JP2024005486A - Bed-leave detection device, bed-leave detection method, and machine learning device - Google Patents

Abstract

To provide a bed-leaving detection device, a bed-leaving detection method, and a machine learning device capable of preventing in advance an object person from overturning when leaving a bed.SOLUTION: A bed-leaving detection device 22 comprises: a sensor unit 20 including a plurality of antennas 10, signal lines 12 connected with the antennas 10, and a connection part 13 connected with the signal lines 12; and a control unit 21 including a signal measurement unit 14 connected with the connection part 13 and measuring signals from the antennas 10 and a control unit 15 for controlling the signal measurement unit 14. The plurality of antennas 10 are aligned in a shape of a grid composed of at least two columns and at least two rows.SELECTED DRAWING: Figure 1

Description

Application for application of Article 30, Paragraph 2 of the Patent Act has been filed. Presented at the 38th "Sensors, Micromachines and Applied Systems" Symposium (held online) on November 9, 2021.

The present invention relates to a bed-leaving detection device, a bed-leaving detection method, and a machine learning device. In particular, the present invention relates to a bed-leaving detection device, a bed-leaving detection method, and a machine learning device that can detect the posture and motion of a subject.

In the past, there have been many reports of falls in medical facilities, nursing care facilities, and general households, where elderly people and other people fall while getting out of bed. , is considered a serious incident.
Currently, there are devices that detect when a subject leaves the bed and issue an alarm in order to prevent the subject from falling when the subject leaves the bed (Patent Documents 1 to 3).

JP 2017-140157 Publication Japanese Patent Application Publication No. 2005-028059 Japanese Patent Application Publication No. 2004-105416

However, conventional bed leaving detection devices only know whether the subject is present or absent, and cannot prevent the subject from falling when leaving the bed.

Therefore, an object of the present application is to provide a bed-leaving detection device, a bed-leaving detection method, and a machine learning device that can prevent a subject from falling when he or she leaves the bed.

In order to solve the above problem, a sensor unit includes a plurality of antenna parts, a signal line connected to the antenna part, and a connection part connected to the signal line,
a control unit that is connected to the connection section and includes a signal measurement section that measures a signal from the antenna section; and a control section that controls the signal measurement section;
including;
A bed leaving detection device is used in which the plurality of antenna units are arranged in a grid pattern in two or more columns and two or more rows.

Further, a receiving step of receiving signals from a plurality of antenna units arranged in a grid pattern in two or more columns and two or more rows;
a processing step of receiving and processing the signal;
A bed leaving detection method is used, which includes a determination step of determining the posture and motion of the subject from the processed signal.

Furthermore, in the bed leaving detection device, a machine learning device that determines the posture and motion of the subject based on the signal from the sensor unit,
a signal measurement unit that receives a signal from the sensor unit;
a learning unit that learns using the signal and training data of the subject's posture or motion;
Using machine learning equipment including.

With the bed leaving detection device, bed leaving detection method, and machine learning device of the present application, it is possible to detect when a subject leaves the bed and prevent a fall when the subject leaves the bed.

(a) Configuration diagram of the bed leaving detection device of Embodiment 1, (b) External view of the bed leaving detection device of Embodiment 1 A diagram illustrating how to use the bed leaving detection device of Embodiment 1 Cross-sectional view of the vicinity of the antenna section of Embodiment 1 (a) to (d) Side views of the bed of Embodiment 2, (e) to (h) Side views showing an example of the usage state of the bed of Embodiment 2, (i) (b) and (f) top view of bed (a) to (d) are side views corresponding to FIGS. 4(a) to 4(d), showing the arrangement of the sensor unit; (e) to (h) are side views corresponding to FIGS. 4(a) to 4(d); h) A side view corresponding to h) and a diagram showing the arrangement of the sensor unit. (a) Top view showing the arrangement of the plates 51a to 51d of the bed in Embodiment 2, (b) to (d) Top view showing the position of the antenna part of the sensor unit in the bed in (a) Configuration diagram of bed leaving detection device according to Embodiment 2 (a) RAW data from the sensor unit of the second embodiment, which is received by the signal measurement section, (b) data obtained by processing the RAW data by the signal processing section, (c) data of the sensor unit of the second embodiment. A diagram showing the number 23 of the antenna part, a plan view of the sensor unit showing the results of (d) to (g), and (b) (a) Diagram showing the process of completing the trained discriminator of Embodiment 2, (b) Diagram illustrating the process using the trained posture discriminator and learned motion discriminator of Embodiment 2 A diagram showing the configuration of the machine learning unit 61 of Embodiment 2

(Embodiment 1)
<Structure>
FIG. 1(a) is a configuration diagram of the bed leaving detection device 22. As shown in FIG. The bed leaving detection device 22 includes a sensor unit 20 and a control unit 21. FIG. 1A shows the configuration of the sensor unit 20 and the configuration of the control unit 21. In the control unit 21, each element is connected by an electric circuit. FIG. 1(b) is an external view of the bed leaving detection device 22.

The sensor unit 20 is a part that includes a sensor that measures the posture and movement of the subject 31 in bed. The target persons 31 include the elderly, patients, and other people with physical disabilities. The target audience may be children, babies, etc.

The control unit 21 is a part that controls the sensor unit 20.

The sensor unit 20 is preferably covered with a cover 41 that is insulating and waterproof. The cover 41 is preferably a belt-shaped waterproof sheet for the purpose of preventing incontinence of the subject 31. It is preferable that the sensor unit 20 and the waterproof sheet are installed together or as an integrated unit. It is preferable that the sensor unit 20 as well as the cover 41 can be discarded. Since the sensor unit 20 is arranged at least around the buttocks, the sensor unit 20 can also be used as part of a sheet for preventing incontinence of the subject.
<Sensor unit 20>
The sensor unit 20 includes a base material 25 , an antenna section 10 , an antenna connection section 11 , a signal line 12 , and a connection section 13 .

The base material 25 is a flexible insulating sheet, and has the antenna section 10, the antenna connection section 11, and the signal line 12 thereon. The base material 25 is, for example, a resin film, cloth, paper, or the like.
The antenna unit 10 measures the degree of proximity between the antenna unit 10 and the subject 31 by measuring the capacitance between the antenna unit 10 and the subject 31 using the signal measuring unit 14 for self-capacitance type capacitance measurement. When the subject 31 approaches the antenna section 10, the measured capacitance increases.

The antenna connection part 11 is a part that connects the antenna part 10 and the signal line 12.

The signal line 12 electrically connects the antenna section 10 and the signal measurement section 14 via the connection section 13 .

The connecting portion 13 is a part of the sensor unit 20 and is a portion connected to the control unit 21 . It is preferable that the connection part 13 can be connected to the control unit 21 with a single touch, and a pin connector type or the like is preferable.

The sensor unit 20 is preferably made of a flexible material and is generally thin, lightweight, and flexible.

<Control unit 21>
The control unit 21 includes a signal measurement section 14 , a control section 15 , a recording section 16 , a display section 17 , a communication section 18 , and a power supply section 19 .

The signal measurement section 14 is a section that measures the capacitance of each of the antenna sections 10 based on the signals of the plurality of antenna sections 10.

The control unit 15 is an IC, a microcomputer, etc. that controls each component of the bed leaving detection device 22.

The recording unit 16 is an IC memory or the like that stores measurement results from the signal measurement unit 14, setting values of each component, and the like. When data is not transmitted to the outside, the recording unit 16 may be a removable recording medium such as an SD card or a USB memory.

The display unit 17 is a display such as liquid crystal or EL, or an indicator such as LED. There are operation buttons, etc., through which instructions for operating the bed leaving detection device 22, operating status, etc. are input. It is also possible to display measurement results. In addition, the display section 17 is a section for setting and instructing the components.

The communication section 18 is a section that sends data etc. from the recording section 16 to the outside. Data can be sent wirelessly to personal computers, smartphones, mobile devices, etc. using Bluetooth (trademark), wireless LAN, etc. In the case of wired communication, the communication section 18 has an external connector part.

<Operation>
With reference to FIG. 2, a method of using the bed leaving detection device 22 will be explained.

FIG. 2 is a plan view showing a posture in which the bed leaving detection device 22 is placed on the bed 30 and the subject 31 is lying on the bed in a supine position.

In the bed leaving detection device 22 of FIG. 2, the sensor unit 20 includes twelve antenna sections 10. In FIG. 2, each antenna section 10 is labeled with an antenna section number 23. The control unit 21 controls the sensor unit 20 and stores data of each antenna section 10 in the recording section 16. Examples of data are explained below. The control unit 21 controls the plurality of antenna sections 10 within the sensor unit 20.

<Antenna section 10>
FIG. 3 shows a cross-sectional view around the antenna section 10. In the antenna section 10, an electrode 10a is formed on a flat base material 10c, and a cover 10b is placed on top of the electrode 10a.

Note that the base material 10c may be used as the entire base material 25 (FIG. 1(a)), and the electrode 10a may be formed directly on the base material 25. Similarly, the cover 10b may also be an entire cover 41 (FIG. 1(b)).

The degree of approach of the object 10d (subject 31) to the antenna section 10 is measured by the change in capacitance at the electrode 10a.

The electrode 10a is, for example, metal foil. The metal foil is, for example, a foil body of gold, silver, copper, aluminum, carbon, etc., and is lightweight, thin, and flexible. The metal foil may be formed by a thin film method such as sputtering or vapor deposition, or may be separately produced and bonded.

Alternatively, the electrode 10a may be formed using a conductive paste by a printing method, spin coating, or the like.

Alternatively, the electrode 10a may be partially or entirely formed of a conductive thread. A conductive thread such as gold thread, silver thread, copper thread, aluminum thread, or carbon thread may be used to form a flat plate. The conductive thread may be a resin fiber coated with each metal. The electrode 10a using conductive thread is preferable because it is lightweight, thin, flexible, and has excellent durability.

The electrode 10a may be formed by combining metal foil, conductive thread, and conductive paste.

The cover 10b is an insulating and waterproof resin sheet made of PET (polyethylene terephthalate), PEN (polyethylene naphthalate), PEI (polyetherimide), PI (polyimide), or the like. Further, cloth or paper that has insulation properties and is treated with waterproofing may be used. The base material 10c may also be made of the same material as the cover 10b.

Although the shape of the antenna section 10 including the electrode 10a is rectangular in FIGS. 1A and 2, it may be circular, trapezoidal, polygonal, linear, or the like. In Embodiment 2, a more preferable shape (arrangement) of the antenna section 10 will be explained.

There are a plurality of antenna sections 10, and the control unit 21 can measure the degree of approach of the object 10d (subject 31) to each antenna section 10.

<Signal line 12>
The signal line 12 electrically connects the antenna connection section 11 and the connection section 13. The signal line 12 may be a metal material patterned on the base material 25 by a printing method or the like, or may be a conductive wire, a conductive thread, or the like.

<Self-capacitance type capacitance sensor>
The antenna section 10 uses a self-capacitance method for measuring capacitance. Therefore, the structure of the antenna section 10 is a simple one-layer structure in which the electrode 10a is not bipolar but monopolar as shown in FIG. 3, and does not require an intermediate layer between the electrodes. The degree of approach of the target object 10d (target person 31) to the antenna section 10 is measured by a change in capacitance at the antenna section 10. This change in capacitance is measured by the control unit 21 via the signal line 12.

Furthermore, it is preferable to provide the signal line 12 with a coating 33 (FIG. 3). This coating 33 is conductive, and by grounding the coating 33 to the control unit 21, electromagnetic noise from the outside can be blocked. It is also possible to use a so-called shielded wire in which the signal line 12 and the covering 33 have a coaxial structure.

In order to quantitatively measure the degree of approach of the object 10d (subject 31) based on the change in capacitance at the electrode 10a, the lengths of the plurality of signal lines 12 may be matched.

<Method for measuring changes in capacitance>
In the antenna section 10 of the first embodiment, the electrode 10a has a monopolar structure, so there are the following two measurement methods.
(1) Measurement using absolute value The degree of approach of the object 10d (subject 31) to the antenna section 10 is measured based on the change over time in the absolute value of the capacitance at the electrode 10a. The closer the antenna unit 10 and the target object 10d (target person 31) are, the more the absolute value of the capacitance increases.

The initial capacitance at each electrode 10a in a state where the object 10d (subject 31) is not close is made equal, the absolute value of the capacitance is divided into levels in advance, and the degree of approach is measured at that level. . For example, by matching the lengths of the covering 33 of the signal line 12 and the plurality of signal lines 12, the initial capacitance at each electrode 10a can be made equal.

As a result, with this method, the degree of approach of the target object 10d (target person 31) to the antenna section 10 can be quantitatively measured.
(2) Measure the difference When the initial capacitances at the electrodes 10a are not equal, the degree of approach of the object 10d (subject 31) to the antenna section 10 is determined by the change in capacitance at the electrodes 10a over time. Measure. The closer the antenna unit 10 and the target object 10d (target person 31) are, the more the amount of change in capacitance increases.

Since the initial capacitance at each electrode 10a is not equal, the capacitance is calculated based on the capacitance at each electrode 10a before measurement (average over a certain period of time before measurement) with respect to that reference. The amount of change (change over time) is calculated for each electrode, the amount of change is determined by a threshold, and approach is detected based on this determination.

As a result, this method can qualitatively detect the approach of the target object 10d (target person 31) to the antenna section 10.

The posture and movement of the subject 31 can be monitored by the bed leaving detection device 22. However, in order to detect the posture and motion of the subject 31 with higher accuracy, the following apparatus and method will be described.

(Embodiment 2)
Embodiment 2 is a modification of the bed leaving detection device and bed leaving detection method of Embodiment 1. In particular, it is a bed-leaving detection device and a bed-leaving detection method that correspond to the structure of the bed 30 and movements such as reclining. Items not explained are the same as those in the first embodiment.

<Bed 30>
An example of the bed 30 will be explained with reference to FIG. 4(a) to 4(d) are side views of the bed 30. FIG. FIGS. 4(e) to 4(h) are side views showing an example of how the bed 30 is used. FIG. 4(i) is a top view of the bed 30 shown in FIGS. 4(b) and 4(f).

The number of plates 51 differs depending on the bed 30. Each bed 30 can change the posture of the subject 31 using it by tilting one of the plates 51.

When placing the bed leaving detection device 22 on these beds 30, if the bed leaving detection device 22 is placed on the entire surface of the bed 30 and a plurality of antenna sections 10 are designed on the entire surface, the posture of the subject 31 on the bed 30 can be adjusted. and the operation can be understood accurately.

However, since the cost would be high, the amount of data to be processed would be large, and the processing time would be long, we considered installing the sensor unit 20 only in the necessary areas, that is, in a part of the entire surface of the bed 30. .

This is shown in FIGS. 5(a) to 5(h).

5(a) to 5(d) are side views corresponding to FIGS. 4(a) to 4(d), respectively, and show the arrangement of the sensor unit 20. FIG.

5(e) to 5(h) are side views corresponding to FIGS. 4(e) to 4(h), respectively, and show the arrangement of the sensor unit 20. FIG.

FIG. 6(a) is a top view showing the arrangement of the plates 51a to 51d of the bed 30. 6(b) to 6(d) are top views showing the position of the antenna section 10 of the sensor unit 20 on the bed 30 of FIG. 6(a). In FIGS. 6(b) to 6(d), the arrangement of the antenna section 10 is different. It is preferable that the sensor unit 20 is not divided into two or more parts.

The antenna portions 10 are preferably arranged in at least two rows and two columns, where the longitudinal direction of the bed 30 is a row and the non-longitudinal direction (vertical direction) is a column. The antenna sections 10 in two or more rows and two or more columns are preferably arranged in a grid in one sensor unit 20. As will be explained below, it is better if there are at least two rows and six columns of antenna sections 10.

It is preferable that two or more rows and two or more columns of antenna sections 10 be arranged within one of the plates 51 of the bed 30. It is preferable that the antenna section 10 is not disposed between any of the plates 51 that are movable zones (flexible zones) for reclining the bed 30. If the antenna section 10 is bent between the plates 51, the antenna section 10 will be bent and deformed and will respond to a signal, which is not preferable.

In FIG. 5(e), the subject's 31 buttocks are located on the plate 51c. In FIG. 5(f), the subject's 31 buttocks are located on the plate 51b. In FIG. 5(g), the subject's 31 buttocks are located on the plate 51b. In FIG. 5(h), the subject's 31 buttocks are located on the plate 51b.
One row of the array of antenna units 10 is arranged in a buttock region located at the buttock of the user (subject 31) on the bed 30. Since the subject 31 moves his or her center of gravity around the buttocks, the posture and motion of the subject 31 can be detected.

One row of the array of antenna sections 10 is arranged in any region other than the buttock region and closer to the head than the buttock region. 5(e), the antenna section 10 is arranged on the plate 51b, FIG. 5(f) on the plate 51a, FIG. 5(g) on the plate 51a, and FIG. 5(h) on the plate 51a. One of the rows is located.
As a result, it is possible to grasp the posture and motion of the subject 31 not only in the buttocks but also in any region on the head side of the buttocks region, and the posture and motion of the subject 31 related to getting off the bed can be detected with high accuracy. Can be done.

As a result, the antenna section 10 of the sensor unit 20 is located in the region (plate 51) where the buttocks of the subject 31 are located, and in the region closer to the head than the buttocks of the subject 31 in the longitudinal direction (row direction) of the bed 30. At least two rows are arranged in another area (another plate 51). The region closer to the head than the buttocks of the subject 31 is, for example, the shoulder, back, waist, and abdomen of the subject 31.

The antenna part 10 may not be provided in the head, legs, and knee parts of the feet.

<Bed-off detection device 22>
FIG. 7 is a configuration diagram of the bed leaving detection device 22. As shown in FIG. The bed leaving detection device 22 includes a sensor unit 20 and a control unit 21. FIG. 7 shows the configuration of the sensor unit 20 and the configuration of the control unit 21. In the control unit 21, each element is connected by an electric circuit.

The control unit 21 includes a signal measurement section 14, a control section 15, a recording section 16, a display section 17, a communication section 18, a power supply section 19, a signal processing section 52, an attitude discrimination section 53a, and an operation discrimination section. It includes a section 53b, a notification determination section 54, and a notification selection section 56.

The signal measurement section 14 is a section that measures the capacitance of each of the antenna sections 10 based on the signals of the plurality of antenna sections 10.

The control unit 15 is an IC or the like that controls each component of the bed leaving detection device 22.

The recording unit 16 is an IC memory or the like that stores measurement results from the signal measurement unit 14, setting values of each component, and the like. When data is not transmitted to the outside, the recording unit 16 may be a removable recording medium such as an SD card or a USB memory.

The display unit 17 is a display such as liquid crystal or EL, or an indicator such as LED. There are operation buttons, etc., and instructions for operating the bed leaving detection device 22 are input. It is also possible to display measurement results. In addition, the display section 17 is a section for setting and instructing the components.

The communication unit 18 is a part that sends a signal to the outside based on instructions from the notification determination unit 54. Data can be sent wirelessly to personal computers, smartphones, mobile devices, etc. using Bluetooth (trademark), wireless LAN, etc. In the case of wired communication, the communication section 18 has an external connector part.

The power supply section 19 is a section that supplies power to other sections. This is a part that receives power from a battery or an external source. You may also produce your own electricity using solar cells.

The signal processing unit 52 processes the data from the signal measurement unit 14. For example, a certain standard is set, and if the signal is above the standard, there is a signal, and if it is smaller than the standard, there is no signal, etc., and the signal is processed into binary data. Data can be processed according to <method for measuring change in capacitance> in Embodiment 1. The processing differs depending on the type and structure of the antenna section 10.

The posture determining section 53a determines the posture of the subject 31 from the data of the signal processing section 52 stored in the recording section 16. For example, the determination is made based on average data for 1 to 2 seconds. It is not essential to store data in the recording unit 16. The average data uses at least two changes.

The motion determining section 53b determines the motion of the subject 31 from the data of the signal processing section 52 stored in the recording section 16. For example, an operation is determined based on three changes in average data for 1 to 2 seconds, that is, three changes in data. The average data uses at least two changes.

The notification determining unit 54 determines whether to notify based on the determination result from the posture determining unit 53a or the motion determining unit 53b.

The notification selection unit 56 is a part that selects in advance what kind of determination result should be used to notify the notification determination unit 54 . That is, the notification determination unit 54 has preset conditions for notification.
For example, for a subject with a high risk of falling, an alarm is issued when movement from near the center of the bed is detected. Furthermore, for subjects with a low risk of falling, it is preferable to issue an alarm when a posture out of bed is detected. In other words, it is preferable to set a standard in the notification determination unit 54 for each target person 31 and to set whether or not to notify. This becomes a bed leaving detection device suitable for the subject 31.

The sensor unit 20 in FIG. 7 has eight antenna parts 10 on one base material 25. They are arranged as shown in FIG. 6(b), FIG. 5(b), and FIG. 5(f).

<Data (signal) from sensor unit 20>
Data (signals) from the sensor unit 20 will be explained using FIG. 8. FIG. 8A shows RAW data from the sensor unit 20 that the signal measurement section 14 receives. The signals from each antenna section 10 change with the passage of time. The in-bed period shown in FIG. 8 is the time period when the subject 31 is stationary on the bed 30, and the bed leaving period is the time period when the subject 31 is moving to get out of the bed 30. The period is a period during which the subject 31 is away from the bed 30.

FIG. 8B shows data obtained by processing RAW data by the signal processing unit 52. In this case, 1 and 0 or ON and OFF are used. It is set as ON (1) when it is higher than the set threshold value, and is set as OFF (0) when it is lower than the set threshold value.

FIG. 8(c) is a plan view showing the number 23 of the antenna section on the sensor unit 20.
When all the signals from the antenna section 10 are OFF, the subject 31 has left the bed 30.
When any of the signals from the antenna unit 10 is ON, the subject 31 is on the bed 30 in any position.

FIGS. 8(d) to 8(g) are plan views of the sensor unit 20 showing the results of FIG. 8(b). Pixel data of the posture of the subject 31 was derived by processing the signals from the plurality of antenna units 10 (ON when above a certain level, OFF when below it).

In FIG. 8D, the antenna portions 23, 3, 4, and 10 are ON, and the subject 31 is in the lying position near the center of the bed 30.

In FIG. 8(e), only the antenna section number 23, 10, is ON, and the subject 31 is in a sitting position near the center of the bed 30.

In FIG. 8(f), antenna portions 23, 2 and 10 are ON, and the subject 31 is in a posture of trying to move from near the center of the bed 30.

In FIG. 8(g), the antenna portions 23, 3, 4, 9, and 10 are ON, and the subject 31 is in a recumbent position near the center of the bed 30.

<Process>
Explain the process.
(1) Receiving process: The signal measurement unit 14 receives the signal from the antenna unit 10 via the connection unit 13 (FIG. 8(a)).

(2) Processing step: The signal processing unit 52 receives the signal from the signal measurement unit 14 and processes it (FIG. 8(b)).

(3) Recording step: The recording section 16 records the signal processed by the signal processing section 52.

(4a) Determination step 1: The posture determination unit 53a determines the posture of the subject 31 from the signals recorded at a certain point in time, for example, the pixel data in FIGS. 8(d) to (g). For example, in FIG. 8(d), it is determined that the subject 31 is in the lying position near the center of the bed 30.

(4b) Determination step 2: The motion determination unit 53b determines the motion of the subject 31 from the signals recorded at three consecutive points in time. For example, if FIG. 8(d) is followed by FIG. 8(e), then FIG. 8(f), the subject 31 changes from a lying position near the center of the bed 30 to a sitting position. It is determined that the user is in a posture of attempting to move toward the end of the bed 30.

(5) The notification determination unit 54 receives either or both of the results of (4a) and (4b) above. The notification determination unit 54 determines whether to notify based on the condition that the determination result is set in advance from the notification selection unit 56 and instructs the communication unit 18.
For example, if the notification selection unit 56 is set in advance to notify when it is determined that the subject 31 has stood up, when receiving the result that the subject 31 has stood up from above (4a), the information is sent to the nurse call system etc. Notify by communication.
For example, if the notification selection unit 56 is set in advance to notify when it is determined that the subject 31 is sitting on the edge of the bed (has become the edge sitting position), from the above (4a), if it is determined that the subject 31 has become the edge sitting position. When the results are received, the information is communicated to the nurse call system, etc.
For example, if the notification selection unit 56 is set in advance to notify when it is determined that the subject 31 is about to move from the center of the bed, in (4b), FIG. 8(d), FIG. 8(e) ), FIG. 8(f), it is determined that the subject 31 is about to move from the center of the bed 30, and this information is notified to a nurse call system or the like via communication.

(6) The communication unit 18 transmits a signal to the outside according to instructions from the notification determination unit 54. For example, it sends information, commands, etc. to a nurse call system, monitoring system, etc.

<AI technology>
AI technology can be used to determine the posture or motion of the subject 31 from the above signals.
This will be explained with reference to FIG. First, in FIG. 9(a), a learning data set is prepared and input to the learning program. As a result, the learning program completes the trained discriminator.
Next, in FIG. 9(b), the processed signal from the recording unit 16 described in FIG. 7 is input to the learned posture discriminator and the learned motion discriminator. The posture discriminator and motion discriminator respectively discriminate signals and output the results as posture labels and motion labels. Each label is sent to the notification determination section 54. The notification determination unit 54 determines whether to notify based on the condition that the determination result is set in advance from the notification selection unit 56 and instructs the communication unit 18.

The learning data set is teacher data. For example, for posture determination, the teacher data is data in which the data in FIGS. 8(d) to 8(g) correspond to the respective postures. For motion determination, the teacher data is data in which the continuous data in FIGS. 8(d) to 8(f) are associated with the motions.

The learning program is one of a linear model, a neural network model, a support vector machine model, a decision tree model, and a random forest model. In particular, the learning program is preferably a neural network model or a random forest model.

<Machine learning department 61>
A machine learning section 61 may be provided as the learning program. This will be explained with reference to FIG. FIG. 10 shows the configuration of the machine learning section 61. The signal measuring section 14 and the signal processing section 52 are the same as described above. The learning section 62 has a learning program, and uses the above model to associate data from the sensor unit 20 with posture or motion, thereby completing a learned discriminator (FIG. 9(a)).
The machine learning section 61 may not be present in the control unit 21. A machine learning section 61 may be provided separately, and the resulting learned classifier (FIG. 9(a)) may be relocated to the control unit 21.
Note that in the case of the embodiment, since the number of antenna units 10 is small, all patterns (patterns like FIGS. 8(d) to 8(g)) are not large in terms of posture. Therefore, the posture may be determined by recording patterns in the recording section 16 in advance without using the machine learning section 61.

<Sensor unit 20>
The above-described specific structure of the antenna section 10 of the sensor unit 20 (particularly, FIG. 3) is one example, and another structure may be used. Various pressure sensors may be used. In that case, it becomes a signal such as voltage. There are characteristics in the number and arrangement, and the structure of the antenna section 10 itself is not limited to the above.

<Effect>
The bed-leaving detection device, bed-leaving detection method, and machine learning device of the embodiment have the following effects.

(1) In medical facilities, nursing care facilities, or general homes, persons with reduced cognitive or physical ability, or those who are at risk of delirium (impaired consciousness), should be advised of how to get up from bed. Posture and motion can be determined, and depending on the determination results, a notification can be output to a nurse call system, monitoring system, audio output device, etc.

(2) As a basic function, the sensor unit 20 has the necessary number of antennas in the necessary arrangement so as to be able to detect the posture and motion related to leaving the bed with high accuracy, so there is no possibility of erroneous alarms of bed leaving detection. Less detected.

(3) As an advanced function, advance detection before leaving the bed is possible.

(4) Economically, the sensor unit 20 has a simple structure and is made of inexpensive materials, so it is inexpensive. Furthermore, since the number of antenna sections 10 is small, the data capacity and calculation load are small.

(5) In terms of operability, since video or images such as cameras are not used, privacy information is not handled and high information management is not required.

(6) In terms of workability, the sensor unit 20 is inexpensive and can be separated from the control unit 21, so it can be immediately removed (replaced) when unnecessary (contamination occurs). In addition, since it is only placed in a necessary part of the bed 30, near the vertical center of the bed 30, rather than on the entire surface of the bed 30, it can be installed easily and quickly when bed leaving detection is required, and the position can be adjusted by reclining motion. No misalignment occurs.

(as a whole)
Each element of the embodiment can be combined.
The bed-leaving detection device, the bed-leaving detection method, and the machine learning device need only have at least one of posture discrimination and motion discrimination, rather than both.

The bed exit detection device, bed exit detection method, and machine learning device of the present invention can be used not only in hospitals and nursing care facilities, but also at home, accommodation facilities, and the like.

10 Antenna section 10a Electrode 10b Cover 10c Base material 10d Object 11 Antenna connection section 12 Signal line 13 Connection section 14 Signal measurement section 15 Control section 16 Recording section 17 Display section 18 Communication section 19 Power supply section 20 Sensor unit 21 Control unit 22 Leave the bed Detection device 23 Antenna section number 25 Base material 30 Bed 31 Subject 33 Covering 41 Cover 51 Plates 51a, 51b, 51c, 51d Plate 52 Signal processing section 53a Posture discrimination section 53b Motion discrimination section 54 Notification determination section 56 Notification selection section 61 Machine Learning Department 62 Learning Department

Claims (13)

a sensor unit having a plurality of antenna parts, a signal line connected to the antenna part, and a connection part connected to the signal line;
a control unit that is connected to the connection section and includes a signal measurement section that measures a signal from the antenna section; and a control section that controls the signal measurement section;
including;
A bed leaving detection device, wherein the plurality of antenna parts are arranged in a grid pattern in two or more columns and two or more rows. The bed leaving detection device according to claim 1, wherein the sensor unit is installed together with or integrated with a waterproof sheet placed around the buttocks and waist of the subject for the purpose of preventing incontinence of the subject. The bed leaving detection device according to claim 1, wherein the array is arranged in at least 2 rows and 6 columns. 2. The bed leaving detection device according to claim 1, wherein, in the arrangement, one antenna section is not arranged across the movable band for a plurality of areas of the seat surface of the bed separated by a movable band for reclining the bed. The bed leaving detection device according to claim 1, wherein one row of the array is arranged in a buttocks region where the buttocks of the user on the bed are located. The bed leaving detection device according to claim 5, wherein one row of the array is arranged in a region other than the buttock region and closer to the head than the buttock region. The control unit includes:
Deriving pixel data of the subject's posture and movement from the plurality of antenna parts arranged in a grid pattern,
The bed-leaving detection device according to claim 1, wherein a preset notification determination section is used to determine a posture or motion regarding bed-leaving, and a notification output is performed in accordance with the determination result. The bed leaving detection device according to claim 7, wherein a criterion is set in the notification determination unit for each subject and outputs the notification. The bed-leaving detection device according to claim 7, wherein a discriminator subjected to machine learning in advance is used as the notification determination section. a receiving step of receiving signals from a plurality of antenna units arranged in a grid pattern in two or more columns and two or more rows;
a processing step of receiving and processing the signal;
A determination step of determining the posture of a subject from the processed signal. 10. The bed leaving detection method according to claim 9, wherein in the determining step, the motion of the subject is determined from the processed signals at a plurality of consecutive times. In the bed leaving detection device, a machine learning device that determines the posture or movement of the subject based on the signal from the sensor unit,
a signal measurement unit that receives a signal from the sensor unit;
a signal processing unit that processes the signal;
a learning unit that learns using teacher data of the signal and the posture or motion of the subject;
machine learning devices including; The bed leaving detection device according to any one of claims 1 to 9, wherein the machine learning device determines the posture or motion of the subject based on the signal from the sensor unit,
a signal measurement unit that receives a signal from the sensor unit;
a signal processing unit that processes the signal;
a learning unit that learns using teacher data of the signal and the posture or motion of the subject;
machine learning devices including;