JP7320222B2 - Bathroom monitoring device and bathroom monitoring method - Google Patents

Description

Application of Article 30, Paragraph 2 of the Patent Act Publication date: March 6, 2018 Publication: 2018 Institute of Electronics, Information and Communication Engineers General Conference, Lecture Proceedings 1, p. Date March 21, 2018 Meeting name General Conference of the Institute of Electronics, Information and Communication Engineers 2018 Venue Tokyo Denki University (5, Senju-Asahi-cho, Adachi-ku, Tokyo) Room 7F, Room 2704

The present disclosure relates to a bathroom monitoring device and a bathroom monitoring method.

In a room, moving between a room with a high room temperature and a room with a low room temperature may cause a sudden change in blood pressure due to a sudden change in temperature. Especially in winter, there is a large temperature difference between the bathroom and the dressing room, between the toilet and the corridor, etc. etc.) are of increasing concern, especially among elderly people with reduced physical strength.

People often go in and out of places where heat shock is likely to occur, such as bathrooms and toilets, by themselves. However, from the viewpoint of privacy, it is not preferable to install a camera in a bathroom, a toilet, or the like to capture an image of a person entering the room. For this reason, various techniques have been proposed for detecting people's behavior in a bathroom or the like using a radio wave sensor instead of a camera, and notifying family members or the like when an abnormality occurs.

For example, Patent Literature 1 discloses a detection system that detects the movement of a person in a room based on the frequency difference between radio waves from a Doppler sensor and reflected radio waves from a human body and a room.

JP 2005-004256 A

However, especially in bathrooms, the Doppler sensor detects movements of objects other than the human body due to the influence of fluctuations in water flow from showers and faucets, the surface of water in bathtubs, steam, and the like. Therefore, it has been difficult to accurately detect only the movement of the human body. In addition, radio waves may be reflected many times indoors, causing multipaths and causing interference between radio signals. Therefore, even though the human body is actually moving, the reflected radio waves from the human body may interfere with other radio waves and be attenuated. may interfere with other radio waves and be strengthened. As a result, detection omissions and erroneous detections of human movements may occur.

Therefore, the present disclosure describes a bathroom monitoring device and a bathroom monitoring method that can accurately detect the state of a bather even in an environment with large disturbances.

A bathroom monitoring device according to one aspect of the present disclosure includes a transmitter configured to transmit broadband radio waves into a bathroom, and a reflected wave of the radio waves transmitted from the transmitter configured to be able to receive. a signal strength calculator configured to calculate the signal strength in each of a plurality of areas in the bathroom based on the reflected waves received by the receiver; and a signal strength calculated from the signal strengths of the plurality of areas. a storage unit configured to store a classification model in which the reference data is classified into one of a plurality of classes by machine learning using a plurality of reference data with associated mean and variance; When the unit receives the reflected wave to be analyzed, the signal strength calculation unit calculates the average and the variance from the signal strengths of the multiple areas, and generates analysis data in which the average and the variance are associated. and a class determination unit configured to determine to which of a plurality of classes the analysis data belongs.

A bathroom monitoring method according to another aspect of the present disclosure includes a first step of transmitting broadband radio waves into a bathroom; Machine learning is performed using a second step of calculating the intensity and a plurality of reference data associated with the average and variance obtained from the signal intensities of a plurality of areas, and the reference data is any one of a plurality of classes. A third step of creating a classification model classified into one, obtaining the average and variance from the signal intensity of a plurality of areas based on the reflected wave to be analyzed, and obtaining the analysis data in which the average and the variance are associated a fourth step of generating; and a fifth step of determining to which of the plurality of classes the analysis data belongs.

According to the bathroom monitoring device and the bathroom monitoring method according to the present disclosure, it is possible to accurately detect the state of the bather even in a highly disturbed environment.

FIG. 1 is a schematic view of the bathroom monitoring system and the surroundings of the bathroom from above. FIG. 2 is a schematic block diagram of the bathroom monitoring device. FIG. 3 is a schematic block diagram mainly showing the control unit. FIG. 4 is a diagram showing an example of a classification model. FIG. 5(a) is a schematic view of the bathroom viewed from the side, and FIG. 5(b) is a graph showing the relationship between distance and signal strength when no bather is present in the bathroom. 5(c) is a graph of distance versus signal strength when a bather is present in the bathroom. FIG. 6 is a block diagram schematically showing the notification device. FIG. 7 is a flow chart for explaining the bathroom monitoring procedure. FIG. 8 is a graph showing an example of changes in received signal strength with respect to humidity. FIG. 9 is a schematic view of the bathroom monitoring system and the surroundings of the bathroom according to another example, viewed from above. FIG. 10(a) is a schematic view of the bathroom viewed from above, and FIG. 10(b) is a graph showing the relationship between distance and signal strength when no bather is present in the bathroom. (c) is a graph showing the relationship between distance and signal strength when a bather is present near the door. FIG. 11(a) is a schematic diagram of the bathroom viewed from above, and FIG. 11(b) is a graph showing the relationship between distance and signal strength when a bather is present in the bathtub.

An example of an embodiment according to the present disclosure will be described below in more detail with reference to the drawings. In the following description, the same reference numerals will be used for the same elements or elements having the same functions, and redundant description will be omitted.

[Configuration of bathroom monitoring system]
A configuration of the bathroom monitoring system 1 will be described with reference to FIG. The bathroom monitoring system 1 is configured to detect the area and state of the bather H and notify the detection result to a place away from the bathroom. A bathroom monitoring system 1 includes a bathroom monitoring device 10 and a notification device 20 .

Here, the configuration of the bathroom 30 will be described. The bathroom 30 includes a washing area 31, a door 32, a shower 33 and a bathtub . The washing area 31 has, for example, a rectangular shape. The door 32 partitions between the washing place 31 and the dressing room (not shown) so that it can be entered and exited. In the form illustrated in FIG. 1, the door 32 is provided at one longitudinal end (the left end in FIG. 1) of the wash area 31. As shown in FIG. A shower 33 is provided on the wall surface of the washing place 31 . In the form illustrated in FIG. 1, the shower 33 is provided at the other longitudinal end of the washing area 31 (the right end in FIG. 1). The bathtub 34 has, for example, a rectangular shape and is positioned adjacent to the washing area 31 .

The bathroom monitoring device 10 has a function of detecting the area and state of the bather H in the bathroom 30 . The bathroom monitoring device 10 is attached to the wall surface of the bathroom 30 extending along the longitudinal direction of the bathtub 34 and does not face the door 32 . The bathroom monitoring device 10 includes a transmitter 11 (transmitting unit), a receiver 12 (receiving unit), a communication device 13, a control unit 14 (controller), and a bus 15, as shown in FIG. .

The transmitter 11 is configured to transmit broadband or ultra wideband (UWB) radio waves into the bathroom 30 based on instructions from the control unit 14 . The directivity angle of the transmitter 11 in the horizontal plane may be, for example, approximately 40° to 80°, or may be approximately 60° to 80°. The transmitter 11 has a detection area R that roughly corresponds to the range of directivity angles. For the purpose of suppressing the influence of fluctuations in the water surface of the bathtub 34, the directivity angle of the transmitter 11 in the vertical plane may be set to, for example, 30° or less.

As used herein, the term "broadband" refers to a frequency bandwidth of 100 MHz or more and 500 MHz or less. Therefore, the frequency bandwidth of the wideband radio waves emitted by the transmitter 11 may be, for example, 100 MHz or more, or may be 300 MHz or more. "Ultra-wideband" shall mean a frequency bandwidth greater than 500 MHz. Therefore, the frequency bandwidth of the ultra-wideband radio waves emitted by the transmitter 11 may be, for example, 3 GHz or more, but may be 5 GHz or less in view of the Radio Law and cost performance. As such, transmitter 11 may be an ultra-wideband wireless sensor with a frequency bandwidth greater than 500 MHz. When broadband or ultra-wideband radio waves are used, the transmission output of the radio waves can be reduced, and the influence of the radio waves on the bather H can be reduced.

The receiver 12 is configured to be able to receive reflected waves of radio waves transmitted from the transmitter 11 . Since the transmitter 11 uses broadband radio waves, the receiver 12 can separate and receive the reflected waves of the radio waves on the time axis corresponding to the reflected path length. That is, as illustrated in FIG. 1, the receiver 12 extracts signals from each of measurement areas (also called range bins, determined by frequency bandwidth) B1 to B8 divided into a plurality of areas in the distance direction of the detection area R. It is possible to The width of each range bin B1 to B8 may be, for example, about 15 cm to 30 cm. When the frequency bandwidth is 750 MHz, the width of one range bin is approximately 20 cm. Although the number of range bins is eight in this embodiment, the number of range bins can be changed according to the width of the range bins and the size of the detection area R. FIG.

In the example shown in FIG. 1, the range bins B1 to B4 approximately cover the bathtub 34, and the range bins B5 to B8 approximately cover the washing area 31. In the example shown in FIG. Therefore, in this specification, the range of range bins B1 to B4 that covers the bathtub 34 in the detection area R is referred to as a bathtub area Rα, and the range bins B5 to B8 that cover the washing place 31 in the detection area R are referred to as a bathtub area Rα. The range is called a wash area Rβ.

Returning to FIG. 2, the communication device 13 is configured to be able to communicate with the communication device 23 of the notification device 20 . The communication method of the communication device 13 with the communication device 23 is not particularly limited. ), W-CDMA (registered trademark), GSM (registered trademark), CDMA2000, UMB (Ultra Mobile Broadband), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, UWB, Bluetooth (registered trademark) or other communication methods may be used.

The control unit 14 is configured to transmit and receive signals to and from the transmitter 11, the receiver 12, and the communication device 13 via the bus 15 and control their operations. Control unit 14 includes, for example, processor 16 , memory 17 , and storage 18 .

When predetermined software (program) is read into the hardware such as the memory 17 and the storage 18, the processor 16 performs predetermined calculations, transmits radio waves from the transmitter 11, and detects reflected waves received by the receiver 12. Analysis, communication by the communication device 13, reading or writing of data in the memory 17 and the storage 18 are executed. Thereby, each function in the bathroom monitoring device 10 is realized.

Each function of the bathroom monitoring device 10 will be described with reference to FIG. The control unit 14 includes, as functional blocks, a storage unit 14a, a transmission/reception processing unit 14b, a signal strength calculation unit 14c, a data generation unit 14d (analysis data generation unit), a classification model generation unit 14e, and a class determination unit 14f. including.

The storage unit 14a has a function of storing various data. The data stored in the storage unit 14a includes, for example, read programs, operation setting data of the transmitter 11, reflected wave data received by the receiver 12, reference data or analysis data generated by the data generation unit 14d, and classification models. Various classification models generated by the generation unit 14e, data on determination results by the class determination unit 14f, and the like are included.

The transmission/reception processing unit 14 b has a function of transmitting and receiving signals among the transmitter 11 , receiver 12 and communication device 13 . Specifically, the transmission/reception processing unit 14b transmits an instruction signal to the transmitter 11 to cause the transmitter 11 to transmit radio waves. The transmission/reception processing unit 14b receives reflected wave data from the receiver 12 and stores the data in the storage unit 14a. The transmission/reception processing unit 14b transmits an instruction signal to the communication device 13 to cause the notification device 20 to transmit the determination result of the class determination unit 14f.

The signal strength calculator 14c has a function of calculating the signal strength in each of the range bins B1 to B8 based on the reflected wave data received by the transmission/reception processor 14b. The signal strength calculator 14c may calculate the signal strength smoothed on the time axis for each of the range bins B1 to B8. As smoothing processing, various processing methods such as simple moving average and weighted moving average may be employed. The signal intensity calculated by the signal intensity calculator 14c may be stored in the storage 14a.

The data generation unit 14d has a function of obtaining at least two parameters using the signal strength calculated by the signal strength calculation unit 14c and generating data in which the parameters are associated with each other. For example, the data generation unit 14d may use the signal intensities p ₁ to p ₄ in the range bins B1 to B4 to obtain the average m ₁ and the variance σ ₁ ² of the signal intensities in the bathtub area Rα by Equations 1 and 2, respectively. . Similarly, the data generation unit 14d may use the signal intensities p ₅ to p ₈ in the range bins B5 to B8 to obtain the average m ₂ and the variance σ ₂ ² of the signal intensities in the washing area Rβ by Equations 3 and 4, respectively. good. In this specification, among the data generated by the data generation unit 14d, the data to be analyzed may be referred to as "analysis data", and the data not to be analyzed (data for machine learning) may be referred to as "reference data". The data generated by the data generation unit 14d may be stored in the storage unit 14a.

The classification model generation unit 14e has a function of classifying the reference data into one of a plurality of classes by machine learning using a plurality of reference data generated in the data generation unit 14d and generating a classification model. have. For example, when a plurality of reference data (a data set including a predetermined number of reference data) generated by the data generation unit 14d are input to the classification model generation unit 14e, the classification model generation unit 14e performs supervised learning. , unsupervised learning, or reinforcement learning. Supervised learning algorithms include, for example, support vector machines, logistic regression, random forests, decision trees, k-nearest neighbors, perceptrons, and neural networks. Algorithms for unsupervised learning include, for example, the k-means method, principal component analysis, self-organizing map, and the like. Algorithms for reinforcement learning include, for example, Q-learning, Monte Carlo method, SARSA, and the like. The classification model generated by the classification model generation unit 14e may be stored in the storage unit 14a.

For example, when the data set DS1 shown in FIG. 4A is input to the classification model generation unit 14e, the classification model generation unit 14e generates standard data marked with a circle and standard data marked with a square in the data set DS1. may be generated. In the example shown in FIG. 4(a), the data set DS1 is a set of reference data in which two parameters, the average signal intensity _m1 in the bathtub area Rα and the average signal intensity _m2 in the washing area Rβ, are associated. is. A plurality of reference data constituting the data set DS1 are classified by a boundary line L1 into one of a class C1a (a set of reference data marked with a circle) and a class C1b (a set of reference data marked with a square). The class C1a may be, for example, a classification indicating that the bather H is present in the washing place 31 . Class C1b may be, for example, a classification indicating that bather H is present in bathtub 34 . The boundary line L1 may be a straight line or a curved line. Hereinafter, the classification model composed of the classes C1a and C1b partitioned by the boundary line L1 may be referred to as a "classification model M1".

For example, when the data set DS2 shown in FIG. 4B is input to the classification model generation unit 14e, the classification model generation unit 14e generates the standard data marked with ○ and the standard data marked with □ in the data set DS2. A boundary line L2 may be generated that separates the reference data marked with Δ. In the example shown in FIG. 4(b), _the data set DS2 is a set of reference data in which two parameters of the mean m2 and the variance _σ22 of the signal intensity in ^the washing area Rβ are associated. A plurality of reference data constituting the data set DS2 are separated by a boundary line L2 into a class C2a (a set of reference data marked with a circle), a class C2b (a set of reference data marked with a square), and a class C2c (a set of reference data marked with a triangle). set). The class C2a may be, for example, a classification indicating that the bather H present in the washing area 31 is normal. The class C2b may be, for example, a classification indicating that the bather H in the washing area 31 has fallen. The class C2c may be, for example, a classification indicating that the bather H present in the washing area 31 has not moved continuously for a predetermined period of time. Boundary line L2 may be a straight line or a curved line. Hereinafter, the classification model composed of classes C2a to C2c partitioned by boundary line L2 may be referred to as "classification model M2".

For example, when the data set DS3 shown in FIG. 4(c) is input to the classification model generation unit 14e, the classification model generation unit 14e generates the reference data marked with ○ and the reference data marked with □ in the data set DS3. A boundary line L3 may be generated that separates the reference data marked with Δ. In the example shown in FIG. 4(c), _{the data set DS2 is a set of reference data in which two parameters of the mean m1 and the variance σ12 of} the ^signal intensity in the bathtub area Rα are associated. The plurality of reference data that make up the data set DS3 are separated by a boundary line L3 into a class C3a (a set of reference data marked with a circle), a class C3b (a set of reference data marked with a square), and a class C3c (a set of reference data marked with a triangle). set). The class C3a may be, for example, a classification indicating that the bather H in the bathtub 34 is normal. The class C3b may be, for example, a classification indicating that the bather H in the bathtub 34 has drowned. The class C3c may be, for example, a classification indicating that the bather H in the bathtub 34 has not moved continuously for a predetermined period of time. The boundary line L3 may be a straight line or a curved line. Hereinafter, the classification model composed of classes C3a to C3c partitioned by boundary line L3 may be referred to as "classification model M3".

The class determination unit 14f has a function of classifying the analysis data generated by the data generation unit 14d into one of a plurality of classes in the classification model generated by the classification model generation unit 14e. That is, the class determination unit 14f applies the analysis data to the classification model and determines to which class the analysis data belongs among a plurality of classes.

The entrance/exit determination unit 14g has a function of determining whether or not the bather H enters/exits the bathroom 30 . First, the room entry/exit determination unit 14g acquires the signal strength corresponding to the distance from the receiver 12 based on the reflected wave data received by the transmission/reception processing unit 14b. Specifically, when the bather H does not exist in the bathroom 30, the signal intensity of the reflected wave from the wall is outstanding (see FIGS. 5(a) and 5(b)). On the other hand, when the bather H is present in the bathroom 30, the bather H is positioned in front of the front wall of the bathroom monitoring device 10, and the distance d1 from the bathroom monitoring device 10 to the bather H is the same as that of the bathroom monitoring device 10. to the front wall (see FIG. 5(a)). At this time, part of the radio wave from the transmitter 11 is reflected by the bather H and does not reach the wall. As a result, the signal strength of the reflected wave from the wall decreases (see arrow Ar1 in FIGS. 5A and 5C), and the signal strength of the reflected wave from the near side (receiver 12 side) of the wall decreases. becomes larger (see arrow Ar2 in FIG. 5(c)). Therefore, the room entry/exit determination unit 14g acquires the signal intensity over time, and determines that the bather H has entered the bathroom 30 when the reflected wave from the front side of the wall becomes greater than a predetermined threshold value TH1. You can judge. The room entry/exit determination unit 14g acquires the signal intensity of the reflected wave from the wall over time, and when the reflected wave from the front side of the wall becomes equal to or less than a predetermined threshold TH1, the bather H exits the bathroom 30. You may decide to leave the room.

When determining whether the bather H enters or leaves the bathroom 30, the room entry/exit determination unit 14g may also consider the signal intensity of the reflected wave from the wall. Specifically, the room entry/exit determination unit 14g acquires the signal strength over time, the reflected wave from the front side of the wall becomes larger than a predetermined threshold TH1, and the signal strength of the reflected wave from the wall reaches a predetermined threshold TH1. It may be determined that the bather H has entered the bathroom 30 when the threshold TH2 or less is reached. The room entry/exit determination unit 14g acquires the signal intensity over time, and when the reflected wave from the front side of the wall becomes equal to or less than a predetermined threshold TH1 and the signal intensity of the reflected wave from the wall becomes greater than a predetermined threshold TH2. Alternatively, it may be determined that the bather H has left the bathroom 30 . By also taking into consideration the time variation of the signal intensity of the reflected wave from the wall, it becomes possible to determine whether the bather H enters or leaves the room with higher accuracy.

Returning to FIG. 1, the notification device 20 is installed in a room separate from the bathroom 30 by a wall W, for example. The notification device 20 includes a monitor 21 (notification section), a speaker 22 (notification section), a communication device 23, a control section 24 (controller), and a bus 25, as shown in FIG.

The monitor 21 and the speaker 22 are configured to notify the persons (eg, family members) in the other room of the determination result by the room entry/exit determination unit 14g received via the communication device 23 by video and/or sound. is configured to The communication device 23 has the same configuration as the communication device 13 of the bathroom monitoring device 10 and functions similarly to the communication device 13 .

The control unit 24 includes a processor 24a, a memory 24b, and a storage 24c, and functions similarly to the control unit 14 of the bathroom monitoring device 10. FIG. The bus 25 has the same configuration as the bus 15 of the bathroom monitor 10 and functions similarly to the bus 15 .

[Bathroom monitoring method]
Next, a method for monitoring the bathroom 30 will be described with reference to FIG. The storage unit 14a of the bathroom monitoring device 10 stores at least one learned classification model obtained by machine learning using a plurality of reference data.

First, the transmission/reception processing unit 14b instructs the transmitter 11 to transmit radio waves into the bathroom 30 from the transmitter 11 . Next, the transmission/reception processing unit 14b receives the reflected wave data of the radio wave through the receiver 12 .

Next, the room entry/exit determination unit 14g temporally acquires the signal strength corresponding to the distance from the receiver 12 based on the reflected wave data received by the transmission/reception processing unit 14b. The room entry/exit determination unit 14g determines whether or not the bather H has entered the bathroom 30 based on the variation in the signal intensity of the reflected wave from the predetermined area (step S1 in FIG. 7). Specifically, the room entry/exit determination unit 14g acquires the signal intensity over time, and determines whether or not the reflected wave from the front side of the wall is greater than a predetermined threshold TH1. When the entry/exit determining unit 14g determines that the bather H has not entered the bathroom 30 (NO in step S1 of FIG. 7), the process of step S1 is repeated.

On the other hand, when the entry/exit determination unit 14g determines that the bather H has entered the bathroom 30 (YES in step S1 in FIG. 7), the signal strength calculation unit 14c calculates the reflected wave received by the signal strength calculation unit 14c. , the signal strength of each of the range bins B1 to B8 is calculated at a predetermined sampling period (step S2 in FIG. 7). The sampling period may be, for example, about 0.01 seconds to 0.1 seconds.

Next, the data generation unit 14d obtains the average and variance of the signal strengths calculated by the signal strength calculation unit 14c, and generates analysis data (step S3 in FIG. 7). Next, the class determination unit 14f applies the analysis data generated by the data generation unit 14d to the classification model M1 to identify the area of the bather H (step S4 in FIG. 7). For example, when the data generation unit 14d generates the analysis data x1 shown in FIG. 4A, the class determination unit 14f determines that the analysis data x1 belongs to class C1a. That is, the class determination unit 14f determines that the bather H is present in the washing area 31. FIG. On the other hand, when the data generation unit 14d generates the analysis data x2 shown in FIG. 4A, the class determination unit 14f determines that the analysis data x2 belongs to class C1b. That is, the class determination unit 14f determines that the bather H is present in the bathtub 34. FIG.

Next, the class determination unit 14f applies the analysis data generated by the data generation unit 14d to the classification model M2 or the classification model M3 to determine the condition of the bather H (step S5 in FIG. 7). For example, when the class determination unit 14f determines that the analysis data x1 belongs to class C1a, the class determination unit 14f applies the analysis data x1 to the classification model M2 so that the analysis data x1 belongs to classes C2a to C2c. Determine which one you belong to.

When it is determined that the analysis data x1 belongs to class C2a, the class determination unit 14f determines that the bather H is normal. If it is determined that the analysis data x1 belongs to class C2b, the bather H has fallen, and the class determination unit 14f determines that the bather H has an abnormality. If it is determined that the analysis data x1 belongs to class C2c (see the analysis data x1 illustrated in FIG. 4(b)), the bather H has not moved continuously for a predetermined period of time. 14f determines that the bather H has an abnormality.

When the analysis data x2 is determined to belong to class C3a (see the analysis data x2 illustrated in FIG. 4(c)), the class determination unit 14f determines that the bather H has no abnormality. When it is determined that the analysis data x2 belongs to class C3b, the bather H is drowning, so the class determination unit 14f determines that the bather H has an abnormality. If it is determined that the analysis data x2 belongs to class C3c, the bather H has not moved continuously for the predetermined time, and the class determination unit 14f determines that the bather H has an abnormality.

Next, if the result of the class determination unit 14f in step S5 is that there is an abnormality in the bather H (YES in step S6 of FIG. 7), the transmission/reception processing unit 14b transmits an instruction signal to the communication device 13, The notification device 20 is caused to transmit the result of determination by the class determination unit 14f. As a result, through the monitor 21 or the speaker 22 of the notification device 20 in a room separate from the bathroom monitoring device 10, the person in the room is notified of the danger of the bather H (step S7 in FIG. 7). ).

On the other hand, if the result of the class determination unit 14f in step S5 is that there is no abnormality in the bather H (NO in step S6 of FIG. 7), the entry/exit determination unit 14g determines the signal strength of the reflected wave from the predetermined position. Based on the change, it is determined whether or not the bather H has left the bathroom 30 (step S8 in FIG. 7). Specifically, the room entry/exit determination unit 14g acquires the signal intensity over time, and determines whether or not the reflected wave from the front side of the wall is equal to or less than a predetermined threshold value TH1. When the entering/leaving room determination unit 14g determines that the bather H has not left the bathroom 30 (NO in step S8 of FIG. 7), the processes from step S2 onward are repeated. On the other hand, when the entry/exit determination unit 14g determines that the bather H has entered the bathroom 30 (YES in step S8 of FIG. 7), the monitoring process of the bathroom 30 ends.

[Action]
In the above embodiment, machine learning using reference data in which at least two parameters (e.g., average, variance, etc.) of the signal strength in each range bin B1 to B8 in the bathroom 30 are associated with each other in the bathroom 30 The state of the bather H is classified into a plurality of classes. Therefore, it is possible to accurately detect the state of the bather H by determining the class to which the analysis data belongs based on such a classification model.

In the above embodiment, the storage unit 14a stores the classification model M1 composed of the class C1a indicating that the bather H is present in the washing area 31 and the class C1b indicating that the bather H is present in the bathtub 34. are doing. Therefore, it is possible to accurately detect in which of the washing area 31 and the bathtub 34 the bather H is present.

In the above embodiment, the classification model M2 is composed of a plurality of classes C2a to C2c representing the situation of the bather H in the washing area 31, and the class C3a to C3c is composed of a plurality of classes C3a to C3c representing the situation of the bather H in the bathtub 34. The storage unit 14a stores the classification model M3. Therefore, it is possible to accurately detect the condition of the bather H in the washing area 31 or the bathtub 34 .

In the above embodiment, the class determination unit 14f determines the area of the bather H by applying the analysis data to the classification model M1. Subsequently, the class determination unit 14f determines the situation of the bather H by applying the analysis data to the classification model M2 or the classification model M3 according to the area of the bather H (washing place 31 or bathtub 34). Therefore, compared with the case where the location and situation of the bather H are determined at the same time, the situation of the bather H can be detected more accurately.

In the above embodiment, the room entry/exit determining unit 14g determines whether the bather H enters or exits the bathroom 30 based on the time variation of the signal intensity of the reflected wave from the predetermined area. Therefore, by comparing the signal intensities, it is possible to determine whether the bather H has entered or left the room very simply.

[Modification]
Although the embodiments according to the present disclosure have been described in detail above, various modifications may be made to the above embodiments without departing from the scope and spirit of the claims.

(1) The bathroom monitoring device 10 is a bathroom remote controller for a water heater, a handrail for a bathroom, etc., in which the transmitter 11, the receiver 12, the communication device 13, the control unit 14, the bath 15, etc. are incorporated. There may be.

(2) The bathroom monitoring device 10 may be connected to the notification device 20 by wire.

(3) In the above embodiment, the notification device 20 is installed in the same building as the bathroom monitoring device 10, but it may be installed in a building different from the bathroom monitoring device 10, and mobile communication is possible. It may be a portable terminal (for example, a mobile phone, a smart phone, etc.). The different buildings include, for example, a house of a family living separately from the bather H, a facility providing emergency medical services (a fire station in Japan), and the like. In these cases, the bathroom monitoring device 10 transmits to the notification device 20 detected information such as the bather H's location, movement, whether or not he is in a dangerous state, etc. , emergency services, etc. can know the condition of the bather H. Therefore, even if the bather H lives alone, it is possible to monitor the condition of the bather H.

(4) In the above embodiment, it is determined in which of the washing area 31 and the bathtub 34 of the bathroom 30 the bather H is present. It may be determined whether the bather H exists in the area. In this case, for example, the brightness of the lighting in the bathroom 30 may be controlled based on the determination result to save energy. An example of a method of controlling the brightness of the lighting is to turn on the lighting in the area where the bather H is present and to turn off or dim the lighting in the other areas.

(5) In the above embodiment, after determining the area where the bather H exists, the classification model corresponding to the area is used to determine the situation of the bather H. and the condition of the bather H may be determined at the same time or at close timings.

(6) Bathroom monitoring device 10 may include notification device 20 . That is, the bathroom monitoring device 10 may incorporate the monitor 21 (informing section), the speaker 22 (informing section), and the like. In this case, the notification device 20 is arranged in the bathroom 30 . Therefore, for example, when the bather H falls asleep in the bathtub 34 and is in a semi-awakened state, the monitor 21 blinks and the speaker 22 issues an alarm (e.g., alarm sound, warning announcement, etc.). , it is possible to prevent the bather H from becoming severely abnormal.

(7) The bathroom monitoring device 10 may include an input unit configured to receive input operations from users including the person H in the bath. Examples of the input unit include a switch and a touch panel. Bathroom monitoring device 10 may include an updating unit configured to update the classification model (boundary line) based on the input operation received by the input unit. For example, when the notification device 20 issues an alarm, the user inputs an input operation indicating that the alarm is a false alarm (when the notification device 20 detects that the bather H is in an abnormal state) If an alarm is issued but the bather H does not actually have an abnormality), the updating unit may update the classification model with the input operation as a trigger. In this case, the classification model is updated by user feedback of false alarms through the input. Therefore, the situation of the bather H in the bathroom 30 can be detected more accurately.

(8) The control unit 14 adds data regarding the state of the bather determined by the class determination unit 14f (for example, the position of the bather H in the bathroom 30, the presence or absence of movement of the bather, the abnormal state of the bather, etc.). Based on this, a control signal generation unit may be included that generates a control signal for controlling other devices (for example, a water heater, lighting in the bathroom 30, etc.). The control signal generation unit may transmit the control signal to the other device through the communication device 13 (communication unit). Examples of the control signal include a control signal to a water heater for adjusting the hot water temperature of the bathtub 34 and a control signal to the lighting for adjusting the lighting in the bathroom 30 .

(9) When the class determination unit 14f determines that the bather H is in the bathtub 34 and that there is an abnormality, the drain plug of the bathtub 34 may be controlled to automatically open the drain plug. good. In this case, even if the bather H faints in the bathtub 34, it is possible to prevent the bather H from drowning.

(10) As shown in FIG. 1, a reflector 35 (for example, a mirror) may be installed in the bathroom 30 with a slight inclination. In FIG. 1, the reflector 35 is attached to the wall surface of the wash area 31 on the side away from the door 32 so as to face the direction of the door 32 . As a result, radio waves emitted from the transmitter 11 are reflected by the reflector 35 and travel toward the door 32 . Also, the reflected wave coming toward the reflector 35 is also reflected by the reflector 35 and goes toward the receiver 12 . Therefore, the dressing room, which is usually a blind spot for radio waves transmitted from the transmitter 11, can also be detected.

(11) The classification model used in the bathroom monitoring device 10 may be a trained model in which a plurality of reference data have been trained in advance, or may be a dynamic model obtained by additionally learning analysis data to the trained model in real time. It may be a classification model.

(12) Even when the bather H is not present in the bathroom 30 (for example, after the bather H leaves the room), the bathroom monitoring device 10 may continue to operate to monitor the inside of the bathroom 30 . Specifically, it may be determined by machine learning that hot water continues to flow from the shower 33, the faucet, etc. even though the bather H is not in the bathroom 30. FIG. In this case, by notifying that the shower 33, faucet, etc. have not been closed, waste of hot water can be suppressed.

(13) By the way, as illustrated in FIG. 8, the signal strength of the reflected wave received by the receiver 12 may change as the humidity of the bathroom 30 changes. This is because the water vapor in the bathroom 30 increases as time passes after the bathtub 34 is filled with hot water, and radio waves are more likely to be absorbed by the water vapor. This is thought to be due to the fact that the radio waves are attenuated between the time they are transmitted from the device 11 and the time they are received by the receiver 12 . In addition, in FIG. 8, the dashed line is the moving average of the signal intensity with respect to humidity. Therefore, the signal intensity of the reflected wave received by the receiver 12 may be corrected according to the humidity. Specifically, the control unit 14 further includes a correction unit as a functional block, and the correction unit is calculated by the signal strength calculation unit 14c based on the attenuation amount of the reflected wave according to the humidity in the bathroom 30. Signal strength may be corrected. The classification model stored in the storage unit 14a may be obtained by machine learning using a plurality of pieces of reference data in which averages and variances obtained from signal intensities after correction by the correction unit are associated with each other. In this case, it is possible to prevent the mean and variance of the signal strengths obtained by the data generator 14d from becoming smaller than they actually are due to the influence of humidity. Therefore, since the analysis data is classified more accurately, it is possible to detect the condition of the bather H more accurately.

Here, as an example of a method of estimating the attenuation amount, the data of the relationship between the humidity and the received signal strength (for example, see the dashed line in FIG. 8) is acquired in advance and stored in the storage unit 14a. An attenuation corresponding to the measured humidity in the bathroom 30 may be calculated based on the data.

Alternatively, as another example of the method of estimating the attenuation amount, since the humidity in the bathroom 30 generally tends to change with the passage of time, may be used to estimate the attenuation of the reflected wave. That is, the time variation of the signal intensity is observed from the reflected wave when the bather H does not exist in the bathroom 30, and the signal intensity calculator 14c calculates the bather H's bathing time based on the time variation characteristic of the signal intensity. Attenuation of the reflected wave at time may be estimated.

(14) As shown in FIG. 9 , the bathroom monitoring device 10 may be attached to the wall surface of the bathroom 30 extending along the width direction of the bathtub 34 so as to face the door 32 . In this case, by controlling the directivity of the radio waves emitted from the transmitter 11, a plurality of detection areas (for example, two detection areas R1 and R2 in FIG. 9) are detected in a direction intersecting the propagation direction of the radio waves. Transmitter 11 may be configured as possible. As a directivity control method, for example, a mechanical scanning method, a monopulse method, a digital beam forming method, or the like can be adopted. In the example shown in FIG. 9, detection other than the location of the bather H (detection of the movement of the bather H and whether the bather H is in a dangerous state) can be performed in the same manner as in the above embodiment. .

(15) When the bathroom monitoring device 10 is attached to the wall surface of the bathroom 30 extending along the width direction of the bathtub 34 so as to face the door 32, the directivity of the radio wave is not controlled, and the bather H may be detected. Specifically, as shown in FIG. 10(a), the distance d3 from the bathroom monitoring device 10 to the front wall of the bathroom 30 and the distance d4 from the bathroom monitoring device 10 to the door 32 are different. It may be used to detect the entry/exit of the bather H and the location of the bather H based on the variation in signal strength according to the distance from the receiver 12 . Details of the detection method will be described below.

When there is no bather H in the bathroom 30, the signal intensity of the reflected wave near the door is relatively small (see FIGS. 10(a) and 10(b)). On the other hand, when the bather H enters the room through the door 32, the signal intensity of the reflected wave from the front wall hardly changes, but the signal intensity of the reflected wave from the vicinity of the door increases (Fig. 10(a) and Fig. 10( c), see arrow Ar3). Therefore, the room entry/exit determination unit 14g acquires the signal intensity over time, and determines that the bather H has entered the bathroom 30 when the reflected wave from the vicinity of the door becomes greater than a predetermined threshold TH3. good too. The room entry/exit determination unit 14g may acquire the signal intensity over time, and determine that the bather H has left the bathroom 30 when the reflected wave from the vicinity of the door becomes equal to or less than a predetermined threshold TH3.

On the other hand, when the bather H moves from the washing area 31 to the bathtub 34, the bather H is positioned in front of the front wall of the bathroom monitoring device 10, and the distance d5 from the bathroom monitoring device 10 to the bather H is the same as that of the bathroom monitoring device. It is shorter than the distance d3 from 10 to the front wall (see FIG. 11(a)). At this time, part of the radio wave from the transmitter 11 is reflected by the bather H and does not reach the wall. Therefore, the signal strength of the reflected wave from the wall decreases (see arrow Ar4 in FIGS. 11A and 11B), and the signal strength of the reflected wave from the near side (receiver 12 side) of the wall decreases. becomes larger (see arrow Ar5 in FIG. 11(b)). Therefore, the room entry/exit determination unit 14g acquires the signal intensity over time, and when the reflected wave from the front side of the wall becomes larger than a predetermined threshold value TH4, the bather H moves from the washing area 31 to the bathtub 34. You can assume that you have moved. The room entry/exit determination unit 14g acquires the signal intensity over time, and determines that the bather H has moved from the bathtub 34 to the washing area 31 when the reflected wave from the front side of the wall becomes equal to or less than a predetermined threshold value TH4. You can judge.

[Example]
Example 1. A bathroom monitoring device (10) according to one example of the present disclosure includes a transmission unit (11) configured to transmit broadband or ultra-wideband radio waves into a bathroom (30), and a transmission unit (11) A receiving unit (12) configured to receive reflected waves of radio waves emitted from a bathroom (30) based on the reflected waves received by the receiving unit (12) in a plurality of areas (B1 to B8 ), and a plurality of reference data associated with at least two parameters obtained from the signal intensities of the plurality of areas (B1 to B8) A storage unit (14a) configured to store a classification model in which reference data is classified into one of a plurality of classes by machine learning using machine learning; At least two parameters are obtained from signal intensities of a plurality of areas calculated by a signal intensity calculator (14c) when a wave is received, and analysis data in which the at least two parameters are associated is generated. and a class determination unit (14f) configured to determine to which of a plurality of classes the analysis data belongs.

By the way, the distance spectrum of the signal intensity in the receiving section changes according to the entrance and exit of the bather, the movement of the bather, the stillness of the bather, the movement of the bather in the bathroom, and the like. Therefore, by machine learning using a plurality of reference data in which at least two parameters of signal strength in each area in the bathroom (for example, the average and variance of signal strength in each area) are associated, bathing in the bathroom A person's condition can be classified into multiple classes. Therefore, by determining the class to which the analysis data belongs based on such a classification model, it is possible to accurately detect the condition of the bather.

Example 2. In the device (10) of Example 1, the storage unit (14a) performs machine learning on the reference data in which at least two parameters obtained from the signal intensities of the plurality of areas (B1 to B8) are associated with the bathroom (30). may be configured to store a first classification model (M1) in which the reference data is classified into any one of a plurality of classes (C1a, C1b) indicating the position of the bather (H) in . In this case, it is possible to accurately detect whether the bather is in the washing area or the bathtub.

Example 3. In the device (10) of Example 1 or Example 2, the storage unit (14a) performs machine learning on reference data in which at least two parameters obtained from signal intensities of a plurality of areas (B1 to B8) are associated with each other. A second classification model (M2, M3) in which the reference data is classified into one of a plurality of classes (C2a to C2c, C2a to C2c) indicating the situation of the bather (H) in (30) may be configured to store. In this case, it is possible to accurately detect the situation of the bather in the bathroom.

Example 4. In the device (10) of Example 2, the storage unit (14a) performs machine learning on the reference data in which at least two parameters obtained from the signal intensities of the plurality of areas (B1 to B8) are associated with the bathroom (30). A second classification model (M2) in which the reference data is classified into one of a plurality of classes (C2a to C2c) indicating the situation of the bather (H) in the washing place (31), and a plurality of areas ( A plurality of classes (C3a to C3c), and further stores a third classification model (M3) in which the reference data is classified into any one of them. A first process of inputting the data into the classification model (M1) and determining to which class (C1a, C1b) the analysis data belongs in the first classification model (M1); When it is determined that the bather (H) exists in the washing place (31) of the bathroom (30) based on the class (C1a) to which the data belongs, the analysis data is input to the second classification model (M2), A second process for determining which class (C2a to C2c) the analysis data belongs to in the second classification model (M2), and in the first process, based on the class (C1b) to which the analysis data belongs When it is determined that the bather (H) is in the bathtub (34) of the bathroom (30), the analysis data is input to the third classification model (M3), and the analysis data is input to the third classification model (M3 ), and a third process of determining which class (C3a to C3c) it belongs to. In this case, it is first determined whether the bather is in the washing area or in the bathtub, and then it is determined what kind of situation the bather is in. Therefore, the classification models are layered and each classification model is simple. Therefore, it is possible to detect the situation of the bather more accurately than when the location and situation of the bather are determined at the same time.

Example 5. In the device (10) of Example 2, the storage unit (14a) performs machine learning on the reference data in which at least two parameters obtained from the signal intensities of the plurality of areas (B1 to B8) are associated with the bathroom (30). A second classification model in which the reference data is classified into any one of a plurality of classes (C2a to C2c, C3a to C3c) indicating the situation of the bather (H) in the washing place (31) and the bathtub (34) (M2, M3) is further stored, and the class determination unit (14f) inputs the analysis data to the first classification model (M1) so that the analysis data is the first classification model (M1 ) to determine which class (C1a, C1b) it belongs to; ) to determine which class (C2a to C2c, C3a to C3c) it belongs to. In this case, whether the bather is in the washing area or the bathtub and what kind of situation the bather is in are determined at the same time or in close timing. Therefore, it is possible to accurately detect the situation of the bather in the bathroom.

Example 6. The device (10) according to any one of Examples 1 to 5, based on the time variation of the signal strength calculated by the signal strength calculator (14c) when the receiver (12) receives the reflected wave to be analyzed, , an entrance/exit determination unit (14g) configured to determine whether the bather (H) enters or exits the bathroom (30). When a bather enters the bathroom, the reception unit receives the reflected wave from the bather, so the signal strength in the washing area increases compared to when the bather is absent. Therefore, by comparing the signal strength, it is possible to determine whether the bather has entered or left the room very simply.

Example 7. The device (10) of any one of Examples 1 to 6 corrects the signal strength calculated by the signal strength calculator (14c) based on the amount of attenuation of the reflected wave according to the humidity in the bathroom (30). and the storage unit (14a) performs machine learning using a plurality of reference data in which at least two parameters obtained from the signal intensities of the plurality of areas (B1 to B8) after correction by the correction unit are associated with each other. , and stores a classification model in which the reference data is classified into one of a plurality of classes, and the analysis data generation unit (14d) generates, when the reception unit (12) receives the reflected wave to be analyzed Further, at least two parameters are obtained from the signal intensities of the plurality of areas (B1 to B8) after correction by the correction unit, and analysis data in which the at least two parameters are associated may be generated. . By the way, as the humidity in the bathroom increases, the radio waves are more likely to be absorbed by water vapor floating in the bathroom, and the signal strength of the reflected wave received by the receiver tends to be attenuated. That is, if parameters (for example, mean, variance, etc.) are obtained from the signal strength after attenuation, these values become small. In this case, when the analysis data is applied to the classification model, there is a concern that the analysis data will be classified into a class different from the actual situation. However, according to example 6, the analysis data is generated based on the signal intensity after the attenuation due to humidity has been corrected. Therefore, since the analysis data is classified more accurately, it is possible to detect the condition of the bather more accurately.

Example 8. In the device (10) of Example 7, the correcting unit changes the time variation characteristic of the signal intensity calculated by the signal intensity calculating unit (14c) from the reflected wave when the bather (H) is not present in the bathroom (30). You may estimate attenuation|damping of the reflected wave at the time of a bather's bathing based on.

Example 9. The apparatus (10) of Example 7 further comprises a humidity sensor configured to measure humidity within the bathroom (30), wherein the compensator, based on the humidity within the bathroom (30) measured by the humidity sensor, determines , the attenuation of the reflected wave may be estimated.

Example 10. The device (10) of any one of Examples 1 to 9 includes a notification unit configured to issue an alarm when the bather (H) is in an abnormal state as a result of the determination by the class determination unit (14f). may be further provided. Abnormal states include, for example, severe abnormalities such as a bather collapsing, sinking in the bathtub, or remaining stationary due to fainting, etc., and, for example, a bather falling asleep in the bathtub. Mild abnormalities such as being in a semi-awake state are present. According to Example 9, by bringing a bather with a mild abnormality into an awake state with an alarm, it is possible to prevent the bather from transitioning to a severe abnormality.

Example 11. The device (10) of Example 10 includes an input unit configured to receive an input operation from a user, and updates a classification model based on the input operation received by the input unit when the notification unit issues an alarm. and an updating unit configured to. In this case, the classification model is updated by user feedback of false alarms through the input. Therefore, it is possible to more accurately detect the situation of the bather in the bathroom.

Example 12. The device (10) of any one of Examples 1-11 may further comprise a communication unit configured to transmit a control signal generated based on the data regarding the condition of the bather to the other device. . In this case, depending on the state of the bather (for example, the position of the bather in the bathroom, the presence or absence of movement of the bather, the abnormal state of the bather, etc.), other equipment (for example, the water heater, the lighting in the bathroom, etc.) ) can be optimized.

Example 13. A bathroom monitoring method according to another example of the present disclosure comprises: a first step of transmitting broadband or ultra-wideband radio waves into a bathroom (30); Machine learning using a plurality of reference data in which at least two parameters obtained from the signal strength of the plurality of areas (B1 to B8) are associated with a second step of calculating the signal strength in each of the plurality of areas within and a third step of creating a classification model in which the reference data is classified into one of a plurality of classes, and the signal intensity of a plurality of areas (B1 to B8) based on the reflected waves to be analyzed A fourth step of obtaining at least two parameters from the at least two parameters to generate analysis data associated with the at least two parameters, and a fifth step of determining to which of a plurality of classes the analysis data belongs . In this case, effects similar to those of the device of Example 1 are obtained.

DESCRIPTION OF SYMBOLS 1...Bathroom monitoring system 10...Bathroom monitoring apparatus 11...Transmitter (transmitting part) 12...Receiver (receiving part) 13...Communicator (communication part) 14...Control part (controller) 14a...Memory Part 14b... Transmission and reception processing part 14c... Signal strength calculation part 14d... Data generation part (analysis data generation part) 14e... Classification model generation part 14f... Class determination part 14g... Room entrance/exit determination part 20... Notification Device 21 Monitor (notification unit) 22 Speaker (notification unit) 30 Bathroom 31 Washing place 34 Bathtub B1 to B8 Range bin (area) C1a, C1b, C2a to C2c, C3a to C3c ... class, DS1 to DS3 ... data set, R ... detection area, Rα ... bathtub area, Rβ ... wash area.

Claims (13)

a transmitter configured to transmit broadband or ultra-wideband radio waves into the bathroom;
a receiving unit configured to receive a reflected wave of a radio wave transmitted from the transmitting unit;
a signal strength calculator configured to calculate signal strength in each of a plurality of areas in the bathroom based on the reflected wave received by the receiver;
A classification model in which the reference data is classified into any one of a plurality of classes by machine learning using a plurality of reference data associated with at least two parameters obtained from the signal intensities of the plurality of areas. a storage configured to store;
When the receiving unit receives the reflected wave to be analyzed, at least two parameters are obtained from the signal strengths of the plurality of areas calculated by the signal strength calculating unit, and the at least two parameters are associated with each other. an analytical data generator configured to generate analytical data;
and a class determination unit configured to determine to which of the plurality of classes the analysis data belongs. The storage unit selects one of a plurality of classes indicating the position of the bather in the bathroom by machine learning with respect to reference data in which at least two parameters obtained from the signal intensities of the plurality of areas are associated. 2. Apparatus according to claim 1, configured to store a first classification model into which the reference data has been classified. The storage unit stores any one of a plurality of classes indicating the state of the bather in the bathroom by machine learning of reference data associated with at least two parameters obtained from the signal intensities of the plurality of areas. 3. Apparatus according to claim 1 or 2, arranged to store a second classification model into which the reference data has been classified into. The storage unit
Machine learning is performed on the reference data associated with at least two parameters obtained from the signal intensities of the plurality of areas, and classifies the reference data into any one of a plurality of classes indicating the state of the bather in the wash area of the bathroom. a second classification model in which
Machine learning is performed on the reference data associated with at least two parameters obtained from the signal intensities of the plurality of areas, and classifies the reference data into one of a plurality of classes indicating the state of the bather in the bathtub of the bathroom. is further configured to store a third classification model with which the is classified;
The class determination unit
a first process of inputting the analysis data into the first classification model and determining to which class of the first classification model the analysis data belongs;
In the first process, when it is determined that a bather is present in the washing area of the bathroom based on the class to which the analysis data belongs, the analysis data is input to the second classification model, and the analysis data a second process of determining which class of the second classification model belongs to;
In the first process, when it is determined that a bather is present in the bathtub of the bathroom based on the class to which the analysis data belongs, the analysis data is input to the third classification model, and the analysis data 3. Apparatus according to claim 2, configured to perform a third operation of determining to which class of said third classification model . The storage unit performs machine learning on reference data associated with at least two parameters obtained from the signal intensities of the plurality of areas, and selects one of a plurality of classes indicating the bather's situation in the washing area and the bathtub of the bathroom. further storing a second classification model in which the reference data is classified into one;
The class determination unit
A process of inputting the analysis data into the first classification model and determining to which class of the first classification model the analysis data belongs;
and inputting the analysis data into the second classification model, and determining to which class the analysis data belongs to among the second classification models. 2. The device according to 2. The entrance/exit of the bather to/from the bathroom is determined based on the time variation of the signal intensity calculated by the signal intensity calculation unit when the reception unit receives the reflected wave to be analyzed. The apparatus according to any one of claims 1 to 5, further comprising an entrance/exit determination unit. Further comprising a correction unit that corrects the signal strength calculated by the signal strength calculation unit based on the attenuation amount of the reflected wave according to the humidity in the bathroom,
The storage unit performs machine learning using a plurality of reference data in which at least two parameters obtained from the signal intensities of the plurality of areas after correction by the correction unit are associated with each other, and the reference data is among the plurality of classes. It stores a classification model classified into one of
The analysis data generation unit obtains at least two parameters from signal intensities of the plurality of areas after correction by the correction unit when the reception unit receives the reflected wave to be analyzed, and calculates the at least two parameters 7. Apparatus according to any one of claims 1 to 6, arranged to generate analytical data associated with the . The correction unit calculates the time variation characteristic of the signal intensity calculated by the signal intensity calculation unit from the reflected wave when the bather is not present in the bathroom, and determines the intensity of the reflected wave when the bather is taking a bath. 8. Apparatus according to claim 7, for estimating attenuation. further comprising a humidity sensor configured to measure humidity within the bathroom;
8. The device according to claim 7, wherein said corrector estimates the amount of attenuation of said reflected wave based on the humidity in said bathroom measured by said humidity sensor. The apparatus according to any one of claims 1 to 9, further comprising a notification unit configured to issue an alarm when the bather is in an abnormal state as a result of the determination by the class determination unit. an input unit configured to receive an input operation from a user;
11. The apparatus according to claim 10, further comprising an updating unit configured to update said classification model based on an input operation received by said input unit when said reporting unit reports an alarm. The device according to any one of claims 1 to 11, further comprising a communication unit configured to transmit a control signal generated based on data relating to the condition of the bather to another device. A first step of transmitting broadband or ultra-wideband radio waves into the bathroom;
a second step of calculating a signal strength in each of a plurality of areas in the bathroom based on reflected radio waves;
A classification model in which machine learning is performed using a plurality of reference data associated with at least two parameters obtained from the signal intensities of the plurality of areas, and the reference data is classified into one of a plurality of classes. a third step of creating
a fourth step of obtaining at least two parameters from the signal intensities of the plurality of areas based on the reflected waves to be analyzed, and generating analysis data in which the at least two parameters are associated;
and a fifth step of determining to which of the plurality of classes the analysis data belongs.

Images