Support surface movement control device

Abstract

This support surface movement control device (100) comprises a sleep stage determination unit (34), a sleep determination unit (35), and a movement control unit (36). The sleep stage determination unit (36) determines the sleep stage of the user on the basis of the classification results of a first classifier (CF11), a second classifier (CF13), and a third classifier (CF14). The sleep determination unit (35) determines whether the user is in a sleep state on the basis of the classification result of a fourth classifier (CF2). The first classifier and the fourth classifier are constructed such that the probability of erroneously classifying the state of the user as a sleep state when the user is in a wakeful state is lower in the fourth classifier than in the first classifier. The movement control unit starts the downward movement of an upper body support part (521) on the basis of the determination result of the sleep stage determination unit and the determination result of the sleep determination unit.

Description

Support surface movement control device

[0001] This disclosure relates to a support surface movement control device.

[0002] Various devices such as beds, sofas, chairs, and wheelchairs are used as support devices for the human body. Furthermore, one type of support device known is one in which the support surface for the human body is movable. For example, in reclining beds, reclining sofas, and reclining chairs, the portion of the support surface that supports the human upper body is pivotable relative to the horizontal plane.

[0003] Patent Document 1 describes a bed device that includes a back-lowering control unit that controls a drive unit to lower the back bottom from a first angle to a second angle that is approximately horizontal and facilitates maintaining the user's sleep state, when the sleep onset determination unit determines that the user has entered a sleep state.

[0004] Patent No. 5665411

[0005] Conventional devices such as the bed device described in Patent Document 1 do not have sufficient accuracy in determining whether or not the user has fallen asleep, resulting in the inconvenience of the bed base being lowered even when the user is awake. Conversely, the inconvenience of the bed base not being lowered even when the user is asleep also occurs.

[0006] The object of this disclosure is to provide a support surface movement control device that can suitably move the support surface that supports the user.

[0007] According to a first aspect of this disclosure, a support surface movement control device moves a support surface that supports the body of a user, comprising: a sleep stage determination unit that determines the user's sleep stage based on the user's heart rate interval on the support surface; a sleep determination unit that determines whether or not the user is in a sleep state based on the user's heart rate interval on the support surface; and a movement control unit that performs support surface movement control to control the movement of the support surface, wherein the support surface has an upper body support unit that supports the upper body of the user on the support surface, the sleep stage determination unit determines the user's sleep stage based on the classification results of a first classifier that classifies the user's state into an awakened state and a sleep state using a feature quantity based on the heart rate interval as an explanatory variable, and a second and third classifier that classifies the user's state into first non-REM sleep, second non-REM sleep, and third non-REM sleep using a feature quantity based on the heart rate interval as an explanatory variable, and the sleep determination unit determines whether or not the user is in a sleep state based on the classification results of a fourth classifier that classifies the user's state into an awakened state and a sleep state using a feature quantity based on the heart rate interval as an explanatory variable. The first and fourth classifiers are constructed such that the probability of misclassifying the user's state as sleep when the user is awake is smaller in the fourth classifier than in the first classifier. The movement control unit is provided with a support surface movement control device that, in the support surface movement control, initiates a downward movement that reduces the inclination angle of the upper body support based on the determination result of the sleep stage determination unit and the determination result of the sleep determination unit.

[0008] A second aspect of the present disclosure relates to a support surface movement control device that moves a support surface supporting the body of a user, comprising: a state determination unit including at least one of a sleep stage determination unit that determines the user's sleep stage based on the user's heart rate interval on the support surface, a sleep determination unit that determines whether or not the user is in a sleep state based on the user's heart rate interval on the support surface, and a body movement determination unit that determines whether or not body movement has occurred in the user based on fluctuations in the load applied by the user to the support surface; and a movement control unit that performs support surface movement control to control the movement of the support surface, wherein the support surface has an upper body support portion that supports the upper body of the user on the support surface, and the movement control unit performs a first descent control to start a descent movement that reduces the inclination angle of the upper body support portion based on the determination result of the state determination unit satisfying a first condition, and if the descent movement of the upper body support portion has not started even after the elapsed time from the start of the support surface movement control exceeds a first elapsed time, the movement control unit performs a second descent control to start the descent movement of the upper body support portion based on the determination result of the state determination unit satisfying a second condition. A support surface movement control device is provided, wherein the first condition and the second condition are, respectively, one and the other of the determination result of the sleep stage determination unit and the conditions relating to the determination result of the sleep determination unit and the conditions relating to the determination result of the body movement determination unit, or, either the first condition or the second condition is the same as the determination result of the sleep stage determination unit and the conditions relating to the determination result of the sleep determination unit and the conditions relating to the determination result of the body movement determination unit, and the second condition is a relaxed version of the first condition.

[0009] A third aspect of the present disclosure is provided, a support surface movement control device for moving a support surface that supports the body of a user, comprising: a body movement determination unit that determines whether or not body movement has occurred in the user based on fluctuations in the load applied by the user to the support surface; and a movement control unit that performs support surface movement control to control the movement of the support surface, wherein the support surface has an upper body support portion that supports the upper body of the user on the support surface, and the movement control unit, in the support surface movement control, starts the downward movement of the upper body support portion based on the determination by the body movement determination unit that no body movement has occurred in the user for a predetermined period of time.

[0010] The support surface movement control device of this disclosure allows for the appropriate movement of the support surface that supports the user.

[0011] Figures 1(a) and 1(b) are side views of a bed to which the bed base movement device is used. In Figure 1(a), the head bed base and the foot bed base are in a horizontal position. In Figure 1(b), the head bed base is in the back-raised position and the foot bed base is in a horizontal position. Figure 2 is a plan view showing the arrangement of load detectors on the bed. Figure 3 is a block diagram showing the configuration of a bed base movement device in one embodiment. Figure 4 is a flowchart of the state determination process. Figure 5 is a graph showing an example of a load signal. The four graphs on the left side of Figure 5 show signals obtained by filtering each of the four load signals based on the output values ​​of the four load detectors. The four graphs on the right side of Figure 5 each show signals obtained by normalizing the four graphs on the left side of Figure 5. Figure 6 shows the waveform of one cycle of the BCG signal. Figure 7(a) is a graph showing an example of the waveform of the load signal before normalization when no body movement occurs from the user during a certain period, and an example of the waveform of the load signal after normalization. Figure 7(b) is a graph showing an example of the waveform of the load signal before normalization when the user experiences body movement during a certain period, and an example of the waveform of the load signal after normalization. Figure 8(a) is an explanatory diagram for explaining the process of extracting multiple BCG signals from the information acquisition signal. Figure 8(b) is a graph showing the waveform of a template signal created by calculating the average of multiple BCG signals extracted from the information acquisition signal. Figure 9 is a flowchart showing the specific procedure of the heart rate information acquisition process. Figure 10 is a conceptual diagram showing the configuration of a sequential two-class classifier that classifies the user's sleep state into one of five sleep stages. Figure 11 is a conceptual diagram showing the configuration of a two-class classifier that classifies the user's sleep state into an awakened state and a sleep state. Figure 12 is a flowchart showing an example of the specific procedure of the bed board movement control process. Figures 13A and 13B are tables summarizing the major categories of feature quantities that are candidates for explanatory variables of a two-class classifier that classifies the user's sleep state. Figures 14A to 14G are tables listing 279 types of features that are candidates for explanatory variables of a two-class classifier used to classify a user's sleep state. Figures 15A and 15B are tables listing the top 40 most effective features for each of the multiple two-class classifiers in a sequential two-class classifier.FIG. 16 is a graph showing the relationship between the number of feature values used as explanatory variables for a two-class classifier and the performance of the two-class classifier. FIG. 17 is a graph showing the top 12 feature values that are highly effective in common among a plurality of two-class classifiers included in a sequential two-class classifier, and the importance levels of the 12 feature values.

[0012] <Embodiment> A floor plate moving apparatus 100 (FIG. 3) according to an embodiment of the present disclosure will be described with reference to FIGS. 1 to 12, taking as an example a case where the floor plate moving apparatus 100 is used for a bed 500 (FIG. 1, which is an example of a "support device").

[0013] [Bed 500] As shown in FIGS. 1 and 2, the bed 500 to which the floor plate moving apparatus 100 is applied includes a base portion 510, a floor plate 520 supported by the base portion 510 (which is an example of a "support surface"), and a moving mechanism 530 that moves the floor plate 520. In the following description, the long side direction (the direction of the Y-axis in FIG. 2) and the short side direction (the direction of the X-axis in FIG. 2) of the bed 500 and the floor plate 520 are respectively referred to as the longitudinal direction and the width direction of the bed 500 and the floor plate 520.

[0014] The base portion 510 includes a frame 511 having a rectangular shape in plan view, and four legs 512 provided at four corners of the frame 511.

[0015] Along the longitudinal direction of the bed 500, the floor plate 520 includes a head-side floor plate 521 on the head side of the bed 500 (the positive side in the Y-axis direction in FIG. 2, which is an example of an "upper body support portion"), and a leg-side floor plate 522 on the leg side of the bed 500 (the negative side in the Y-axis direction in FIG. 2).

[0016] The head-side floor plate 521 is positioned near the leg-side floor plate 522 and is pivotable about an axis AX1 extending in the width direction of the bed 500. The leg-side floor plate 522 is fixed to the base portion 510.

[0017] The moving mechanism 530 is a mechanism for moving the head-side floor plate 521, and is an electric cylinder in the present embodiment. However, the moving mechanism 530 may be any arbitrary actuator.

[0018] By the operation of the moving mechanism 530, the head-side floor plate 521 pivots about an axis AX1, and is displaced between a horizontal position where the top surface of the head-side floor plate 521 coincides with a horizontal plane (FIG. 1(a)) and a back-raising position where the top surface of the head-side floor plate 521 is inclined with respect to the horizontal plane (FIG. 1(b)). In the horizontal position, the top surface of the head-side floor plate 521 and the top surface of the leg-side floor plate 522 are flush with each other. An inclination angle of the top surface of the head-side floor plate 521 with respect to the top surface of the leg-side floor plate 522 (i.e., the horizontal plane) is defined as an angle θ (FIG. 1(b)).

[0019] [Configuration of Floor Plate Moving Device 100] As shown in FIG. 3, the floor plate moving device 100 of the present embodiment mainly includes a load detection unit 10, a control unit 30, and a storage unit 40. The load detection unit 10 and the control unit 30 are connected via an A / D conversion unit 20. A display unit 50, a notification unit 60, and an input unit 70 are further connected to the control unit 30. A moving mechanism 530 of the bed 500 is connected to the control unit 30.

[0020] The load detection unit 10 includes four load detectors 11, 12, 13, and 14. Each of the load detectors 11, 12, 13, and 14 is a load detector that detects a load using, for example, a beam-type load cell. The load detectors 11, 12, 13, and 14 are each connected to the A / D conversion unit 20 via wiring or wireless communication.

[0021] As shown in FIG. 2, in the present embodiment, the four load detectors 11 to 14 of the load detection unit 10 are respectively arranged below casters C1, C2, C3, and C4 attached to lower ends of four legs 512 of the bed 500. The arrangement of the load detectors 11 to 14 is not limited thereto. The load detectors 11 to 14 can be arranged in the vicinity of the four legs 512 or the four casters C1 to C4 in any manner that allows detection of a load applied to the bed 500. Specifically, for example, any one of the four load detectors 11 to 14 may be incorporated into each of the four legs 512.

[0022] The A / D conversion unit 20 includes an A / D converter that converts an analog signal from the load detection unit 10 into a digital signal, and is connected to the load detection unit 10 and the control unit 30 respectively via wiring or wireless communication.

[0023] The control unit 30 is a dedicated or general-purpose computer and includes a load signal acquisition unit 31, a body movement determination unit 32, a heart rate information acquisition unit 33, a sleep stage determination unit 34, a sleep determination unit 35, and a floor plate movement control unit 36.

[0024] The storage unit 40 is a storage device that stores data used in the floor plate moving device 100, and can use a hard disk (magnetic disk), for example. The display unit 50 is a monitor such as a liquid crystal monitor that displays information output from the control unit 30. The notification unit 60 is a device that provides predetermined notifications audibly based on information from the control unit 30, for example, a speaker. The input unit 70 is an interface for providing predetermined input to the control unit 30, and may be an operation terminal with operation buttons, for example. A general-purpose terminal such as a smartphone can also be used as the input unit 70.

[0025] [Operation of the bed base moving device 100] The bed base moving device 100 constantly performs a state determination process that includes determining whether or not the user U on the bed 500 is moving, determining the sleep stage, and determining sleep-wake status. Then, it moves the bed base 520 by control based on the results of each determination included in the state determination process.

[0026] [State Determination Process] The state determination process performed by the floor plate moving device 100 mainly includes the load signal acquisition process S101, the body movement determination process S102, the heart rate information acquisition process S103, the sleep stage determination process S104, and the sleep determination process S105, as shown in the flowchart of Figure 4.

[0027] [Load signal acquisition process S101] In the load signal acquisition process S101, the load signal acquisition unit 31 acquires a plurality of load signals via the load detection unit 10, each containing a component that fluctuates according to the heart rate of the user U.

[0028] The load signal acquisition unit 31 detects the load of the user U on the bed 500 using load detectors 11, 12, 13, and 14. The load of the user U on the bed 500 is distributed and applied to the load detectors 11 to 14 located under the legs 512 at the four corners of the bed 500, and is detected in a distributed manner by these detectors.

[0029] Each of the load detectors 11 to 14 detects a load (load change) and outputs it as an analog signal to the A / D conversion unit 20. The A / D conversion unit 20 converts the analog signal into a digital signal with a sampling period of, for example, 5 milliseconds (0.005 seconds), and outputs the digital signal (hereinafter referred to as "load signal") to the load signal acquisition unit 31. Hereinafter, the load signals obtained by digitally converting the analog signals output from load detectors 11, 12, 13, and 14 in the A / D conversion unit 20 will be referred to as load signals S1, S2, S3, and S4, respectively.

[0030] The load signal acquisition unit 31 applies a bandpass filter to each of the load signals S1 to S4 and acquires load signals LS1 to LS4 (Figure 5) from which noise has been removed from the load signals S1 to S4.

[0031] Human respiration is typically 12 to 20 breaths per minute (approximately 0.2 Hz to 0.33 Hz) at rest, and human heart rate is typically 60 to 100 beats per minute (approximately 1 Hz to 1.7 Hz) at rest. Therefore, for example, by removing low-frequency components of the load signal (for example, components below 0.5 Hz) using a bandpass filter, it is possible to reduce components in the load signal that are different from the component that shows temporal variation corresponding to the user U's heart rate (for example, components that show temporal variation corresponding to the user U's body movement, components that show temporal variation corresponding to the user U's respiration, and various types of noise such as noise caused by the surrounding environment).

[0032] In this embodiment, the load signal acquisition unit 31 applies a bandpass filter to each of the load signals S1 to S4 acquired from the A / D conversion unit 20, allowing only components in the frequency band of 1 Hz to 8.5 Hz to pass through. This results in the acquisition of load signals LS1 to LS4. The passband of the bandpass filter can be arbitrarily set to remove frequency components that are not involved in the acquisition of heart rate information.

[0033] Next, the load signal acquisition unit 31 performs normalization processing on each of the load signals LS1 to LS4. The load signal acquisition unit 31 can normalize the load signals LS1 to LS4 using any known method. Specifically, for example, normalization is performed by adjusting the amplitude of each load signal LS1 to LS4 so that the average value of the amplitude during period PR is 0 and the standard deviation of the signal during period PR is 1. By performing normalization processing on the load signals LS1 to LS4, normalized signals NS1 to NS4 are obtained. The length of period PR is 30 seconds in this embodiment, but is not limited to this.

[0034] As shown in the graphs on the left side of Figure 5, the amplitudes of the load signals LS1 to LS4 may differ from each other. This is because the amplitude changes depending on several factors, such as the position and posture of the user U on the bed 500, the method of setting up the load detectors, and the magnitude of the passband frequency components of the bandpass filter corresponding to the user U's breathing. The graphs on the left side of Figure 5 show a situation where the user U is located in the positive region in the X direction (width direction) of the bed 500, and the amplitudes of the load signals LS1 and LS4 based on load detectors 11 and 14 are larger than the amplitudes of the load signals LS2 and LS3 based on load detectors 12 and 13.

[0035] Normalization eliminates the amplitude variations of the load signals LS1 to LS4. As shown in the graphs on the right side of Figure 5, the amplitudes of the normalized signals NS1 to NS4 become similar in magnitude.

[0036] Each of the load signals LS1 to LS4 and normalized signals NS1 to NS4 acquired in the load signal acquisition process S101 exhibits a BCG (BallistoCardioGram) waveform as shown in Figure 6. The BCG waveform is a waveform that arises in response to slight body movements corresponding to the heartbeat, and for each heartbeat, the J peak PK, which has the largest amplitude, is generated. Ｊ And, J Peak PK Ｊ Occurs before and J-peak PK Ｊ H-peak PK has a smaller amplitude than Ｈ and I-peak PK Ｉ And, J Peak PK Ｊ Occurring later than and J-peak PK ＪK peak PK with a smaller amplitude than Ｋ and L peak PK Ｌ are shown.

[0037] [Body Motion Determination Step S102] In the body motion determination step S102, the body motion determination unit 32 determines whether or not body motion has occurred in the user U during the period PR every time the period PR elapses. The principle of body motion determination executed by the body motion determination unit 32 is as follows.

[0038] When no body motion occurs in the user U within the period PR, as shown in the graph on the left side of FIG. 7(a), the amplitude of the load signal LS (which is any one of the load signals LS1 to LS4) is substantially the same over the entire period within the period PR. In this case, as shown in the graph on the right side of FIG. 7(a), the amplitude of the normalized signal NS (which is any one of the normalized signals NS1 to NS4) is also substantially the same over the entire period of the period PR.

[0039] On the other hand, when body motion occurs in the user U within the period PR, as shown in the graph on the left side of FIG. 7(b), the amplitude of the load signal LS is greater in the body motion occurrence period PR ＭＶ than in other periods. In this case, since the normalization of the load signal LS is affected by the amplitude in the body motion occurrence period PR ＭＶ , as shown in the graph on the right side of FIG. 7(b), the amplitude of the normalized signal NS becomes smaller in a period different from the body motion occurrence period PR ＭＶ .

[0040] As described above, the amplitude of the normalized signal NS changes depending on the presence or absence of body motion of the user U. Therefore, when the standard deviation is calculated for each period obtained by dividing the period PR into a plurality of periods (for example, the periods PR1, PR2, and PR3 in FIG. 7), if no body motion occurs in the user U within the period PR, the standard deviations of the normalized signal NS in each period are substantially the same as each other (for example, approximately 1 each). On the other hand, when body motion occurs in the user U within the period PR, the standard deviation of the normalized signal NS in a period that does not include the body motion occurrence period PR ＭＶ (for example, the period PR1 in FIG. 7) is relatively small (for example, less than 1).

[0041] Therefore, based on the magnitude of the standard deviation of each period obtained by dividing the normalized signal NS, it is possible to determine whether or not body movement occurred in the user U.

[0042] Based on the above principle, the motion determination unit 32 first calculates Q standard deviations for each of the normalized signals NS1 to NS4 of period PR acquired by the load signal acquisition unit 31, divided into Q periods at regular intervals. This results in 4Q standard deviations σ ＮＳ１（１） ~σ ＮＳ１（Ｑ） , σ ＮＳ２（１） ~σ ＮＳ２（Ｑ） , σ ＮＳ３（１） ~σ ＮＳ３（Ｑ） , σ ＮＳ４（１） ~σ ＮＳ４（Ｑ） The following is calculated. Then, the standard deviation σ of the four calculated Qs is calculated. ＮＳ１（１） ~σ ＮＳ１（Ｑ） , σ ＮＳ２（１） ~σ ＮＳ２（Ｑ） , σ ＮＳ３（１） ~σ ＮＳ３（Ｑ） , σ ＮＳ４（１） ~σ ＮＳ４（Ｑ） The minimum value among them σ ＭＩＮ threshold TH σ Compare it to this.

[0043] The body movement determination unit 32 determines the minimum value σ ＭＩＮ is threshold TH σ If it is smaller than σ, it is determined that there was body movement during period PR, and the minimum value σ ＭＩＮ is threshold TH σ If the above conditions are met, it will be determined that there was no physical movement during the PR period.

[0044] The body movement determination unit 32 determines whether or not body movement is occurring in the user U for each of the sequentially occurring periods PR. The body movement determination unit 32 sequentially stores the determination results for each of the periods PR in the storage unit 40.

[0045] [Heart rate information acquisition process S103] In the heart rate information acquisition process S103, the heart rate information acquisition unit 33 acquires heart rate information based on the normalization signals NS1 to NS4.

[0046] Specifically, the heart rate information acquisition unit 33 selects one of the normalization signals NS1 to NS4 as the information acquisition signal IS (Figure 8(a)) and acquires the heart rate interval based on the information acquisition signal IS. Specifically, in selecting the information acquisition signal IS, the heart rate information acquisition unit 33 selects the normalization signals NS1 to NS4 with the highest kurtosis during the period in which the body movement determination unit 32 determined that no body movement occurred in the user U. In addition, the heart rate information acquisition unit 33 may select any one of the normalization signals NS1 to NS4 as the information acquisition signal IS.

[0047] A specific example of the process for acquiring heart rate intervals based on the information acquisition signal IS is shown in the flowchart of Figure 9, and includes a BCG signal extraction step S31, a template creation step S32, and a heart rate interval acquisition step S33.

[0048] In the BCG signal extraction process S31, the heart rate information acquisition unit 33 extracts a BCG signal BS, which represents the BCG waveform, from the information acquisition signal IS for period PR.

[0049] The heart rate information acquisition unit 33 first performs peak detection and identifies the J peak PK of the BCG waveform included in the information acquisition signal IS. Ｊ Estimated J-peak EPK that is estimated to show Ｊ The heart rate information acquisition unit 33 detects the estimated J-peak EPK by performing the following (1) to (3). Ｊ It detects.

[0050] (1) A peak detection process is performed on the information acquisition signal IS for the period PR, and all peaks included in the information acquisition signal IS for the period PR are detected.

[0051] (2) Calculate the average value of the amplitude of all peaks detected in (1), and set a threshold value that is twice the average value.

[0052] (3) The peak detection process is performed again on the information acquisition signal IS for the period PR, and the peak with an amplitude larger than the threshold set in (2) is estimated as the J peak EPK. Ｊ It is detected as such.

[0053] Next, the heart rate information acquisition unit 33 detects the estimated J-peak EPK ＪA predetermined period centered around (for example, EPK) Ｊ The signal for the average heart rate interval (0.7 seconds to 2.5 seconds), calculated from the median of the intervals, is extracted as the BCG signal BS. The heart rate information acquisition unit 33 then extracts one estimated J-peak EPK from the detected data. Ｊ One BCG signal BS is extracted for each. Therefore, the heart rate information acquisition unit 33 extracts the estimated J peak EPK detected within the period PR. Ｊ Multiple BCG signals BS are extracted according to the number.

[0054] In the template creation process S32, the heart rate information acquisition unit 33 creates a template signal TS (Figure 8(b)) based on the multiple BCG signals BS extracted in the BCG signal extraction process S31. Specifically, the heart rate information acquisition unit 33 uses the multiple BCG signals BS extracted in the BCG signal extraction process S31 to calculate the average value of the amplitudes of the multiple BCG signals BS for each time point, and sets the signal represented by the average value of the amplitudes of the multiple BCG signals BS at each time point as the template signal TS. The waveform shape represented by the template signal TS is the average shape of the multiple waveforms represented by the multiple BCG signals BS extracted in the BCG signal extraction process S31.

[0055] In the heart rate interval acquisition process S33, the heart rate information acquisition unit 33 performs template matching between the template signal TS created in the template creation process S32 and the information acquisition signal IS for the period PR. Specifically, for example, it calculates the cross-correlation function between the template signal TS and the information acquisition signal IS. The calculated cross-correlation function shows a peak at the timing when the degree of agreement between the template signal TS and the information acquisition signal IS is highest. Therefore, based on the peak of the calculated cross-correlation function, the heart rate information acquisition unit 33 selects the J peak PK included in the information acquisition signal IS. Ｊ Identify the J-peak PK. Ｊ The heart rate interval (HBI) is obtained based on this. The heart rate interval (HBI) obtained here is a JJ interval (J-J Interval: JJI).

[0056] The peak interval of the calculated cross-correlation function is the J peak PK of the IS signal used for information acquisition. ＪThis becomes equal to the interval. Therefore, the heart rate information acquisition unit 33 may acquire the peak interval of the calculated cross-correlation function as the user U's heart rate interval HBI. The heart rate information acquisition unit 33 may calculate the average heart rate interval based on the acquired multiple heart rate interval HBIs.

[0057] [Sleep Stage Determination Process S104] In the sleep stage determination process S104, the sleep stage determination unit 34 determines the sleep stage of user U based on the heart rate interval HBI acquired in the heart rate information acquisition process S103.

[0058] In this embodiment, the sleep stages are five stages based on the standards of The American Academy of Sleep Medicine (AASM), and include wakefulness (W), REM sleep, first non-REM sleep (N1), second non-REM sleep (N2), and third non-REM sleep (N3).

[0059] In this specification, when user U's sleep stage is wakefulness W, user U is considered to be in a waking state. Also, when user U's sleep stage is any of REM sleep, first non-REM sleep N1, second non-REM sleep N2, or third non-REM sleep N3, user U is considered to be in a sleep state. When user U's sleep stage is REM sleep, user U is considered to be in a REM sleep state. When user U's sleep stage is any of first non-REM sleep N1, second non-REM sleep N2, or third non-REM sleep N3, user U is considered to be in a non-REM sleep state. When user U's sleep stage is first non-REM sleep N1 or second non-REM sleep N2, user U is considered to be in a light non-REM sleep state, and when user U's sleep stage is third non-REM sleep N3, user U is considered to be in a deep non-REM sleep state. The depth of user U's sleep increases in the order of first non-REM sleep N1, second non-REM sleep N2, and third non-REM sleep N3.

[0060] The sleep stage determination unit 34 determines the sleep stage of user U using the sequential two-class classifier CF1 shown in Figure 10. The sequential two-class classifier CF1 includes a two-class classifier CF11 that classifies user U's state into either "awake state" or "sleep state", a two-class classifier CF12 that classifies user U's state into either "REM sleep state" or "non-REM sleep state", a two-class classifier CF13 that classifies user U's state into either "third non-REM sleep N3" or "first non-REM sleep N1 or second non-REM sleep N2", and a two-class classifier CF14 that classifies user U's state into either "first non-REM sleep N1" or "second non-REM sleep N2".

[0061] In this embodiment, the explanatory variables of the sequential two-class classifier CF1 for determining sleep stages over a certain period PR (hereinafter referred to as the "target period") are the following features I to VIII.

[0062] (I) NN20: The number of adjacent heart rate interval HBIs (JJI in this embodiment) with a difference of 20 [ms] or more during the target period. (II) RRI range minus previous median: The difference between the maximum and minimum heart rate interval HBIs (JJI in this embodiment) during the target period, minus the median heart rate interval HBI in the past period PR closest to the target period for which at least one heart rate interval could be obtained.

[0063] (III) pNN20: The percentage of adjacent heart rate intervals (HBI) with a difference of 20 ms or more during the study period.

[0064] (IV) Median heart rate interval: The median heart rate interval (HBI) during the study period. (V) Log max power of HF (90s): The logarithm of the maximum power of HF during a 90-second period including three consecutive period PRs centered on the study period. Specifically, the logarithm of the maximum power spectral density (PSD) of the high-frequency component HF (0.15–0.4 Hz) of heart rate variability during three consecutive period PRs. (VI) Log normalized LF ＊(90s): Normalized, frequency-shifted low-frequency component (LF) of heart rate variability over a 90-second period including three consecutive periods (PR) centered around the target period. ＊ The logarithm of . LF ＊ The frequency domain is defined as a 0.11 Hz range centered on the point of greatest power within the 0.04 to 0.15 Hz range. For example, if the power is greatest at 0.12 Hz, the LF is the sum of the PSD values ​​within the range of 0.065 to 0.175 Hz. ＊ This is the result. Normalization is then performed by dividing this by (sum of PSD across all frequency bands) - (VLF (0.003 to 0.04 Hz)), and the natural logarithm of this value is obtained as Log normalized LF. ＊ This is the result. (VII) Frequency of max modulus LF pole (90s): The frequency of the pole with the maximum modulus within the LF frequency range (0.04 to 0.15 Hz) in a 90-second period including three consecutive periods PR centered on the target period, calculated using an AR (autoregressive) model. (VIII) Log LF (90s): The logarithm of the low-frequency component LF of heart rate variability in a 90-second period including three consecutive periods PR centered on the target period.

[0065] The sleep stage determination unit 34 calculates feature quantities I to VIII based on the heart rate interval HBI acquired by the heart rate information acquisition unit 33 and sequentially inputs them to the two-class classifier CF1. As a result, the two-class classifier CF11 first classifies the state of user U into either "awake state" or "sleep state". If the classification result of the two-class classifier CF11 is "awake state", the sleep stage determination unit 34 determines that user U's sleep stage is wakeful W.

[0066] When the classification result of the two-class classifier CF11 is "sleep state", the sequential two-class classifier CF1 inputs features I to VIII to the two-class classifier CF12. As a result, the two-class classifier CF12 classifies the state of user U as either "REM sleep state" or "non-REM sleep state". When the classification result of the two-class classifier CF12 is "REM sleep state", the sleep stage determination unit 34 determines that user U's sleep stage is REM sleep.

[0067] When the classification result of the two-class classifier CF12 is "non-REM sleep state", the sequential two-class classifier CF1 inputs features I to VIII to the two-class classifier CF13. As a result, the two-class classifier CF13 classifies the state of user U into either "third non-REM sleep N3" or "first non-REM sleep N1 or second non-REM sleep N2". When the classification result of the two-class classifier CF13 is "third non-REM sleep N3", the sleep stage determination unit 34 determines that user U's sleep stage is third non-REM sleep N3.

[0068] The sequential two-class classifier CF1 inputs features I to VIII to the two-class classifier CF14 when the classification result of the two-class classifier CF13 is "first non-REM sleep N1 or second non-REM sleep N2". As a result, the two-class classifier CF14 classifies the state of user U into either "first non-REM sleep N1" or "second non-REM sleep N2". When the classification result of the two-class classifier CF14 is "first non-REM sleep N1", the sleep stage determination unit 34 determines that user U's sleep stage is first non-REM sleep N1. When the classification result of the two-class classifier CF14 is "second non-REM sleep N2", the sleep stage determination unit 34 determines that user U's sleep stage is second non-REM sleep N2.

[0069] In this embodiment, each of the two-class classifiers CF11 to CF14 is a Support Vector Machine (SVM). Two-class classifier CF11 can be constructed, for example, by supervised learning using training data with features I to VIII as explanatory variables and "awake state" and "sleep state" as the target variables. Two-class classifier CF12 can be constructed, for example, by supervised learning using training data with features I to VIII as explanatory variables and "REM sleep state" and "non-REM sleep state" as the target variables. Two-class classifier CF13 can be constructed, for example, by supervised learning using training data with features I to VIII as explanatory variables and "third non-REM sleep N3" and "first non-REM sleep N1 or second non-REM sleep N2" as the target variables. The two-class classifier CF14 can be constructed, for example, by supervised learning using training data with features I to VIII as explanatory variables and "first non-REM sleep N1" and "second non-REM sleep N2" as the target variables.

[0070] The sleep stage determination unit 34 determines the user U's sleep stage for each sequentially occurring period PR based on the heart rate interval HBI acquired during that period PR. The sleep stage determination unit 34 sequentially stores the determination results for each period PR in the memory unit 40.

[0071] [Sleep determination process S105] In the sleep determination process S105, the sleep determination unit 35 determines whether or not the user U is in a sleep state based on the heart rate interval HBI acquired in the heart rate information acquisition process S103.

[0072] The sleep determination unit 35 uses the two-class classifier CF2 shown in Figure 11 to determine whether user U is in a sleep state. The two-class classifier CF2 is a classifier that classifies user U's state into either "awake state" or "sleep state," similar to the two-class classifier CF11 of the sequential two-class classifier CF1. However, compared to the two-class classifier CF11, the two-class classifier CF2 has a smaller probability of misclassification (hereinafter referred to as "false positive"), where user U is actually in an awake state (negative) but is classified as being in a sleep state (positive). In other words, the two-class classifiers CF11 and CF2 are constructed such that the probability of misclassifying user U's state as being in a sleep state when user U is awake is smaller in the two-class classifier CF2 than in the two-class classifier CF11. In other words, the two-class classifier CF2 is a classification model with higher specificity compared to the two-class classifier CF11. Specifically, for example, the false positive rate in the two-class classifier CF2 may be 20% or less. If the false positive rate is 20%, the false negative rate may be up to about 40%.

[0073] In this embodiment, the explanatory variables of the two-class classifier CF2 are the same feature quantities I to VIII as the explanatory variables of the sequential two-class classifier CF1. The sleep determination unit 35 inputs the feature quantities I to VIII calculated by the sleep stage determination unit 34 to the two-class classifier CF2. As a result, the two-class classifier CF2 classifies the state of user U into either "awake state" or "sleep state". When the classification result of the two-class classifier CF11 is "sleep state", the sleep determination unit 35 determines that user U is in a sleep state.

[0074] In this embodiment, the two-class classifier CF2 is a Support Vector Machine (SVM). The two-class classifier CF2 can be constructed, for example, by supervised learning using training data with features I to VIII as explanatory variables and "awake state" and "sleep state" as the target variables. In constructing the two-class classifier CF2, in order to reduce the false positive rate (i.e., increase the specificity) in the two-class classifier CF2, the ratio of positives to negatives in the dataset is changed, the penalty for misclassification for each class is adjusted, and the algorithm is selected.

[0075] The sleep determination unit 35 determines whether the user U is in a sleep state for each of the sequentially occurring periods PR, based on the heart rate interval HBI acquired during that period PR. The sleep determination unit 35 sequentially stores the determination results for each period PR in the memory unit 40.

[0076] [Bed board movement control process] In the bed board movement control process, the bed board movement control unit 36 ​​moves the head-side bed board 521 of the bed 500 and changes the angle θ of the head-side bed board 521 based on the determination results obtained in the state determination process. In this embodiment, the bed board movement control unit 36 ​​reduces the angle θ of the head-side bed board 521 based on the determination result in the body movement determination process S102, the determination result in the sleep stage determination process S104, and the determination result in the sleep determination process S105.

[0077] A specific example of the floorboard movement control process performed by the floorboard movement control unit 36 ​​will be explained according to the flowchart in Figure 12. Specifically, the floorboard movement control unit 36 ​​starts executing the floorboard movement control process based on, for example, the presence of a user U on the floorboard 520 and the angle θ of the head-side floorboard 521 being greater than or equal to a predetermined value. Specifically, the floorboard movement control unit 36 ​​can determine that a user U is present on the floorboard 520 based on at least one of the following: the total load applied to the floorboard 520 calculated based on load signals S1 to S4 being greater than or equal to a predetermined value, and the body movement determination unit 32 determining that body movement is occurring in the user U. In addition, the floorboard movement control unit 36 ​​may start executing the floorboard movement control process based on the instruction to start floorboard movement control being received via the operation unit 70.

[0078] In the floorboard movement control process, the floorboard movement control unit 36 ​​first executes the first determination process S201. In the first determination process S201, the floorboard movement control unit 36 ​​determines whether or not to start moving the head-side floorboard 521 from the back-raised state to the back-lowered state (hereinafter referred to as "back-lowering movement" as appropriate) based on the determination result in the sleep stage determination process S104 and the determination result in the sleep determination process S105.

[0079] Specifically, the floorboard movement control unit 36 ​​determines whether the determination result in the sleep stage determination step S104 and the determination result in the sleep determination step S105 satisfy the following conditions (1) to (3). In conditions (1) to (5) below, the "target period" may be, but is not limited to, the period PR immediately preceding the determination (i.e., the latest period PR at the time of determination).

[0080] (1) The sleep determination unit 35 determined that user U was in a sleep state during six consecutive periods including the target period. (2) The sleep stage determination unit 34 determined that user U's sleep stage was third non-REM sleep N3 during the target period. (3) The average of the determination results of the sleep stage determination unit 34 over three consecutive periods including the target period was deeper than second non-REM sleep N2.

[0081] Here, whether or not "the average of the sleep stage determination results of the sleep stage determination unit 34 over a period of K times (K=3 in condition (3)) is deeper than the second non-REM sleep N2" can be determined by, for example, setting predetermined values ​​corresponding to each sleep stage and comparing the average value obtained by dividing the sum of these predetermined values ​​over K determination results by K with the predetermined value corresponding to the second non-REM sleep N2. More specifically, for example, the bed plate movement control unit 36 ​​can set the predetermined value corresponding to the awake state to "5", the predetermined value corresponding to REM sleep to "4", the predetermined value corresponding to the first non-REM sleep to "3", the predetermined value corresponding to the second non-REM sleep to "2", and the predetermined value corresponding to the third non-REM sleep to "1", and determine that condition (3) is met when the average value obtained by dividing the sum of these predetermined values ​​over three consecutive periods PR by 3 is less than 2. The values ​​of the predetermined values ​​corresponding to each sleep stage can be set as appropriate.

[0082] If the floorboard movement control unit 36 ​​determines that all of conditions (1) to (3) are met (step S201: YES), it drives the movement mechanism 530 to start the backrest lowering movement and moves the head-side floorboard 521 until the angle θ of the head-side floorboard 521 becomes 0°, that is, until the head-side floorboard 521 is in the backrest lowered position (backrest lowering step S202). After that, the floorboard movement control unit 36 ​​terminates the floorboard movement control step.

[0083] If the floorboard movement control unit 36 ​​determines that at least one of conditions (1) to (3) is not met (step S201: NO), it determines whether the elapsed time T since the start of the floorboard movement control process is equal to or greater than the first threshold TH1 (an example of "first elapsed time") (step S203). The first threshold TH1 may be approximately 2 to 3 hours, or 2.5 hours. If the floorboard movement control unit 36 ​​determines that the elapsed time T is less than the first threshold TH1 (step S203: NO), it performs the first determination step S201 again.

[0084] If the floorboard movement control unit 36 ​​determines that the elapsed time T is equal to or greater than the first threshold TH1 (step S203: YES), it executes the second determination step S204. In the second determination step S204, the floorboard movement control unit 36 ​​determines whether or not to start the backrest lowering movement based on the determination result in the body movement determination step S102.

[0085] Specifically, the floor plate movement control unit 36 ​​determines whether the determination result in the body movement determination step S102 satisfies the following condition (4).

[0086] (4) In the judgment result of the body motion determination step S102, there is no judgment of body motion for X consecutive periods, including the judgment result of the target period (i.e., the judgment result of the body motion determination step S102 is no body motion for X or more consecutive times, including the judgment result of the target period). Note that in condition (4), X can be set arbitrarily, but as an example it may be around 12 to 24, or it may be 18. The product of the length of period PR and X is an example of the "first predetermined period" and is also an example of the "second predetermined period".

[0087] Furthermore, when condition (4) is met, user U can be considered to be in a deep sleep state. The frequency of large body movements correlates with the sleep stage, being lowest when user U is in the third non-REM sleep stage (N3), and increasing in the following order: third non-REM sleep stage (N3), second non-REM sleep stage (N2), REM sleep, first non-REM sleep stage (N1), and wakefulness (W). Therefore, the longer the period without large body movements, the higher the likelihood of being in a deep sleep state.

[0088] If the floorboard movement control unit 36 ​​determines that condition (4) is met (step S204: YES), it drives the movement mechanism 530 to perform back-down movement (back-down movement step S202). If the floorboard movement control unit 36 ​​determines that condition (4) is not met (step S204: NO), it determines whether the elapsed time T since the start of the floorboard movement control process is greater than or equal to the second threshold TH2 (an example of "second elapsed time") (step S205). The second threshold TH2 is a value greater than the first threshold TH1, and may be 3 to 4 hours or 3.5 hours as an example. If the floorboard movement control unit 36 ​​determines that the elapsed time T is less than the second threshold TH2 (step S205: NO), it performs the second determination step S204 again.

[0089] If the floorboard movement control unit 36 ​​determines that the elapsed time T is greater than the second threshold TH2 (step S205: YES), it executes the third determination step S206. In the third determination step S206, the floorboard movement control unit 36 ​​determines whether or not to start lowering the backrest based on the determination result in the body movement determination step S102.

[0090] Specifically, the floor plate movement control unit 36 ​​determines whether the determination result in the body movement determination step S102 satisfies the following condition (5).

[0091] (5) In the body movement determination step S102, there is no determination of body movement for Y consecutive periods, including the most recent determination result (i.e., the determination result in the body movement determination step S102 is no body movement for Y or more consecutive periods, including the determination result for the target period. In this case, user U can be considered to be in a deep sleep state). Note that in condition (5), Y is a value smaller than X. Y can be set arbitrarily, but as an example it may be around 5 to 15, or it may be 10. The product of the length of period PR and Y is an example of the "second predetermined period" and is also an example of the "third predetermined period".

[0092] If the floorboard movement control unit 36 ​​determines that condition (5) is met (step S206: YES), it drives the movement mechanism 530 to perform back-lowering movement (back-lowering step S202). If the floorboard movement control unit 36 ​​determines that condition (5) is not met (step S204: NO), it performs the third determination step S206 again.

[0093] The effects of the floor plate moving device 100 of this embodiment are summarized below.

[0094] In the floor plate moving device 100 of this embodiment, the sequential two-class classifier CF1 used by the sleep stage determination unit 34 classifies the sleep stage of user U using feature quantities I to VIII as explanatory variables. Therefore, the sleep state of user U can be classified with high accuracy. Based on the classification results of the sequential two-class classifier CF1, the sleep stage determination unit 34 can determine the sleep stage of user U with high accuracy.

[0095] In the bed base moving device 100 of this embodiment, the sleep determination unit 35 determines whether the user U is in a sleep state using a two-class classifier CF2, which has a lower rate of false positives than the two-class classifier CF11 of the sequential two-class classifier CF1. The bed base moving control unit 36 ​​then starts lowering the head bed base 521 based on the determination result of the sleep determination unit 35. Therefore, the bed base moving device 100 of this embodiment can suitably move the bed base 520 by suppressing the occurrence of the event in which the lowering of the head bed base 521 is started even though the user U is awake.

[0096] In the floorboard moving device 100 of this embodiment, the floorboard moving control unit 36 ​​also starts the downward movement of the head-side floorboard 521 based on the determination result of the sleep stage determination unit 34. Therefore, the floorboard moving device 100 of this embodiment can suitably move the floorboard 520 while suppressing the occurrence of the event in which the downward movement of the head-side floorboard 521 is started even though the user U is not in a sufficiently deep sleep state, causing the user U to wake up.

[0097] In the floorboard moving device 100 of this embodiment, the floorboard moving control unit 36 ​​first executes a first determination step S201 based on the determination results of the sleep stage determination unit 34 and the sleep determination unit 35 after the start of floorboard moving control, and determines whether or not to start the back-down movement of the head-side floorboard 521. If the back-down movement has not started even after a first elapsed time has elapsed since the start of floorboard moving control, the second determination step S204 based on the determination result of the body movement determination unit 32 is executed, and it is determined whether or not to start the back-down movement of the head-side floorboard 521.

[0098] As described above, the bed base movement control unit 36 ​​of this embodiment changes the type of input used to determine the start of back lowering movement, i.e., the type of information that forms the basis of the determination, if back lowering movement has not started even after a first elapsed time has elapsed since the start of bed base movement control. Therefore, the bed base movement device 100 of this embodiment can suitably move the bed base 520 by suppressing the occurrence of the event in which back lowering movement of the head bed base 521 does not start even though the user U is in a sleeping state. Changing the type of input used to determine the start of back lowering movement in this way is particularly advantageous when the accuracy of sleep stage determination and sleep determination may decrease, such as when the user U has a lot of body movement related to respiratory disorders or restless legs syndrome, and the period during which heart rate interval acquisition and sleep stage determination can be performed is short.

[0099] In the floor plate moving device 100 of this embodiment, if the backrest lowering movement has not started even after a second elapsed time, which is longer than the first elapsed time, the floor plate moving control unit 36 ​​determines whether or not to start the backrest lowering movement of the head-side floor plate 521 based on a third determination step S206 that uses conditions that are more relaxed than those of the second determination step S204.

[0100] Thus, in this embodiment, if the backrest lowering movement has not started even after a second elapsed time has elapsed since the start of the floorboard movement control, the floorboard movement control unit 36 ​​does not change the type of input for determining the start of the backrest lowering movement, but relaxes the conditions that serve as the threshold for the determination. Therefore, the floorboard movement device 100 of this embodiment can suitably move the floorboard 520 by suppressing the occurrence of the event in which the backrest lowering movement of the head-side floorboard 521 does not start even though the user U is in a sleeping state. In this disclosure and the present invention, relaxing the conditions means changing the threshold included in the conditions so that the determination to start the backrest lowering movement is more likely to occur.

[0101] <Modified Version> In the floor plate moving device 100 of the above embodiment, the following modified form can also be used.

[0102] [Modification of the sleep stage determination unit 34] In the floor plate moving device 100 of the above embodiment, the explanatory variables of the sequential two-class classifier CF1 used by the sleep stage determination unit 34 are not limited to features I to VIII, but can be changed as appropriate. The method for determining the explanatory variables in the creation of the sequential two-class classifier CF1 will be described below.

[0103] In the following description, heart rate interval HBI refers to the time interval between one heartbeat and the next. An example of heart rate interval HBI is the RR interval (R-R Interval: RRI), which is the interval between R waves on an electrocardiogram. The JJ interval (J-J Interval: JJI), which is the interval between J peaks in the BCG signal, is also an example of heart rate interval HBI and is based on the acceleration generated when the heart beats. It has been shown that RRI and JJI have a very high correlation. In this specification, features related to RRI can be considered to be substantially the same even if RRI is replaced with JJI or other heart rate interval HBIs. That is, in implementing the techniques described herein, RRI, JJI, and other heart rate interval HBIs can be used without distinction.

[0104] The explanatory variables in the creation of the sequential two-class classifier CF1 in the above embodiment were determined as follows.

[0105] In a dataset of 400 subjects with various conditions, including healthy individuals and those with sleep disorders, including PSG (polysomnography) measurements and sleep labels determined by sleep specialists, the RRI for each epoch (30-second period) was calculated using a heart rate peak extraction algorithm based on electrocardiograms. To ensure good sleep stage determination even when using data where the period for which heart rate interval HBI (specifically, JJI, for example) can be continuously acquired is relatively short, relatively short periods of 30 seconds and 90 seconds were used as feature calculation periods. For each epoch in which a sleep stage was to be determined, the RRI data for the 30-second period of each epoch and the 90-second period centered on each epoch (i.e., three epochs centered on each epoch) were normalized using the following (Equation 1).

[0106] In (Equation 1), RRI ｎｏｒｍ is the normalized RRI, and median(RRI) is the median of the RRI data for one night. By normalizing the data in this way to match the median of each subject's data for one night to 1, the accuracy of model learning performed by combining data from different subjects can be improved.

[0107] Next, the calculated RRI ｎｏｒｍ Based on this, 279 types of features included in the 23 groups shown in Figures 13A and 13B were calculated as candidates for explanatory variables. Of these 23 groups, the features included in Groups 1 to 9 are time-domain features, the features included in Groups 10 to 19 are frequency-domain features, and the features included in Groups 20 to 23 are nonlinear-domain features. Further details of the 279 types of features included in these 23 groups are shown in the tables in Figures 14A to 14G.

[0108] In the tables in Figures 14A to 14G, features with "(90s)" in their name are features for the 90-second interval, while others are features for the 30-second interval. Features with "(AR)" in their name are features calculated using the AR (Autoregressive) model. Features with "(Welch)" in their name are features calculated using the Welch method. Features with "LF" in their name ＊In features containing "HF", LF is the total power in a frequency domain defined with a width of 0.11 Hz, centered on the point with the greatest power within the conventional LF frequency domain. ＊ In features containing "", HF is the total power in a frequency domain defined with a width of 0.10 Hz, centered on the point with the highest power within the conventional HF frequency domain. Features containing "(detrended)" in their name are features calculated by removing trends. In this disclosure and the present invention, features containing "(ANY)" in their name are features whose calculation interval length is not limited to 30 seconds, 90 seconds, etc., but can be any length. Features that do not contain "(AR)" or "(Welch)" in their name are features calculated by any method.

[0109] Next, from the 279 types of features mentioned above, we selected the most effective features for each of the two-class classifiers CF11 to CF14. The selection of the most effective features was performed using SVM-RFE, which ranks the effectiveness of features based on the change in the optimization function when the SVM learns the decision boundary. SVM-RFE ranks the effectiveness of features by utilizing the property that features that cause a larger change from the original optimization function when a certain feature is removed and the SVM is trained have a greater influence on classification. Note that if there are many correlated features, the influence may be underestimated, so we used SVM-RFE+CBR, which adds a correction function to SVM-RFE.

[0110] A Standardized Variable Model (SVM) was used for the classification model. Furthermore, to ensure accurate classification even for a small number of classes, a SVM format was employed that applied penalties based on the ratio of positive to negative values ​​in the training data.

[0111] Next, the 400 subjects included in the dataset were divided into four groups of 100 each, and each group was further divided into three subsets for 3-fold cross-validation. Then, for each of the 12 subsets obtained by dividing each of the four groups into three subsets, features that showed good performance for the two-class classifier were extracted, and the features were ranked based on the performance of the two-class classifier corresponding to each feature. Finally, the ranking of the features in each of the 12 subsets was averaged to determine the particularly effective features for each of the two-class classifiers CF11 to CF14. The top 40 features that showed particularly good performance for each of the two-class classifiers CF11 to CF14 are shown in Figures 15A and 15B.

[0112] The performance of the two-class classifier based on each feature was evaluated using the geometric mean G-mean value calculated based on (Equation 2) below, with the recall fixed.

[0113] Next, we plotted the number of features N used as explanatory variables against the performance of the two-class classifier created using N features, as shown in Figure 16. In Figure 16, the number of features used being N means that the features used are the features ranked from 1st to Nth, as shown in Figures 15A and 15B.

[0114] As shown in Figure 16, the performance of a two-class classifier (i.e., the geometric mean G-mean) decreases in slope as the number of features N used as explanatory variables increases, similar to the graph of a logarithmic function. This means that influential features are concentrated at the top of the ranking, and features at the bottom of the ranking contribute less to the performance of the two-class classifier.

[0115] To determine the most influential features for each of the two-class classifiers CF11 to CF14, the average of the ranking positions for the two-class classifiers CF11 to CF14 shown in Figures 15A and 15B is calculated, and when arranged in descending order of ranking (smallest numerical value), the result is as shown in Figure 17. Feature quantities I to VIII, used as explanatory variables in the above embodiment, correspond to the top eight features shown in Figure 17.

[0116] Thus, the explanatory variables of the sequential two-class classifier CF1 in the above embodiment use features that enable sleep stage determination with particularly high accuracy. However, it is not limited to this. Specifically, any features shown in Figures 15A and 15B can be used as explanatory variables for the sequential two-class classifier CF1. Alternatively, any features understood based on the features shown in Figures 15A and 15B can be used.

[0117] More specifically, for example, at least one of the following features (i) to (v) may be used as an explanatory variable for the sequential two-class classifier CF1: (i) The number or proportion of pairs of adjacent heart rate intervals (HBI) during the target period in which the difference between the two HBIs is greater than or equal to a threshold. (ii) The value obtained by subtracting the median of HBI in past periods prior to the target period from the percentile range of HBI during the target period. (iii) The median of HBI during the target period. (iv) A feature based on the maximum power of HF. (v) A feature based on Log LF.

[0118] The above features (i) to (v) are evaluation indices for heart rate variability (HRV). Heart rate variability reflects the autonomic nervous system's adjustments that allow a person to adapt to environmental and psychological stress. In a typical sleep pattern, one progresses from wakefulness (W) to non-REM sleep and then to REM sleep. Non-REM and REM sleep alternate until waking. Because the balance between the sympathetic and parasympathetic nervous systems changes depending on the sleep stage, using heart rate variability evaluation indices as features (explanatory variables) allows for effective differentiation of sleep stages.

[0119] Feature (i) represents short-term heart rate variability and is related to the activity of the parasympathetic nervous system. Heart rate variability, which represents fluctuations in the parasympathetic nervous system, tends to be smaller when the sympathetic nervous system is active and larger when the parasympathetic nervous system is active. Specific examples of feature (i) include NN20, NN50, pNN20, pNN50, etc. The numerical values ​​for each feature indicate thresholds.

[0120] Feature (ii) is the value obtained by subtracting the median of heart rate interval HBI for past periods prior to the target period from the percentile range of heart rate interval HBI variability. Since sleep data is a time series and past sleep states are correlated with current sleep states, the effectiveness of the features can be enhanced by reflecting the properties of past periods in the features of the target period. In particular, the effectiveness of features can be further enhanced in LSTM and other methods that use information from preceding and succeeding intervals through feedback. A specific example of feature (ii) is RRI range minus previous median. Furthermore, although not shown in the tables in Figures 14A to 14G, Percentile range (2.5-97.5) minus previous medium, calculated using the 2.5-97.5% percentile range instead of the RRI range (i.e., the 0-100% percentile range), and Percentile range (5-95) minus previous medium, calculated using the 5-95% percentile range, are also examples of feature quantities (ii). Note that the calculation of RRI range minus previous medium, etc., is performed using a method equivalent to trend removal. Therefore, for example, the feature name of RRI range minus previous Median may be RR range (detrended), RRI range (detrended), etc.

[0121] Regarding feature (ii), the median of heart rate interval HBI for past periods is not limited to the method of using the median of heart rate interval HBI for the period closest to the target period among past periods in which at least one heart rate interval could be obtained. The calculation of feature (ii) can be appropriately performed using the median of heart rate interval HBI for the period immediately preceding the target period, the median of heart rate interval HBI for the period two periods prior to the target period, the median of heart rate interval HBI for several periods prior to the target period, etc. In this disclosure and the present invention, periods in which no heart rate interval could be obtained are not included in the "past period". In the above embodiments and modifications, the past period is a period PR prior to the period PR corresponding to the target period, and the length of the past period and the length of the target period are the same. However, it is not limited to this, and the length of the past period and the length of the target period may be different.

[0122] Feature (iii) is a value that reduces the influence of outliers on the average heart rate interval (HBI). The more active the sympathetic nervous system is, the higher the heart rate and the lower the HBI. Therefore, by using feature (iii), it is possible to effectively distinguish between wakefulness (W), REM sleep, and non-REM sleep. Specific examples of feature (iii) include Median of RRI, Median of RRI (detended) (90s), etc.

[0123] Feature(iv) and Feature(v) are frequency domain features. The respiratory frequency is included within the HF frequency band (0.15 Hz to 0.40 Hz), and due to respiratory arrhythmias (RSA) based on the activity of the parasympathetic nervous system, the heart rate interval (HBI) fluctuates with the respiratory frequency. In non-REM sleep, where the parasympathetic nervous system is dominant, the HF peak due to respiratory arrhythmias is more prominent compared to wakefulness (W) and REM sleep (REM), where the sympathetic nervous system is relatively dominant, and the maximum power in the HF band of Feature(iv) reflects this. For Feature(v), LF is the sum of the powers from 0.04 Hz to 0.15 Hz. LF is related to the blood pressure reflex and is influenced by the activity of both the sympathetic and parasympathetic nervous systems.

[0124] Examples of feature vectors (iv) include Log max power of HF (Welch) (90s) and Log max power of HF (AR). Examples of feature vectors (v) include Log normalized LF* (AR) (90s) and Log LF (AR) (90s).

[0125] In addition, any feature based on heart rate interval HBI can be used as an explanatory variable for the sequential two-class classifier CF1 of the above embodiment. Since the feature based on heart rate interval HBI changes according to heart rate variability, sleep stages can be distinguished based on the feature based on heart rate interval HBI.

[0126] Furthermore, the explanation above regarding the determination of explanatory variables for the sequential two-class classifier CF1 of the sleep stage determination unit 34 also applies similarly to the determination of explanatory variables for the two-class classifier CF2 used by the sleep determination unit 35.

[0127] In the sequential two-class classifier CF1 of the above embodiment, the explanatory variables of the two-class classifiers CF11 to CF14 may be different from each other. Also, the explanatory variables of the sequential two-class classifier CF1 and the explanatory variables of the two-class classifier CF2 may be different.

[0128] In the above embodiment, the order in which the sequential two-class classifier CF1 classifies the state of user U is arbitrary. Specifically, for example, the sequential two-class classifier may classify the state of user U into "third non-REM sleep N3" and "other state," and if the state of user U is classified as "other state," it may classify the state of user U into "awake W or REM sleep REM" and "first non-REM sleep N1 or second non-REM sleep N2." In this way, even if the order of classification and the divisions of classification differ from the above embodiment, the same feature quantities as in the above embodiment and its modifications can be used as explanatory variables.

[0129] In the above embodiment, the length of the period PR is not limited to 30 seconds and can be set arbitrarily. In the above embodiment, the length of one epoch in the feature calculation and the length of the period PR are both 30 seconds and coincide. However, it is not limited to this.

[0130] In the above embodiment, the sequential two-class classifier CF1 may consist of only at least one of the two-class classifiers CF1 to CF4. Furthermore, in the above embodiment, a support vector machine is used to construct the two-class classifier (two-class classification model), but this is not the only option. Instead of a support vector machine, any method such as logistic regression, decision tree, random forest, CNN, LSTM, etc., can be used to construct the two-class classifier (two-class classification model).

[0131] [Modified Version of Floorboard Movement Control Unit 36] In the floorboard movement control process of the above embodiment, the contents of the first determination step S201, the second determination step S204, and the third determination step S206 in the flowchart of Figure 12 can be changed as appropriate.

[0132] As an example, the floor plate movement control unit 36 ​​may, in the first determination step S201, determine whether or not to start the backrest lowering movement based on whether or not the following conditions (1') to conditions (3') are met. In the following conditions (1') to conditions (3'), the "target period" may be the latest period PR, but is not limited to this.

[0133] (1') The sleep determination unit 35 determined that user U was in a sleep state for at least A consecutive periods (A is an integer of 2 or more) including the target period. (2') The sleep stage determination unit 34 determined that user U's sleep stage was third non-REM sleep N3 during the target period. (3') The average of the determination results of the sleep stage determination unit 34 for at least B consecutive periods (B is an integer of 2 or more) including the target period is deeper than second non-REM sleep N2.

[0134] As another example, the floorboard movement control unit 36 ​​may determine whether or not to start the backrest lowering movement in each of the first determination step S201, the second determination step S204, and the third determination step S206, based on whether or not conditions (1') to (3') are met. In this case, for example, "A" in condition (1') and "B" in condition (3') may be set to the largest value in the first determination step S201, and then decrease in the order of the second determination step S204 and the third determination step S206.

[0135] As another example, the floor plate movement control unit 36 ​​may, in the first determination step S201, determine whether or not to start lowering the backrest based on the determination result in the body movement determination step S102. Specifically, the floor plate movement control unit 36 ​​may determine whether or not the determination result in the body movement determination step S102 satisfies the following condition (6).

[0136] (6) The judgment result in the body movement judgment step S102 is that there has been no body movement for Z consecutive times, including the judgment result for the target period (which may be the most recent period PR) (i.e., the judgment result in the body movement judgment step S102 does not indicate any body movement for Z consecutive periods, including the judgment result for the target period. In this case, the user U can be considered to be in a deep sleep state).

[0137] In condition (6), Z can be any value greater than X in condition (4), but as an example, it may be around 36 to 60, or it may be 48. The product of the length of period PR (30 seconds in this embodiment) and Z is an example of the "first predetermined period".

[0138] In this embodiment, the floor plate movement control unit 36 ​​makes a determination in each of the first determination step S201, the second determination step S204, and the third determination step S206 based on the determination result in the body movement determination step S102.

[0139] In the body movement determination step S102 of the above embodiment, the body movement determination unit 32 determines whether or not body movement is occurring in the user U based on the normalized signal NS, but is not limited to this. For example, the body movement determination unit 32 may acquire a time integral value IV at predetermined intervals (for example, 1 second) of any one output value (absolute value) of load signals S1 to S4 or the average value of at least two output values ​​(absolute values) of load signals S1 to S4, or any one output value (absolute value) of load signals LS1 to LS4 or the average value of at least two output values ​​(absolute values) of load signals LS1 to LS4, and determine whether or not body movement is occurring in the user U based on a comparison of the acquired time integral value IV with a threshold. As another example, the body movement determination unit 32 may acquire a time integral value IVσ at predetermined intervals (for example, 1 second) of the standard deviation of any one of the load signals S1 to S4 or the average value of at least two standard deviations of the load signals S1 to S4, or the standard deviation of any one of the load signals LS1 to LS4 or the average value of at least two standard deviations of the load signals LS1 to LS4, and determine whether or not body movement is occurring in the user U based on a comparison of the acquired time integral value IVσ with a threshold. The time integral values ​​IV and IVσ are examples of parameters that indicate "the magnitude of fluctuations in the load applied by the user to the support surface".

[0140] In this case, the floor plate movement control unit 36 ​​may use the following condition (7) instead of the above conditions (4) to (6).

[0141] (7) The average values ​​of the time integral values ​​IV and IVσ over a predetermined period are less than or equal to a predetermined threshold. Note that in condition (7), the predetermined period (an example of each of the "first predetermined period," "second predetermined period," and "third predetermined period") can be set arbitrarily, but as an example, when used in place of condition (4), it may be a period corresponding to a period PR of about 12 to 24 times, when used in place of condition (5), it may be a period corresponding to a period PR of about 5 to 15 times, and when used in place of condition (6), it may be a period corresponding to a period PR of about 36 to 60 times.

[0142] Furthermore, the conditions in the first determination step S201, the second determination step S204, and the third determination step S206 may be conditions based on the determination result of the body movement determination unit 32, a combination of conditions based on the determination result of the sleep stage determination unit 34 and conditions based on the determination result of the sleep determination unit 35, or a combination of conditions based on the determination result of the body movement determination unit 32, conditions based on the determination result of the sleep stage determination unit 34, and conditions based on the determination result of the sleep determination unit 35. By making a determination using a combination of conditions based on the determination result of the sleep stage determination unit 34 and conditions based on the determination result of the sleep determination unit 35, it is possible to suppress the start of back-down movement when the user U is awake and to start back-down movement when the user U is in a better sleep state. Even if the determination is made based on conditions based on the determination result of the body movement determination unit 32, it is possible to make a determination to start back-down movement according to the sleep state of the user U based on the decrease in body movement occurrence corresponding to the user U being in a sleep state. At least one of the first determination step S201, the second determination step S204, and the third determination step S206 may be omitted. Alternatively, at least one additional determination step may be added, which has different conditions from the first determination step S201, the second determination step S204, and the third determination step S206.

[0143] In the above embodiment, if the floorboard movement control unit 36 ​​determines that the user U is awake during the execution of the floorboard movement control process, it may reset the elapsed time from the start of the floorboard movement control process and return the process to the first determination process S201. Here, "user U is awake" means the state in which user U is awake, excluding the state in which user U's sleep stage is temporarily awake W. Specifically, for example, if user U's sleep stage is wakeful W for a predetermined period of time or longer, user U can be said to be awake. According to this modification, even if user U is awake, the initiation of back-lowering movement based on relatively relaxed judgment criteria used in the second determination process S204, the third determination process S206, etc., is suppressed.

[0144] Specifically, the floorboard movement control unit 36 ​​determines that user U is awake based on, for example, that user U has been out of bed for a predetermined period of time (e.g., 30 minutes) or longer. Furthermore, the floorboard movement control unit 36 ​​can determine that user U is out of bed based on, for example, that the total load applied to the floorboard 520, calculated based on load signals S1 to S4, is below a predetermined value. As another example, the floorboard movement control unit 36 ​​determines that user U is awake based on the operation of the input unit 70 to increase the angle θ of the head-side floorboard 521 (back-raising operation).

[0145] [Calculation of Sleep Score] In the floor plate moving device 100 of the above embodiment, the control unit 30 may calculate the sleep score of the user U based on the determination result of the sleep stage determination unit 34. The control unit 30 is an example of a "sleep score calculation unit". Specifically, the sleep score can be calculated, for example, by the following formula (3).

[0146] The total sleep time [s] is specifically the time from the start to the end of measurement, minus the time during which the sleep stage determination unit 34 determined the patient to be awake (W), and the time during which the sleep stage determination unit 34 could not obtain a determination result for any reason. The measurement start time may be, for example, the time when the floorboard movement control process is started. The measurement end time may be, for example, the time when the lights are turned on. In this case, the control unit 30 can obtain the time when the lights are turned on based on a light sensor (not shown).

[0147] Sleep latency [s] is specifically the time from the start of measurement until the first time the sleep stage determination result of the sleep stage determination unit 34 becomes REM sleep, first non-REM sleep N1, second non-REM sleep N2, or third non-REM sleep N3 (epoch).

[0148] Post-sleep awakening [s] is specifically the sum of the time during which the sleep stage determination unit 34 determined wakefulness W, from the time of sleep onset (for example, the time when the sleep stage determination unit 34 first determined REM sleep, first non-REM sleep N1, second non-REM sleep N2, or third non-REM sleep N3) to the end of measurement (for example, the time when the lights are turned on).

[0149] The N3 occurrence rate [%] is the ratio of the time during which the sleep stage determination unit 34 determined the sleep stage to be the third non-REM sleep stage (N3) to the total sleep time [s]. The sleep efficiency [%] is the ratio of the total sleep time [s] to the total recording time.

[0150] a1-a5 and b1-b5 are coefficients, which can be determined based on regression. The sleep score can be calculated using any formula that can quantify the sleep state of user U, in addition to the above formula (Equation 3).

[0151] [Modification of Load Signal Acquisition Process S101] In the load signal acquisition process S101 of the above embodiment, the load signal acquisition unit 31 may invert the phase of at least one of the load signals LS1 to LS4. The load signals LS1 and LS3 from the load detectors 11 and 13, which are positioned with the user U in between, are often signals with opposite phases to each other. Similarly, the load signals LS2 and LS4 from the load detectors 12 and 14, which are positioned with the user U in between, are often signals with opposite phases to each other. Therefore, for example, by inverting the phases of the load signals LS3 and LS4 to match the phases of the load signals LS1 and LS2 and performing the heart rate information acquisition process S103, heart rate information can be acquired with higher accuracy.

[0152] [Modification of the body movement determination step S102] In the body movement determination step S102 of the above embodiment, the body movement determination unit 32 can determine whether or not body movement has occurred in the user U in any manner. The body movement determination unit 32 may determine whether or not body movement has occurred in the user U based on fluctuations in the load applied by the user U to the support surface 520 in any manner.

[0153] The body movement determination unit 32 may divide one of the normalized signals NS1 to NS4 of period PR into multiple signals along the time axis, and calculate the standard deviation for each of these multiple signals. Then, it may compare the minimum value of the multiple standard deviations corresponding to each of these multiple signals with a threshold value, and determine, for example, that body movement occurred in user U during period PR if the minimum value is smaller than the threshold value.

[0154] [Modification of heart rate information acquisition step S103] In the heart rate information acquisition step S103 of the above embodiment, the heart rate information acquisition unit 33 can acquire the heart rate information of the user U in any manner based on the information acquisition signal IS.

[0155] In the heart rate information acquisition step S103 of the above embodiment, the heart rate information acquisition unit 33 may extract feature quantities related to the waveform of the peak of the information acquisition signal IS, identify a peak corresponding to the user U's heart rate from the peak of the information acquisition signal IS based on the classification of said feature quantities, and acquire user U's heart rate information based on the identified peak.

[0156] Specifically, the extraction of feature quantities related to the waveform of the peaks of the information acquisition signal IS is performed, for example, by extracting the following feature quantities A to E for each peak indicated by the information acquisition signal IS.

[0157] (1) Feature A: The difference in amplitude between the negative maximum value just before the peak and the peak itself. (2) Feature B: The amplitude of the negative maximum value just before the peak. (3) Feature C: The amplitude of the peak. (4) Feature D: The time from the negative maximum value just before the peak to the peak. (5) Feature E: The time from the peak to the negative maximum value just after the peak.

[0158] Feature classification can be performed, for example, by k-means clustering. The heart rate information acquisition unit 33, for example, classifies a dataset containing feature quantities A to E for each peak into three classes by k-means clustering, and J peak PK Ｊ The heart rate information acquisition unit 33 then identifies a dataset that shows the J peak PK of the information acquisition signal IS based on the identified dataset. Ｊ Identify the identified Jpeak PK Ｊ Based on this, the user U's heart rate interval (HBI) is obtained.

[0159] In the heart rate information acquisition step S103 of the above embodiment, the heart rate information acquisition unit 33 may acquire heart rate interval HBI (specifically, for example, RRI) based on an electrocardiogram (ECG). In this case, the sleep stage determination unit 34 and the sleep determination unit 35 input the feature quantities based on the heart rate interval HBI based on the electrocardiogram to the sequential two-class classifiers CF1 and CF2.

[0160] [Other Modifications] In the floor plate moving device 100 of the above embodiment, the control unit 30 may perform a display step in which it displays at least one of the following on the display unit 60: the determination result of the body movement determination step S102, the heart rate information acquired in the heart rate information acquisition step S103, the determination result of the sleep stage determination step S104, and the determination result of the sleep determination step S105.

[0161] The sequential two-class classifier CF1 of the above embodiment can also be used as a sleep state classifier independent of the bed base moving device 100. In this case, feature quantities based on heart rate interval HBI are acquired by a device separate from the sleep state classifier, and the acquired feature quantities are input to the sleep state classifier as explanatory variables. This separate device is not limited to a device that acquires BCG signals and JJI and acquires feature quantities based on JJI. Specifically, this separate device may be, for example, a device that acquires electrocardiograms and RRI and acquires feature quantities based on RRI.

[0162] Furthermore, the sleep state classifier may have only one of the two-class classifiers CF11 to CF14. The floor plate moving device 100 of the above embodiment can also be used as a sleep state determination device equipped with the sleep state classifier. In this case, components other than the sequential two-class classifier CF1 may be omitted as appropriate.

[0163] The floor plate moving device 100 in the above embodiment does not necessarily need to include all of the load detectors 11 to 14, and may include only at least one of the load detectors 11 to 14. Furthermore, the load detectors do not necessarily need to be placed at the four corners of the bed, and can be placed at any position so as to be able to detect the load of the user on the bed and its fluctuations. In addition, the load detectors 11 to 14 are not limited to load sensors using beam-type load cells, but can also be used, for example, force sensors.

[0164] In the floor plate moving device 100 of the above embodiment, the load detection unit 10 may be a plurality of pressure sensors arranged in a matrix shape beneath the sheet.

[0165] The load detectors 11 to 14 may be integrated with the bed BD or detachably combined to form a bed system consisting of the bed BD and the floor plate moving device 100 of this embodiment.

[0166] In the present invention, there is no need to strictly distinguish between the phrase "above the threshold" and the phrase "greater than the threshold," nor is there a need to strictly distinguish between the phrase "below the threshold" and the phrase "less than the threshold (less than the threshold)."

[0167] In the above embodiment, the use of the floorboard moving device 100 with respect to a bed 500 was described as an example, but the use of the floorboard moving device 100 is not limited to a bed 500. The floorboard moving device 100 can be used as a support surface moving device to control the movement of the support surface of any support device in which the support surface that supports the human body is configured to be movable, such as a bed, sofa, chair, or wheelchair. In this case, the support surface moving device moves the upper body support portion of the support surface that supports the upper body of the user on the support surface.

[0168] The above embodiments and their respective modifications can be used in combination as appropriate.

[0169] As long as the features of the present invention are maintained, the present invention is not limited to the embodiments described above, and other forms conceivable within the scope of the technical idea of ​​the present invention are also included within the scope of the present invention.

[0170] 10: Load detection unit, 30: Control unit, 31: Load signal acquisition unit, 32: Body movement determination unit, 33: Heart rate information acquisition unit, 34: Sleep stage determination unit, 35: Sleep determination unit, 36: Bed base movement control unit, 40: Memory unit, 50: Display unit, 60: Notification unit, 70: Input unit, 500: Bed, CF1: Sequential 2-class classifier, CF11-CF14, CF2: 2-class classifier, U: User

Claims

1. A support surface movement control device for moving a support surface that supports the body of a user, comprising: a sleep stage determination unit that determines the user's sleep stage based on the user's heart rate interval on the support surface; a sleep determination unit that determines whether or not the user is in a sleep state based on the user's heart rate interval on the support surface; and a movement control unit that performs support surface movement control to control the movement of the support surface, wherein the support surface has an upper body support unit that supports the upper body of the user on the support surface, the sleep stage determination unit determines the user's sleep stage based on the classification results of a first classifier that classifies the user's state into a wakeful state and a sleep state using a feature quantity based on the heart rate interval as an explanatory variable, and a second and third classifier that classifies the user's state into first non-REM sleep, second non-REM sleep, and third non-REM sleep using a feature quantity based on the heart rate interval as an explanatory variable, and the sleep determination unit determines whether or not the user is in a sleep state based on the classification results of a fourth classifier that classifies the user's state into a wakeful state and a sleep state using a feature quantity based on the heart rate interval as an explanatory variable. The first classifier and the fourth classifier are constructed such that the probability of misclassifying the user's state as sleep when the user is awake is smaller in the fourth classifier than in the first classifier, and the movement control unit is a support surface movement control device that, in the support surface movement control, initiates a downward movement that reduces the inclination angle of the upper body support based on the determination result of the sleep stage determination unit and the determination result of the sleep determination unit.

2. The support surface movement control device according to claim 1, wherein the sleep stage determination unit determines the user's sleep stage for each of a plurality of periods based on the user's heart rate interval on the support surface; the sleep determination unit determines whether the user is in a sleep state for each of the plurality of periods based on the user's heart rate interval on the support surface; the movement control unit performs a first descent control in the support surface movement control; the first descent control includes: the sleep determination unit determining that the user is in a sleep state for at least A consecutive periods (A is an integer of 2 or more) that are included in the plurality of periods and include the target period; the sleep stage determination unit determining that the user's sleep stage is third non-REM sleep for the target period; and the average of the determination results of the sleep stage determination unit for at least B consecutive periods (B is an integer of 2 or more) that are included in the plurality of periods and include the target period is deeper than second non-REM sleep.

3. The support surface movement control device according to claim 2, further comprising a body movement determination unit that determines whether or not body movement has occurred in the user based on fluctuations in the load applied by the user to the support surface, wherein the movement control unit, in the support surface movement control, executes a second descent control instead of the first descent control if the elapsed time from the start of the support surface movement control exceeds a first elapsed time but the descent movement of the upper body support has not started, and the movement control unit, in the second descent control, determines based on the determination result of the body movement determination unit that no body movement has occurred in the user for a first predetermined period of time, and initiates the descent movement of the upper body support.

4. The support surface movement control device according to claim 3, wherein the body movement determination unit determines whether or not body movement has occurred in the user for each of a plurality of periods, and the movement control unit determines that no body movement has occurred in the user for the first predetermined period based on the fact that the body movement determination unit has determined that no body movement has occurred in the user for a predetermined number of consecutive periods.

5. The support surface movement control device according to claim 3, wherein the body movement determination unit determines whether or not body movement is occurring in the user based on the magnitude of the fluctuation in the load applied by the user to the support surface, and the movement control unit determines that no body movement has occurred in the user over the first predetermined period based on a comparison between the average value of the magnitude of the load fluctuation over the first predetermined period and a threshold value.

6. The support surface movement control device according to any one of claims 3 to 5, wherein, in the support surface movement control, if the downward movement of the upper body support has not started even after a second elapsed time, which is longer than the first elapsed time, the movement control unit executes a third downward control instead of the second downward control, and in the third downward control, the movement control unit determines, based on the determination result of the body movement determination unit, that no body movement has occurred in the user for a second predetermined period which is shorter than the first predetermined period, and initiates the downward movement of the upper body support.

7. A support surface movement control device for moving a support surface that supports the body of a user, comprising: a state determination unit including at least one of a sleep stage determination unit that determines the user's sleep stage based on the user's heart rate interval on the support surface, a sleep determination unit that determines whether or not the user is in a sleep state based on the user's heart rate interval on the support surface, and a body movement determination unit that determines whether or not body movement has occurred in the user based on fluctuations in the load applied by the user to the support surface; and a movement control unit that performs support surface movement control to control the movement of the support surface, wherein the support surface has an upper body support portion that supports the upper body of the user on the support surface, and the movement control unit, in the support surface movement control, performs a first descent control to start a descent movement that reduces the inclination angle of the upper body support portion based on the determination result of the state determination unit that satisfies a first condition, and if the descent movement of the upper body support portion has not started even after the elapsed time from the start of the support surface movement control exceeds a first elapsed time, it performs a second descent control to start the descent movement of the upper body support portion based on the determination result of the state determination unit that satisfies a second condition. A support surface movement control device wherein the first condition and the second condition are, respectively, one and the other of the determination result of the sleep stage determination unit and the conditions relating to the determination result of the sleep determination unit and the conditions relating to the determination result of the body movement determination unit, or both the first condition and the second condition are the same one of the determination result of the sleep stage determination unit and the conditions relating to the determination result of the sleep determination unit and the conditions relating to the determination result of the body movement determination unit, and the second condition is a relaxed version of the first condition.

8. The state determination unit comprises at least the sleep stage determination unit and the sleep determination unit, wherein the sleep stage determination unit determines the user's sleep stage for each of a plurality of periods based on the user's heart rate interval on the support surface, the sleep stage determination unit determines the user's sleep stage based on the classification results of a first classifier that classifies the user's state into a wakeful state and a sleep state using a feature quantity based on the heart rate interval as an explanatory variable, and a second and third classifier that classifies the user's state into first non-REM sleep, second non-REM sleep, and third non-REM sleep using a feature quantity based on the heart rate interval as an explanatory variable, the sleep determination unit determines whether the user is in a sleep state for each of the plurality of periods based on the user's heart rate interval on the support surface, and the sleep determination unit determines whether the user is in a sleep state based on the classification results of a fourth classifier that classifies the user's state into a wakeful state and a sleep state using a feature quantity based on the heart rate interval as an explanatory variable. The first classifier and the fourth classifier are constructed such that the probability of misclassifying the user's state as a sleep state when the user is awake is smaller in the fourth classifier than in the first classifier, and the movement control unit, in each of the first descent control and the second descent control, starts the downward movement of the upper body support when the sleep determination unit determines that the user's state is a sleep state for at least A consecutive period (A is an integer of 2 or more) that is included in the plurality of periods and includes the target period, the sleep stage determination unit determines that the user's sleep stage is third non-REM sleep during the target period, and the average of the determination results of the sleep stage determination unit for at least B consecutive periods (B is an integer of 2 or more) that are included in the plurality of periods and include the target period is deeper than second non-REM sleep, and A in the second descent control is smaller than A in the first descent control, and B in the second descent control is smaller than B in the first descent control.

9. The support surface movement control device according to claim 7, wherein the state determination unit has the body movement determination unit, which determines whether or not body movement has occurred in the user based on fluctuations in the load applied by the user to the support surface, the movement control unit starts the downward movement of the upper body support unit in the first downward control based on the determination by the body movement determination unit that no body movement has occurred in the user for a first predetermined period of time, and the movement control unit starts the downward movement of the upper body support unit in the second downward control based on the determination by the body movement determination unit that no body movement has occurred in the user for a second predetermined period of time shorter than the first predetermined period.

10. The support surface movement control device according to claim 9, wherein, in the support surface movement control, if the downward movement of the upper body support portion has not started even after a second elapsed time, which is longer than the first elapsed time, the movement control unit executes a third downward control instead of the second downward control, and in the third downward control, the movement control unit determines, based on the determination result of the body movement determination unit, that no body movement has occurred in the user for a third predetermined period which is shorter than the second predetermined period, and initiates the downward movement of the upper body support portion.

11. The support surface movement control device according to any one of claims 7 to 10, wherein in the support surface movement control, if the period during which the user is away from the support surface exceeds a predetermined period or if an upward movement is performed that increases the inclination angle of the upper body support, the elapsed time from the start of the support surface movement control is reset and the first downward control is executed.

12. A support surface movement control device for moving a support surface that supports the body of a user, comprising: a body movement determination unit that determines whether or not body movement has occurred in the user based on fluctuations in the load applied by the user to the support surface; and a movement control unit that performs support surface movement control to control the movement of the support surface, wherein the support surface has an upper body support portion that supports the upper body of the user on the support surface, and the movement control unit, in the support surface movement control, starts the downward movement of the upper body support portion based on the determination by the body movement determination unit that no body movement has occurred in the user for a predetermined period of time.

13. The support surface movement control device according to claim 12, wherein the body movement determination unit determines whether or not body movement has occurred in the user for each of a plurality of target periods, and the movement control unit determines that no body movement has occurred in the user over the predetermined period based on the fact that the body movement determination unit has determined that no body movement has occurred in the user for a predetermined number of consecutive target periods.

14. The support surface movement control device according to claim 12, wherein the body movement determination unit determines whether or not body movement is occurring in the user based on the magnitude of fluctuations in the load applied by the user to the support surface, and the movement control unit determines that no body movement has occurred in the user over the predetermined period based on a comparison between the average value of the magnitude of fluctuations in the load over the predetermined period and a threshold value.

Images