JP7805271B2 - Sweat rate estimation system and bathroom warning system - Google Patents

Description

The present invention relates to a sweat rate estimation system that estimates the amount of sweat produced by a bathroom user, and a bathroom warning system that warns of possible changes in physical condition based on the amount of sweat estimated by the sweat rate estimation system.

Bathroom users are prone to sweating due to the temperature and humidity of the bathroom, as well as the environment while bathing. If users sweat too much, they may experience health problems such as heatstroke.

Patent Document 1 discloses a technology that uses a biosensor attached to the human body to measure changes in activity level, body temperature, and other data while bathing and not bathing, and obtains the amount of sweating.

JP 2018-26006 A

However, in order to improve the accuracy of predicting abnormalities in physical condition in advance, there is a need for an easier and more accurate way to estimate sweat rate.

The present invention aims to easily and accurately estimate sweat rate.

To achieve the above object, one embodiment of the present invention provides a sweat rate estimation system for estimating the amount of sweat of a user using a bathroom, and includes: an information acquisition unit for acquiring environmental information for estimation; an estimation feature generation unit for generating estimation features from the user information of the user and the environmental information for estimation; a model acquisition unit for acquiring a trained model that outputs an estimated amount of sweat when the estimation features are input, by machine learning multiple training input data linked to training features and training data that represent the amount of sweat generated when bathing in specified environmental information for learning; and a sweat rate estimation unit that inputs the estimation features into the trained model to output the estimated amount of sweat. The training features include the user information and the environmental information for learning, and the user information includes age, gender, and weight. The environmental information for estimation includes bathroom temperature, bathroom humidity, bathtub temperature, total bathing time, continuous bathing time, and bathroom usage time. The environmental information for learning includes the bathroom temperature, bathroom humidity, bathtub temperature, total bathing time, continuous bathing time, and bathroom usage time.

This configuration makes it possible to easily and accurately estimate the amount of sweat produced by a user while bathing or in the bathroom based on the user's characteristics and the bathroom environment.

In particular, the amount of sweat is estimated based on the user's age, gender, and weight, as well as bathroom temperature, bathroom humidity, bathtub temperature, total bathing time, continuous bathing time, and bathroom usage time, which have a significant impact on human sweating, allowing for more accurate estimation of sweating.

It is also preferable that the estimation environmental information is continuously acquired at predetermined time intervals, and the sweat rate estimation unit continuously outputs the estimated sweat rate.

Users sweat from time to time while bathing, and sweating is strongly influenced by environmental information that changes in a variety of ways. With the above configuration, environmental information for estimation is continuously acquired, and the estimated amount of sweating is also continuously output accordingly. As a result, it is possible to check the changing or increasing amount of sweating, and to accurately estimate the amount of sweating.

It is also preferable that the device further includes a feature calculation unit, and in order to output the estimated sweat rate using the trained model, the degree of influence of each of the estimation feature quantities is measured in advance, and the feature calculation unit assigns a weight to each of the estimation feature quantities according to the degree of influence.

With this configuration, sweat rate can be estimated more accurately by using estimation features that take into account the degree of influence on sweating.

The information acquisition unit may also include a human presence sensor, a temperature sensor, a humidity sensor, a water temperature sensor, a water pressure sensor, a weight scale, and a timer, wherein the bathroom temperature is measured by the temperature sensor, the bathroom humidity is measured by the humidity sensor, and the bathtub temperature is measured by the water temperature sensor, the total bathing time and the continuous bathing time are calculated from the change in the bathtub water level calculated from the change in the detected value of the water pressure sensor and the time the user is immersed in the bathtub determined by the timer, and the bathroom usage time may be determined by the human presence sensor and the timer as the time the user is in the bathroom.

This configuration makes it easy to obtain environmental information for estimation and environmental information for learning, making it easy to estimate sweat rate.

Furthermore, the information acquisition unit may include an imaging device, and at least one of the estimation environment information and the learning environment information may include an image captured by the imaging device.

The images make it easy to check the movements of the user's head, shoulders, eyelids, etc. Because user movement can affect the amount of sweat produced, checking the user's movement allows for a more accurate estimation of the amount of sweat produced.

The model acquisition unit may also generate the trained model by performing machine learning on the training input data.

This configuration eliminates the need to prepare a separate trained model, making it easy to obtain a trained model.

Furthermore, the information acquisition unit may be provided in the bathroom, and the model acquisition unit, the estimation feature generation unit, and the sweat rate estimation unit may be provided on a server that communicates data with the bathroom.

With this configuration, the model acquisition unit, estimation feature generation unit, and sweat rate estimation unit, which may require large processing power, can be placed on a server where processing power can be easily increased, thereby simplifying the configuration within the bathroom.

The trained model may also be a gradient boosting decision tree model.

This configuration makes it possible to generate a trained model with high estimation accuracy.

Furthermore, a bathroom warning system according to one embodiment of the present invention includes the sweat rate estimation system, a notification unit, a sweat rate determination unit that determines whether the estimated sweat rate is equal to or greater than a predetermined threshold, and a warning control unit that causes the notification unit to issue a predetermined warning when the sweat rate determination unit determines that the estimated sweat rate is equal to or greater than the threshold.

If the amount of sweating becomes too much, the user's physical condition may change due to heatstroke or other issues. With the above configuration, it is possible to predict the possibility of a change in physical condition from the amount of sweating. Then, by issuing a warning before a change in physical condition occurs, it is possible to alert the user.

The warning may also include at least one of the following: emitting a warning sound, illuminating a warning light, displaying a message on a display unit, or sending a notification to a pre-registered mobile device.

This configuration makes it easy to draw attention to the user or those who may be able to assist the user.

The threshold may also be determined based on the user information.

The amount of sweating that leads to changes in physical condition varies from person to person. With the above configuration, it is possible to predict the possibility of a change in physical condition from the amount of sweating, based on the impact on physical condition of an individual identified by user information or the amount of sweating estimated from the user information. This makes it possible to issue a warning with greater accuracy before a change in physical condition occurs.

The system may further include an air conditioning unit installed in the bathroom and a bathroom control unit that controls the air conditioning unit, and the bathroom control unit may operate the air conditioning unit when the sweat rate determination unit determines that the estimated sweat rate is equal to or greater than the threshold.

With this configuration, when the user's sweating rate increases, the air conditioning equipment can take action to reduce the amount of sweating by ventilating, blowing air, cooling, etc. This reduces the amount of sweating by the user and prevents any abnormalities in the user's physical condition.

FIG. 1 is a schematic diagram illustrating the configuration of a bathroom. FIG. 1 is a diagram illustrating an outline of a configuration for estimating the amount of sweating. FIG. 1 is a diagram illustrating a schematic configuration of a bathroom warning system. 3 is a diagram illustrating information stored in a storage unit included in the remote controller. FIG. FIG. 2 is a diagram illustrating information stored in a storage unit included in the server. FIG. 10 is a diagram illustrating a flow for generating a trained model. FIG. 10 is a diagram illustrating a flow for estimating an estimated amount of sweat. FIG. 10 is a diagram illustrating a flow for issuing a warning. FIG. 10 is a diagram illustrating an example of the estimation accuracy of the amount of sweating. FIG. 10 is a diagram illustrating an example of the degree of influence of a feature amount.

The sweat rate estimation system and bathroom warning system of this embodiment will be described with reference to the drawings. The sweat rate estimation system estimates the amount of sweat produced by a user using the bathroom. The bathroom warning system predicts in advance that there may be a change in the user's physical condition based on the estimated amount of sweat, and issues a warning.

[Overview of sweat rate estimation system and bathroom warning system]
As shown in FIGS. 1 to 3, the bathroom is provided with a bathtub 3, a remote control 1, an air conditioning unit 5, an information acquisition unit 6, and the like.

The bathtub 3 holds hot water for users to bathe in. The air conditioning unit 5 can ventilate the bathroom, cool and heat it, and also function as a bathroom dryer.

The information acquisition unit 6 includes a human presence sensor 12, a temperature sensor 13, a humidity sensor 14, a water temperature sensor 15, a water pressure sensor 16, a weight scale 17, a timer 18, and other devices installed in the bathroom. The information acquisition unit 6 acquires estimation environment information 9 and learning environment information 8, which will be described later.

As shown in Figure 2, the sweat rate estimation system estimates the user's estimated sweat rate 58 by inputting estimation features 57 into a pre-generated trained model 55.

The trained model 55 is generated by inputting training input data 53 into AI 60 (Artificial Intelligence) and performing machine learning. The training input data 53 is information that links training features 51 and teacher data 52. The training features 51 are generated from user information 7 and training environment information 8 that are acquired in advance.

The learning input data 53 used to generate the trained model 55 is generated based on multiple pieces of information collected in advance about sweating by a certain user when actually bathing in a bathroom in a certain environment. When a piece of information is collected, the user information 7 of the user, learning environment information 8 which is the bathing environment, and training data 52 which is the amount of sweating at that time are linked to each other and stored.

The learning input data 53 may be information including multiple sets of linked user information 7, learning environment information 8, and training data 52, or the user information 7 and learning environment information 8 may be learning features 51 to which weighting is assigned taking into account the effect on sweating, as described below. In this case, the learning input data 53 is information including multiple sets of linked learning features 51 and training data 52.

The estimation feature quantity 57 is generated from the user information 7 and the estimation environment information 9. The estimation environment information 9 is various environmental information about the time when the user, whose sweat rate is to be estimated, is bathing.

The bathroom warning system predicts the possibility of a change in the user's physical condition, such as the user developing heatstroke, based on the estimated sweat rate 58 estimated by the sweat rate estimation system. If the bathroom warning system determines that there is a possibility of a change in the user's physical condition, it issues a predetermined warning in advance.

The sweat rate estimation system uses the trained model 55 to output the user's estimated sweat rate 58, allowing the estimated sweat rate 58 to be estimated easily and accurately.

The bathroom warning system can predict the possibility of a change in the user's physical condition based on the accurately estimated amount of sweat 58, and can therefore issue a warning before a change in the user's physical condition occurs, thereby preventing the user's physical condition from changing.

[Information acquisition department]
Next, the information acquisition unit 6 and the information acquired by the information acquisition unit 6 will be described with reference to FIG. 1 and FIGS.

The human presence sensor 12 detects people in the bathroom, and the timer 18 measures the time the user is in the bathroom to obtain bathroom usage time 48. The temperature sensor 13 obtains bathroom temperature 41, which is the temperature in the bathroom. The humidity sensor 14 obtains bathroom humidity 42, which is the humidity in the bathroom. For example, the temperature sensor 13 and humidity sensor 14 may be installed in any position in the bathroom, or in the air conditioning unit 5. The water temperature sensor 15 obtains bathtub temperature 43, which is the temperature of the water in the bathtub 3. The water temperature sensor 15 may be installed in any position as long as it can measure the temperature of the water in the bathtub 3, but may also be installed in the piping between the bathtub 3 and the water heater, for example. If the water temperature sensor 15 is installed in the piping between the bathtub 3 and the water heater, it can also measure the temperature of the water supplied to the bathtub 3.

The water pressure sensor 16 measures the water pressure of the hot water in the bathtub 3, and the water level of the hot water in the bathtub 3 (bathtub water level 44) is calculated from the water pressure (detected value). The water level may be calculated by the water pressure sensor 16, or by the remote control 1, a processor in the remote control 1, or any functional block. The water pressure sensor 16 may be installed in any position as long as it can measure the water pressure of the hot water in the bathtub 3, but it may also be installed in the piping between the bathtub 3 and the water heater, for example.

The weighing scale 17 measures the user's weight 7D. The weighing scale 17 may be installed in any location as long as it is capable of measuring the weight 7D of the user using the bathroom, but it may also be installed near the entrance to the bathroom, such as on the outside of the bathroom door. This allows the user's weight 7D to be measured immediately before entering the bathroom.

In addition, the water pressure sensor 16 determines whether the user is soaking in the bathtub 3 based on changes in the water level, and the timer 18 measures the time the user is soaking in the bathtub 3. This makes it possible to obtain the continuous bathing time 47 during which the user is continuously bathing in the bathtub 3, and the total bathing time 46.

[Remote control]
Next, the remote control 1 constituting the sweat rate estimation system and the bathroom warning system will be described with reference to FIGS. 3 and 4 as well as FIGS. 1 and 2. FIG.

Remote control 1 accepts and controls settings for hot water temperature and volume, accepts and controls settings for air conditioning equipment 5, and accepts and controls various other settings for the bathroom. Remote control 1 is equipped with a processor (computer) such as a CPU, which controls each functional block of remote control 1. Multiple remote controls 1 may be installed, and they may be installed in any location, such as inside the bathroom, near the bathroom entrance, or elsewhere.

The remote control 1 includes an operation unit 1A, a display unit 1B, a speaker 1C, a warning light 1D, a communication unit 31, an estimation feature generator 33 (corresponding to a feature calculator), a bathroom controller 34, a sweat rate determiner 35, a warning controller 36, and a memory unit 38.

Operation unit 1A accepts various operations on remote control 1. Operation unit 1A may be a combination of various switches such as a numeric keypad, levers, dials, or other operating devices, and may also include software switches displayed on display unit 1B.

The display unit 1B is a monitor that displays various information. The speaker 1C issues audio information, warning sounds, etc. The warning light 1D displays various information such as warnings by emitting light. The communication unit 31 performs data communication with the server 2 (described below) and other remote controls 1. The communication unit 31 may also be configured to perform data communication with a mobile device such as a smartphone.

The estimation feature generator 33 generates estimation feature values 57 for estimating the estimated sweat rate 58 and stores them in the memory unit 38. The bathroom controller 34 adjusts the water temperature, amount of water, etc., in accordance with the operations received by the operating unit 1A, and controls the air conditioning equipment 5.

The sweat rate determination unit 35 predicts the possibility of an abnormality in the user's physical condition, such as heat stroke, based on the estimated sweat rate 58.

The warning control unit 36 issues a predetermined warning if there is a possibility that an abnormality will occur in the user's physical condition. The warning control unit 36 controls the display unit 1B, speaker 1C, and warning light 1D, and issues the warning by displaying the display unit 1B, issuing a warning sound or voice from the speaker 1C, illuminating the warning light 1D, etc.

The memory unit 38 stores a personal information database 10, weight 7D, user information 7, estimated environmental information 9, estimated feature values 57, and estimated sweat rate 58.

The personal information database 10 is a database that stores information on multiple users, and each user is linked to user information 7. The user information 7 includes age 7A, gender 7B, weight 7D, and may also include height 7C. In other words, the personal information database 10 can identify the user information 7 of an individual user by identifying that user.

Weight 7D is the user's weight measured by the scale 17 when the user enters or leaves the bathroom. Weight 7D does not have to be measured by the scale 17, but may also be input by the user via the operation unit 1A.

The estimated environmental information 9 includes bathroom temperature 41, bathroom humidity 42, bathtub temperature 43, total bathing time 46, continuous bathing time 47, and bathroom usage time 48, all of which are acquired by the information acquisition unit 6. Furthermore, the estimated environmental information 9 may also include bathtub water level 44, which is acquired by the information acquisition unit 6. In other words, the information acquisition unit 6 stores the acquired information as estimated environmental information 9 in the memory unit 38.

It is preferable that the estimated environmental information 9 be continuously acquired at predetermined time intervals (at predetermined intervals) while the user is in the bathroom (while bathing). This allows the amount of sweat to be continuously estimated while the user is bathing. For example, the estimated environmental information 9 is acquired every 5 seconds.

The estimation feature 57 is a feature weighted according to the degree of influence that the user information 7 and the estimation environmental information 9 have on estimating the amount of sweating. The trained model 55 is analyzed in advance to determine the degree of influence that each piece of information in the user information 7 and the estimation environmental information 9 has on estimating the amount of sweating. For example, the weighting can be determined by adjusting the parameters (weighting) of each piece of information in the user information 7 and the estimation environmental information 9 so as to minimize the root mean square error of the estimation result. The estimation feature generation unit 33 performs a calculation to assign a weight according to the degree of influence to each piece of information in the user information 7 and the estimation environmental information 9, generates the estimation feature 57, and stores it in the memory unit 38.

The estimated sweat rate 58 is the amount of sweat estimated by the server 2, as described below, and is obtained from the server 2 via the communication unit 31.

〔server〕
Next, the server 2 constituting the sweat rate estimation system and the bathroom warning system will be described with reference to FIGS. 3 and 5 as well as FIGS. 1 and 2. FIG.

The server 2 is configured to be able to communicate data with the remote control 1 via any communication line such as the Internet. Data communication may also be performed via a mobile terminal such as a smartphone that communicates data with the remote control 1 via a wireless LAN or the like. The server 2 also includes a processor (computer) such as a CPU, which controls each functional block of the server 2.

The server 2 includes a communication unit 21, a learning feature generation unit 23, a learning input data generation unit 24, a learning unit 25, a sweat rate estimation unit 27, and a memory unit 28.

The communication unit 21 communicates data with the remote control 1 via the communication unit 31 of the remote control 1. Specifically, the server 2 receives estimation features 57 from the remote control 1 via the communication units 21 and 31, and transmits an estimated sweat rate 58 to the remote control 1. The learning feature generation unit 23 calculates learning features 51 from a previously acquired learning dataset 11, as described below.

The learning input data generation unit 24 links the learning features 51 with the training data 52 to generate learning input data 53, which is then stored in the memory unit 28.

The learning unit 25 is equipped with AI 60, and generates a trained model 55 by inputting learning input data 53 into the AI 60 and performing machine learning, and stores the trained model 55 in the memory unit 28.

For example, machine learning is performed using linear regression models such as gradient boosting decision tree models.

The sweat rate estimation unit 27 inputs the estimation feature 57 into the trained model 55 to estimate the estimated sweat rate 58 and stores it in the memory unit 28.

The memory unit 28 stores the training dataset 11, training features 51, training data 52, training input data 53, trained model 55, estimation features 57, and estimated sweat rate 58.

The learning dataset 11 includes multiple sets of user information 7 and learning environment information 8 that are linked to each other. The user information 7 includes age 7A, gender 7B, and weight 7D, and may also include height 7C. The learning environment information 8 includes bathroom temperature 41, bathroom humidity 42, bathtub temperature 43, bathtub water level 44, total bathing time 46, continuous bathing time 47, and bathroom usage time 48.

Like the estimation features 57, the learning features 51 are features to which weighting has been assigned according to the degree of influence that the user information 7 and learning environment information 8 have on estimating the amount of sweat. The magnitude of the weighting may be the same as that of the estimation features 57, or may be determined separately. The learning feature generation unit 23 performs a calculation to assign weighting according to the degree of influence to each piece of information in the user information 7 and learning environment information 8, and generates learning features 51, which are then stored in the memory unit 28.

The teacher data 52 is the amount of sweat actually produced by a user under the conditions of a set of linked user information 7 and learning environment information 8. In other words, a user corresponding to the user information 7 previously bathes in an environment corresponding to the learning environment information 8, and the amount of sweat produced during this bathing becomes the teacher data 52. The amount of sweat produced when multiple users bathe in various environments is then determined, and multiple sets of teacher data 52 corresponding to the user information 7 and learning environment information 8 are obtained and stored in the memory unit 28.

The learning input data 53 is a collection of multiple pairs of learning features 51 and training data 52 linked to the learning features 51. The learning input data generation unit 24 links the learning features 51 and training data 52 corresponding to each piece of user information 7 and learning environment information 8 to generate learning input data 53, and stores the data in the memory unit 28.

The trained model 55 is a model generated by the training unit 25, and outputs an estimated sweat rate 58 in response to input of the estimation feature 57. The sweat rate estimation unit 27 inputs the estimation feature 57 into the trained model 55, thereby estimating the estimated sweat rate 58 and storing it in the memory unit 28.

The estimation feature 57 is information generated and stored in the remote control 1 and acquired via the communication unit 21.

Since machine learning is performed using such detailed user information 7 and detailed learning environment information 8 to generate a trained model 55, it is possible to estimate the estimated sweat rate 58 with high accuracy. In particular, since learning features 51 are generated from the user information 7 and learning environment information 8, with weighting assigned according to the degree of influence they have on estimating the sweat rate, and machine learning is performed using these learning features 51 to generate a trained model 55, it is possible to estimate the estimated sweat rate 58 with high accuracy.

The sweat rate estimation system is composed of the information acquisition unit 6, the estimation feature generation unit 33 of the remote control 1, and the learning feature generation unit 23, learning input data generation unit 24, learning unit 25, and sweat rate estimation unit 27 of the server 2. In addition to the sweat rate estimation system, the bathroom warning system also includes a notification unit including at least one of the display unit 1B, speaker 1C, and warning light 1D, and the sweat rate determination unit 35 and warning control unit 36 of the remote control 1.

[Sweat rate estimation system]
Next, with reference to FIGS. 1 to 5, the processing of the sweat rate estimation system, which includes generating the trained model 55 and estimating the estimated sweat rate 58, will be described using FIGS.

First, a training dataset 11 is collected to generate a trained model 55 (step #1 in Figure 6). Specifically, a user (verifier) actually takes a bath, and the amount of sweat is obtained using a predetermined method. This amount of sweat is used as training data 52. At this time, training environment information 8 during bathing is obtained. In addition, user information 7 of the user is stored. The user information 7, training environment information 8, and training data 52 are then linked and stored. Such information is obtained in advance for different users and in different environments, and stored in the memory unit 28 of the server 2.

Next, the learning feature generation unit 23 calculates (generates) learning features 51 from a learning dataset 11 consisting of multiple pieces of user information 7 and learning environment information 8 that are linked to each other, and stores them in the memory unit 28 (step #2 in Figure 6).

At this time, the degree of influence of each piece of information constituting the user information 7 and the learning environment information 8 on estimating the amount of sweating (estimated amount of sweating 58) is calculated in advance using any method. For example, a previously generated trained model 55 is verified to determine the degree of influence of each piece of information.

The learning feature generator 23 generates learning features 51 by performing a calculation to assign weights according to the magnitude of this influence to each piece of information constituting the user information 7 and learning environment information 8, and stores these in the memory unit 28. The learning features 51 are generated for each set of linked user information 7 and learning environment information 8.

Next, the learning input data generation unit 24 links the training features 51 to the training data 52 linked to the user information 7 and learning environment information 8 that form the basis of the training features 51, to generate training input data 53, and stores the training input data 53 in the memory unit 28 (step #3 in Figure 6). This generates training input data 53 that includes all of the training datasets 11 and training data 52 that were previously collected.

Next, the learning unit 25 inputs the generated learning input data 53 into the AI 60 for machine learning, thereby generating a trained model 55 (step #4 in Figure 6). The trained model 55 is generated prior to estimating the amount of sweat and is stored in the memory unit 28.

At this time, the generated trained model 55 may be analyzed to determine the magnitude of the influence (importance) on estimating the estimated sweat rate 58, and the magnitude of the influence may be updated. Furthermore, the trained model 55 may be generated again using the updated magnitude of the influence.

With this trained model 55 stored in the memory unit 28, when a user uses the bathroom, the amount of sweat produced by the user while bathing is estimated.

First, the user operates the operation unit 1A of the remote control 1 to read their own user information 7 from the personal information database 10 previously stored in the memory unit 38 (step #1 in Figure 7). The read user information 7 is stored in the memory unit 38. The user information 7 includes age 7A, gender 7B, weight 7D, and may also include height 7C. Weight 7D may be measured using a scale 17 when entering the bathroom, or may be input using the operation unit 1A. Age 7A, gender 7B, and height 7C may also be input using the operation unit 1A.

Next, the information acquisition unit 6 acquires estimated environmental information 9 and stores it in the memory unit 38 (step #2 in Figure 7). The estimated environmental information 9 includes the above-mentioned bathroom temperature 41, bathroom humidity 42, bathtub temperature 43, total bathing time 46, continuous bathing time 47, and bathroom usage time 48, and may also include the bathtub water level 44.

The information acquisition unit 6 may acquire the estimation environment information 9 at a predetermined timing while the user is bathing, but it is preferable to continuously acquire the estimation environment information 9 every 5 seconds (a predetermined time). Simply estimating the amount of sweat at a certain timing while bathing will not be able to detect a subsequent increase in sweat amount and a deterioration in physical condition. By continuously acquiring the estimation environment information 9, it is possible to continuously estimate the amount of sweat and determine the possibility of an abnormality in the user's physical condition while the user is bathing, as described below. As a result, it is possible to more accurately estimate the amount of sweat and determine the possibility of an abnormality in the user's physical condition.

The actual measured value of the amount of sweat can be obtained by subtracting the weight measured after the user leaves the bathroom from the weight 7D measured before entering the bathroom. In other words, the estimated amount of sweat 58 is estimated from the estimation environment information 9 and the user information 7, and various pieces of information used to estimate the amount of sweat are added to the learning input data 53. The actual measured value of the amount of sweat is added as training data 52, and the trained model 55 is updated, thereby improving the accuracy of the amount of sweat estimation.

As described above, it is not necessarily necessary to add the actual measured sweat rate and the environmental information at the time the actual measured sweat rate was obtained as learning environmental information 8 to the learning input data 53 and update the trained model 55, but it is preferable to do so at any time to improve the accuracy of sweat rate estimation.

Next, the estimation feature generation unit 33 generates estimation feature 57 by performing a calculation to assign weights according to the magnitude of the aforementioned influence to each piece of information constituting the user information 7 and the estimation environment information 9, and outputs the estimation feature 57 to the storage unit 38 (step #3 in Figure 7). If estimation environment information 9 is continuously acquired, the estimation feature generation unit 33 generates estimation feature 57 each time it is acquired.

Next, the remote control 1 transmits the generated feature for estimation 57 to the server 2 via the communication unit 31. The server 2 receives the transmitted feature for estimation 57 via the communication unit 21 and stores it in the storage unit 28 (step #4 in FIG. 7). The remote control 1 transmits the feature for estimation 57 to the server 2 each time a feature for estimation 57 is generated.

Next, the sweat rate estimation unit 27 of the server 2 estimates the amount of sweat from the feature for estimation 57. Specifically, the sweat rate estimation unit 27 inputs the received feature for estimation 57 into the trained model 55 stored in the memory unit 28, thereby outputting an estimated amount of sweat 58 (step #5 in Figure 7). The sweat rate estimation unit 27 of the server 2 estimates the estimated amount of sweat 58 each time it receives the feature for estimation 57.

The server 2 then transmits the output estimated sweat rate 58 to the remote control 1 via the communication unit 21 (step #6 in Figure 7). Each time the estimated sweat rate 58 is output, the server 2 transmits the estimated sweat rate 58 to the remote control 1.

[Bathroom warning system]
Next, the processing of the bathroom warning system will be described using FIG. 8 while referring to FIGS. 1 to 5.

First, the remote control 1 receives the transmitted estimated sweat rate 58 via the communication unit 31 and stores it in the memory unit 38 (step #1 in Figure 8).

Next, the sweat rate determination unit 35 of the remote control 1 determines, from the estimated sweat rate 58, whether the user is in a state where there is a risk of developing an abnormality in their physical condition due to heat stroke or the like (step #2 in FIG. 8). Specifically, the sweat rate determination unit 35 checks whether the estimated sweat rate 58 is equal to or greater than a predetermined threshold, and if the estimated sweat rate 58 is equal to or greater than the threshold, determines that there is a risk of developing an abnormality in their physical condition. Each time the estimated sweat rate 58 is received, the sweat rate determination unit 35 determines whether there is a risk of developing an abnormality in their physical condition. The threshold is the amount of sweat at which there is a risk of developing heat stroke or the like, and can be, for example, the amount of sweat that is 3% of body weight 7D.

It is possible to prepare multiple thresholds, and determine the likelihood of an abnormality occurring based on the threshold. The threshold may also be determined based on the user information 7, or a predetermined function may be used as the threshold.

The warning control unit 36 then issues a warning via a specified notification unit as necessary based on the determination result. If the estimated sweat rate 58 is less than the threshold value (No in step #2 of Figure 8), the warning control unit 36 does not issue a warning and waits for the next determination result.

If the estimated sweat rate 58 is equal to or greater than the threshold (step #2 in Figure 8, Yes), the warning control unit 36 causes the notification unit to issue a warning (step #3 in Figure 8).

The notification unit is, for example, at least one of a display unit 1B, a speaker 1C, and a warning light 1D. The display unit 1B displays warning text, pictures, etc. as a warning. The speaker 1C issues a warning sound or voice as a warning. The warning light 1D emits a predetermined light, such as a steady light or flashing light, as a warning. It is also preferable that these warnings be issued on a remote control 1 other than the remote control 1 in the bathroom. The warning may also be issued on a mobile device such as a smartphone carried by another person, such as a family member of the user. By having a person other than the user receive the warning, the person who receives the warning can take appropriate action even if the user is unable to deal with the situation on their own.

In addition, if a high probability of an abnormality occurring is determined, the display content, warning sound, light color, and light pattern will be changed according to the probability.

In this way, a warning is given in advance of the possibility of an abnormality occurring in the user's physical condition, before the abnormality occurs. This allows the user or others to take appropriate action before an abnormality occurs in the user.

Furthermore, if the estimated sweat rate 58 is equal to or greater than the threshold value (Step #2 in Figure 8: Yes), the bathroom control unit 34 may control the air conditioning equipment 5 to start ventilation, air circulation, or cooling, or may lower the hot water temperature. This can prevent the user's physical condition from worsening.

[Prediction accuracy]
As shown in FIG. 9, as a result of cross-validation of the learning input data 53, the estimation accuracy of the estimated sweat rate 58 using the trained model 55 was good.

In other words, the results of the residual analysis showed that regardless of the estimated sweat rate (estimated amount), no significant estimation error was observed, and the distribution of errors was concentrated near zero.

Furthermore, analysis of the trained model 55 revealed the influence (importance) of each feature on the estimation, as shown in Figure 10. As shown in Figure 10, the influence, in descending order, is as follows: bathroom temperature at the time of judgment, height 7C, initial bathroom humidity (described below), weight 7D, etc.

The training features 51 and estimation features 57 are weighted according to their importance as shown in Figure 10.

[Another embodiment]
(1) Either the learning environment information 8 or the estimation environment information 9 may include at least one of the user's volume, the size of the bathtub 3, and the hot water supply temperature. The user's volume is calculated from changes in the water level obtained by changes in the water pressure sensor 16. The size of the bathtub 3 is at least one of the amount of hot water contained in the bathtub 3, the depth of the bathtub 3, the length of the bathtub 3, and the width of the bathtub 3. The hot water supply temperature is the temperature of the hot water supplied to the bathtub 3 while the user is bathing. Furthermore, either the learning environment information 8 or the estimation environment information 9 may include various other information. By including this information in either the learning environment information 8 or the estimation environment information 9, the trained model 55 can be generated and the estimated sweat rate 58 can be estimated with greater accuracy.

(2) The bathroom temperature 41 in either the learning environment information 8 or the estimation environment information 9 may include the bathroom temperature measured at a predetermined timing, for example, at predetermined intervals (bathroom temperature at determination), as well as the bathroom temperature when the user enters the bathroom (initial bathroom temperature). Similarly, the bathroom humidity 42 in either the learning environment information 8 or the estimation environment information 9 may include the bathroom humidity measured at a predetermined timing, for example, at predetermined intervals (bathroom humidity at determination), as well as the bathroom humidity when the user enters the bathroom (initial bathroom humidity). Furthermore, the bathtub temperature 43 in either the learning environment information 8 or the estimation environment information 9 may include the bathtub temperature 3 measured at a predetermined timing, for example, at predetermined intervals (bathtub temperature at determination), as well as the bathtub temperature 3 when the user enters the bathroom (initial bathtub temperature). This enables the generation of a trained model 55 and the estimation of an estimated sweat rate 58 to be performed with greater accuracy.

(3) The information acquisition unit 6 may include an imaging device. Either the learning environment information 8 or the estimation environment information 9 may include an image captured by the imaging device. The image makes it easy to confirm the movement of the user's head, shoulders, eyelids, etc. Because the user's movement can affect the amount of sweating, and the amount of sweating can affect the user's movement, by confirming the user's movement, it is possible to generate a trained model 55 and estimate the estimated amount of sweating 58 with greater accuracy.

(4) At least one of the estimation feature generation unit 33, the sweat amount determination unit 35, and the warning control unit 36 may be provided in the server 2 rather than in the remote control 1. That is, the remote control 1 may transmit the user information 7 and the estimation environment information 9 to the server 2. The server 2 may generate the estimation feature 57, estimate the estimated sweat amount 58, and then generate a control signal for a warning in accordance with the determination result of the estimated sweat amount 58, and transmit the control signal to the remote control 1. The remote control 1 then receives the control signal and issues a predetermined warning in accordance with the control signal.

This configuration simplifies the configuration of the remote control 1, allowing for efficient estimation of the estimated sweat rate 58 and the necessary warnings.

Conversely, the trained model 55 may be stored in the remote control 1 rather than the server 2. In this case, the sweat rate estimation unit 27 is also provided in the remote control 1 rather than the server 2. Then, in the remote control 1, the estimation feature 57 is input to the trained model 55 to estimate the estimated sweat rate 58.

This configuration makes it easy to create an optimal system for the remote control 1.

(5) The learning feature generation unit 23, the learning input data generation unit 24, and the learning unit 25 may be provided in the remote control 1 rather than in the server 2. In other words, the trained model 55 may be generated in the remote control 1 rather than in the server 2.

This configuration makes it possible to easily generate a trained model 55 that is specialized for bathrooms and users of the bathroom, thereby enabling more accurate generation of the trained model 55 and estimation of the estimated sweat rate 58.

(6) The trained model 55 does not have to be generated by the server 2, and the server 2 may be configured to acquire a trained model 55 that has been generated separately. In other words, the learning unit 25 of the server 2 may generate the trained model 55, or may function as a model acquisition unit that acquires the trained model 55. In this case, the sweat rate estimation system includes a model acquisition unit instead of the training feature generation unit 23, training input data generation unit 24, and training unit 25. Furthermore, the trained model 55 and the sweat rate estimation unit 27 may be provided in the remote control 1, and the trained model 55 may be generated by the remote control 1 or may be generated separately and transmitted to the remote control 1.

This configuration simplifies the system configuration.

(7) At least one of the remote control 1 and the server 2 is not limited to being composed of functional blocks as shown in FIG. 3, but may be composed of any functional blocks. For example, each functional block of at least one of the remote control 1 and the server 2 may be further subdivided, or conversely, some or all of the functional blocks may be combined. Furthermore, the functions of at least one of the remote control 1 and the server 2 may be realized by a method executed by any functional block, not limited to the above functional blocks. Furthermore, some or all of the functions of the remote control 1 and the server 2 may be composed of software. A program related to the software is stored in any storage device such as the memory unit 28 or the memory unit 38, and is executed by a processor such as a CPU provided in the remote control 1 or the server 2, or by a separately provided processor.

The configurations disclosed in the above embodiments (including other embodiments, the same applies below) can be applied in combination with configurations disclosed in other embodiments, provided that no contradictions arise. Furthermore, the embodiments disclosed in this specification are merely examples, and the present invention is not limited to these embodiments. They can be modified as appropriate within the scope of the purpose of the present invention.

The present invention can be applied to estimating the amount of sweat produced by bathroom users, or to issuing warnings of possible changes in physical condition based on the amount of sweat produced.

REFERENCE SIGNS LIST 1 remote control 2 server 6 information acquisition unit 7 user information 8 learning environment information 9 estimation environment information 23 learning feature generation unit 24 learning input data generation unit 25 learning unit 27 sweat rate estimation unit 33 estimation feature generation unit (feature calculation unit)
35 Sweat amount determination unit 36 Warning control unit 51 Learning feature amount 52 Teacher data 53 Learning input data 55 Trained model 57 Estimation feature amount 58 Estimated sweat amount 60 AI

Claims (12)

A sweat rate estimation system for estimating the amount of sweat of a user using a bathroom, comprising:
an information acquisition unit that acquires environmental information for estimation;
an estimation feature generating unit that generates estimation feature values from the user information of the user and the estimation environment information;
a model acquisition unit that acquires a trained model that is generated by machine learning training data that is the amount of sweating when bathing in predetermined learning environment information and a plurality of training input data linked to training features, and that outputs an estimated amount of sweating when the estimation features are input;
a sweat rate estimation unit that outputs the estimated sweat rate by inputting the estimation feature amount to the trained model;
the learning features include the user information and the learning environment information,
The user information includes age, sex, and weight,
the environmental information for estimation includes bathroom temperature, bathroom humidity, bathtub temperature, total bathing time, continuous bathing time, and bathroom usage time;
The learning environment information includes the bathroom temperature, the bathroom humidity, the bathtub temperature, the total bathing time, the continuous bathing time, and the bathroom usage time. the environmental information for estimation is continuously acquired at predetermined time intervals;
The sweat rate estimation system according to claim 1 , wherein the sweat rate estimation unit continuously outputs the estimated sweat rate. Further comprising a feature amount calculation unit,
In order to output the estimated sweat rate using the trained model, the magnitude of the influence of each of the estimation features is measured in advance;
The sweat rate estimation system according to claim 1 , wherein the feature amount calculation unit assigns a weight to each of the estimation feature amounts according to the magnitude of the degree of influence. The information acquisition unit includes a human sensor, a temperature sensor, a humidity sensor, a water temperature sensor, a water pressure sensor, a weight scale, and a timer,
the bathroom temperature is measured by the temperature sensor;
The bathroom humidity is measured by the humidity sensor,
The bathtub temperature is measured by the water temperature sensor,
The total bathing time and the continuous bathing time are calculated from the change in the bathtub water level calculated from the change in the detected value of the water pressure sensor and the time the user is immersed in the bathtub calculated by the timer,
The sweat rate estimation system according to claim 1 , wherein the bathroom usage time is determined by the human presence sensor and the timer as the time the user is in the bathroom. an imaging device is included as the information acquisition unit,
The sweat rate estimation system according to claim 1 , wherein at least one of the estimation environment information and the learning environment information includes an image captured by the imaging device. The sweat rate estimation system of claim 1, wherein the model acquisition unit generates the trained model by machine learning the training input data. the information acquisition unit is provided in the bathroom,
The sweat rate estimation system according to claim 1 , wherein the model acquisition unit, the estimation feature generation unit, and the sweat rate estimation unit are provided in a server that communicates data with the bathroom. The sweat rate estimation system of claim 1, wherein the trained model is a gradient boosting decision tree model. The sweat rate estimation system according to any one of claims 1 to 8,
The notification department,
a sweat rate determination unit that determines whether the estimated sweat rate is equal to or greater than a predetermined threshold;
and a warning control unit that causes the notification unit to issue a predetermined warning when the sweat rate determination unit determines that the estimated sweat rate is equal to or greater than the threshold. The bathroom warning system of claim 9, wherein the warning includes at least one of the following: emitting a warning sound, illuminating a warning light, displaying a message on a display unit, or sending a notification to a pre-registered mobile device. The bathroom warning system of claim 9, wherein the threshold value is determined based on the user information. an air conditioning device provided in the bathroom;
Further, a bathroom control unit that controls the air conditioning equipment is provided.
The bathroom warning system according to claim 9 , wherein the bathroom control unit activates the air conditioning device when the sweat rate determination unit determines that the estimated sweat rate is equal to or greater than the threshold value.