Training data generation device, training data generation method, model generation device, estimation device, and program
The system uses odor sensors and sub-sensors to generate training data and model excretion types, addressing the inefficiency in existing methods and providing accurate, automated excretion type identification in nursing and medical facilities.
Patent Information
- Application Number
- JP2023071603
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-04-25
- Publication Date
- 2024-11-07
AI Technical Summary
In facilities providing nursing care and medical services, there is a need to efficiently generate training data for machine learning models to determine the type of excretion by service users, as current methods place a heavy burden on service providers by requiring manual inspection.
A system utilizing odor sensors and sub-sensors to generate teacher data, including a wearing determination process, signal acquisition, and type information input by service providers, followed by model generation and estimation using machine learning to identify excretion types.
Efficient generation of training data and accurate estimation of excretion types, reducing the burden on service providers and enabling timely notification of excretion events.
Smart Images

Figure 2024157309000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to an estimation device and an estimation method for estimating the type of excretion of a service user, a model generation device and a model generation method for generating a model used for such estimation by machine learning, and a training data generation device and a training data generation method for generating training data used for such machine learning. [Background technology]
[0002] In facilities that aim to provide nursing care services and / or medical services, it is necessary for service providers to be informed of the excretion of service users in a timely manner. However, in such facilities, there are quite a few service users (including service users who are asleep at night) for whom it is impossible or difficult to inform the service provider of their excretion. In order for the service provider to know the excretion of such service users, it is necessary to have the service user remove their underwear to check whether or not they have excreted, which is a heavy burden for the service provider.
[0003] In order to reduce such burdens, there is a demand for utilizing information processing technology in nursing care sites. For example, Patent Document 1 is an example of a document disclosing information processing technology that can be utilized in nursing care sites. Patent Document 1 discloses a biological monitor device that monitors the health condition of a patient based on an output signal from a gas sensor attached to a T-belt. [Prior art documents] [Patent documents]
[0004] [Patent Document 1] Japanese Patent Application Publication No. 9-43182 Summary of the Invention [Problem to be solved by the invention]
[0005] In facilities that aim to provide nursing care services and / or medical services, there is a demand for a technology that notifies a service provider of the type of excretion of a service user in addition to the presence or absence of excretion by the service user. For example, if the type of excretion is flatus, the service provider does not need to take measures, but if the type of excretion is feces, the service provider needs to take measures. This is because the service provider needs to change its behavior depending on the type of excretion. In this case, it is considered that the type of excretion of the service user is determined by a model (trained model) generated by machine learning. However, a technology for efficiently generating training data required for generating a model by machine learning, especially supervised learning, has not been established.
[0006] One aspect of the present invention has been made in view of the above problems, and has an object to realize a technique capable of efficiently generating training data. Another object of the present invention is to realize a technique for generating a model using the training data thus generated, and a technique for estimating the type of excretion using such a model. [Means for solving the problem]
[0007] A teacher data generation device according to one embodiment of the present invention has at least one processor, and the processor executes an attachment determination process for determining whether or not an odor sensor is worn by a service user based on an output signal of a sub-sensor installed near an odor sensor associated with the service user, a first acquisition process for acquiring an output signal of the odor sensor, a second acquisition process for acquiring type information indicating a type of excretion of the service user determined by a service provider, and a teacher data generation process for generating teacher data including sensor information indicating the output signal acquired in the first acquisition process from the odor sensor determined to be worn by the service user in the attachment determination process and the type information acquired in the second acquisition process.
[0008] A teacher data generation method according to one embodiment of the present invention includes an attachment determination process in which at least one processor determines whether an odor sensor is worn by a service user based on an output signal of a sub-sensor installed near an odor sensor associated with the service user; a first acquisition process in which the processor acquires the output signal of the odor sensor; a second acquisition process in which the processor acquires type information indicating a type of excretion of the service user, the type of excretion determined by a service provider; and a teacher data generation process in which the processor generates teacher data including sensor information indicating the output signal acquired in the first acquisition process from the odor sensor determined in the attachment determination process to be worn by the service user and the type information acquired in the second acquisition process.
[0009] A model generation device according to one embodiment of the present invention has at least one processor, and the processor executes a model generation process to generate a model by machine learning using teacher data generated by a teacher data generation device according to one embodiment of the present invention, the model having sensor information representing an output signal of an odor sensor as input and type information representing a type of excretion as output.
[0010] An estimation device according to one embodiment of the present invention has at least one processor, and the processor executes an estimation process to estimate the type of excretion of a service user from the output signal of an odor sensor associated with the service user using a model generated by a model generation device according to one embodiment of the present invention.
[0011] Each of the teacher data generation device, model generation device and estimation device according to each aspect of the present invention may be realized by a computer. In this case, the control program for the device, which causes the computer to operate as each part (software element) of the device, and the computer-readable recording medium on which the control program is recorded, also fall within the scope of the present invention. Effect of the Invention
[0012] According to one aspect of the present invention, it is possible to efficiently generate training data. Also, according to another aspect of the present invention, it is possible to generate a model using the training data thus generated. Also, according to another aspect of the present invention, it is possible to estimate the type of excretion using such a model. [Brief description of the drawings]
[0013] [Figure 1] 1 is a schematic diagram showing a configuration of a notification system according to an embodiment of the present invention; [Diagram 2] 2 is a block diagram showing a configuration of a main device included in the notification system shown in FIG. 1. [Diagram 3] 1A to 1C are diagrams illustrating an overview of a teacher data generation method, a model generation method, and an estimation method according to one embodiment of the present invention. [Figure 4] 4 is a flow diagram showing the operation (including the teacher data generating method shown in FIG. 3) of the notification system shown in FIG. 1 during a provisional operation period. [Diagram 5] 11A and 11B are diagrams illustrating an example of a notification screen and an input screen displayed on a notification terminal during a provisional operation period. [Figure 6] FIG. 4 is a flowchart showing the flow of the model generation method shown in FIG. [Figure 7] 4 is a flow diagram showing the operation (including the estimation method shown in FIG. 3) of the notification system shown in FIG. 1 during the actual operation period. [Figure 8] FIG. 13 is a diagram showing an example of a notification screen displayed on a notification terminal during the actual operation period. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0014] (Configuring the notification system) A notification system S according to one embodiment of the present invention will be described with reference to Fig. 1. Fig. 1 is a schematic diagram showing the configuration of the notification system S.
[0015] The notification system S is a system for notifying a service provider Ub of the excretion of a service user Ua in a facility that aims to provide nursing care services and / or medical services.
[0016] 1, the notification system S includes a main device 1, multiple notification terminals 2, multiple relay terminals 3, and multiple sensor devices 4. The main device 1 and each notification terminal 2 perform one-to-many communication via a network, the main device 1 and each relay terminal 3 perform one-to-many communication via the network, and the relay terminal 3 and the sensor device 4 perform one-to-one communication (e.g., short-range wireless communication) without using a network.
[0017] The sensor device 4 is attached to the service user Ua. As an example, the sensor device 4 is attached inside the underwear (diaper, pants, etc.) of the service user Ua. The sensor device 4 has an odor sensor (main sensor) associated with the service user Ua and a sub-sensor provided close to the odor sensor, and transmits output signals of the odor sensor and the sub-sensor together with the ID (identification information) of the service user Ua to the relay terminal 3. In this embodiment, a temperature sensor is used as the sub-sensor.
[0018] The relay terminal 3 is placed in the vicinity of the sensor device 4. As an example, the relay terminal 3 is placed in a room used by the service user Ua. The relay terminal 3 receives output signals of the odor sensor and the temperature sensor from the sensor device 4 together with the ID of the service user Ua. The relay terminal 3 also transmits the output signals of the odor sensor and the temperature sensor received from the sensor device 4 to the main device 1 together with the ID of the service user Ua. In this embodiment, a smartphone is used as the relay terminal 3. A stationary wireless repeater may also be used as the relay terminal 3.
[0019] The main device 1 is placed outside the facility described above. As an example, the main device 1 is a workstation placed in a data center. The main device 1 receives output signals from the odor sensor and the temperature sensor together with the ID of the service user Ua from the relay terminal 3. The main device 1 also detects excretion of the service user Ua by referring to the output signal from the odor sensor received from the relay terminal 3. The main device 1 then (1) transmits data representing a notification screen σ1a and an input screen σ1b in a teacher data generation method M1 described later, and (2) transmits data representing a notification screen σ2 in an estimation method M3 described later. Note that a personal computer or the like placed in the facility may be used as the main device 1.
[0020] The notification terminal 2 is carried by the service provider Ub. The notification terminal 2 (1) receives data representing a notification screen σ1a and an input screen σ1b from the main device 1 in a teacher data generation method M1 described later, and (2) receives data representing a notification screen σ2 from the main device 1 in an estimation method M3 described later. The notification terminal 2 also includes a touch panel display, and (1) displays the notification screen σ1a and the input screen σ1b on the display based on data acquired from the main device 1 in a teacher data generation method M1 described later, and (2) displays the notification screen σ2 on the display based on data acquired from the main device 1 in an estimation method M3 described later. In this embodiment, a smartphone is used as the notification terminal 2. In the following description, transmitting data representing various screens is also described as transmitting the screen, and receiving data representing various screens is also described as receiving the screen.
[0021] (Main unit configuration) The configuration of the main device 1 included in the notification system S will be described with reference to Fig. 2. Fig. 2 is a block diagram showing the configuration of the main device 1.
[0022] 2, the main unit 1 includes a processor 11, a primary memory 12, a secondary memory 13, a communication interface 14, and a bus 15. The processor 11, the primary memory 12, the secondary memory 13, and the communication interface 14 are connected to each other via the bus 15. An example of a device that can be used as the main unit 1 is a workstation that constitutes a cloud server.
[0023] The secondary memory 13 stores a teacher data generation program P1, a model generation program P2, and an estimation program P3. The processor 11 expands the teacher data generation program P1 stored in the secondary memory 13 onto the primary memory 12. Then, the processor 11 executes each process included in a teacher data generation method M1 described later according to an instruction included in the teacher data generation program P1 expanded onto the primary memory 12. The teacher data D generated by the teacher data generation method M1 is also stored in the secondary memory 13. The processor 11 also expands the model generation program P2 stored in the secondary memory 13 onto the primary memory 12. Then, the processor 11 executes each process included in a model generation method M2 described later according to an instruction included in the model generation program P2 expanded onto the primary memory 12. The model M generated by the model generation method M2 is also stored in the secondary memory 13. The processor 11 also expands the estimation program P3 stored in the secondary memory 13 onto the primary memory 12. Then, the processor 11 executes each process included in an estimation method M3, which will be described later, in accordance with instructions included in the estimation program P3 deployed on the primary memory 12.
[0024] An example of a device that can be used as the processor 11 is a CPU (Central Processing Unit). An example of a device that can be used as the primary memory 12 is a semiconductor RAM (Random Access Memory). An example of a device that can be used as the secondary memory 13 is a HDD (Hard Disk Drive).
[0025] The communication interface 14 is an interface for communicating with the notification terminal 2 and the relay terminal 3 via a network. An example of an interface that can be used as the communication interface 14 is an Ethernet (registered trademark) interface. Examples of networks that can be used include a personal area network (PAN), a local area network (LAN), a campus area network (CAN), a metropolitan area network (MAN), a wide area network (WAN), a global area network (GAN), or an internetwork including these networks. The internetwork may be an intranet, an extranet, or the Internet.
[0026] The teacher data generation program P1 for causing the processor 11 to execute the teacher data generation method M1 may be recorded in a computer-readable, non-transitory, tangible recording medium. This recording medium may be the secondary memory 13 or other recording media. For example, a tape, a disk, a card, a semiconductor memory, a programmable logic circuit, etc. may be used as other recording media. The same applies to the model generation program P2 and the estimation program P3.
[0027] In addition, in this embodiment, a configuration is adopted in which the teacher data generation method M1 is implemented using a single processor (processor 11), but the present invention is not limited to this. That is, a configuration may be adopted in which the teacher data generation method M1 is implemented using multiple processors. In this case, the multiple processors that implement the teacher data generation method M1 may be provided in a single computer and configured to be able to communicate with each other via a bus, or may be provided in a distributed manner in multiple computers and configured to be able to communicate with each other via a network. For example, a mode in which the processors provided in each of the multiple computers that constitute the cloud server cooperate to implement the teacher data generation method M1, or a mode in which the processor 11 of the main device 1 and the processor of the notification terminal 2 cooperate to implement the teacher data generation method M1, etc. are also conceivable. The same applies to the model generation method M2 and the estimation method M3.
[0028] In addition, in this embodiment, a configuration is adopted in which the teacher data generation method M1, the model generation method M2, and the estimation method M3 are implemented by the main device 1, but the present invention is not limited to this. A teacher data generation device that implements at least the teacher data generation method M1, a model generation device that implements at least the model generation method M2, and an estimation device that implements at least the estimation method M3 are also included in the scope of the present invention. Here, the teacher data generation device only needs to implement the teacher data generation method M1, and may or may not implement the model generation method M2 and the estimation method M3. Also, the model generation device only needs to implement the model generation method M2, and may or may not implement the teacher data generation method M1 and the estimation method M3. Also, the estimation device only needs to implement the estimation method M3, and may or may not implement the teacher data generation method M1 and the model generation method M2.
[0029] (Overview of training data generation method, model generation method, and estimation method) The teacher data generation method M1, the model generation method M2, and the estimation method M3 will be outlined below with reference to Fig. 3. Fig. 3 is a schematic diagram illustrating the teacher data generation method M1, the model generation method M2, and the estimation method M3.
[0030] The teacher data generation method M1 is a method for generating teacher data D that includes sensor information representing an output signal of an odor sensor that is associated with a service user Ua and that is determined to be attached to the service user Ua, and type information representing the type of excretion of the service user Ua. The type information can be considered as a label attached to the sensor information.
[0031] Here, the sensor information may be, for example, numerical data representing an output signal of an odor sensor. Alternatively, the sensor information may be image data including a graph representing an output signal of an odor sensor. In this embodiment, image data including a graph representing an output signal of an odor sensor is used as the sensor information. Furthermore, the type information may be a character string representing a type of excretion (such as "feces", "urination", "flatus" etc.). Alternatively, the type information may be a code representing a type of excretion (such as C1, C2, C3 etc.). In this embodiment, a code representing at least one type of excretion among feces, urination and flatus is used as the type information. The type information may represent multiple types, for example, it may represent both feces and urination.
[0032] The model generation method M2 is a method for generating a model M by supervised learning, in which sensor information representing an output signal of an odor sensor is input and type information representing a type of excretion is output. In this embodiment, machine learning using the teacher data D generated by the teacher data generation method M1 is adopted as the supervised learning.
[0033] Here, the model M may be any model that can be applied to a classification problem and can be generated by supervised learning. The model M may be, for example, a k-nearest neighbor method, a decision tree (classification tree), a random forest, a support vector machine, or a neural network. In this embodiment, a neural network is used as the model M.
[0034] The estimation method M3 is a method for estimating the type of excretion of the service user Ua from the output signal of the odor sensor associated with the service user Ua, using a model generated by supervised learning. In this embodiment, the model M generated by the model generation method M2 is adopted as the model generated by supervised learning.
[0035] The main device 1 operates differently during the provisional operation period and the actual operation period. That is, during the provisional operation period, the main device 1 notifies the service provider Ub of the excretion of the service user Ua without specifying the type of excretion. On the other hand, during the actual operation period, the main device 1 notifies the service provider Ub of the excretion of the service user Ua by specifying the type of excretion.
[0036] In this embodiment, the teacher data generation method M1 is incorporated into the operation of the main device 1 during the provisional operation period. That is, the teacher data D is generated during the provisional operation period. Furthermore, the model generation method M2 is implemented by the main device 1 after the provisional operation period ends and before the actual operation period starts. That is, the model M is generated after the provisional operation period ends and before the actual operation period starts. The estimation method M3 is incorporated into the operation of the main device 1 during the actual operation period. That is, the model M is used during the actual operation period.
[0037] Below, we will explain (1) the operation of main unit 1 during the provisional operation period (including teacher data generation method M1), (2) model generation method M2, and (3) the operation of main unit 1 during the actual operation period (including estimation method M3).
[0038] (Main unit operation during provisional operation period) The operation of the main device 1 during the provisional operation period will be described with reference to Fig. 4 and Fig. 5. Fig. 4 is a flow diagram showing the flow of the operation of the main device 1 during the provisional operation period. Fig. 5 is a diagram showing an example of a notification screen σ1a and an input screen σ1b displayed on the notification terminal 2 during the provisional operation period.
[0039] During the provisional operation period, the main device 1 executes an attachment determination process M11, a first acquisition process M12, an excretion determination process M13, a notification process M14, a second acquisition process M15, and a teacher data generation process M16 shown in Fig. 4. The teacher data generation method M1 is composed of the attachment determination process M11, the first acquisition process M12, the excretion determination process M13, the notification process M14, the second acquisition process M15, and the teacher data generation process M16.
[0040] The wearing determination process M11 is a process for determining whether or not the odor sensor associated with the service user Ua is worn by the service user Ua. In this embodiment, the processor 11 acquires an output signal of a temperature sensor installed near the odor sensor from the relay terminal 3 using the communication interface 14. The processor 11 determines that the temperature represented by the output signal of the temperature sensor is "worn" when it is equal to or higher than a predefined threshold, and determines that the temperature is "not worn" when it is lower than the predefined threshold. In this embodiment, the predefined threshold is 36.0°C, which is defined by the teacher data generation program P1.
[0041] If the wearing determination process M11 determines that the sensor device 4 is "worn," a first acquisition process M12 is executed. If the wearing determination process M11 determines that the sensor device 4 is "not worn," the wearing determination process M11 is executed repeatedly. As a result, the processor 11 does not obtain an output signal from the odor sensor during a period when the service user Ua is not wearing the sensor device 4, and thus noise-reduced teacher data for generating a model with high estimation accuracy is generated in the teacher data generation method M1.
[0042] The first acquisition process M12 is a process for acquiring an output signal of the odor sensor associated with the service user Ua together with the ID of the service user Ua. In this embodiment, the processor 11 acquires the output signal of the odor sensor associated with the service user Ua together with the ID of the service user Ua from the relay terminal 3 using the communication interface 14.
[0043] In this embodiment, the processor 11 acquires the output signal of the odor sensor in the first acquisition process M12 regardless of whether or not the odor sensor is determined to be worn by the service user Ua in the wearing determination process M11. However, the present invention is not limited to this. For example, the processor 11 may acquire only the output signal of the odor sensor that is determined to be worn by the service user Ua in the wearing determination process M11 in the first acquisition process M12.
[0044] The excretion determination process M13 is a process for determining whether or not the service user Ua has excreted based on the output signal of the odor sensor associated with the service user Ua. In this embodiment, the processor 11 determines whether or not the service user Ua has excreted based on the output signal of the odor sensor associated with the service user Ua, which is acquired in the first acquisition process M12. If the excretion determination process M13 determines that "excretion has occurred," a notification process M14 is executed.
[0045] The notification process M14 is a process for notifying the notification terminal 2 used by the service provider Ub that excretion of the service user Ua has been detected, without specifying the type of excretion. In this embodiment, the processor 11 transmits a notification screen σ1a indicating that excretion of the service user Ua has been detected to the notification terminal 2 using the communication interface 14.
[0046] The notification terminal 2 displays on the display a notification screen σ1a (see FIG. 5) indicating that the service user Ua's excretion has been detected. After the confirmation button on the notification screen σ1a is tapped, the service provider Ub processes the service user Ua's excretion. When the confirmation button on the notification screen σ1a is tapped, the notification terminal 2 displays on the display an input screen σ1b (see FIG. 5) for inputting the type of excretion of the service user Ua. After processing the excretion of the service user Ua, the service provider Ub inputs the type of excretion of the service user Ua using the input screen σ1b. In this embodiment, if the excretion in the underwear of the service user Ua is feces or urine, the service provider Ub selects "feces" or "urination", respectively, and if there is no excretion in the underwear, selects "no feces or urine". In this embodiment, if "no feces or urine" is selected, the processor 11 processes the type of excretion as being flatus. When the type of excretion of the service user Ua is input using the input screen σ1b, the notification terminal 2 transmits to the main apparatus 1 type information indicating the input type.
[0047] The second acquisition process M15 is a process for acquiring type information indicating the type of excretion of the service user Ua. In this embodiment, the processor 11 acquires the type information indicating the type of excretion of the service user Ua, which is input by the service provider Ub using the input screen σ1b, from the notification terminal 2 using the communication interface 14.
[0048] The teacher data generation process M16 is a process for generating teacher data including an output signal of an odor sensor associated with the service user Ua and type information indicating the type of excretion of the service user Ua, which is determined by the service provider Ub. In this embodiment, the processor 11 generates teacher data D including an output signal of the odor sensor acquired in the first acquisition process M12 from the odor sensor determined to be attached to the service user Ua in the wearing determination process M11, and type information acquired in the second acquisition process M15. In the teacher data D generated in this embodiment, the output signal of the odor sensor acquired during a predetermined period based on the time point at which it was determined that "excretion has occurred" in the excretion determination process M13 is associated with the type information acquired in the second acquisition process M15.
[0049] The generated teacher data D is stored in the secondary memory 13 in association with the ID of the service provider Ub acquired in the first acquisition process M12.
[0050] (Implementation of model generation method) When the provisional operation period ends, the main device 1 executes the model generation method M2. Fig. 6 is a flow diagram showing the flow of the model generation method M2.
[0051] The model generation method M2 includes a model generation process M21, as shown in Fig. 6. The model generation process M21 is a process for generating a model M that receives sensor information representing an output signal of an odor sensor as an input and outputs type information representing a type of excretion, by machine learning using training data D generated during the provisional operation period. In this embodiment, the processor 11 receives sensor information representing an output signal of an odor sensor as an input and generates a model M that outputs type information representing a type of excretion, by machine learning using training data D generated during the provisional operation period.
[0052] In the model generation process M21, the processor 11 may generate, as the model M, a generic model that does not specify the service user Ua to which the model is to be applied, or may generate a dedicated model that specifies the service user to which the model is to be applied. In the former case, the processor 11 generates the model M by machine learning using all the training data D generated during the provisional operation period. In the latter case, the processor 11 generates the model M corresponding to each service user Ua by machine learning using the training data D associated with that service user Ua from among the training data D generated during the provisional operation period. In the latter case, when estimating the type of excretion of each service user Ua, by using the model M corresponding to that service user Ua, highly accurate estimation results can be obtained.
[0053] (Main unit operation during this operation period) The operation of the main device 1 during the regular operation period will be described with reference to Fig. 7 and Fig. 8. Fig. 7 is a flow diagram showing the flow of the operation of the main device 1 during the regular operation period. Fig. 8 is a diagram showing an example of a notification screen σ2 displayed on the notification terminal 2 during the provisional operation period.
[0054] In this operation period, the main apparatus 1 executes an acquisition process M31, a determination process M32, an estimation process M33, and a notification process M34 shown in Fig. 7. The estimation method M3 is constituted by the estimation process M33.
[0055] The acquisition process M31 is a process for acquiring an output signal of an odor sensor associated with a service user Ua together with the ID of the service user Ua. In this embodiment, the processor 11 acquires an output signal of an odor sensor associated with a service user Ua together with the ID of the service user Ua from the relay terminal 3 using the communication interface 14.
[0056] The determination process M32 is a process for determining whether or not the service user Ua has excreted based on the output signal of the odor sensor associated with the service user Ua. In this embodiment, the processor 11 determines whether or not the service user Ua has excreted based on the output signal of the odor sensor associated with the service user Ua, which is acquired in the acquisition process M31. If the determination process M32 determines that "excretion has occurred," the estimation process M33 is executed.
[0057] The estimation process M33 is a process for estimating the type of excretion of the service user Ua from the output signal of the odor sensor associated with the service user Ua using a model M generated after the provisional operation period ends and before the start of the actual operation period. In this embodiment, the processor 11 estimates the type of excretion of the service user Ua from the output signal of the odor sensor associated with the service user Ua using a model M generated after the provisional operation period ends and before the start of the actual operation period.
[0058] In the estimation process M33, the processor 11 may estimate the type of excretion of the service user Ua using a general-purpose model M that does not specify the service user Ua to which the model is applied, or may estimate the type of excretion of the service user Ua to which the model is applied using a dedicated model M that specifies the service user Ua to which the model is applied. In the latter case, a highly accurate estimation result can be obtained.
[0059] The notification process M34 is a process for identifying the type of excretion and notifying the notification terminal 2 used by the service provider Ub that excretion by the service user Ua has been detected. In this embodiment, the processor 11 transmits a notification screen σ2 indicating that excretion by the service user Ua has been detected and including the estimation result obtained in the estimation process M33 to the notification terminal 2 using the communication interface 14.
[0060] The notification terminal 2 displays on the display a notification screen σ2 (see FIG. 8) indicating that a specific type of excretion by the service user Ua has been detected, the notification screen σ2 including the inference result obtained in the inference process M33. This allows the service provider Ub to know that the service user Ua has performed excretion together with the type of excretion.
[0061] (Modification) The operation of the main device 1 during the actual operation period in the present invention is not limited to the operation of the above-mentioned embodiment. For example, during the actual operation period, the processor 11 may display an input screen for inputting the type of excretion on the display in the notification process M34 as in the notification process M14. After the notification process M34, the processor 11 may obtain the type of excretion input by the service provider Ub, generate teacher data including sensor information and type information, and re-learn the model using the generated teacher data. In such a modified example, the processor 11 may subject a general-purpose model that does not specify the service user Ua to which the model is applied to re-learning using teacher data that specifies the service user Ua. According to such a modified example, a model with high estimation accuracy for a specific service user Ua who has not yet passed through the provisional operation period can be generated by re-learning while applying the estimation process M33 during the actual operation period.
[0062] Furthermore, during this operation period, the processor 11 may determine whether the odor sensor is attached to a service user Ua other than the service user Ua associated with the odor sensor. In such a modified example, if the matching rate (correct answer rate) between the estimation result of the estimation process M33 and the excretion type input by the service provider Ub is less than a predetermined threshold, the processor 11 may determine that the odor sensor is attached to a service user Ua other than the service user Ua associated with the odor sensor.
[0063] In the present invention, the sub-sensor is not limited to a temperature sensor. Any sensor whose output signal changes depending on whether the sensor device 4 is attached to the service user Ua can be used as the sub-sensor. Examples of the sub-sensor include a temperature sensor and a proximity sensor. Examples of the proximity sensor include a capacitance type proximity sensor.
[0064] (summary) The teacher data generation device of aspect 1 of the present invention has at least one processor, and the processor executes an attachment determination process for determining whether or not the odor sensor is attached to the service user based on an output signal of a sub-sensor installed near the odor sensor associated with the service user, a first acquisition process for acquiring the output signal of the odor sensor, a second acquisition process for acquiring type information indicating the type of excretion of the service user, which type is determined by the service provider, and a teacher data generation process for generating teacher data including sensor information indicating the output signal acquired in the first acquisition process from the odor sensor determined to be attached to the service user in the attachment determination process and the type information acquired in the second acquisition process.
[0065] The teacher data generation device of aspect 2 of the present invention is characterized in that, in the above-mentioned aspect 1, the processor further executes a notification process to notify a notification terminal used by the service provider that excretion of the service user has been detected, and the type information is type information input by the service provider using the notification terminal.
[0066] The teacher data generation device according to a third aspect of the present invention is characterized in that, in the above-mentioned first or second aspect, the sensor information is image data including a graph of the output signal of the odor sensor.
[0067] A teacher data generation method according to aspect 4 of the present invention includes an attachment determination process in which at least one processor determines whether or not an odor sensor is worn by a service user based on an output signal of a sub-sensor installed near an odor sensor associated with the service user; a first acquisition process in which the processor acquires the output signal of the odor sensor; a second acquisition process in which the processor acquires type information indicating the type of excretion of the service user, the type of excretion determined by the service provider; and a teacher data generation process in which the processor generates teacher data including sensor information indicating the output signal acquired in the first acquisition process from the odor sensor determined in the attachment determination process to be worn by the service user and the type information acquired in the second acquisition process.
[0068] A model generation device according to aspect 5 of the present invention includes at least one processor, and the processor executes a model generation process to generate a model by machine learning using teacher data generated by the teacher data generation device according to any one of aspects 1 to 3 above, the model having sensor information representing an output signal of an odor sensor as input and type information representing the type of excretion as output.
[0069] An estimation device according to aspect 6 of the present invention includes at least one processor, and the processor executes an estimation process to estimate the type of excretion of a service user from the output signal of an odor sensor associated with the service user using a model generated by the model generation device according to aspect 5 above.
[0070] A program according to a seventh aspect of the present invention is a program for causing a computer to function as the teacher data generation device according to any one of the first to third aspects described above.
[0071] A model generation device according to aspect 8 of the present invention includes at least one processor, and the processor executes a model generation process to generate a model having sensor information representing an output signal of an odor sensor as input and type information representing a type of excretion as output by machine learning using training data including sensor information representing the output signal of an odor sensor that is associated with a service user and determined to be attached to the service user, and type information indicating the type of excretion of the service user, the type of excretion being determined by the service provider.
[0072] The model generation device of aspect 9 of the present invention is characterized in that, in the model generation process, the processor generates a model corresponding to each service user by machine learning using training data including sensor information corresponding to the output signal of an odor sensor determined to be attached to the service user and type information indicating the type of excretion of the service user, which is a model in which sensor information representing the output signal of an odor sensor and type information indicating the type of excretion determined by the service provider is used as input and type information corresponding to the service user is used as output.
[0073] A model generating device according to a tenth aspect of the present invention, in the above eighth or ninth aspect, is characterized in that the sensor information is image data including a graph of an output signal of the odor sensor.
[0074] A model generation device according to aspect 11 of the present invention, in any one of aspects 8 to 10 above, further executes an attachment determination process for determining whether or not the odor sensor is attached to the service user based on an output signal of a sub-sensor installed near an odor sensor associated with the service user, a first acquisition process for acquiring an output signal of the odor sensor, a second acquisition process for acquiring type information representing the type of excretion of the service user determined by the service provider, and a teacher data generation process for generating teacher data including sensor information representing the output signal acquired in the first acquisition process from the odor sensor determined to be attached to the service user in the attachment determination process and the type information acquired in the second acquisition process, and is characterized in that in the model generation process, the processor generates a model by machine learning using the teacher data generated in the teacher data generation process, in which sensor information representing the output signal of the odor sensor is used as input and type information representing the type of excretion is used as output.
[0075] A model generation method according to aspect 12 of the present invention includes a model generation process in which at least one processor generates a model that takes sensor information representing an output signal of an odor sensor as input and outputs type information representing a type of excretion by machine learning using training data that includes sensor information representing an output signal of an odor sensor that is associated with a service user and determined to be attached to the service user, and type information indicating the type of excretion of the service user, the type of excretion being determined by the service provider.
[0076] An estimation device according to aspect 13 of the present invention includes at least one processor, and the processor executes an estimation process to estimate the type of excretion of a service user from the output signal of an odor sensor associated with the service user using a model in which sensor information representing the output signal of an odor sensor is input and type information representing the type of excretion is output.
[0077] The estimation device of aspect 14 of the present invention is characterized in that, in the estimation process, the processor estimates the type of excretion of the service user from the output signal of the odor sensor associated with the service user using a model associated with the service user, the model receiving sensor information representing the output signal of an odor sensor and outputting type information representing the type of excretion.
[0078] The estimation device according to a fifteenth aspect of the present invention is characterized in that, in the thirteenth or fourteenth aspect above, the sensor information is image data including a graph of the output signal of the odor sensor.
[0079] The estimation device of aspect 16 of the present invention is characterized in that, in any one of aspects 13 to 15 above, the processor further executes a model generation process to generate a model having sensor information representing the output signal of an odor sensor as input and type information representing a type of excretion as output by machine learning using training data including sensor information representing the output signal of an odor sensor associated with a service user and determined to be attached to the service user, and type information indicating the type of excretion of the service user, which type is determined by the service provider, and in the estimation process, the processor estimates the type of excretion of the service user from the output signal of the odor sensor associated with the service user using the model generated in the model generation process.
[0080] The estimation device of aspect 17 of the present invention is characterized in that, in the above-mentioned aspect 16, the processor further executes an attachment determination process for determining whether or not the odor sensor is attached to the service user based on an output signal of a sub-sensor installed near the odor sensor associated with the service user, a first acquisition process for acquiring the output signal of the odor sensor, a second acquisition process for acquiring type information representing the type of excretion of the service user determined by the service provider, and a teacher data generation process for generating teacher data including sensor information representing the output signal acquired in the first acquisition process from the odor sensor determined to be attached to the service user in the attachment determination process and the type information acquired in the second acquisition process, and in the model generation process, the processor generates a model by machine learning using the teacher data generated in the teacher data generation process, which takes sensor information representing the output signal of the odor sensor as input and outputs type information representing the type of excretion.
[0081] An estimation method according to aspect 18 of the present invention includes an estimation process in which at least one processor estimates the type of excretion of a service user from the output signal of an odor sensor associated with the service user using a model in which sensor information representing the output signal of an odor sensor is input and type information representing the type of excretion is output.
[0082] (Additional Notes) The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in the above-described embodiments are also included in the technical scope of the present invention. [Explanation of symbols]
[0083] 1 Main device 2 Notification terminal 3. Relay terminal 4 Sensor device S Notification System Ua Service User Ub Service Provider 11 Processors 12 Primary Memory 13 Secondary Memory 14 Communication Interface 15 Bus
Claims
1. At least one processor, The processor, a wearing determination process for determining whether or not the odor sensor associated with the service user is worn by the service user based on an output signal of a sub-sensor provided near the odor sensor; A first acquisition process for acquiring an output signal of the odor sensor; A second acquisition process for acquiring type information indicating a type of excretion of the service user, the type being determined by a service provider; a teacher data generation process for generating teacher data including sensor information representing an output signal acquired in the first acquisition process from the odor sensor determined to be worn by the service user in the wearing determination process and type information acquired in the second acquisition process; A teacher data generating device comprising:
2. The processor further executes a notification process of notifying a notification terminal used by the service provider that excretion of the service user has been detected; the type information is type information input by the service provider using the notification terminal; 2. The teacher data generating device according to claim 1 .
3. The sensor information is image data including a graph of an output signal of the odor sensor.
2. The teacher data generating device according to claim 1 .
4. a wearing determination process in which at least one processor determines whether or not the odor sensor associated with the service user is worn by the service user based on an output signal of a sub-sensor provided near the odor sensor; a first acquisition process in which the processor acquires an output signal of the odor sensor; a second acquisition process in which the processor acquires type information indicating a type of excretion of the service user determined by a service provider; and a teacher data generation process in which the processor generates teacher data including sensor information representing an output signal acquired in the first acquisition process from the odor sensor determined to be worn by the service user in the wearing determination process and type information acquired in the second acquisition process. A teacher data generating method comprising:
5. At least one processor, The processor, A model generation process is executed to generate a model by machine learning using teacher data generated by the teacher data generation device according to any one of claims 1 to 3, the model having sensor information representing an output signal of an odor sensor as an input and type information representing a type of excretion as an output. A model generating device comprising:
6. At least one processor, The processor, executing an estimation process for estimating the type of excretion of the service user from an output signal of an odor sensor associated with the service user, using a model generated by the model generation device according to claim 5; An estimation device comprising:
7. A program for causing a computer to function as the teacher data generation device according to any one of claims 1 to 3.
Citation Information
Patent Citations
Organism monitoring device
JP1997043182A
