Method for controlling and adjusting bedding, bed system, program, and recording medium.

By integrating user feedback and machine learning, the method optimizes bedding firmness and height to improve sleep quality based on individual assessments, addressing the limitations of existing technologies.

JP2026182723APending Publication Date: 2026-11-19川瀬 信夫
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
JP2025279693
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-08
Filing Date
2025-12-23
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Existing bedding technologies do not effectively address the subjective quality of sleep experienced by users, lacking methods to adjust firmness and height based on individual sleep quality assessments.

Method used

A method involving self-assessment of sleep quality, combined with machine learning and statistical processing, to adjust the local height and hardness of bedding in response to user feedback, using sensors and actuators to optimize sleep quality.

Benefits of technology

The method provides bedding adjustments tailored to individual user needs, enhancing sleep quality by correlating with subjective user evaluations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2026182723000001_ABST
    Figure 2026182723000001_ABST
Patent Text Reader

Abstract

To enable the maintenance and improvement of the user's sleep quality, and to provide a method for controlling and adjusting bedding, a bed system, a program, and a recording medium for that purpose. [Solution] Using a bed / mattress capable of acquiring local height and local hardness data and adjusting these parameters via external electrical signals, time-series data of changes in local height and local load of the bed / mattress in response to changes in sleeping posture is acquired. This data is then used to learn and classify time-series patterns, and the results are statistically processed to control the bed / mattress based on time-series patterns that correlate highly with the user's evaluation of sleep quality. By repeating this process, a high quality of sleep that provides satisfaction to the user is provided.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention generally relates to a method for improving sleep quality, and particularly relates to, for example, a bedding control and adjustment method, a bed system, a program, and a recording medium. In the present invention, bedding refers to the entire body support, which is a general term for all equipment and appliances that can be typically referred to as sleeping bedding, body supports for rest and relaxation, relaxation furniture, body pressure dispersion tools, and the like, and is not limited to those used only at bedtime, and may include supports that support and assist the body in an active exercise state used in gym facilities for exercise and the like.

Background Art

[0002] As this type of technical idea, for example, Patent Document 1 describes a bedding that includes a large number of support point elements capable of adjusting the height value of the support points, a support element for supporting the body of a supported person, and an adjustment element for adjusting at least one value of at least one of the large number of support point elements according to at least one of a change in posture and a change over time, as well as an invention that automatically adjusts the sleeping position and monitors sleep and wake habits, body temperature, heart rate, respiratory rate, and the like.

Prior Art Literature

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Patent Document 3

Patent Document 4

Summary of Invention

Problems to be Solved by the Invention

[0004] Patent Document 1 describes bedding comprising a support element that supports the body of the person being supported, with a large number of support point elements whose height values ​​can be adjusted, and an adjustment element that adjusts the value of at least one of the large number of support point elements in response to changes in posture and changes in their time. It also describes an invention that automatically adjusts the sleeping position and monitors sleep and wakefulness habits, body temperature, heart rate, respiratory rate, etc. However, it does not describe a direct relationship with the quality of sleep perceived by the user, nor does it describe an adjustment method based on the user's own sleep quality.

[0005] Patent Document 2 describes an invention of bedding that includes a support element that supports the body of the person being supported, comprising a large number of support point elements whose hardness and height values ​​can be adjusted, and an adjustment element that adjusts the value of at least one of the large number of support point elements in response to changes in at least one of postural changes and changes in time, and which is controlled by software using a computer. However, although it is written that such control equalizes and distributes pressure, there is no description of the direct relationship with the quality of sleep that the user feels.

[0006] Patent Document 3 describes an invention of bedding that includes a support element that supports the body of the person being supported by having a large number of support point elements whose height values ​​can be adjusted, and an adjustment element that adjusts the value of at least one of the numerous support point elements in response to changes in at least one of postural changes and changes in time, thereby adjusting various local positions of the bedding based on parameters such as the height, weight, and body type of the human body, and an invention of bedding that can adjust the height of the sleep pad to the most comfortable range. However, there is no specific description of what algorithm is used to perform this adjustment, in particular how the adjustment is performed based on what aspects of the user, and there is no description of the direct relationship with the quality of sleep that the user feels.

[0007] Patent Document 4 deals with the surrounding environment (brightness, scent) as the control target, and does not mention the bed or mattress, which directly affect sleep quality, nor does it describe a direct relationship with the sleep quality perceived by the user. Thus, none of the patent documents provide specific descriptions of how to determine the firmness of the bed or mattress, or the local height to match the body shape, taking into account the sleep quality as evaluated by the user.

[0008] In view of the aforementioned problems, the present invention aims to provide a method for controlling and adjusting bedding, a bed system, a program, and a recording medium that enable the maintenance and improvement of the user's sleep quality. [Means for solving the problem]

[0009] Self-assessment of sleep quality is essential because sleep satisfaction is highly subjective. Therefore, self-assessment and the setting and control of the bed and mattress based on that assessment are indispensable for achieving high sleep quality. This invention was made with this in mind.

[0010] In solving the above problems, the inventor focused on a means of automatically adjusting the bed and mattress in response to changes in sleeping posture over time, by allowing the user to provide feedback on sleep quality to the bed and mattress to adjust it, thereby maintaining and improving the user's sleep quality. Specifically, the inventor obtained time-series data of changes in the local height and local hardness of the bed and mattress in response to changes in sleeping posture from a bed and mattress that can acquire data on local height and local hardness and can adjust these by external electrical signals. This data is then used to learn and classify time-series patterns using machine learning, and the results are statistically processed to control the bed and mattress based on time-series patterns that have a high correlation with the user's evaluation of sleep quality. By repeating this process, it is believed that a high quality of sleep that gives the user satisfaction can be provided.

[0011] The present invention provides a means to improve the quality of sleep tailored to each individual user by adjusting the local height and hardness of the bed mattress according to the user's needs. This involves acquiring time-series data of the temporal changes in local height and local weight during sleep, relative to the initial state of the bed mattress's local height and local hardness, and obtaining the user's evaluation of sleep quality. The local height and local hardness are then adjusted to be optimal based on the user's evaluation. To achieve this, machine learning and statistical processing are combined, with time-series data from height sensors and / or weight sensors as input data, and time-series data that allows for setting the moment-by-moment height and hardness for each local part of the bed or mattress as output data. The patterns of time-series data obtained from the height sensors and / or weight sensors are automatically classified, and the correlation between these patterns and the user's own sleep evaluation is determined. The time-series pattern that gives the highest correlation is selected as the time-series pattern that provides higher sleep quality. The height and / or hardness of each local part of the bed mattress are controlled using this selected time-series pattern. By repeating this process, the bed mattress is controlled to provide a high quality of sleep that gives the user satisfaction.

[0012] More specifically, in order to solve the above problems, the first aspect of the present invention provides a method for controlling and adjusting bedding, characterized in that, for a bed or mattress having a divided upper surface which is an actuator mechanism that independently electronically adjusts the height and / or hardness based on an external signal, and a height sensor and / or weight sensor capable of measuring height and / or weight, the method acquires moment-by-moment height and / or weight measurements from each sensor during sleep, and after waking up, the user provides their own satisfaction level with sleep, thereby setting moment-by-moment height and / or hardness for each local part of the bed and / or mattress in accordance with the user's physical movements during sleep.

[0013] In a second aspect of the present invention, in the first aspect, the localized setting of the new height and hardness values ​​of the bed or mattress may be set once when going to sleep, set multiple times during sleep, or a time-varying setting value may be provided that corresponds to continuous physical movements during sleep.

[0014] In a third aspect of the present invention, in the first embodiment, the calculation and determination of the local height and hardness of the top surface of the bed or mattress may be performed by a machine learning method, a deterministic method based on statistical processing, or a combination of both.

[0015] In a fourth aspect of the present invention, in the first embodiment, the user may directly set the height adjustment mechanism and / or hardness adjustment mechanism for each or divided area of ​​the bed or mattress via a control computer to calculate and determine the local height and hardness of the upper surface of the bed or mattress.

[0016] In a fifth aspect of the present invention, in the first embodiment, the input data may be time-series data from a height sensor and / or a weight sensor, either by machine learning or by statistical processing.

[0017] In a sixth aspect of the present invention, in the first embodiment, the initial values ​​of height and / or hardness may be set using the user as the subject, or using average data such as age, gender, and weight, or they may be set so that the pressure on each part of the body is the same.

[0018] In a seventh aspect of the present invention, in the first embodiment, a machine learning method may be used to input multiple time-series data as patterns, classify them into multiple time-series patterns, and determine the correlation with sleep quality as evaluated by the user, thereby obtaining a time-series pattern of local height and / or hardness of a bed or mattress that provides better sleep quality, and providing this as a control value to the height adjustment mechanism and / or hardness adjustment mechanism of the bed or mattress.

[0019] As an eighth aspect of the present invention, in the first aspect, in the method by statistical processing, the correlation between the pattern of the entire plurality of time-series data and the user's sleep evaluation may be calculated, a time-series pattern having a high correlation with high sleep quality may be extracted, and the extracted pattern may be used as a control value to be supplied to the height adjustment mechanism and / or the firmness adjustment mechanism of the bed or mattress.

[0020] As a ninth aspect of the present invention, in the first aspect, a sleeping posture including supine, lateral, and prone positions may be detected from the values of the height sensor and / or the weight sensor, and control based thereon may be performed.

[0021] As a tenth aspect of the present invention, in the first aspect, by calculating the temporal change (velocity and acceleration) of the signal from the weight sensor, the intensity and quickness of physical movements during sleep including tossing and turning may be calculated, and control for providing high sleep satisfaction may be performed.

[0022] As an eleventh aspect of the present invention, in the first aspect, preferences for the local height and firmness of the upper surface of the bed or mattress, for example, high / low, soft / firm, etc., may be specified by either a method based on machine learning or a method based on statistical processing.

[0023] As a twelfth aspect of the present invention, in the first aspect, the set value of firmness may be calculated based on the corresponding value of the height sensor and the value of the weight sensor.

[0024] As a thirteenth aspect of the present invention, in the first aspect, calculation for obtaining new adjustment mechanism control data may be performed using feedback from the user as a trigger, whether by a method based on machine learning or a method based on statistical processing.

[0025] As a fourteenth aspect of the present invention, in the first aspect, the values from the respective sensors may be handled individually, or may be grouped, and the average value within the group may be used as the grouped sensor value.

[0026] In a fifteenth aspect of the present invention, in the first embodiment, REM sleep and non-REM sleep may be determined based on time-varying data acquired from a height sensor and / or a weight sensor, and this may be used as an evaluation value for sleep quality in combination with the user's self-assessment of sleep quality.

[0027] In a sixteenth aspect of the present invention, in the first embodiment, dreams may be classified and used to evaluate the quality of sleep.

[0028] In a 17th aspect of the present invention, in the first aspect, the quality of sleep may be evaluated by taking into account the time of waking up and going to bed in the middle of the night, and the time of waking up and going to bed.

[0029] In an eighteenth aspect of the present invention, in the first embodiment, the type of activity on the previous day, including desk work and fieldwork, and indicators of activity on the previous night, including drinking, dinner time, and bath time, may be added to the quality of sleep.

[0030] In a 19th aspect of the present invention, in the first embodiment, learning and statistical processing may be performed that takes into account seasonal variations (sunlight, sunset, sunrise) and room environment (temperature, humidity).

[0031] In a 20th aspect of the present invention, the first embodiment may also be used as an exploratory means for obtaining higher quality sleep by adding a slight perturbation to the determined control pattern.

[0032] In a 21st aspect of the present invention, in the first embodiment, a perturbed control pattern may be used to perform control that is adapted to changes in the user's sleep due to aging.

[0033] In a 22nd aspect of the present invention, in the first embodiment, the comfort of sleeping may be evaluated rather than the quality of sleep in beds and mattresses for purposes other than sleep.

[0034] Furthermore, in order to solve the above problems, a bed system according to a 23rd aspect of the present invention comprises a bed housing, a mattress disposed on the upper part of the bed housing, a sensor disposed between the bed housing and the mattress capable of measuring local height and / or load values, an actuator disposed between the bed housing and the mattress capable of adjusting local height and / or hardness with an external signal, and a control mechanism for controlling the actuator. The control mechanism is characterized in that, a sensor signal obtained from the sensor is input to a feature extraction layer, the feature extraction layer performs segmentation and Fourier transform of the input time series data to extract features from the time series data, the features of the obtained time series pattern are input to a convolutional neural network layer, the temporal relationship of the features obtained in the convolutional neural network layer is learned in a recurrent neural network layer, the importance of the time series pattern is taken into account in an attention layer, the importance of the time series pattern is evaluated, classified and output in an output layer, and an external signal for adjusting the actuator is generated based on the output of this output layer.

[0035] Furthermore, in order to solve the above problems, a program according to a 24th aspect of the present invention is a program for causing a computer to function as a control mechanism for controlling the actuator in a bed device comprising a bed housing, a mattress disposed on the upper part of the bed housing, a sensor disposed between the bed housing and the mattress capable of measuring local height and / or load values, and an actuator disposed between the bed housing and the mattress capable of adjusting local height and / or hardness with an external signal, wherein the control mechanism is characterized in that a sensor signal obtained from the sensor is input to a feature extraction layer, the feature extraction layer performs segmentation and Fourier transform of the input time series data to extract features from the time series data, the features of the obtained time series pattern are input to a convolutional neural network layer, the temporal relationship of the features obtained in the convolutional neural network layer is learned in a recurrent neural network layer, the importance of the time series pattern is taken into account in an attention layer, the importance of the time series pattern is evaluated, classified and output in an output layer, and an external signal for adjusting the actuator is generated based on the output of this output layer.

[0036] A 24th aspect of the present invention may be implemented as a recording medium storing a program according to the 24th aspect. [Effects of the Invention]

[0037] According to each aspect of the present invention, it is possible to adjust the firmness and localized height of the bed mattress to improve sleep quality based on the user's own evaluation, thereby providing a bed mattress that achieves sleep highly correlated with the user's subjective sleep quality. Furthermore, it is possible to adjust the firmness and localized height of the bed mattress in accordance with the user's changes in sleep over time. [Brief explanation of the drawing]

[0038] [Figure 1]As an example of a height and firmness adjustable bed mattress 100 according to one embodiment of the present invention, this is a cross-sectional view of such bed mattress 100. [Figure 2] This is a block diagram showing the hardware configuration of a computer system 200 for controlling a height and firmness adjustable bed mattress 100 according to one embodiment of the present invention. [Figure 3] This is a conceptual diagram illustrating a method for automatically classifying time-series data of sensor values ​​from a bed mattress 100 using machine learning means according to one embodiment of the present invention. [Figure 4] This is a conceptual diagram illustrating a method for controlling the local height and / or hardness of a bed mattress using candidate time-series setting values ​​obtained using machine learning means according to one embodiment of the present invention as time-series setting values ​​or candidate values. [Modes for carrying out the invention]

[0039] Embodiments of the present invention will be described below with reference to the drawings. In the following, the scope necessary to achieve the objectives of the present invention will be schematically shown, and the explanation will mainly focus on the relevant parts of the present invention, with any omitted parts being covered by prior art.

[0040] Figure 1 is a cross-sectional view of a bed mattress 100 with adjustable height and firmness, as an example of one embodiment of the present invention. The right-hand view in the same figure is an enlarged portion of the left-hand view. The methods shown here are those described in or similar to those described in Reference Documents 1, 2, 3, and 4, but the present invention is not limited to these methods and may encompass any bed mattress that can obtain height and load signals and whose height and firmness can be adjusted by external signals.

[0041] Using the same figure, the structure of the bed mattress 100 to which one embodiment of the present invention applies will be explained. As shown in the figure, in the height and hardness adjustable bed mattress 100 according to one embodiment of the present invention, multiple sensor actuators 130 capable of measuring local height and / or load values, and whose local height and / or hardness can be adjusted by an external signal, are arranged between the bed housing 110 and the mattress 120, and the structure has such that the sleeping posture that changes over time can be adjusted by the actuators. The coordinates of the sensor actuators are fixed by the support columns 140. Signals from the sensors (hereinafter referred to as "sensor signals") and signals that control the actuators (hereinafter referred to as "control signals") are exchanged with the outside via connection cables 160 through the control mechanism interface 150. The sensor actuators 130 and the control mechanism interface 150 are connected via control mechanism interface connection cables 141. In the bed housing 110 of Figure 1, only three control mechanism interface connection cables 141 are shown. In this embodiment, the sensor-actuator 130 is shown as having both sensor and actuator functions. However, instead of these configurations, or in addition to these configurations, the sensor and actuator may be provided as separate components rather than as a single unit.

[0042] In this embodiment, conventional techniques (ideas) for adjusting hardness can also be incorporated. Specifically, for example, any of the following forms, or a combination of several such forms, can be adopted. <First form> The system uses air pressure (air cells / multi-chamber) to change the firmness. In other words, the support points are air cells (small airbags), and the support reaction force (firmness) is continuously variable by changing the internal pressure. Multiple chambers are suitable for configurations that have "many support points". More specifically, for example, the following can be adopted: 1) Zoning (by body part): A system that changes the pressure setting for each area such as the shoulders, waist, and buttocks to adjust the firmness according to the body pressure distribution and sleeping posture; 2) Active control (alternating pressure): A system that periodically changes the firmness and pressure of the support points by inflating and deflating multiple cells over time; 3) Multi-cell precision control: A system that controls individual cells in a closed loop using solenoid valves + pumps + pressure sensors. <Second form> The stiffness will be changed using a mechanical method (spring, slat, cam / variable preload). In other words, the support point = spring / slat (leaf spring) / link mechanism, and the spring constant itself or the effective spring constant will be changed mechanically. More specifically, for example, the following methods can be adopted: 1) Variable preload: A motor changes the initial compression amount of the spring, so that the reaction force changes even with the same displacement (shifting the range in which it "feels stiff"), 2) Variable point of application / lever ratio: A method that changes the effective stiffness by moving the pivot point position of the link (cam plate, eccentric mechanism, etc.), 3) Variable slat support conditions: A method that changes the bending stiffness (perceived stiffness) by changing the support spacing of the slats or the position of the support blocks, and 4) Clutch-type step switching: A method that switches the stiffness in steps by connecting / disconnecting multiple springs. <3rd form> The system uses a fluid (hydraulic / hydrostatic / gel) to change the stiffness. In other words, the support points are fluid cells, and the "ease of sinking (dynamic stiffness)" is controlled by flow resistance (orifice) and the opening and closing of partitions. More specifically, for example, 1) passive type (self-equilibrium): a system that distributes body pressure through fluid movement (distributing rather than "varying" the stiffness), and 2) semi-active / active type: a system that uses valves or pumps to transfer fluid between cells and adjust pressure to match the stiffness of different parts of the body. <4th form> The method involves changing the material rigidity itself using smart materials. Specifically, it involves combining support elements with materials whose viscoelasticity changes in response to magnetic and electric fields, thereby changing the material properties (elastic modulus and yield stress) and making the hardness variable. More specifically, for example, 1) a method using MR (magnetic fluid): a group of materials whose rheological properties change significantly in a magnetic field and which are being put into practical use in dampers / clutches, etc., and 2) a method in which the support points are "MR fluid-filled springs" or "MR fluid damper-equipped support points" to simultaneously vary "hardness + damping". <5th form> The method involves changing the hardness through jamming (particle / layer structure) (increasing rigidity with vacuum). In other words, by changing the degree of vacuum, the system can switch between soft and hard by using particles or layer material within a membrane (bag) as support points. More specifically, for example, 1) a method using the "variable rigidity" principle where particles are constrained by vacuum and their rigidity increases as particle jamming, and 2) a method that increases hardness locally by creating a matrix of small jamming cells (multiple support points) as support points. <6th form> This method utilizes time-dependent (periodic) changes in "stiffness." In other words, it's not just about "adjusting the stiffness to a constant value," but also about varying it over time. More specifically, it's a method that reduces "sustained local pressure" by periodically changing the support points through alternating pressure, thereby achieving effects such as pressure ulcer prevention.

[0043] Figure 2 is a block diagram showing the hardware configuration of a computer system (control mechanism) 200 for controlling a bed mattress 100 according to one embodiment of the present invention. As shown in the figure, the computer system (control mechanism) 200 is configured to include, for example, a CPU (computer main unit) 210, a display device 220, an HDD 230, a keyboard or keypad 240, and a bed mattress system interface 250. The bed mattress 100 is controlled by a control program running on the CPU (computer main unit) 210. The sensor signal is a local height value and / or weight value from the bed mattress 100 (hereinafter referred to as "sensor value"), and the output control signal is a local height setting value and / or hardness setting value (hereinafter referred to as "setting value"), which is sent to the control mechanism interface 150 of the bed mattress 100 via the bed mattress system interface 250, and the local height and / or hardness of the bed mattress is set by this setting value.

[0044] After waking, the user performs sleep evaluations, etc., via the keyboard or keypad 240. In one embodiment of the present invention, the optimal local height and / or hardness, which changes moment by moment, is set by the method described below. The display 220 (hereinafter also referred to as "display device 220") is used for various settings, including sensor values ​​and set values ​​during sleep, and the user's bed and mattress preferences. These control programs, as well as past sensor signals and control signals, are recorded in the HDD 230 and displayed on the display device 220 as needed.

[0045] Figure 3 is a conceptual diagram showing a method for automatically classifying time-series data of sensor values ​​from a bed mattress 100 using machine learning means according to one embodiment of the present invention. This method uses the method of Yamamuro et al. (Summarization and Classification of Time-Series Data Using Deep Learning, DEIM Forum 2018, C3-3) and calculates importance from multiple time-series sensor values. The above report by Yamamuro et al. describes a method for basketball free throws using 3-axis accelerometers attached to the elbow and knee of the dominant hand, and a glasses-type IoT device. By analyzing time-series data capturing free throws and eye movements using a device, the system automatically identifies time-series data such as "releasing the ball" and "routine work" from a given time-series sequence, and successfully learns the data important for classifying the success or failure of the shot.

[0046] The following outlines the operation of one embodiment of the present invention. Similar to the free throw example by Yamamuro et al., in the control of bed mattresses during sleep, not all sensor signals from bed mattresses 100, and not all of their sensor signals at all times, are important; there are parts and situations where they are not important. This is because some sensor signals contribute significantly to sleep, while others do not. To achieve high-quality sleep, it is first necessary to classify the sensor signals and extract the time-series data portions that are important for high-quality sleep. A machine learning device 300 is used for this purpose. The input to the machine learning device 300 is the sensor signals 310 from the bed mattresses, and the output is classification "1 (circled character)" to "5 (circled character)". The five classifications are just an example and are not limited to them. The sensor signals that have been classified and extracted as important are subsequently used for controlling the bed mattresses 100.

[0047] To automatically classify sensor signals, the method of Yamamuro et al. is executed in the machine learning device 300. First, the input time-series signal is divided into segments, and a Fourier transform is performed in the feature extraction layer 320 to divide the time-series sensor signal into patterns and extract features. In the case of sleep, this division is classified into patterns based on whether there is little or significant body movement, or which part of the body is moving most intensely. Next, the convolutional neural network layer 330 extracts the relationships between sensors in each of the divided features. Then, the recurrent neural network layer 340 performs feature extraction on signals with temporal relationships. Finally, the attention layer 350 calculates the importance α of the segments using attention. Attention is calculated using the output vector of the recurrent neural network layer 340 based on the method of Yamamuro et al. Linear coupling is performed on the output of the recurrent neural network layer 340 to obtain the output, and the classifier 360 outputs the classification result. Further details can be found in the paper by Yamamuro et al. (Summarization and Classification of Time Series Data Using Deep Learning, DEIM Forum 2018, C3-3).

[0048] The following is an example of bed / mattress control based on user evaluation. The sensor signal 310 shows an example of time-series data of local setting values ​​at multiple locations on the bed / mattress 100. These are input to the feature extraction layer 320 of the machine learning device 300. The feature extraction layer 320 performs segmentation and Fourier transform on the input time-series data and extracts features from the time-series data using multidimensional time-series data, etc. Here, segmentation that divides the time-series data into time domains may also be performed. The obtained time-series pattern features are input to the convolutional neural network layer 330, and the relationship between each sensor value 310 for each feature is extracted. Next, the recurrent neural network layer 340 learns the temporal relationships of the features obtained in the convolutional neural network layer 330. Next, the attention layer 350 takes into account the importance of the time-series pattern. The output layer 360 evaluates and classifies the importance of time series patterns and outputs them (for more details, see "Summarizing and Classifying Time Series Data Using Deep Learning," DEIM Forum 2018, C3-3).

[0049] Next, refer to Figure 4. Figure 4 is a conceptual diagram showing a method for controlling the local height and / or hardness of a bed mattress using candidate time-series setting values ​​obtained using machine learning means according to one embodiment of the present invention as time-series setting values ​​or candidate values. As shown in the figure, the correlation coefficient calculation means 410 calculates the correlation coefficient between the segmented time-series data group 420 and the user's evaluation value 430, and outputs the correlation coefficient 440. Based on this, the time-series data with high evaluation values ​​is extracted from the time-series pattern by the highest correlation coefficient holding pattern selection unit 450, and this is used as a candidate time-series setting value for obtaining high-quality sleep. The local height and / or hardness of the bed mattress 100 is controlled using the time-series setting value candidate obtained in this way as time-series setting values ​​or candidate values. The user evaluates the quality of sleep for the new setting values ​​upon waking the next morning, and the above is repeated. This repetition makes it possible to provide the user with higher quality sleep. As an initial state of control, it is also possible to start with control that makes the values ​​of the weight sensor as similar as possible, that is, control that makes the pressure on each part of the body uniform. [Explanation of symbols]

[0050] 100 Bed and Mattress Systems 110 bed cabinet 120 mattress 130 Height / Hardness Sensor / Adjustable Actuator 140 Post 141 Control mechanism interface connection cable 150 Control Mechanism Interface 160 connection cables 200 Control mechanism 210 CPU (computer unit) 220 Display device 230 HDD 240 Keyboards & Keypads 250 Bed and Mattress System Interface 300 Machine Learning Devices 310 Input time series data 320 Feature Extraction Layers 330 Convolutional Neural Network Layers 340 Recurrent Neural Network Layers 350 Attention layer 360 output layers 400 Optimal Control Pattern Determination Device 410 Correlation Coefficient Calculation Unit 420 Time Series Patterns 430 User ratings 440 Correlation coefficients between time series patterns and evaluation values 450 Selection section for patterns with the highest correlation coefficient [Industrial applicability]

[0051] Each aspect of the present invention contributes broadly to improving sleep quality and is not limited to use in ordinary households. It provides significant value not only to residents but also to staff in hospitals and care facilities for the elderly and people with physical disabilities (especially those who are bedridden). In addition, by providing high-quality sleep to healthy individuals of all ages who are actively engaged in social activities, it promotes improvements in the quality and quantity of their daily activities and supports further contributions to social development.

Claims

1. A method for controlling and adjusting bedding, characterized in that, for a bed or mattress having a divided upper surface which is an actuator mechanism that independently electronically adjusts the height and / or hardness based on an external signal, and a height sensor and / or load sensor capable of measuring height and / or load, the method acquires moment-by-moment height and load measurements from each sensor during sleep, and after waking up, the user provides their own sleep satisfaction, thereby setting the moment-by-moment height and / or hardness for each local part of the bed and / or mattress in accordance with the user's physical movements during sleep.

2. The method for controlling and adjusting bedding according to claim 1, wherein the localized settings of the new height and hardness values ​​of the bed or mattress may be set once when going to sleep, set multiple times during sleep, or set to a time-varying value that corresponds to continuous physical movements during sleep.

3. The method for controlling and adjusting bedding according to claim 1, wherein the calculation and determination of the local height and hardness of the top surface of the bed or mattress may be by a machine learning method, a deterministic method based on statistical processing, or a method combining both.

4. The bedding control and adjustment method according to claim 1, wherein the user themselves may directly set the height adjustment mechanism and / or hardness adjustment mechanism for each or divided area of ​​the bed or mattress via a control computer to calculate and determine the local height and hardness of the top surface of the bed or mattress.

5. The bedding control and adjustment method according to claim 1, characterized in that the input data is time-series data from a height sensor and / or a weight sensor, whether by machine learning or statistical processing.

6. The bedding control and adjustment method according to claim 1, wherein the initial values ​​of height and / or hardness may be set using the user as a subject, or using average data such as age, gender, and weight, or the initial values ​​may be set so that the pressure on each part of the body is the same.

7. The bedding control adjustment method according to claim 1, characterized in that a machine learning method inputs multiple time series data as patterns, classifies them into multiple time series patterns, and seeks a correlation with sleep quality as evaluated by the user to obtain a time series pattern of local height and / or hardness of the bed or mattress that provides better sleep quality, and provides this as a control value to the height adjustment mechanism and / or hardness adjustment mechanism of the bed or mattress.

8. The method for controlling and adjusting bedding according to claim 1, characterized in that, using a statistical processing method, the correlation between the patterns of multiple time series data and the sleep evaluation by the user is calculated, time series patterns that correlate highly with high sleep quality are extracted, and these are given as control values ​​to the height adjustment mechanism and / or hardness adjustment mechanism of the bed or mattress.

9. A method for controlling and adjusting bedding according to claim 1, which may detect sleeping positions, including supine, side-lying, and prone, from the values ​​of a height sensor and / or a weight sensor, and perform control based on that.

10. A method for controlling and adjusting bedding according to claim 1, which may calculate the intensity and speed of physical movements during sleep, including turning over, by calculating the time change (velocity and acceleration) of signals from a weight sensor, and perform control to provide high sleep satisfaction.

11. The bedding control and adjustment method according to claim 1, which may specify local height and firmness preferences for the top surface of a bed or mattress, such as high / low, soft / hard, by machine learning or by statistical processing.

12. The bedding control and adjustment method according to claim 1, wherein the hardness setting value may be calculated based on the corresponding height sensor value and the load sensor value.

13. The bedding control adjustment method according to claim 1, characterized in that calculations are performed to obtain new adjustment mechanism control data triggered by user feedback, whether by machine learning or statistical processing.

14. The bedding control and adjustment method according to claim 1, wherein the values ​​from each sensor may be handled individually, or they may be grouped, and the average values ​​within the group may be grouped together to obtain the appropriate sensor values.

15. A method for controlling and adjusting bedding according to claim 1, wherein REM sleep and non-REM sleep are determined based on time-varying data obtained from a height sensor and / or a weight sensor, and these are used as an evaluation value for sleep quality in combination with the user's self-assessment of sleep quality.

16. A method for controlling and adjusting bedding according to claim 1, wherein dreams may be classified and used to evaluate the quality of sleep.

17. The bedding control and adjustment method according to claim 1, which may also evaluate the quality of sleep by taking into account the time of waking up in the middle of the night, the time of going to bed, the time of waking up, and the duration of sleep.

18. The method for controlling and adjusting bedding according to claim 1, which may also take into account the type of activity on the previous day, including desk work and fieldwork, and indicators of activity on the previous night, including drinking, dinner time, and bath time, in order to improve sleep quality.

19. A method for controlling and adjusting bedding according to claim 1, which may also perform learning and statistical processing that takes into account seasonal fluctuations comprising at least one of sunlight, sunset, or sunrise, and room environmental conditions comprising at least one of temperature or humidity.

20. The bedding control adjustment method according to claim 1, which may also be used as an exploratory means for obtaining higher quality sleep by adding a slight tactile element to the determined control pattern.

21. The bedding control adjustment method according to claim 1, which may use a perturbed control pattern to perform control in accordance with changes in the user's sleep due to aging.

22. A method for controlling and adjusting bedding according to claim 1, wherein in beds and mattresses other than those used for sleeping, the comfort of sleeping may be evaluated rather than the quality of sleep.

23. Bed frame and A mattress is placed on top of the aforementioned bed frame, A sensor is disposed between the bed frame and the mattress and is capable of measuring local height and / or load values. An actuator is disposed between the bed housing and the mattress, and whose local height and / or hardness can be adjusted by an external signal. A control mechanism for controlling the actuator and of The control mechanism is provided with, A bed system characterized in that the sensor signal obtained from the sensor is input to a feature extraction layer, the feature extraction layer performs segmentation and Fourier transform on the input time series data to extract features from the time series data, the obtained time series pattern features are input to a convolutional neural network layer, the recurrent neural network layer learns the temporal relationships of the features obtained in the convolutional neural network layer, the Attention layer takes into account the importance of the time series pattern, the output layer evaluates and classifies the importance of the time series pattern and outputs it, and an external signal for adjusting the actuator is generated based on the output of this output layer.

24. A bed device comprising a bed frame, a mattress disposed on top of the bed frame, a sensor disposed between the bed frame and the mattress capable of measuring local height and / or load values, and an actuator disposed between the bed frame and the mattress capable of adjusting local height and / or hardness with an external signal, wherein a program is provided to cause a computer to function as a control mechanism for controlling the actuator, In the control mechanism, A program characterized in that the sensor signal obtained from the sensor is input to a feature extraction layer, the feature extraction layer performs segmentation and Fourier transform on the input time series data to extract features from the time series data, the obtained time series pattern features are input to a convolutional neural network layer, the recurrent neural network layer learns the temporal relationships of the features obtained in the convolutional neural network layer, the Attention layer takes into account the importance of the time series pattern, the output layer evaluates and classifies the importance of the time series pattern and outputs it, and an external signal for adjusting the actuator is generated based on the output of this output layer.

25. A recording medium storing the program described in claim 24.

Citation Information

Patent Citations

  • Method for controlling mattress with adjustment function

    JP2020175115A

  • Sleep state improving system, sleep state improving instrument and sleep state improving program

    JP2024145337A

  • An intelligent lift mattress that balances the force acting on the body

    JP3244473U

  • Automated bed and method of operation thereof

    US20150351982A1