JP7497497B2 - system - Google Patents

Description

The present invention relates to electrically operated furniture.

Inventions for determining the state of a patient on a bed device have been known for some time. For example, Patent Document 1 discloses an invention in which a bed occupancy detection device is provided with a vibration detection unit that is placed on a bed and that separately detects components of vibrations applied to the bed that run along the longitudinal direction of the bed and components that run along the vertical direction to the floor surface of the bed, a signal extraction unit that extracts a heartbeat vibration signal derived from the heartbeat of a living body present on the bed from the vibrations detected by the vibration detection unit, and a state estimation unit that estimates the posture of the living body on the bed based on the heartbeat vibration signal extracted by the signal extraction unit.

Patent Document 2 also discloses an alarm device that includes a user state acquisition unit that acquires a user state including the user's posture and/or the user's position on or off the bed, a danger judgment unit that judges whether the user state corresponds to a danger state, a safety judgment unit that judges whether the user state corresponds to a safe state, a state setting unit that sets either a valid state or an invalid state, and an alarm unit that judges that the user state corresponds to a danger state and is in a valid state and performs an alarm process, the state setting unit switching from the valid state to the invalid state when the alarm process is performed, and switching from the invalid state to the valid state when it judges that the user state corresponds to a safe state in the invalid state.

JP 2011-120667 A JP 2016-192998 A

An object of the present invention is to provide a system capable of appropriately determining a patient's posture from a body vibration waveform.

In order to solve the above-mentioned problems, the patient posture determination device of the present invention is a posture determination device having at least two vibration sensors and a control unit, and the control unit detects vibrations of a subject lying down using the vibration sensors, calculates a waveform from the detected vibrations, and recognizes the characteristics of the waveform to determine the posture of the subject from the characteristics.

The posture determination device of the present invention can appropriately determine the posture of a subject as one of the subject's conditions from the body vibration waveform in combination with other biological information such as heart rate, respiratory rate, and activity level, as well as sleep state.

FIG. 2 is a diagram for explaining the whole of the first embodiment. FIG. 2 is a diagram for explaining a configuration in the first embodiment. FIG. 2 is a diagram for explaining a configuration in the first embodiment. FIG. 2 is a diagram for explaining a sensor in the first embodiment. 10A to 10C are diagrams for explaining a patient posture determination process in the first embodiment. 11 is a diagram for explaining a patient posture determination process in the first embodiment. FIG. FIG. 11 is a diagram showing an example of a waveform in a prone position in the first embodiment. FIG. 11 is a diagram showing an example of a waveform in a supine position in the first embodiment. FIG. 11 is a diagram showing an example of a waveform in the left lateral position in the first embodiment. FIG. 11 is a diagram showing an example of a waveform in a right lateral position in the first embodiment. FIG. 4 is a diagram illustrating an example of frequency components in the first embodiment. 13A and 13B are diagrams for explaining a patient posture determination process in the second embodiment. FIG. 11 is a diagram for explaining a posture determination condition in the second embodiment. FIG. 11 is a diagram for explaining the state of a waveform in the second embodiment. FIG. 13 is a diagram for explaining the functional configuration of a patient condition inferring unit in the third embodiment. FIG. 13 is a diagram for explaining a neural network in the fourth embodiment. 11A and 11B are diagrams for explaining a bed apparatus as an application example. FIG. 13 is a diagram for explaining a process in the case of an application example.

One embodiment of the present invention will be described below with reference to the drawings. In order to prevent a patient from falling when getting out of bed, in comparative examples, it is common to detect whether the patient is out of bed (in bed). However, there is no detection of the posture of the patient when in bed.

For example, when detecting a patient's posture before getting out of bed, a large number of sensors may be placed over a wide area of the bed to distinguish between lying down and sitting, but it has not been possible to determine the direction in which the patient is sleeping (e.g., supine, prone, lateral, etc.).

In order to determine the patient's posture, it was necessary to either install a separate camera device to monitor the patient or to perform image analysis. In this case, the burden on the person monitoring the patient was heavy. In addition, because the camera had to be constantly recording, there were cases where patients objected to this from the standpoint of protecting their privacy.

Therefore, with the posture determination device of this embodiment, it is possible to detect the posture of the patient simply by detecting the vibrations while the patient is sleeping on the bed device.

In this specification, a patient refers to a person who uses a bed device (mattress), and is not limited to those who are receiving treatment for an illness, but can be anyone who receives care in a facility or sleeps on a bed device.

[1. First embodiment]
[1.1 Overall System]
Fig. 1 is a diagram for explaining an overall outline of a system 1 to which a posture determination device of the present invention is applied. As shown in Fig. 1, the system 1 is configured with a detection device 3 placed between a floor part of a bed device 10 and a mattress 20, and a processing device 5 for processing values output from the detection device 3. The detection device 3 and the processing device 5 configure a system for determining the posture of a patient.

When a subject (hereinafter, referred to as "patient P" as an example) is present on the mattress 20, the detection device 3 detects body vibrations (vibrations emitted from the human body) as a biosignal of the subject, patient P. In addition, the bioinformation value of patient P can be calculated based on the detected vibrations.

The calculated bioinformation values (e.g., respiratory rate, heart rate, activity level) may be output and displayed as the bioinformation values of patient P.

The detection device 3 may be provided with a storage unit, a display unit, etc., so that it is formed integrally with the processing device 5. The processing device 5 may be a general-purpose device, and is not limited to an information processing device such as a computer, and may be configured as a device such as a tablet or smartphone. If the detection device 3 has a communication function, it may be connected to a server device instead of the processing device 5.

The subjects may be people undergoing medical treatment or those in need of nursing care. They may also be healthy people who do not require nursing care, elderly people, children, people with disabilities, or even non-human animals.

Here, the detection device 3 is configured in a sheet shape so that it is thin. This means that even if it is placed between the bed device 10 and the mattress 20, it can be used without causing discomfort to the patient P, making it possible to detect the patient's condition in bed for a long period of time.

The detection device 3 only needs to be able to detect the vibrations of the patient P. For example, an actuator with a strain gauge or a load cell that measures the load placed on the legs of the bed may be used. Also, by using a built-in acceleration sensor, the detection device 3 may be realized by, for example, a smartphone or tablet placed on the bed device 10.

In addition, in FIG. 1, the head side of the bed device 10 (mattress 20) is direction H, the foot side is direction F, and when patient P is in the supine position, the left side is direction L and the right side is direction R.

1.2 Configuration
Next, the configuration of the system 1 will be described with reference to Fig. 2 to Fig. 4. The system 1 in this embodiment includes a detection device 3 and a processing device 5, and each functional unit (processing) may be realized by either unit except for the vibration detection unit 110. In other words, by combining these devices, the system functions as a posture determination device.

System 1 (posture determination device) includes a control unit 100, a vibration detection unit 110, a bioinformation calculation unit 120, a waveform calculation unit 130, a frequency distribution calculation unit 135, a patient condition determination unit 140, a memory unit 150, an input unit 160, and an output unit 170.

The control unit 100 controls the operation of the system 1. For example, it is a control device such as a CPU (Central Processing Unit). The control unit 100 realizes various processes by reading and executing various programs stored in the storage unit 150. Note that in this embodiment, the control unit 100 operates as a whole, but it can also be provided in each of the detection device 3 and the processing device 5, as shown in FIG. 4, which will be described later.

The vibration detection unit 110 detects vibrations on the detection device 3 and acquires vibration data. In this embodiment, as an example, a sensor that detects pressure changes is used to detect vibrations (body vibrations) based on the patient's movements, etc. Vibration data is acquired based on the detected vibrations and is output to the bioinformation calculation unit 120, the waveform calculation unit 130, and the frequency distribution calculation unit 135 for processing. Note that the output vibration data may be analog vibration data or digital vibration data.

The vibration detection unit 110 may, for example, detect the patient's vibrations using a pressure sensor to obtain vibration data, or may provide a microphone instead of a pressure sensor to obtain a biosignal based on the sound picked up by the microphone and obtain vibration data from the biosignal. Vibration data may also be obtained from the output value of an acceleration sensor, capacitance sensor, or load sensor. In this way, it is sufficient to use any method to obtain a biosignal (body vibration indicating the patient's body movement).

The bioinformation calculation unit 120 acquires the biosignals of the patient P from the vibration data, and calculates bioinformation values (respiratory rate, heart rate, activity level, etc.). In this embodiment, the respiratory component and heart rate component may be extracted from the vibration (body vibration) data acquired by the vibration detection unit 110, and the respiratory rate and heart rate may be calculated based on the respiratory interval and heart rate interval. In addition, the periodicity of body movement may be analyzed (Fourier transform, etc.), and the respiratory rate and heart rate may be calculated from the peak frequency.

The waveform calculation unit 130 converts the analog vibration data input from the vibration detection unit 110 into a digital voltage signal at a predetermined sampling interval and calculates the waveform of the body vibration (vibration waveform).

In this embodiment, for convenience of explanation, the waveform calculation unit 130 calculates the waveform of the vibration data (vibration waveform) from the vibration data output by the vibration detection unit 110, and processing is performed based on the calculated waveform. However, various processing may be performed based on the vibration data without having the waveform calculation unit 130.

Furthermore, the frequency distribution calculation unit 135 calculates the frequency distribution from the vibration (waveform). For example, the frequency distribution calculation unit 135 calculates the frequency components by performing a fast Fourier transform (FFT) on the waveform calculated by the waveform calculation unit 130.

The patient condition determination unit 140 determines the condition of the patient. For example, the patient condition is determined based on the vibration data acquired by the vibration detection unit 110, the bio-information value calculated by the bio-information calculation unit 120, a load sensor separately provided in the bed device 10, etc.

In this embodiment, the patient state determination unit 140 mainly determines the patient's posture (supine, prone, lateral). The patient state determination unit 140 may also determine other patient postures, such as sitting on the edge of the bed, or postures other than posture (for example, getting out of bed, staying in bed, etc.).

The storage unit 150 stores various data and programs for the operation of the system 1. The control unit 100 realizes various functions by reading and executing the programs stored in the storage unit 150. Here, the storage unit 150 is composed of a semiconductor memory (e.g., an SSD (Solid State Drive) or an SD card (registered trademark)), a magnetic disk device (e.g., an HDD (Hard Disk Drive)), etc. The storage unit 150 may be a built-in storage device or a removable external storage device. It may also be a storage area of an external server such as a cloud.

The memory unit 150 has a vibration data memory area 152 and a waveform data memory area 154.

The vibration data storage area 152 stores the vibration data output from the vibration detection unit 110. Here, the interval at which the vibration data is stored may be every predetermined time, and may be a short interval such as every second or every five seconds, or a relatively long interval such as 30 seconds, one minute, or five minutes.

The waveform data storage area 154 stores data on the vibration waveform (waveform data) calculated by the waveform calculation unit 130 based on the vibration data output from the vibration detection unit 110 or the vibration data stored in the vibration data storage area 152. Note that, in this embodiment, the waveform data is described as being stored in the waveform data storage area 154, but the waveform calculation unit 130 may calculate the waveform data each time as necessary.

Note that the waveform data storage area 154 is reserved as a storage area for the convenience of explanation, but it is sufficient if it is calculated from the vibration data as appropriate and temporarily stored.

The input unit 160 accepts operational input from the user. For example, the user may perform various operational inputs, such as starting vibration acquisition or adjusting the sensitivity of the vibration detection unit 110.

The output unit 170 outputs various types of information. For example, it may be configured with a liquid crystal display device, a light-emitting element such as an LED, a speaker that outputs sound or voice, an interface that outputs data to another recording medium, etc. Also, the input unit 160 and the output unit 170 may be configured as a touch panel by being integrated together.

Of the above-mentioned components, the bioinformation calculation unit 120, the waveform calculation unit 130, the frequency distribution calculation unit 135, and the patient condition determination unit 140 are mainly realized by software. For example, the control unit 100 reads and executes software stored in the memory unit 150. When the software is executed, the control unit 100 is realized as each component.

In other words, the control unit 100 loads and executes the programs that realize the bioinformation calculation unit 120, the waveform calculation unit 130, the frequency distribution calculation unit 135, and the patient condition determination unit 140, so that the control unit 100 has the functions of each component.

Figure 2 also illustrates a conceptual configuration of a posture determination device as system 1. This configuration may be realized, for example, by a single device capable of detecting vibration, or may be configured separately as detection device 3 and processing device 5 as in Figure 1. Also, instead of processing device 5, it may be realized by an external server capable of providing the same service.

The case where the system 1 in FIG. 2 is realized by the detection device 3 and processing device 5 in FIG. 1 will be described with reference to FIG. 3. The detection device 3 includes a control unit 300, a vibration detection unit 320 which is a sensor, a storage unit 330, and a communication unit 390.

The control unit 300 also functions as a bioinformation calculation unit 310 by executing software (programs) stored in the storage unit 330. The vibration detection unit 320 outputs vibration data based on the detected vibration.

The biometric information calculation unit 310 calculates the biometric information based on the vibration data. The calculated biometric information (value) is then stored in the biometric information data 340 or transmitted to the processing device 5 via the communication unit 390. Additionally, the vibration data detected by the vibration detection unit 320 is also transmitted to the processing device 5 via the communication unit 390.

The timing of transmitting vibration data from the detection device 3 to the processing device 5 and the timing of storing the bioinformation value (biometric information) in the bioinformation data 340 may be in real time or at predetermined intervals.

The vibration detection unit 320 is the vibration detection unit 110 in FIG. 2. Here, the vibration detection unit 320 will be described with reference to FIG. 4.

Figure 4 is a view of the bed device 10 (mattress 20) as seen from above. The upper side of Figure 4 corresponds to direction H in Figure 1, and the lower side corresponds to direction F in Figure 1. The right side of Figure 4 corresponds to direction F in Figure 1, and the left side of Figure 4 corresponds to direction R in Figure 1.

The detection device 3 is placed between the bed device 10 and the mattress 20 or on the mattress 20. The place where it is placed is preferably near the patient's back, so it is at least in a direction closer to the H side than the center of the bed device 10 (mattress 20).

The detection device 3 also has a built-in sensor (vibration detection unit 110/320). The sensor is, for example, a vibration sensor, and can detect the patient's vibrations (body vibrations). At least two sensors are provided, and for example, in FIG. 4, two sensors (vibration sensor 320a, vibration sensor 320b) are provided on the left and right sides of the detection device 3.

These vibration sensors 320a and 320b are placed at a predetermined distance. The distance between the two sensors may be, for example, approximately the lateral width of the patient, and preferably the distance between the two sensors is approximately 15 to 60 cm.

The biometric information calculation unit 310 in FIG. 3 is the biometric information calculation unit 120 in FIG. 2. The communication unit 390 is, for example, a communication interface that can be connected to a network (for example, a LAN/WAN).

The processing device 5 includes a control unit 500, a storage unit 530, an input unit 540, an output unit 550, and a communication unit 590. The processing device 5 receives vibration data from the detection device 3 via the communication unit 590. The received vibration data is stored in the vibration data storage area 532.

The control unit 500 executes software (programs) stored in the storage unit 530 to function as a waveform calculation unit 502, a frequency distribution calculation unit 504, and a patient condition determination unit 506. In addition, the waveform data calculated by the waveform calculation unit 502 is stored in the waveform data storage area 534.

The waveform calculation unit 502 is the waveform calculation unit 130 in FIG. 2. The frequency distribution calculation unit 504 is the frequency distribution calculation unit 135 in FIG. 2. The patient condition determination unit 506 is the patient condition determination unit 140 in FIG. 2. The input unit 540 is the input unit 160 in FIG. 2. The output unit 550 is the output unit 170 in FIG. 2. The memory unit 530 is the memory unit 150 in FIG. 2.

[1.3 Processing flow]
Next, the posture determination process in this embodiment will be described with reference to Fig. 5. The posture determination process is a process executed by the patient condition determiner 140.

First, the control unit 100 (patient condition determination unit 140) acquires vibration data (step S102). The patient condition determination unit 140 acquires the vibration data by reading it from the vibration data storage area 152 or by receiving it from the vibration detection unit 110.

Next, the patient's posture is determined based on the correlation between sensors and the correlation within the sensor based on the vibration data. In this embodiment, for example, the waveform calculation unit 130 calculates a waveform and outputs it as waveform data (step S104). The waveform data is calculated for each vibration sensor. For example, as shown in FIG. 4, in this embodiment, two vibration sensors (vibration sensor 320a, vibration sensor 320b) are provided, and therefore the waveform data is calculated for each of them.

The waveform data may be stored in the waveform data storage area 154 or may be output by the output unit 170. For example, if the output unit 170 is a display device, the waveform data is displayed on the display device.

Next, the patient condition determination unit 140 determines whether there is a correlation between the sensors from the waveform data (step S106). Here, the correlation between the sensors is one method of determining whether there is a similarity between waveforms calculated based on vibration data acquired from two sensors (e.g., vibration sensor 320a and vibration sensor 320b in FIG. 4).

As an example, the cross-correlation function is used for two waveforms. The cross-correlation function outputs a value normalized between "0" and "1" based on the similarity between the two waveforms. This value changes depending on the similarity between the two waveforms. For example, when the value of the cross-correlation function is "1", the two waveforms match perfectly and the similarity is at its maximum. On the other hand, when the value of the cross-correlation function is "0", the two waveforms do not match at all and the similarity is at its minimum.

When the patient condition determination unit 140 determines whether or not two waveforms are correlated, it does so based on whether or not the output value of the cross-correlation function exceeds the cross-correlation function threshold. For example, if the cross-correlation function threshold is "0.7", then if the output value of the cross-correlation function is "0.7" or less, it is determined that the two waveforms are not correlated. On the other hand, if the output value of the cross-correlation function exceeds "0.7", it is determined that the two waveforms are correlated.

Next, if there is a correlation between the sensors (step S106; Yes), the patient state determination unit 140 determines that the patient's position is "supine or prone" (steps S116 to S120). Note that the time spent in the prone position during sleep is extremely short, so the patient's position may simply be determined to be "supine or prone", but because the prone position has a high risk of suffocation and has been reported to be associated with sudden infant death syndrome, and because automatic operation of the electric bed may be prohibited when the patient is in the prone position, the patient may be further determined to be in the supine or prone position.

That is, it is determined whether or not there is correlation within the sensor (step S106; Yes -> step S116). Here, to determine whether or not there is correlation within the sensor, for example, the strength of periodicity in the waveform is evaluated. As an example, an autocorrelation function is used to determine whether or not there is correlation for the waveform of one sensor. The autocorrelation function outputs a value normalized to "0" to "1" based on the strength of the periodicity of the waveform within the same sensor. For example, when the value of the autocorrelation function is "1", it is clear that the waveform data is output perfectly periodically and there is perfect correlation within the sensor.

The strength of periodicity may also be calculated as a normalized value between "0" and "1" using a Fourier transform or chi-squared periodogram.

When determining whether or not a waveform is correlated, the patient condition determination unit 140 may make the determination based on whether or not the output value of the autocorrelation function exceeds an autocorrelation function threshold. For example, if the autocorrelation function threshold is set to "0.7", then if the output value of the autocorrelation function is "0.7" or less, it is determined that the calculated waveform (a single piece of detected vibration data) is not correlated. On the other hand, if the output value of the autocorrelation function exceeds 0.7, it is determined that the waveform is correlated.

If the patient state determination unit 140 determines that there is a correlation within the sensor, it determines that the patient's position is "supine" (step S116; Yes → step S118). On the other hand, if the patient state determination unit 140 determines that there is no correlation within the sensor, it determines that the patient's position is prone (step S116; Yes → step S118).

If the patient state determination unit 140 determines in step S106 that there is no correlation between the sensors (step S106; No), the patient's position is determined to be "lateral position" (step S110). Note that while the patient's position may be determined to be simply "lateral position", in order to check for a change in position, it may also be determined to be right lateral position or left lateral position (step S114) because, for example, if the patient has a paralyzed side, it is necessary to be careful that the patient is not sleeping with the paralyzed side facing down.

In this case, the input of the heartbeat signal is greater in the left lateral position than in the right lateral position. Therefore, if a heartbeat signal is extracted that is greater than a predetermined level (if a high-frequency signal is extracted), the position is determined to be left lateral.

There are various methods for determining the magnitude of the heart rate signal input, but for example, the ratio of the frequency components corresponding to the heart rate signal to the frequency components corresponding to the respiratory signal, the signal strength of data that has been subjected to high-pass filtering, etc. can be used.

In this way, according to this embodiment, the patient's posture (sleeping posture) can be determined from the vibration data.

For convenience of explanation, the correlation between sensors and the presence or absence of correlation within a sensor are determined based on the waveform data calculated by the waveform calculation unit 130. However, the correlation between sensors and the correlation within a sensor may be determined simply based on vibration data. In this case, the process of step S104 does not need to be executed.

In addition, in the above-described embodiment, the patient's posture is determined by calculating a waveform from the patient's vibration data, but the posture may also be determined by evaluating the shape of the frequency distribution. In this case, higher accuracy is achieved by comprehensively evaluating the shape of the frequency distribution of vibrations acquired by two or more sensors, but the number of sensors acquiring vibrations may be one.

For example, the process of determining the patient's posture using frequency analysis will be described with reference to FIG. 6. As in FIG. 5, vibration data is acquired from the vibration detection unit 110 (step S152), and waveform data is calculated (output) by the waveform calculation unit 130 (step S154).

Next, the frequency distribution calculation unit 135 calculates the frequency distribution of the waveform data output by the waveform calculation unit 130 (step S156). For example, if no high-frequency components are calculated from the frequency distribution, the patient's posture is identified as "lateral position."

The relationship between the position of the heart and the position of the sensor varies depending on the posture, and so does the body tissue between them (muscle, fat, bone, internal organs, etc.). This means that the way vibrations are transmitted also changes, resulting in differences in the frequency components that are measured.

The direction of movement of the heart and the chest and abdomen due to breathing is fixed. For example, when the patient is in the supine position, the chest and abdomen move significantly in a direction perpendicular to the sensor (vibration detection unit 110). When the patient is in the lateral position, the movement is greater in a direction parallel to the sensor (vibration detection unit 110). Therefore, the frequency distribution differs for each posture of the patient. Therefore, the posture can be determined by storing a frequency distribution for each posture and comparing the actually extracted frequency distribution with the frequency distribution stored for each posture.

For example, in the lateral position, components with higher frequencies than the heartbeat component are difficult to detect (except for integer multiples of the frequency corresponding to the heartbeat component). Therefore, if there are few higher frequency components than the heartbeat component, the patient's posture can be determined to be "lateral position."

In the above description, the frequency distribution calculation unit 135 calculates the frequency distribution from the waveform data, but the frequency distribution may be calculated directly from the vibration data. In this case, step S154 is not executed.

[1.4 Conditions for posture judgment]
Here, the posture determination process described above may be executed normally or when a predetermined condition is met. Hereinafter, the patient's body movement is used as a condition for posture determination. Here, the patient's body movement refers to the patient's posture being changed, such as when the patient turns over in bed.

(1) Determining a posture for each section For example, a continuous section that does not include body movement is determined to be a continuation of the same posture, and the same posture is output.

In this case, the sections that do not include body movement may be analyzed together (applying the results of the posture determination process). In addition, the results of the posture determination process are tallied for a certain period of time (e.g., every 3 minutes). If there is a period within that period that does not include body movement, the determination result of the posture that is most prevalent within the period that does not include body movement is output as the result of the posture determination process for that period.

(2) Posture determination is performed only in sections where no body movement occurs Posture determination is performed only in sections where no body movement occurs. For example, when the vibration detection unit 110 detects the body movement of the patient, execution of the posture determination process is stopped. In addition, the posture determination process may be performed again after a predetermined time has elapsed since the body movement was detected and the body movement was no longer detected (for example, 10 seconds after the body movement was no longer detected).

[1.5 Waveform-based explanation]
Here, the determination of the patient's posture based on the waveform will be described. Each waveform is based on vibration detected based on two vibration sensors. The horizontal axis indicates time, and the vertical axis indicates voltage value, and each waveform based on the detected vibration is shown. For convenience of explanation, the determination of the patient's posture will be described based on the waveform, but the patient's posture may be determined by detecting a trend from time-series vibration data.

In Figure 7, the waveforms in the sensor are not correlated, and the same shape is not repeated. In other words, this waveform is a waveform that indicates a prone position (step S120 in Figure 5).

Figure 8 shows that there is a correlation between the waveforms of the sensors, and also between the waveforms within the sensor. In other words, this waveform is a waveform that indicates a supine posture (step S118 in Figure 5).

Figures 9 and 10 show waveforms with no correlation between the sensors. In other words, this waveform is a waveform that indicates a lateral position (step S110 in Figure 5). Comparing Figures 9 and 10, the high-frequency signal due to heartbeat is more prominent in Figure 10, which shows the waveform in the left lateral position.

Figure 11 is a graph showing the frequency distribution obtained from the waveform. Figure 11 is a graph showing the frequency distribution obtained in step S156 of Figure 6, with Figure 11(a) showing the supine position and Figure 11(b) showing the lateral position. In this way, the patient's posture can be determined by further extracting the frequency distribution from the waveform (vibration data).

[2. Second embodiment]
A second embodiment will now be described. The second embodiment is an embodiment in which the posture of a patient is determined based on a plurality of posture determination conditions.

Note that this embodiment replaces the operational flow of FIG. 5 of the first embodiment with the operational flow of FIG. 12, and explanations of the same configurations and other parts will be omitted.

First, the patient condition determining unit 140 acquires vibration data (step S202). Then, the waveform calculating unit 130 calculates a waveform (step S204). Next, the patient condition determining unit 140 calculates an index (value) for each condition (steps S206 to S214), and calculates a total value from the calculated index values (step S216).

The total value is calculated for each of the "supine position," "prone position," and "lateral (left and right) positions." The patient state determination unit 140 then outputs the position with the largest value as the patient's position.

Here, the patient condition determination unit 140 calculates an index value for each condition in steps S206 to S214, and the method for calculating the index value will be described with reference to FIG. 13.

(1) Calculation of correlation index in sensor (step S206)
The value of the intra-sensor correlation index is calculated by correlation based on the waveform in the sensor. First, the patient condition determining unit 140 calculates the value of the autocorrelation function based on the waveform in the sensor by the method described in the first embodiment.

The patient condition determination unit 140 then weights the output value of the autocorrelation function and outputs it as the value of the intra-sensor correlation index. Here, the weighting method is explained.

For example, referring to Figure 13, the intra-sensor correlation index is "present" for "supine position" and "side-lying position," but "absent" for "prone position."

The output value of the autocorrelation function is between 0 and 1. Here, in FIG. 13, where there is "Yes" ("Supine" and "Prone"), the output value is used as is as the intra-sensor correlation index value. Also, in FIG. 13, where there is "No", the intra-sensor correlation index value is the maximum value minus the output value of the autocorrelation function.

To give a specific example, if the output value of the autocorrelation function is "0.8", the patient condition determination unit 140 outputs the following values for the intra-sensor correlation index: "supine position = 0.8", "prone position = 0.2", and "lateral position = 0.8".

(2) Calculation of inter-sensor mutual index (step S208)
The inter-sensor correlation index value is calculated based on the correlation between the waveforms of the sensors. First, the patient condition determining section 140 calculates an output value using the cross-correlation function of the two waveform data by the method described above.

The patient condition determination unit 140 then weights the output value and outputs it as the intra-sensor correlation index value. Here, the weighting method is explained.

For example, referring to FIG. 13, the inter-sensor correlation index is "Yes" for "supine position," "Yes/No" for "prone position," and "No" for "lateral position."

The output value of the cross-correlation function is between 0 and 1. Here, in FIG. 13, where it is "present" ("supine position"), the output value is used as is as the intra-sensor correlation index value. Also, in FIG. 13, where it is "absent" ("lateral position"), the intra-sensor correlation index value is the maximum value minus the output value. Also, in FIG. 13, where it is "present/absent" ("prone position"), the output value is halved as the inter-sensor cross-index value.

To give a specific example, if the output value of the cross-correlation function is "0.9", the patient condition determination unit 140 outputs the inter-sensor correlation index values as "supine position = 0.9", "prone position = 0.45", and "lateral position = 0.1".

(3) Calculation of Heart Rate Waveform Index Value (Step S210)
A heartbeat waveform index value indicating the degree to which the heartbeat waveform is present on the output waveform is calculated. For example, when the frequency distribution calculation unit 135 determines that the heartbeat waveform is present strongly when the ratio of the power spectrum density of the heartbeat component frequency and its integer multiple frequency in high frequency components equal to or higher than the respiratory component frequency is equal to or higher than a certain value.

The patient condition determination unit 140 then outputs "1" if the waveform contains a heartbeat waveform, and "0" if it does not. Furthermore, the patient condition determination unit 140 weights the output value and outputs it as the heartbeat waveform index value. Here, the weighting method is explained.

For example, referring to FIG. 13, the heart rate waveform is smaller in the "supine" and "prone" positions. Also, it is larger in the "lateral position." Also, in the "lateral position," it is larger in the "left lateral position," but in the "right lateral position," it is larger than the "supine" and "prone positions," but smaller than the "left lateral position."

Therefore, in the case of "lateral position", the output value is output as is as the heart rate waveform index value. In addition, in the "supine position" and "prone position", a small value (for example, "0.1" times or "0") is output.

(4) Sensor-to-sensor respiration index value (step S212)
The patient condition determining unit 140 calculates an index for the shape of peaks and valleys in the waveform. The portion of the waveform from a valley to a peak is, for example, the portion from time t1 to t2 in FIG. 14, and the portion from a peak to a valley is, for example, the portion from time t2 to t3 in FIG. 14. These portions indicate a transition of vibration (a transition of pressure), which usually corresponds to exhalation/inhalation. The patient condition determining unit 140 determines whether the time from a valley to a peak or the time from a peak to a valley is the same between the sensors, and calculates the proportion of pairs having the same correspondence relationship of the times included in the posture determination section as the value of the inter-sensor respiration index. Referring to FIG. 13, the respiration index is the same between the sensors for the "supine position" and the "prone position".

Therefore, the value of the sensor-to-sensor respiration index is output as is. Also, in the case of the "lateral position," the values may be the same or opposite, so the rate of same pairs described above is multiplied by "0.5" and output as the value of the sensor-to-sensor respiration index.

(5) Calculation of respiration index value (step S214)
The patient condition determining section 140 calculates an index of the result of comparison between the time from a trough to a peak in the waveform and the time from a peak to a trough.

At this time, the patient condition determination unit 140 compares the length of the valley-to-peak time and the peak-to-trough time included in the posture determination section from the waveform, and determines whether or not a portion where the lengths of the two sections have a "short to long" relationship can be detected. The patient condition determination unit 140 then outputs the proportion of pairs with a "short to long" relationship as an output value.

Referring to FIG. 13, in the case of the "supine position," since "short to long" is often possible, the patient condition determination unit 140 subtracts 0.5 from the calculated output value and outputs it as is as the respiratory index value. In addition, since it is difficult to distinguish features in the "prone position" and "lateral position," the patient condition determination unit 140 does not output a respiratory index value (outputs "0").

In this way, it is possible to determine the patient's posture using each index value, making it possible to more appropriately determine the patient's posture.

In this embodiment, the waveform is calculated based on the vibration data, but the patient's posture may be determined simply based on the vibration data. That is, step S204 in FIG. 12 may not be executed, and the process may transition from step S202 to step S206.

For example, the heart rate waveform index can be determined by frequency analysis of the vibration data. Also, in the case of respiratory indexes, the peaks and valleys of the waveform can be interpreted in the same way as the waveform if the maximum value (or its vicinity) and minimum value (or its vicinity) of the vibration data can be extracted.

[3. Third embodiment]
Next, a third embodiment will be described. In this embodiment, a case will be described in which the patient condition determining unit 140 determines the posture of the patient using artificial intelligence (machine learning).

In this embodiment, instead of the process of FIG. 5, the patient's posture, which is one of the patient's conditions, is estimated based on the patient condition estimation unit 700 of FIG. 15.

Here, the operation of the patient condition estimation unit 700 in this embodiment will be described. The patient condition estimation unit 700 estimates the patient's posture by using vibration data and the patient's condition as input values (input data) and utilizing artificial intelligence and various statistical indicators.

As shown in FIG. 15, the patient condition estimation unit 700 includes a feature extraction unit 710, a classification unit 720, a classification dictionary 730, and a patient condition output unit 740.

First, various parameters are input and used as input data to the patient condition estimation unit 700. For example, in this embodiment, vibration data as vibration data and waveform data calculated from the vibration data are used.

Then, the feature extraction unit 710 extracts each feature point and outputs it as a feature vector. Here, examples of features that the feature extraction unit 710 extracts as feature points include the following:

(1) Is there a correlation within the sensor? (2) Is there a correlation between sensors? (3) Is the heart rate waveform included in the waveform data? (4) Is the time from the trough to the peak of the respiratory waveform shorter or longer than the time from the peak to the trough? (5) Is the occurrence rate of two-peak waveforms high or low? (6) Is there a difference in the area above and below the center line in the waveform data? (7) Are there differences in the heart rate waveforms and how they are included between sensors?

The feature extraction unit 710 outputs a feature vector by combining one or more of these feature points. Note that the feature points described above are just examples, and are not limited to these values. Also, as described above, each value is for the convenience of explanation. A "1" may be output for a corresponding feature point, and a "0" for a non-corresponding feature point, or a random variable may be output.

When all of the above feature points are included, the feature space is seven-dimensional, and a seven-dimensional feature vector is output to the identification unit 720.

The identification unit 720 identifies a class corresponding to the patient's condition from the input feature vector. At this time, the class is identified by comparing it with multiple prototypes prepared in advance as the identification dictionary 730. The prototypes may be stored as feature vectors corresponding to each class, or a feature vector representing the class may be stored.

When a feature vector representing a class is stored, the class to which the closest prototype belongs is determined. In this case, the identification unit 720 may determine the class using a nearest neighbor decision rule or may identify the class using the k-nearest neighbor method.

The identification dictionary 730 used by the identification unit 720 may store prototypes in advance, or may be stored using machine learning.

Then, the patient state output unit 740 outputs the (lying) posture as one of the patient states corresponding to the class identified by the identification unit 720. The patient state to be output may be identified as "supine position," "prone position," "lateral position," etc., or a random variable may be output as is.

As a result, according to this embodiment, it is possible to obtain vibration data output from the sensor and infer the patient's posture from this information.

[4. Fourth embodiment]
A fourth embodiment will be described. The fourth embodiment is an embodiment in which the patient condition estimation unit 700 of the third embodiment estimates the posture of a patient based on waveform data by using deep learning that uses a neural network.

In this embodiment, the patient's waveform data is input to the patient condition estimation unit 700. The patient condition estimation unit 700 estimates the patient's condition (posture) from the input waveform data, and recently, deep learning (deep neural network) has been used as a method for this estimation, which has achieved high accuracy, especially in image recognition. This embodiment also uses this method as an example. The deep learning process will be briefly explained using FIG. 16.

First, the patient condition estimation unit 700 inputs the waveform data (image data) signal output by the waveform calculation unit 130 to a neural network composed of multiple layers and the neurons contained in each layer. Each neuron receives signals from multiple other neurons, performs calculations on the signal, and outputs the result to multiple other neurons. When the neural network has a multi-layer structure, the layers are called the input layer, intermediate layer (hidden layer), and output layer in the order in which the signals flow.

Neural networks that have multiple intermediate layers are called deep neural networks (for example, convolutional neural networks that use convolutional operations), and machine learning techniques that use them are called deep learning.

The waveform data undergoes various operations (convolution, pooling, normalization, matrix operations, etc.) in the neurons of each layer of the neural network, changing its shape as it flows, and multiple signals are output from the output layer.

Each of the multiple output values from the neural network is associated with a patient's posture, and the patient's posture associated with the largest output value is inferred. Alternatively, instead of directly outputting the posture, which is the patient's state, one or more output values may be passed through a classifier, and the patient's posture may be inferred from the output of the classifier.

The parameters, which are coefficients used in various calculations of the neural network, are determined in advance by inputting a large amount of waveform data and the patient's posture corresponding to that waveform data into the neural network, propagating the error between the output value and the correct value backwards through the neural network using the error backpropagation method, and repeatedly updating the parameters of the neurons in each layer. This process of updating and determining parameters is called learning.

The structure of neural networks and the individual calculations are publicly known technologies explained in books and papers, so it is sufficient to use any of these technologies.

In this way, by utilizing the patient condition estimation unit 700, the patient's posture is output by referring to the vibration wave data calculated from the vibration data output from the sensor.

In the present embodiment, an example of making an estimation using a neural network as waveform image data has been described, but it is also possible to simply input vibration data (time series voltage output values) and have the system learn to estimate the patient's posture. It is also possible to input data converted into a frequency domain signal using a Fourier transform or discrete cosine transform and have the system learn to estimate the patient's posture.

5. Application Examples
The above-described posture determination device can be incorporated into other devices in the following application examples.

5.1 Bed Apparatus
The configuration of the bed apparatus is shown in Fig. 17. The bed apparatus 10 has a back bottom 12, a waist bottom 14, a knee bottom 16, and a foot bottom 18. The upper body of a user P is supported by the back bottom 12, and the waist is supported by the waist bottom 14.

The drive control unit 1000 controls the driving of the bed device. Here, the drive control unit 1000 operates the bottom to realize the functions of the bottom control unit 1100 for controlling the back-raising, knee-raising (leg-lowering) functions, etc.

To realize the back-raising function, the bottom control unit 1100 is connected to a back bottom drive unit 1110 and a knee bottom drive unit 1120. The back bottom drive unit 1110 is, for example, an actuator, and is connected to a link for back-raising via a link mechanism. Then, under the control of the back bottom drive unit 1110, the back bottom 12 placed by the link operates, and back-raising and back-lowering control is performed.

The knee bottom drive unit 1120 is, for example, an actuator, and is connected to a link for knee lifting via a link mechanism. The knee bottom 16 placed on the link and the foot bottom 18 further connected to it are operated by the control of the knee bottom drive unit 1120, and knee lifting/knee lowering (foot lowering/foot lifting) control is performed.

When the bed device attempts to perform a back-raising operation, the bottom control unit 1100 will not perform the back-raising operation if the patient's position is determined to be "prone position" by the patient state determination unit 140. In other words, even if the user selects the back-raising operation, the bottom control unit 1100 will not drive the back-bottom drive unit 1110, and the back-raising operation will not be performed. Also, even if the bed device is in automatic operation, if the patient's position is determined to be "prone position" by the patient state determination unit 140, the bottom control unit 1100 will not perform the back-raising operation.

The operation in this case will be described with reference to FIG. 18. First, it is determined whether or not an operation has been selected by the user (step S302). For example, the user selects the back-raising button via the input unit 160 (operation remote control). This causes the control unit 100 to determine that the back-raising operation has been selected.

Next, the control unit 100 (patient state determination unit 140) executes a posture determination process (step S304). In the posture determination process, the patient state determination unit 140 executes any one of the posture determination processes described above to determine the posture of the patient on the bed apparatus.

Here, the control unit 100 determines whether the patient is in a specific position (step S306). In this application example, if the patient is in the "prone position", the control unit 100 does not perform the back-raising operation (step S306; Yes). If the patient is in any other position, the control unit 100 issues an instruction to the bottom control unit 1100 (back bottom drive unit 1110) to perform the back-raising operation (step S306; Yes → step S308).

Then, when the user cancels the back-raising operation (for example, the user/patient cancels the operation or the back-raising button is released), the control unit 100 stops the back-raising operation (step S310; Yes → step S312).

The control unit 100 also stops the back-raising operation if the patient assumes a specific position during the back-raising operation (for example, if the patient assumes a prone position during the back-raising operation) (step S306; Yes → step S312).

[5.2 Posture Change]
In order to grasp the risk of bedsores, the frequency of position changes (posture changes) and the proportion of each posture are automatically stored. In other words, the posture of the patient determined by the patient condition determination unit 140 is automatically stored and utilized for care and treatment.

In addition, if the posture determined by the patient condition determination unit 140 remains the same for a predetermined time, an alert is issued or a posture change is automatically performed. For example, as shown in FIG. 17, the drive control unit 1000 controls the posture change drive unit 1200. The posture change drive unit 1200 changes the posture of the patient, for example, by expanding and contracting air cells provided on the left and right sides of the patient or by raising and lowering the left and right bottoms.

The drive control unit 1000 controls the position change drive unit 1200 according to the posture determined by the patient condition determination unit 140, and performs control to change the posture of the patient P.

For example, the process in FIG. 18 will be described. When a position change operation is selected (step S302; Yes), the patient condition determination unit 140 determines the patient's posture using one of the methods described above (step S304).

Here, it is determined whether the determined posture is a specific posture. For example, if the patient is in a right lateral position, the posture change drive unit 1200 changes the posture by controlling the air cell provided on the right side. If the patient is in a left lateral position, the posture change drive unit 1200 changes the posture by controlling the air cell provided on the left side. If the patient is in a prone position, no posture change is performed.

In addition, if the posture continues for a predetermined period of time, a change in posture may be performed. For example, if the posture determined by the posture stabilization process continues for 10 minutes or more, a process of changing the posture may be performed.

[5.3 Notification device]
A configuration is provided for notifying the patient according to the posture of the patient determined by the patient condition determining unit 140. The notification method may be a voice output device or a display device, or may be a method of notifying another terminal (for example, a mobile terminal device carried by a medical staff member).

As for the timing of the notification, for example, if the patient is paralyzed, the risk of bedsores increases if the paralyzed side is placed on the bottom. Therefore, a notification is issued when the posture determined by the patient condition determination unit 140 is a lateral position in which the paralyzed side is placed on the bottom.

In addition, to prevent infants from suffocating to death when sleeping on their stomachs, an alarm is issued if the patient state determination unit 140 determines that the patient is in the prone position.

[6. Modifications]
Although an embodiment of the present invention has been described in detail above with reference to the drawings, the specific configuration is not limited to this embodiment, and designs that do not deviate from the gist of the present invention are also included in the scope of the claims.

In addition, in this embodiment, the patient's posture is determined in the processing device 5 based on the results output by the detection device 3, but it is also possible to perform all the determinations in one device. Also, instead of implementing this by installing an application on a terminal device (e.g., a smartphone, tablet, or computer), for example, the processing may be performed on the server side and the processing results may be returned to the terminal device.

For example, the detection device 3 may upload vibration data to a server, thereby realizing the above-mentioned processing on the server side. The detection device 3 may be realized by a device such as a smartphone with a built-in acceleration sensor and vibration sensor.

In the above embodiment, two vibration sensors are described, but more than two may be provided. Also, the method of calculating the frequency distribution of the first embodiment to determine posture can be realized with only one sensor.

In addition, in the embodiments, the programs that run on each device are programs that control the CPU and the like (programs that make the computer function) so as to realize the functions of the above-described embodiments. Information handled by these devices is temporarily stored in a temporary storage device (e.g., RAM) during processing, and is then stored in various ROM, HDD, and SSD storage devices, and is read, modified, and written by the CPU as necessary.

When distributing the program on the market, the program can be stored on a portable recording medium and distributed, or transferred to a server computer connected via a network such as the Internet. In this case, the storage device of the server computer is of course included in the present invention.

1 System 3 Detection device 5 Processing device 10 Bed device 12 Back bottom 14 Waist bottom 16 Knee bottom 18 Foot bottom 20 Mattress 100 Control unit 110 Vibration detection unit 120 Biometric information calculation unit 130 Waveform calculation unit 135 Frequency distribution calculation unit 140 Patient condition determination unit 150 Memory unit 152 Vibration data storage area 154 Waveform data storage area 160 Input unit 170 Output unit 700 Patient condition estimation unit 710 Feature extraction unit 720 Identification unit 730 Identification dictionary 740 Patient condition output unit

Claims (6)

At least two vibration sensors are installed at a plurality of positions on the left and right of the mattress at a predetermined distance apart ;
a control unit that detects vibrations of the subject when the subject is lying down using the vibration sensor and determines the posture of the subject from a waveform of the detected vibrations;
a drive unit that does not operate a button of an input unit even if the button is selected when the posture of the subject is in a specific posture, and that operates the button when the button of the input unit is selected when the posture of the subject is not in a specific posture;
Equipped with
The control unit determines that the subject's posture is in a lateral position if there is no correlation between the two vibration sensors. At least two vibration sensors are installed at a plurality of positions on the left and right of the mattress at a predetermined distance apart ;
a control unit that detects vibrations of the subject when the subject is lying down using the vibration sensor and determines the posture of the subject from a waveform of the detected vibrations;
a drive unit that does not operate a button of an input unit even if the button is selected when the posture of the subject is in a specific posture, and that operates the button when the button of the input unit is selected when the posture of the subject is not in a specific posture;
Equipped with
The control unit determines that the subject's posture is supine or prone if there is correlation between the two vibration sensors. The system of claim 2 , wherein the control unit determines that the subject's posture is prone if there is correlation between the two vibration sensors and no correlation within the vibration sensor. The system of claim 2 , wherein the control unit determines that the subject's posture is supine if there is correlation between the two vibration sensors and correlation within the vibration sensor. The system according to claim 1 , wherein the specific position is a prone position and the button is a back-raising button. The system according to any one of claims 1 to 5, wherein the at least two vibration sensors are installed at intervals of 15 to 60 cm in the left-right direction.

Images