experience as an occupational therapist, has completed an invention to estimate activity of daily living data based on sleep data and basic data, which are objective data independent of the evaluator, in order to solve the problem of obtaining an objective occupational therapy evaluation that is independent of the evaluator (see Patent Document 3). It was predicted that there would be a complex correlation between a set of sleep data and basic data and activity of daily living data. Therefore, the inventor believed that, even if it would be too much of a burden and unrealistic for a human to perform it by itself, it would be possible in principle to estimate activity of daily living data based on sleep data and basic data. The inventor then realized that obtaining such estimated data for occupational therapy evaluation, which exceeds human intellect, could be made practical by using artificial intelligence, and completed the invention (see paragraph 0046 of the specification of Patent Document 3). The present invention utilizes the above-mentioned patented invention by the same applicant to obtain activity of daily living data for a protected person 11.

[0104] The sleep data preferably includes at least some of the following data obtained by the sleep sensor 1: respiration, heart rate (used as a concept including pulse), sleep time, sleep rhythm, number of turns in sleep, number of body movements, number of toilet visits, and toilet time. In particular, heart rate and respiration rate are desirable features for making estimations using the artificial intelligence 23. In addition, the basic data preferably includes at least some of age, sex, height, weight, medical history, level of care required, and BMI (Body Mass Index). In particular, age, sex, and BMI are desirable features for making estimations using the artificial intelligence 23. The more items included in both the sleep data and the basic data, the more accurately the estimated data of daily living activity data can be obtained.

[0105] Of the data on activities of daily living, the data on motor function desirably includes at least a part of the 13 motor function items defined in the FIM, and the data on cognitive function desirably includes at least a part of the 5 cognitive function items defined in the FIM. The more items included in both the data on motor function and the data on cognitive function, the more accurate the determination by the wandering possibility determination unit 41 can be.

[0106] FIG. 6 is a schematic diagram illustrating the conceptual configuration of the artificial intelligence used by the wandering alert device 101. The artificial intelligence provided by the server 9 also has a similar configuration, for example. The illustrated artificial intelligence 23 is a neural network, and has an input layer 33 where nodes that receive data input are arranged, an output layer 37 where nodes that output data resulting from calculations are arranged, and an intermediate layer 35 where nodes connecting the input layer 33 and the output layer 37 are arranged. In the illustrated example, there is only one intermediate layer 35, but there may be multiple layers. The value of the previous node is transmitted to the next node, reflecting the parameters assigned to each node, i.e., the weight and bias value of each node.

[0107] The input layer 33 receives input data accepted by the input data accepting unit 15, i.e., pairs of sleep data and basic data items. The input data is transmitted to the output layer 37 via the intermediate layer 35 while reflecting the parameters of each node. The data transmitted to the output layer 37 becomes estimated data for pairs of activity of daily living data items. The estimation unit 19 (see Figure 3) inputs pairs of sleep data and basic data items of the protected person 11 to the input layer 33 of the artificial intelligence 23, and causes the output layer 37 to generate estimated data for the activity of daily living data of the protected person 11. The wandering possibility determination unit 41 uses the generated estimated data to determine the possibility of wandering, with or without any conversion, such as rounding.

[0108] In order for the estimated data appearing in the nodes of the output layer 37 to be highly accurate estimates of the activity of daily living data, it is necessary to train the artificial intelligence 23 using actually measured activity of daily living data. Learning is performed by inputting a set of sleep data and basic data items of a certain protected person 11, the input of which is accepted by the input data accepting unit 15, to the input layer 33, and inputting teacher data for the same protected person 11, i.e., a set of actually measured activity of daily living data items, accepted by the teacher data accepting unit 17, to the output layer 37 as teacher data. The learning unit 21 (see FIG. 3 ) inputs such data to the artificial intelligence 23.

[0109] The artificial intelligence 23 calculates estimated data of the activities of daily living data based on the input sleep data and basic data, generates the estimated data in the output layer 37, and calculates the error between the generated estimated data and the activities of daily living data input as training data. The artificial intelligence 23 then changes the parameters of each node from the output layer 37 to the input layer 33, for example, using a well-known error backpropagation algorithm, so that error-free estimated data is generated. This function is provided by the artificial intelligence 23 itself. By preparing many pairs of input data and training data and repeating learning, the artificial intelligence 23 can generate highly accurate estimated data. When training the artificial intelligence 23, it is also possible to adjust the number of intermediate layers 35 and the number of nodes in each layer to optimal values. Such techniques are also well known.

[0110] To obtain estimated data on the activities of daily living of the protected person 11, not only can the latest data be input to the artificial intelligence 23 for the sleep data and basic data of the protected person 11, but also data from multiple time points, including earlier data, can be input to the artificial intelligence 23 along with the respective time data. This allows for the estimation of the activities of daily living of the same protected person 11, taking into account the past history of the protected person 11's sleep data and basic data. This allows for more accurate estimation. The time data may be expressed, for example, as the date and time of each time point, or as the date and time difference from the latest time point. The input data receiving unit 15 receives data from multiple time points along with the respective time data, and the estimation unit 19 inputs the data from the multiple time points and the respective time data to the input layer 33 of the artificial intelligence 23. The greater the number of time points of the input data, the greater the number of nodes in the input layer 33 that receive the data input.

[0111] To obtain estimated data based on data from multiple time points, the artificial intelligence 23 must be trained using data from multiple time points, the respective time data, and the corresponding training data. For example, to obtain estimated data on activities of daily living using sleep data and basic data from three past time points, including the most recent time point, the sleep data and basic data from the three time points for various protected persons 11 and the respective time data can be input to the input layer 33, and the most recent actual measurement data of activities of daily living for each protected person 11 can be input to the output layer 37, thereby training the artificial intelligence 23. Because the time data is input simultaneously, the multiple time points at which the sleep data and basic data are collected may differ between different protected persons 11. For example, for one protected person 11, data from the most recent time point, one week ago, and five weeks ago may be input, while for another subject, data from the most recent time point, three weeks ago, and 15 weeks ago may be input. By learning from a large amount of data, the artificial intelligence 23 adjusts node parameters so as to calculate estimated data that also reflects the influence of the temporal distance from the most recent time point.

[0112] 6, other types of artificial intelligence may be used as the artificial intelligence 23, such as a decision tree-based LGBM (Light GBM; manufactured by Microsoft). LGBM has the advantage of being able to easily analyze which variables in the input data play an important role in the estimated data to be output, and is therefore particularly useful in the process of building the artificial intelligence 23.

[0113] (Second embodiment) Fig. 7 is a block diagram illustrating the configuration of a wandering alert device according to another embodiment of the present invention. Fig. 8 is a flowchart illustrating the processing flow of the wandering alert method implemented by this wandering alert device 102. The wandering alert device 102 differs from the wandering alert device 101 (see Figs. 3 and 4) in that the wandering possibility determination unit 51 determines the possibility of wandering using artificial intelligence (steps S51 and S53). In the illustrated example, the wandering possibility determination unit 51 has an estimation unit 53, artificial intelligence 55, and a learning unit 57.

[0114] The estimated data calculated by the artificial intelligence 23, i.e., the estimated data on the daily living activity data of the protected person 11, is read by the estimation unit 53. The estimation unit 53 inputs the read-out estimated data of the artificial intelligence 23 into the artificial intelligence 55, causing the artificial intelligence 55 to calculate estimated data on whether or not the protected person 11 is likely to wander. The estimation unit 53 performs conversions such as rounding on the read-out estimated data of the artificial intelligence 23, as with the wandering possibility determination unit 41 (see FIG. 3), or inputs the estimated data to the artificial intelligence 55 without any conversion. If the artificial intelligence 55 has already been trained, it will calculate estimated data with a high degree of accuracy on whether or not the protected person is likely to wander. For example, "no possibility of wandering" can be represented by the number "0", and "possibility of wandering" can be represented by the number "1". The wandering possibility determination unit 51 inputs the estimated data calculated by the artificial intelligence 55, after or without conversion such as rounding, to the alert output unit 45 as a determination result as to the possibility of the protected person 11 wandering. In this way, the trained artificial intelligence 55 is made to estimate whether or not there is a possibility of wandering, so it can be easily determined whether or not there is a possibility of the protected person 11 wandering.

[0115] The estimation unit 53 may input the read-out estimation data of the artificial intelligence 23 to the artificial intelligence 55, thereby causing the artificial intelligence 55 to calculate estimated data on the reference bed getting-out time 46 for the protected person 11, along with estimated data on whether or not the protected person 11 is likely to wander (see the description in parentheses in step S53). In this case, the wandering possibility determination unit 51 also inputs the estimated data on the reference bed getting-out time 46 calculated by the artificial intelligence 55 to the alert output unit 45. The alert output unit 45 records the input estimated data on the reference bed getting-out time 46 in an internal memory (not shown) or the like as the reference bed getting-out time 46 for the protected person 11, and uses it for comparison with the bed getting-out time of the protected person 11 (step S41). In this way, since the reference bed getting-out time 46 set for each protected person 11 is used, a wandering alert notification targeted only at protected persons who are likely to wander can be realized with higher accuracy.

[0116] Like the artificial intelligence 23, the artificial intelligence 55 is able to output highly accurate estimation data by undergoing machine learning. The wandering alert device 102 has an input data receiving unit 16, a teacher data receiving unit 18, and a learning unit 57, which makes it possible for the wandering alert device 102 to train the artificial intelligence 55 by itself, without using an external artificial intelligence learning device. In other words, the wandering alert device 102 also has a built-in artificial intelligence learning device that trains the artificial intelligence 55 through machine learning.

[0117] When the wandering alert device 102 causes the artificial intelligence 55 to perform machine learning, the input data receiving unit 16 receives data on activities of daily living related to the motor and cognitive functions of the protected person 11, and the teacher data receiving unit 18 receives input of teacher data corresponding to the input data, including data indicating whether the protected person 11 is likely to wander or whether they have actually wandered. The teacher data may also include an actual value of the standard time out of bed 46. The learning unit 57 inputs the input data received by the input data receiving unit 16 and the teacher data received by the teacher data receiving unit 18 to the artificial intelligence 55, thereby training the artificial intelligence 55 to estimate the teacher data from the input data. By inputting a large number of pairs of input data and teacher data that are associated with each other into the wandering alert device 102, the learning of the artificial intelligence 55 progresses and the accuracy of estimation improves.

[0118] In the past, data on activities of daily living related to motor and cognitive functions collected for various protected persons 11, and data on the likelihood or history of wandering for the protected persons 11, are associated with each other and recorded, for example, in storage medium 31. This allows the input data receiving unit 16 and the teacher data receiving unit 18 to sequentially read out a large amount of data required for learning from storage medium 31, and the learning unit 57 to repeatedly learn the artificial intelligence 55 for each piece of read data. In this way, the wandering alert device 102 can also switch between two operating modes: an estimation mode in which estimated data is calculated using the artificial intelligence 55, and a learning mode in which the artificial intelligence 55 undergoes machine learning. The switching of the operating modes can be instructed, for example, by the input device 27.

[0119] In the example of FIG. 7 , the artificial intelligence 55 is incorporated into the computer 10 as part of the wandering alert device 102. Alternatively, an artificial intelligence built on an external server 59, etc., may be used, as illustrated by the dotted line in FIG. 7 . In this case, the estimation unit 53 and the learning unit 57 operate the external artificial intelligence via the network 5 or the like. The estimation unit 53 inputs the input data received by the input data receiving unit 16 to the external trained artificial intelligence, causing it to output estimated data. The estimated data is sent to the alert output unit 45, for example, via the interface 13. The learning unit 57 also inputs the input data received by the input data receiving unit 16 and the training data received by the training data receiving unit 18 to the external artificial intelligence, thereby training the external artificial intelligence to estimate the training data from the input data. When an external artificial intelligence is used, the artificial intelligence 55, which is part of the wandering alert device 102, becomes unnecessary.

[0120] (Third embodiment) Fig. 9 is a block diagram illustrating the configuration of a wandering alert device according to yet another embodiment of the present invention. Fig. 10 is a flowchart illustrating the processing flow of the wandering alert method implemented by this wandering alert device 103. The wandering alert device 103 differs from the wandering alert devices 101 and 102 in that the artificial intelligence 61 estimates whether or not the protected person 11 is likely to wander based directly on the sleep data and basic data of the protected person 11 (steps S55 and S57). Just as the artificial intelligence 23 in the wandering alert devices 101 and 102 is accompanied by an estimation unit 19 and a learning unit 21, the artificial intelligence 61 is accompanied by an estimation unit 63 and a learning unit 65.

[0121] When the input data receiving unit 15 receives input data including sleep data of the protected person 11 and basic data of the protected person 11 (step S1), the estimation unit 63 inputs the input data received by the input data receiving unit 15 to the artificial intelligence 61, causing the artificial intelligence 61 to calculate estimated data regarding whether or not there is a possibility that the protected person 11 has wandered (steps S55, S57). If the artificial intelligence 61 has already been trained, it calculates estimated data with high accuracy regarding whether or not there is a possibility that the protected person 11 has wandered. The estimated data calculated by the artificial intelligence 61 is input to the alert output unit 45. The alert output unit 45 treats the input estimated data in the same way as the determination result regarding the possibility that the protected person 11 has wandered made by the wandering possibility determination units 41, 51 (see FIGS. 3 and 7). In this way, the trained artificial intelligence 61 directly estimates whether or not there is a possibility of wandering from input data including the sleep data and basic data of the protected person 11, so there is no need to separately provide wandering possibility determination units 41, 51 (see Figures 3 and 7), simplifying the configuration of the device.

[0122] The estimation unit 63 may input input data including the sleep data and basic data of the protected person 11 to the artificial intelligence 61, and cause the artificial intelligence 61 to calculate estimated data on the reference bed getting out time 46 for the protected person 11, along with estimated data on whether or not the protected person 11 is likely to wander (see the description in parentheses in step S57). In this case, the artificial intelligence 61 also inputs the calculated estimated data on the reference bed getting out time 46 to the alert output unit 45. As with the wandering alert device 102, the alert output unit 45 records the input estimated data on the reference bed getting out time 46 in an internal memory (not shown) or the like as the reference bed getting out time 46 of the protected person 11, and uses it for comparison with the bed getting out time of the protected person 11 (step S41).

[0123] Like the artificial intelligence 23, the artificial intelligence 61 undergoes machine learning and is thereby able to output highly accurate estimation data. The wandering alert device 103 has a learning unit 65, which allows the wandering alert device 103 to learn the artificial intelligence 61 by itself, without using an external artificial intelligence learning device. In other words, the wandering alert device 103 also has a built-in artificial intelligence learning device that trains the artificial intelligence 61 through machine learning.

[0124] When the wandering alert device 103 causes the artificial intelligence 61 to perform machine learning, the input data receiving unit 15 receives input data including the sleep data and basic data of the protected person 11, and the teacher data receiving unit 17 receives input of teacher data corresponding to the input data, including data indicating whether the protected person 11 is likely to wander or whether they have actually wandered. The teacher data may also include an actual value of the standard bed getting-out time 46. The learning unit 65 inputs the input data received by the input data receiving unit 15 and the teacher data received by the teacher data receiving unit 17 to the artificial intelligence 61, thereby training the artificial intelligence 61 to estimate the teacher data from the input data. By inputting a large number of pairs of associated input data and teacher data into the wandering alert device 103, the learning of the artificial intelligence 61 progresses and the accuracy of estimation improves.

[0125] By correlating the sleep data and basic data collected in the past for various protected persons 11 with the wandering possibility data or actual data for the protected persons 11 and recording them, for example, in the storage medium 31, the input data receiving unit 15 and the teacher data receiving unit 17 can sequentially read out a large amount of data required for learning from the storage medium 31, and the learning unit 65 can repeatedly learn the artificial intelligence 61 for each piece of read data. In this way, the wandering alert device 103 can also switch between two operating modes: an estimation mode in which estimated data is calculated using the artificial intelligence 61, and a learning mode in which the artificial intelligence 61 undergoes machine learning. The switching of the operating modes can be instructed, for example, by the input device 27.

[0126] In the example of FIG. 9 , the artificial intelligence 61 is incorporated into the computer 10 as part of the wandering alert device 103. Alternatively, an artificial intelligence built on an external server 67, etc., may be used, as illustrated by the dotted line in FIG. 9 . In this case, the estimation unit 63 and the learning unit 65 operate the external artificial intelligence via the network 5 or the like. The estimation unit 63 inputs input data received by the input data receiving unit 15 to the external trained artificial intelligence, causing it to output estimated data. The estimated data is sent to the alert output unit 45, for example, via the interface 13. Furthermore, the learning unit 65 inputs the input data received by the input data receiving unit 15 and the teacher data received by the teacher data receiving unit 17 to the external artificial intelligence, thereby training the external artificial intelligence to estimate the teacher data from the input data. When an external artificial intelligence is used, the artificial intelligence 61, which is part of the wandering alert device 103, becomes unnecessary.

[0127] (Artificial intelligence learning methods) FIG. 11 is a flowchart illustrating the process flow of the artificial intelligence learning method implemented by the wandering alert devices 101, 102, and 103 illustrated in FIGS. 3, 7, and 9. When the process starts, the input data receiving units 15 and 16 receive input data (step S61). Furthermore, the teacher data receiving units 17 and 18 receive teacher data (step S63). Either process S61 or process S63 may be performed first, or they may be performed simultaneously. Next, the learning units 21, 57, and 65 input the input data and teacher data to the artificial intelligences 23, 55, and 61, thereby training the artificial intelligences 23, 55, and 61 to estimate the teacher data from the input data (step S65).

[0128] Next, when the wandering alert devices 101, 102, 103 determine that the process should be repeated based on user instructions or the like (Yes in step S67), they return the process to S61. As a result, the input data receiving units 15, 16 receive new input data (step S61), and the teacher data receiving units 17, 18 receive new teacher data (step S63). When the wandering alert devices 101, 102, 103 determine that the process should not be repeated (No in step S67), they end the process.

[0129] As already mentioned, in the example shown in Fig. 2, the wandering alert devices 101, 102, and 103 are incorporated into the computer 10. By installing and running a specific application, i.e., a program, on the computer 10, the computer 10 functions as the wandering alert devices 101, 102, and 103. This program may be supplied via the network 5, or may be supplied by a storage medium 31 such as a CD-ROM (see Figs. 2, 7, and 9).

[0130] (Example of a computer screen) FIG. 12 is a screen diagram illustrating an example of an image displayed on the screen of the computer 10 (see FIG. 2). In the illustrated example, the computer 10 is a smartphone. On the screen in the illustrated example, the status of multiple protected persons 11, such as whether they are lying down or out of bed, is displayed simultaneously. By scrolling the screen appropriately, the status of multiple protected persons 11 that cannot fit on one screen can be visually grasped.

[0131] In the illustrated example, when it is detected that the protected person 11 has gotten out of bed, the pictogram representing the protected person 11 changes from lying down to getting out of bed. In the illustrated example, an alarm sounds at the same time. The alert output unit 45 (see Figures 3, 7, and 9) may output a signal notifying the protected person 11 of getting out of bed by image and sound in this way. In the illustrated example, by tapping the icon representing the speaker, it is possible to select whether or not to notify each protected person 11 of getting out of bed. The tap operation on the screen corresponds to the operation of the input device 27 (see Figures 3, 7, and 9). The selection result is communicated to the alert output unit 45 via the interface 13.

[0132] Furthermore, in the illustrated example, if the protected person 11 remains out of bed for 15 minutes at night, wandering is detected, and a "footprint" icon is displayed to indicate wandering, and an alarm sounds to notify the user of wandering. In this way, the alert output unit 45 (see FIGS. 3, 7, and 9) can limit the output of wandering alerts to specific times, such as at night, based on information from a clock (not shown) built into the computer 10. The time limit can be set by tapping on the smartphone screen, and this is communicated to the alert output unit 45 via the interface 13. By tapping the "footprint" icon, the user can select whether or not to notify the protected person 11 of wandering (corresponding to step S9 in FIG. 4, etc.). The selection result is similarly communicated to the alert output unit 45 via the interface 13.

[0133] As already described in the description of the first embodiment, the smartphone in the illustrated example may be a part of the computer 10 instead of being an example of the computer 10. This can reduce the hardware and processing load on the smartphone. [Explanation of symbols]

[0134] 1 Sleep sensor, Wi-Fi router 3, 5 Network, 7, 9 Server, 10 Computer, 11 Protected person, 13 Interface, 15 Input data reception unit, 16 Input data reception unit, 17 Teacher data reception unit, 18 Teacher data reception unit, 19 Estimation unit, 21 Learning unit, 23 Artificial intelligence, 27 Input device, 29 Output device, 31 Storage medium, 33 Input layer, 35 Intermediate layer, 37 Output layer, 41 Wandering possibility determination unit, 43 Sensor signal receiving unit, 45 Alert output unit, 46 Reference bed getting-out time, 47 Determination output unit, 49 Alert necessity reception unit, 51 Wandering possibility determination unit, 53 Estimation unit, 55 Artificial intelligence, 57 Learning unit, 59 Server, 61 Artificial intelligence, 63 Estimation unit, 65 Learning unit, 67 Server, 100 Wandering alert system, 101, 102, 103 Wandering alert device.

Claims

1. an input data receiving unit that receives input data including sleep data that is data related to the sleep of the protected person and basic data that is data related to the body of the protected person; an estimation unit that inputs the input data received by the input data receiving unit into a trained artificial intelligence, thereby causing the artificial intelligence to calculate estimated data of the daily living activities data, which is data related to the daily living activities of the protected person; A wandering possibility determination unit that determines whether or not there is a possibility that the protected person will wander based on the estimated data calculated by the artificial intelligence; a sensor signal receiving unit that receives a signal from a bed exit sensor that detects the protected person getting out of bed; and an alert output unit that outputs a wandering alert when the time that the protected person continues to get out of bed exceeds a standard bed getting out time based on the signal received by the sensor signal receiving unit, on the condition that the wandering possibility determination unit has determined that there is a possibility that the protected person may wander. the daily living activity data includes data on motor function and data on cognitive function; the sleep data includes data expressed in numerical values ​​based on data measured by a sleep sensor; The basic data includes data expressed in numerical values, A wandering alert device, wherein the daily living activity data includes data evaluated in stages using numerical values.

2. a determination output unit that outputs the determination that the wandering possibility determination unit has determined that the protected person has a wandering possibility; An alert necessity receiving unit that receives an input of whether or not a wandering alert is necessary for the protected person who is likely to wander and is output by the determination output unit, 2. The wandering alert device according to claim 1, wherein the alert output unit outputs the wandering alert on a further condition that a wandering alert is required for the protected person.

3. 2. The wandering alert device of claim 1, wherein the wandering possibility determination unit determines that there is a possibility of wandering when the estimated motor function data for the protected person exceeds a reference motor function level and the estimated cognitive function data does not reach a reference cognitive function level.

4. 2. The wandering alert device of claim 1, wherein the wandering possibility determination unit has another estimation unit that inputs the estimated daily living activity data into another trained artificial intelligence to estimate whether or not there is a possibility of wandering, and the result of the estimation by the other artificial intelligence is used as the result of determining whether or not there is a possibility of wandering.

5. The other estimation unit causes the other artificial intelligence to estimate the reference time to get out of bed, The wandering alert device according to claim 4, wherein the alert output unit compares the standard bed leaving time estimated by the other artificial intelligence with the time the protected person continues to leave bed.

6. 2. The wandering alert device of claim 1, wherein the data on activities of daily living is data on items related to activities of daily living, including data evaluated on a multiple-stage basis for at least some of the evaluation items defined in the Functional Independence Assessment Method.

7. 2. The wandering alert device according to claim 1, wherein the sleep data is data on items related to sleep, including at least some of breathing, pulse, sleep time, sleep rhythm, number of times turning over in bed, number of body movements, number of times going to the toilet, and toilet time.

8. 2. The wandering alert device according to claim 1, wherein the basic data is data on basic items including at least some of age, sex, height, weight, medical history, level of care required, and BMI (Body Mass Index).

9. The wandering alert device according to claim 1 , wherein the sleep sensor also functions as the bed exit sensor.

10. an input data receiving process in which a computer receives input data including sleep data relating to the sleep of the protected person and basic data relating to the body of the protected person; an estimation process in which the computer inputs the input data received by the input data reception process into a trained artificial intelligence, thereby causing the artificial intelligence to calculate estimated data of the daily living activity data, which is data related to the daily living activity of the protected person; A wandering possibility determination process in which the computer determines whether or not there is a possibility that the protected person will wander based on the estimated data calculated by the artificial intelligence; a sensor signal receiving process in which the computer receives a signal from a bed exit sensor that detects the protected person getting out of bed; The computer includes an alert output process for outputting a wandering alert when the time that the protected person continues to get out of bed exceeds a reference bed time based on the signal received by the sensor signal receiving process, on the condition that the wandering possibility determination process has determined that there is a possibility that the protected person may wander; the daily living activity data includes data on motor function and data on cognitive function; the sleep data includes data expressed in numerical values ​​based on data measured by a sleep sensor; The basic data includes data expressed in numerical values, A wandering alert method, wherein the daily living activity data includes data evaluated in stages using numerical values.

11. a determination output process in which the computer outputs a determination that the wandering possibility determination process has determined that there is a possibility that the protected person has wandered; The computer further includes an alert necessity reception process for receiving an input of whether or not a wandering alert is necessary for the protected person who is likely to wander and is output by the judgment output process, The wandering alert method according to claim 10 , wherein the alert output process outputs the wandering alert on a further condition that a wandering alert is required for the protected person.

12. The wandering alert method of claim 10, wherein the wandering possibility determination process determines that there is a possibility of wandering when the estimated data regarding the motor function of the protected person exceeds a reference motor function level and the estimated data regarding the cognitive function does not reach a reference cognitive function level.

13. 11. The wandering alert method of claim 10, wherein the wandering possibility determination process includes another estimation process that inputs the estimated daily living activity data into another trained artificial intelligence to estimate whether or not there is a possibility of wandering, and the result of the estimation by the other artificial intelligence is used as the result of determining whether or not there is a possibility of wandering.

14. The other estimation process also causes the other artificial intelligence to estimate the reference time to get out of bed, The wandering alert method according to claim 13, wherein the alert output process compares the standard bed leaving time estimated by the other artificial intelligence with the time the protected person continues to leave bed.

15. A wandering alert program that, when read by a computer, causes the computer to execute the wandering alert method according to any one of claims 10 to 14.

16. An artificial intelligence learning device for a wandering alert device, which learns the other artificial intelligence used by the wandering alert device according to claim 4, a separate input data receiving unit that receives, as separate input data, data on activities of daily living including data on motor function and data on cognitive function of the protected person; A teacher data receiving unit that receives teacher data including data indicating whether or not the protected person has a possibility of wandering or has actually wandered, corresponding to the other input data; a learning unit that inputs the other input data received by the other input data receiving unit and the teacher data received by the teacher data receiving unit to the other artificial intelligence, thereby training the other artificial intelligence to estimate the teacher data from the other input data, An artificial intelligence learning device, wherein the daily living activity data includes data that is evaluated in stages using numerical values.

17. 17. The artificial intelligence learning device according to claim 16, wherein the teacher data also includes performance data on the standard bed-getting time for the protected person, which is used by the wandering alert device according to claim 5.

18. An artificial intelligence learning method for a wandering alert method, which trains the other artificial intelligence used in the wandering alert method according to claim 13, comprising: a separate input data receiving process in which the computer receives, as separate input data, activity of daily living data including data on motor function and data on cognitive function; A teacher data reception process in which the computer receives teacher data including data indicating whether or not the protected person has a possibility of wandering or has actually wandered, corresponding to the other input data; a learning process in which the computer inputs the other input data received by the other input data receiving process and the teacher data received by the teacher data receiving process to the other artificial intelligence, thereby training the other artificial intelligence to estimate the teacher data from the other input data; An artificial intelligence learning method, wherein the daily living activity data includes data that is evaluated in stages using numerical values.

19. The artificial intelligence learning method according to claim 18, wherein the training data also includes performance data on the standard bed time for the protected person, which is used in the wandering alert method according to claim 14.

20. An artificial intelligence learning program that, when read by a computer, causes the computer to execute the artificial intelligence learning method according to claim 18 or 19.

Images