Air Cell Mattress System with Inter-Air Cell Coupling Compensation Control Function

Abstract

An air cell mattress system having a coupling compensation control function between air cells according to the present invention comprises: a zone in which a mattress body equipped with a plurality of air cells is divided into a plurality of regions; a driving unit including a plurality of valves connected to the air cells to perform charging and discharging, and a pump that supplies air to the air cells through the valves; and a pressure sensor unit that measures the pressure of each of the air cells. The controller comprises: an identification execution condition determination module that determines an identification execution condition in which at least one of the pressure change rate and the user's body movement index is maintained for each of the above zones in a state smaller than a stable threshold for a stable threshold time; an identification signal application module that, when the identification execution condition is satisfied, applies an identification signal having an amplitude limited to a user's perceived threshold to an input zone among the plurality of zones, and fixes the pressure control of the driving unit for other zones during the application period of the identification signal and the response stabilization period after the application ends; a combination matrix calculation module that calculates a combination matrix having an element defined as the ratio of the pressure change amount of the output zone to the magnitude of the identification signal applied to the input zone at the time of response extraction when the pressure change rate of the output zone after the application of the identification signal ends is smaller than a response stable threshold; and a combination compensation module that calculates a compensation input to reduce pressure interference between zones using the combination matrix and controls the driving unit with the compensation input.

Description

Technology Field

[0001] The present invention relates to an air cell mattress system having a coupling compensation control function between air cells, and more specifically, to an air cell mattress system that divides a mattress body composed of a plurality of air cells into a plurality of zones, sequentially applies identification signals within a range that does not induce user sensation to each zone to automatically identify a coupling matrix that quantifies pressure interference caused by mechanical coupling between zones, and compensates for pressure interference between zones in real time using a decoupling matrix derived from the identified coupling matrix, thereby precisely tracking the target pressure of each of the plurality of zones. Background Technology

[0002] The air cell mattress system is a system equipped with multiple air cells capable of inflating and defusing air internally, which provides different support pressures to different body parts according to the user's body type, weight, and posture by independently adjusting the air pressure of each air cell.

[0003] Air cell mattress systems are used in medical and home applications for purposes such as preventing bedsores, correcting sleeping posture, and improving sleep quality. Recently, there has been an increasing demand for smart mattress systems that automatically provide optimal support pressure by responding in real-time to changes in the user's body shape and sleeping posture.

[0004] In an air cell mattress system, a proportional-integral controller or a proportional-integral-derivative controller is typically applied independently to each air cell to maintain the pressure of each air cell at a target value.

[0005] However, since multiple air cells are mechanically connected through physical connection structures such as foam, covers, and frames, a change in the pressure of a specific air cell causes the volume and load distribution of adjacent air cells to change in tandem, thereby interfering with the pressure of that air cell.

[0006] Pressure interference caused by mechanical coupling between these air cells is a major cause of the degradation of the independent pressure-following performance of each air cell.

[0007] Conventional air cell mattress systems design controllers by treating each air cell as a completely independent control target without considering pressure interference between air cells; consequently, a problem arises where the pressure of adjacent air cells fluctuates unintentionally during the process of adjusting the pressure of a specific air cell.

[0008] In addition, conventional technology does not have means to quantify the size and direction of the coupling between air cells, so it cannot identify the coupling characteristics between air cells that vary depending on the structure, material, and weight of the mattress in an actual usage environment, and it does not have means to calculate a compensation input to offset pressure interference between air cells and apply it to the drive unit.

[0009] Therefore, it is necessary to develop a novel and advanced air cell mattress system that precisely tracks the target pressure of each of the multiple zones by dividing a mattress body composed of multiple air cells into multiple zones, sequentially applying identification signals limited to below a user’s perceived threshold to the multiple zones to automatically calculate a coupling matrix that quantifies pressure interference caused by mechanical coupling between zones, and calculating a compensation input using a decoupling matrix derived from the calculated coupling matrix and applying it to a driving unit. Prior art literature

[0010] Korean Registered Patent No. 10-2878283 The problem to be solved

[0011] The present invention was devised to overcome the problems of the above technology, and the main purpose of the present invention is to provide a system that precisely tracks the target pressure of each of the multiple zones by dividing a mattress body composed of a plurality of air cells into multiple zones, sequentially applying identification signals within a range that does not induce user sensation to each zone, automatically calculating a coupling matrix that quantifies pressure interference caused by mechanical coupling between zones, and calculating a compensation input in real time using a decoupling matrix derived from the calculated coupling matrix.

[0012] Another objective of the present invention is to monitor the amount of pressure change, the rate of pressure change, and the user's body movement indicators in real time during the application period of the identification signal, and to immediately stop the application of the identification signal when the interruption threshold is exceeded, and to re-apply the identification signal with reduced amplitude, thereby excluding contaminated data and simultaneously achieving the accuracy of the combined matrix estimation and user-perceived protection.

[0013] Another objective of the present invention is to preferentially suppress cross-user pressure interference occurring in a multi-user environment by calculating the occupancy area and membership weight of each of a plurality of users based on the coordinates and long-term average pressure of each of a plurality of zones, and by relatively increasing the compensation gain between zones of different users based on a cross-user mask. means of solving the problem

[0014] To achieve the above objective, an air cell mattress system having a coupling compensation control function between air cells according to the present invention comprises: a zone in which a mattress body equipped with a plurality of air cells is divided into a plurality of regions; a driving unit including a plurality of valves connected to the air cells to perform charging and exhausting, and a pump that supplies air to the air cells through the valves; and a pressure sensor unit that measures the pressure of each of the air cells. The controller comprises: an identification execution condition determination module that determines an identification execution condition in which at least one of the pressure change rate and the user's body movement index is maintained for each of the above zones in a state smaller than a stable threshold for a stable threshold time; an identification signal application module that, when the identification execution condition is satisfied, applies an identification signal having an amplitude limited to a user's perceived threshold to an input zone among the plurality of zones, and fixes the pressure control of the driving unit for other zones during the application period of the identification signal and the response stabilization period after the application ends; a combination matrix calculation module that calculates a combination matrix having an element defined as the ratio of the pressure change amount of the output zone to the magnitude of the identification signal applied to the input zone at the time of response extraction when the pressure change rate of the output zone after the application of the identification signal ends is smaller than a response stable threshold; and a combination compensation module that calculates a compensation input to reduce pressure interference between zones using the combination matrix and controls the driving unit with the compensation input.

[0015] In addition, the coupling compensation module is characterized by including a decoupling matrix derivation unit that derives a decoupling matrix from the coupling matrix, and a compensation input calculation unit that calculates the compensation input by applying the decoupling matrix to a virtual input vector calculated based on the zone-specific pressure error and performing a saturation operation.

[0016] In addition, the identification signal application module further comprises a calibration unit that applies a test signal with a stepwise increasing amplitude to a user in an awake state, and calculates the perceived threshold from the amplitude of the test signal at the point in time when at least one of the user's feedback and the user's body movement indicator is detected. Effects of the invention

[0017] According to the air cell mattress system having an air cell coupling compensation control function according to the present invention,

[0018] 1) By applying an identification signal limited to below the user's perceived threshold only when the identification execution condition is satisfied, automatically calculating a combination matrix, and controlling the driving unit with a compensation input calculated based thereon, it has the advantage of reducing inter-zone pressure interference that inevitably occurs in conventional independent pressure control and precisely tracking the target pressure of each of multiple zones,

[0019] 2) By monitoring pressure change amount, pressure change rate, and body movement indicators in real time during the application period of the identification signal, and immediately suspending the application of the identification signal, discarding contamination data, and reapplying an identification signal with reduced amplitude when an abnormal condition exceeding the suspension threshold is detected, the accuracy of the combined matrix estimation is maintained while protecting the user's perception and safety even in an environment where identification is performed during sleep, and

[0020] 3) By relatively increasing the compensation gain between zones of different users based on the cross-user mask calculated from the membership weights of each of the multiple users, it has the effect of preferentially suppressing cross-user pressure interference in which one user's pressure adjustment interferes with another user's sleep in an environment where multiple users use the air cell mattress system simultaneously. Brief explanation of the drawing

[0021] FIG. 1 is a block diagram illustrating the configuration of the system of the present invention. FIG. 2 is a control block diagram of the coupling compensation module of the present invention. FIG. 3 is a flowchart illustrating the operation flow of the controller of the present invention. FIG. 4 is a timing diagram illustrating the timing of the application of the identification signal and the extraction of the response. FIG. 5 is a flowchart illustrating the re-identification trigger and decoupling matrix update operation of the present invention. FIG. 6 is an example of a top view of a mattress and zone division in a multi-user environment according to the present invention. Specific details for implementing the invention

[0022] Preferred embodiments of the present invention will be described in detail below with reference to the attached drawings. The attached drawings are not drawn to scale, and the same reference numerals in each drawing refer to the same components.

[0023] The air cell mattress system having an air cell coupling compensation control function according to the present invention (hereinafter referred to as the "system") is a system that precisely tracks the target pressure for each zone while identifying and compensating for pressure interference caused by mechanical coupling between multiple zones in real time, whereas conventional systems independently controlled the pressure of each of the multiple air cells.

[0024] In other words, the main purpose is to sequentially apply identification signals that do not cause user sensation to each zone of the mattress body (100) to automatically estimate the coupling matrix between zones, derive a decoupling matrix from the estimated coupling matrix, and calculate a compensation input that reduces pressure interference between zones.

[0025] The specific configuration and operation of the system of the present invention will be described below with reference to the attached drawings.

[0026] FIG. 1 is a block diagram illustrating the configuration of the air cell mattress system of the present invention.

[0027] Referring to FIG. 1, it can be seen that the system of the present invention is based on including a mattress body (100), a driving unit (120), a pressure sensor unit (130), and a controller (200).

[0028] The mattress body (100) is a cushion structure that supports the body from below while a person is lying down, and can be applied to various pneumatic support structures that the body comes into contact with for a long time, such as a bed mattress, a pneumatic seat for a vehicle, a wheelchair cushion, and a medical pressure sore prevention cushion.

[0029] The mattress body (100) includes a support frame in which a plurality of air cells (110) are arranged internally, and an outer shell that covers the upper part of the support frame and comes into direct contact with the user's skin.

[0030] The support frame is a rigid or semi-rigid structure that structurally accommodates and supports a plurality of air cells (110), and, for example, an aluminum profile, a rigid plastic, or a wooden frame may be used. The outer shell may be made of, for example, a medical fabric or a synthetic resin film having breathability, water resistance, and antibacterial properties.

[0031] In the present invention, the specific shape, material, and size of the mattress body (100) are not specifically limited, and it can be implemented in various forms as long as it is a structure capable of accommodating the filling and exhaust passages for a plurality of air cells (110) arranged inside.

[0032] The air cell (110) is a pneumatic support member in which the internal pressure is regulated by the filling and exhaust of air, and a plurality of them are arranged within the mattress body (100).

[0033] Since the support force applied by the air cell (110) to the user's body is controlled by the change in internal pressure of the air cell (110), the pressure distribution to each part of the user's body can be precisely controlled by individually controlling multiple air cells (110).

[0034] The shape of the air cell (110) can be implemented as, for example, a rectangular prism, a cylinder, or a tube, and the material can be made of, for example, thermoplastic polyurethane (TPU), silicone, or an elastic material of the natural rubber family.

[0035] The arrangement method of the air cells (110) can be implemented as a grid-type arrangement of N rows and M columns as an example, or as a non-uniform arrangement corresponding to the shape of each body part as another example.

[0036] The size and spacing of the air cells (110) are determined according to the target pressure distribution resolution, the number of valves (121) of the driving unit (120), and the overall dimensions of the mattress body (100), and for example, the planar size of each air cell (110) can be set to a range of 100mm x 100mm to 400mm x 400mm.

[0037] In the present invention, the air cell (110) is a basic unit that constitutes a zone.

[0038] That is, in the present invention, the term "zone" refers to a pressure-independent control zone within a mattress body (100) in which an air cell (110) is installed, which is logically defined by the controller (200) as a unit for applying an identification signal and combining compensation.

[0039] The zone may be a section determined by the physical separation structure of the mattress body (100), or it may be a software-based division unit in which a controller (200) groups multiple air cells (110) and treats them as a single control unit.

[0040] The reason for dividing a plurality of air cells (110) into a plurality of zones in the present invention is to reduce the size of the combination matrix and alleviate the burden of identification and compensation operations by performing collective control at the zone level instead of individual control at the air cell (110) level, while increasing the effectiveness of pressure control for each body part.

[0041] For example, if divided into three zones—head, torso, and leg—pressure interference between zones can be represented by a 3x3 combination matrix, and as another example, if divided into two zones—left and right—it can be represented by a 2x2 combination matrix.

[0042] The controller (200) logically divides a plurality of air cells (110) into a plurality of zones, and one zone includes at least one air cell (110) controlled to the same pressure.

[0043] For example, one zone may be composed of a single air cell (110), and for another example, one zone may be composed of multiple air cells (110) connected to the same manifold (123).

[0044] The driving unit (120) is a pneumatic driving device that performs charging and exhausting for each of the plurality of air cells (110) according to a control command of the controller (200), and includes a plurality of valves (121) and a pump (122), and further may include a manifold (123) that fluidly connects the pump (122) and the plurality of valves (121).

[0045] The valve (121) is connected to each air cell (110) to perform charging (charging of air) and exhaust (discharging of air).

[0046] The implementation method of the valve (121) can be implemented as a 2-valve method, for example, by having a valve dedicated to charging and a valve dedicated to exhaust for each air cell (110), or as another example, by a 3-port solenoid valve method that performs charging, exhaust, and flow path blocking with a single valve.

[0047] The driving method of the valve (121) can be implemented, for example, by a solenoid method that opens and closes by an electric signal, or, for other examples, by a method using a piezoelectric element or a shape memory alloy.

[0048] The controller (200) controls the internal pressure of the air cell (110) by adjusting the duty ratio (0 to 1) of the valve (121) corresponding to each air cell (110), and as will be described later, the control command value u of the controller (200) in the present invention is defined by this duty ratio.

[0049] The pump (122) is a pneumatic generating device that supplies air to a plurality of air cells (110) through a valve (121), and can be implemented as a diaphragm pump or piston pump, as an example, and as a rotary vane pump.

[0050] The pump (122) can be implemented as a single pump that supplies air to a plurality of air cells (110) in common, or as a plurality of independent pumps for each group of air cells (110).

[0051] The manifold (123) is a pneumatic distribution structure that distributes air supplied from a pump (122) to a plurality of valves (121), and can be implemented as a block-type structure with a plurality of flow paths formed inside, or can be implemented in a manner that branches a plurality of tubes.

[0052] If a manifold (123) is not provided, the pump (122) and a plurality of valves (121) may be implemented in a manner where they are directly connected to individual tubes or hoses.

[0053] The pressure sensor unit (130) is a pressure measuring device that measures the pressure of each of the plurality of air cells (110) and transmits it to the controller (200).

[0054] The pressure sensor unit (130) can be implemented in various configurations, including a structure consisting only of a pressure sensor (131) and a structure combining a pressure sensor (131) and a scanning valve (132).

[0055] For example, the pressure sensor unit (130) can be implemented by placing an independent pressure sensor (131) for each air cell (110). In this case, the pressure of multiple air cells (110) can be measured simultaneously, which has the advantage of having no measurement delay, but the number of pressure sensors (131) and wiring increases in proportion to the number of air cells (110).

[0056] As another example, the pressure sensor unit (130) can be implemented by combining a single pressure sensor (131) and a scanning valve (132).

[0057] The scanning valve (132) is a switching valve that sequentially connects a plurality of air cells (110) to a single pressure sensor (131), and can be implemented as a combination of a plurality of 2-port solenoid valves, for example, or as a multiplexer type pneumatic switching valve.

[0058] In this case, the pressure of multiple air cells (110) can be measured sequentially using only the pressure sensor (131), which simplifies the system cost and configuration, but since the measurement is performed sequentially for each air cell (110), a measurement delay occurs.

[0059] The time it takes for the scanning valve (132) to circulate through a plurality of air cells (110) once is called the scanning period, and the combination matrix calculation module (230) of the present invention determines the time for extracting the response by taking into account this scanning period. A detailed explanation of this will be provided later.

[0060] The type of pressure sensor (131) may be, for example, a capacitive pressure sensor or a piezoresistive pressure sensor, and the measurement range may be set to, for example, a range of 0 kPa to 10 kPa.

[0061] The controller (200) is a device that performs the core control function of the present invention, which controls the driving unit (120) based on the pressure measurement value received from the pressure sensor unit (130) to track the pressure of each of the plurality of zones to a target value, while identifying and compensating for the coupling matrix between zones.

[0062] The installation location of the controller (200) can be implemented in a form embedded inside the lower frame of the mattress body (100), or in a form mounted in a separate control box separated from the mattress body (100) and connected by a cable, or as another example, in a form mounted on a central server or edge computing device within the hospital and connected to the driving unit (120) and the pressure sensor unit (130) via wireless communication.

[0063] The controller (200) is implemented as a combination of hardware and software.

[0064] In terms of hardware, the controller (200) may be configured, for example, around a microcontroller unit (MCU), a digital signal processor (DSP), or a field programmable gate array (FPGA), or for other examples, in the form of an embedded system including an ARM Cortex-family processor, an analog-to-digital converter (ADC), a digital-to-analog converter (DAC), and a communication interface.

[0065] In terms of software, the controller (200) is implemented as a control algorithm that operates in a real-time operating system (RTOS) or bare-metal firmware environment, and can be implemented as embedded software written in C or C++, for example, or as a MATLAB / Simulink-based automatic code generation algorithm.

[0066] Specifically, the controller (200) is based on including an identification execution condition determination module (210), an identification signal application module (220), a combination matrix calculation module (230), and a combination compensation module (240).

[0067] The identification execution condition determination module (210) determines the identification execution condition for each of the plurality of zones and performs the role of authorizing the operation of the identification signal application module (220) only when the identification execution condition is satisfied.

[0068] Here, the identification execution condition is to confirm that the pressure state of multiple zones and the user's movement state are simultaneously in a stable state.

[0069] In other words, to accurately estimate the inter-zone coupling matrix by applying the identification signal described later, the pressure of the zone must reach a steady state before the identification signal is applied.

[0070] In other words, if an identification signal is applied when the zone pressure is in an overactive state or when the user is moving, the pressure response caused by the identification signal and the pressure change caused by user movement or pressure overactive phenomena may overlap, leading to an incorrect estimation of the coupling matrix. Since the decoupling matrix derived from the incorrectly estimated coupling matrix can cause overcompensation that actually amplifies pressure interference between zones, determining the identification execution condition is a prerequisite for guaranteeing the accuracy of the coupling matrix estimation.

[0071] In addition, the steady-state condition is also intended to ensure an operating range in which the static linear approximation model of the combination matrix ΔP_i = Σ_j C_ij·Δu_j can be validly applied.

[0072] The identification execution condition determination module (210) is based on determining the pressure change rate condition and the body movement index condition, and determines that the identification execution condition is satisfied when both of these conditions are satisfied simultaneously.

[0073] The pressure change rate condition is a condition in which the pressure change rate for each of the multiple zones is maintained at a state smaller than the stability threshold for the stability threshold time.

[0074] The pressure change rate can be calculated by differentiating the pressure measurement value of the zone received from the pressure sensor unit (130) with respect to time.

[0075] For example, the rate of change in pressure can be calculated by dividing the difference between two consecutive measurements by the sampling period, or, as another example, by applying a moving average filter or a low-pass filter and then differentiating.

[0076] As an example, the pressure change rate condition can be defined by the following formula.

[0077] formula.

[0078]

[0079] Here, dP_i / dt represents the rate of change of pressure in the i-th zone, ε_p represents the stability threshold, and n represents the number of zones.

[0080] At this time, the stability threshold ε_p is a reference value for determining whether the pressure of the zone has reached a steady state, and can be set, for example, in the range of 0.05 kPa / s to 0.2 kPa / s.

[0081] In addition, the stable critical time T_stable is a minimum holding time to confirm that the pressure change rate condition is a continuous stable state rather than a temporary phenomenon, and can be set, for example, in the range of 5 to 20 seconds.

[0082] The body movement index condition is a condition in which the body movement index (J_move), which reflects the user's body movement, remains smaller than the body movement threshold.

[0083] A body movement indicator is an indicator that shows whether the user is engaging in physical activity, such as tossing and turning, changing posture, or fine movements, while sleeping.

[0084] When the body movement index is high, pressure changes in the air cell (110) caused by the user's body movement are mixed into the identification signal response, causing a combination matrix estimation error; therefore, the identification execution condition is permitted only when the body movement index is sufficiently low.

[0085] For example, the body movement index can be defined as the sum of the variances of the components to which a high-pass filter is applied to each of the pressure signals of the plurality of air cells (110) (J_move = Σ_i Var(HPF(P_i))), and for other examples, it can be defined as a separate acceleration sensor or a load distribution change rate.

[0086] At this time, the cutoff frequency of the high-pass filter can be set to, for example, a range of 0.3 Hz to 1.0 Hz.

[0087] If either of these pressure change rate conditions and body movement index conditions is not satisfied in at least one of the multiple zones, the identification performance condition is determined to be unsatisfied.

[0088] If the identification execution condition is determined not to be satisfied, the identification execution condition determination module (210) re-evaluates the condition after a predetermined waiting time has elapsed.

[0089] In addition, the identification execution condition determination module (210) can optionally link the sleep stage estimation result to the determination of the identification execution condition.

[0090] For example, if the user is presumed to be in deep sleep (N3 stage) or REM sleep stage, the risk of user awakening increases, so the authorization of the identification signal may be suppressed even if the identification performance conditions are met, and if the user is presumed to be in light sleep or awake state, the authorization of the identification signal may be permitted. The sleep stage may be presumed, for example, by a combination of body movement indicators (J_move) and breathing frequency.

[0091] The identification signal application module (220) applies an identification signal to an input zone among a plurality of zones when the identification execution condition is satisfied, and performs the role of fixing the pressure control of the driving unit (120) for another zone during the period of application of the identification signal and the response stabilization period after the end of application.

[0092] The identification signal is a change in the control command value applied to the driving unit (120) of the input zone to estimate the inter-zone coupling matrix, and has an amplitude limited to be smaller than the user's perceived threshold.

[0093] The amplitude of the identification signal is defined as a ratio to the base pressure (P_bar) of the input zone, where the base pressure (P_bar) in the present invention refers to the normal state pressure of each zone before the application of the identification signal, and the amplitude of this identification signal is limited to below the user's perceived threshold (A_comfort).

[0094] There are two reasons why amplitude limiting is necessary for the identification signal.

[0095] If the amplitude of the identification signal is too small, the pressure response due to the identification signal becomes equal to the measurement noise level of the pressure sensor unit (130), and the combined matrix estimation error increases rapidly, and if the amplitude of the identification signal is too large, the user may feel the pressure change and sleep may be disturbed.

[0096] Based on an environment where the pressure sensor resolution is 5 Pa to 20 Pa and the base pressure is 8 kPa to 12 kPa, at an amplitude of less than 0.5% relative to the base pressure, the signal-to-noise ratio decreases sharply, and the combined matrix estimation error may increase to more than 50%, and at an amplitude exceeding 5%, the user's perceived risk and non-linear error that exceeds the valid range of the linear approximation model may increase.

[0097] Accordingly, the practical design range can be set to a range of 0.5% to 5% relative to the base pressure, but in actual implementation, it may be adjusted according to the results of the perceived threshold calibration and the specifications of the pressure sensor unit (130). The method for calculating the perceived threshold (A_comfort) will be described later.

[0098] The waveform of the identification signal can be implemented in the form of a pulse or a step, and the rising and falling sections can be processed in the form of a gentle ramp to further limit the rate of pressure change.

[0099] The waveform of the identification signal can be implemented in the form of a pulse or a step, and the rising and falling sections can be processed in the form of a gentle ramp to further limit the rate of pressure change. In the case of the pulse form, the duration (T_probe) can be set to a range of 0.2 seconds to 2.0 seconds.

[0100] Pseudo-Random Binary Sequence (PRBS) waveforms may also be used for the purpose of improving the signal-to-noise ratio and simultaneously estimating multiple combined parameters.

[0101] When applying the PRBS waveform, the number of bits (N_bit), the bit period (T_bit), and the amplitude (A_probe) are designed to match the scanning period of the pressure sensor unit (130) and the bandwidth of the air cell mattress system. The total duration of the PRBS (N_bit × T_bit) may be set so as not to exceed the maximum duration (T_hold_max) for user-perceived protection, or may be applied by dividing it into multiple short intervals, and T_hold_max may be set in the range of 5 seconds to 30 seconds.

[0102] During the period of application of the identification signal and the response stabilization period after the end of application, the identification signal application module (220) fixes the pressure control of the driving unit (120) for zones other than the input zone.

[0103] Fixing the pressure control of the drive unit means maintaining the drive unit (120) control command value at the point immediately before the identification signal is applied without changing it during the corresponding period.

[0104] By fixing the control command value of the driving unit (120) of another zone by the identification signal application module (220), the pressure tracking operation is blocked from being transmitted to the driving unit (120) even if a new control command value is calculated in response to the pressure response caused by the identification signal, and a pure inter-zone combined response caused by the identification signal can be extracted.

[0105] Specific implementation forms for fixing pressure control of the drive unit can be implemented in various forms, including an output fixing structure that fixes the output value of a PI or PID controller to a value immediately before the application of an identification signal, and an integral term freezing structure that freezes the integral term of a PI or PID controller.

[0106] In addition, it can be implemented in a form where an identification signal is added to a normal pressure-following control input, but the amount of change in the pressure-following control input is corrected when extracting the response to separate only the pure identification signal response. In this case, the control command value applied to the driving unit (120) is defined by the following formula.

[0107] formula.

[0108]

[0109] Here, u is the final control command value applied to the driving unit (120), u_PI is the pressure-following control command value calculated by the PI or PID controller, and Δu_probe represents the magnitude of the identification signal.

[0110] The identification signal application module (220) can apply an identification signal to the j-th zone as an input zone in an air cell mattress system composed of n zones, and simultaneously measure the pressure response of all output zones to calculate the element corresponding to the j-th column of the combination matrix C all at once.

[0111] Additionally, the identification signal application module (220) can produce all elements of an n×n combined matrix by sequentially repeating from j = 1 to n.

[0112] For example, in an air cell mattress system composed of two zones, a 2×2 combination matrix can be completed by applying two identification signals from the first zone to the second zone, and then from the second zone to the first zone.

[0113] The waiting time (T_wait) between each identification signal application is set as the time for the pressure response from the previous identification signal to sufficiently dissipate so that the pressure in the zone returns to a normal state before the next identification signal application.

[0114] As an example, the waiting time (T_wait) can be set to a range of 3 to 5 times the equivalent time constant (τ_eq) determined by the pneumatic characteristics of the air cell (110).

[0115] Additionally, the identification signal application module (220) can reduce the estimated variance caused by measurement noise by repeating the identification signal application K times for the same zone and using the average or median value of the measured response.

[0116] In summary, the identification signal application module (220) sequentially applies an identification signal with an amplitude limited to below the user's perceived threshold to multiple zones, and by fixing the control command value of the driving unit (120) of another zone during the identification signal application period and response stabilization period, it blocks the cancellation of the identification signal response by closed-loop pressure control and enables the elements of the combination matrix to be accurately calculated from the pure inter-zone combination response.

[0117] In an air cell mattress system containing multiple zones, each zone is mechanically connected through physical connection structures such as foam, covers, and frames. Consequently, when the pressure in a specific zone changes, the volume and load distribution of adjacent zones change, causing interference with the pressure in that zone as well.

[0118] When the controller (200) controls the pressure of each zone independently, pressure interference caused by mechanical coupling between these zones causes the target pressure tracking performance to degrade.

[0119] In particular, this problem is exacerbated in multi-user environments where the pressure adjustment of one user affects the support pressure of other users.

[0120] The combination matrix calculation module (230) provides the function of calculating a combination matrix that quantitatively expresses the magnitude and direction of the mechanical combination between these zones.

[0121] The combination matrix is ​​applied to the combination compensation module (240) to calculate the compensation input, and the accuracy of the combination matrix can be considered a key factor in determining the compensation effect.

[0122] The combination matrix calculation module (230) defines the static linear relationship between the minute pressure change around the base pressure and the minute control command value change as a combination matrix.

[0123] In an air cell mattress system composed of n zones, if we define ΔP as a vector consisting of pressure change amounts for each zone and Δu as a vector consisting of control command value change amounts for each zone, the coupling relationship between zones can be expressed by the following formula.

[0124] formula.

[0125]

[0126] Here, ΔP is an n-dimensional vector consisting of pressure change amounts for each of n zones, Δu is an n-dimensional vector consisting of control command value change amounts for each of n zones, and C is an n×n combination matrix.

[0127] The element C_ij in the i-th row and j-th column of the combination matrix C is defined as the ratio of the change in pressure of the i-th zone to the change in the control command value applied to the j-th zone. This can be expressed mathematically as follows.

[0128] formula.

[0129]

[0130] Here, ΔP_i represents the pressure change amount of the i-th zone, Δu_j represents the magnitude of the identification signal applied to the j-th zone, and C_ij represents the element of the i-th row and j-th column of the combination matrix.

[0131] The diagonal element C_ii of the combination matrix C represents the self-pressure response sensitivity to the control input of the i-th zone itself, and has a positive value as the pressure in the corresponding zone increases due to charging.

[0132] The non-diagonal element C_ij (i ≠ j) represents the magnitude and direction of cross-interference where the control input of the j-th zone affects the pressure of the i-th zone, and may have a positive or negative value depending on the characteristics of the mechanical coupling between the air cells (110).

[0133] In addition, C_ij and C_ji can generally have different values, allowing for asymmetric combinations. For example, in an air cell mattress system composed of two zones, the combination matrix can be expressed by the following formula.

[0134] formula.

[0135]

[0136] Here, C_11 and C_22 represent the magnetic pressure response sensitivity of each zone, and C_12 and C_21 represent the cross-interference magnitude between zones.

[0137] The combination matrix calculation module (230) determines the time when the rate of change of pressure in the output zone after the end of application of the identification signal becomes smaller than the response stability threshold as the response extraction time, and calculates the elements of the combination matrix by measuring the amount of pressure change in each output zone at the response extraction time.

[0138] If the pressure response is extracted immediately after the identification signal is terminated, the pneumatic transient response of the air cell (110) is not sufficiently stabilized, so an underestimation may occur in which the elements of the combination matrix are estimated to be smaller than the actual steady state value.

[0139] To prevent this, the combination matrix calculation module (230) continuously monitors the pressure change rate of the output zone and determines the first time point at which the pressure change rate satisfies the condition that it becomes smaller than the response stability threshold (ε_resp) as the response extraction time point.

[0140] The response stability threshold (ε_resp) is a separate reference value distinct from the stability threshold (ε_p) of the identification execution condition determination module (210), and is a value for determining whether the pressure response has reached a steady state.

[0141] The combination matrix calculation module (230) calculates elements of the combination matrix using the magnitude (Δu_j) of the identification signal applied to the input zone and the amount of pressure change (ΔP_i) of the output zone at the time of response extraction. The combination matrix calculation module (230) calculates all elements of the combination matrix from the input and response data collected by the identification signal application module (220) sequentially applying identification signals to multiple zones.

[0142] For example, in an air cell prototype consisting of two zones, when a 75 kg human body model is used and an identification signal of 80 Pa, which is 1% of the base pressure under a base pressure condition of 8.0 kPa, is applied to the first zone, the pressure change amount in the first zone at the time of response extraction is measured to be an average of 72.3 Pa, and the pressure change amount in the second zone is measured to be an average of -9.6 Pa.

[0143] At this time, the reason the pressure change amount in the second zone has a negative value is that the deformation of the foam and cover due to the filling of the first zone predominantly acts as a mechanical coupling mechanism that increases the volume of the second zone and reduces the pressure in the second zone.

[0144] If the identification signal application module (220) additionally applies an identification signal using the second zone as an input zone in the same way, the combination matrix calculation module (230) can calculate the following 2×2 combination matrix.

[0145]

[0146] Here, rows represent output zones and columns represent input zones, diagonal elements 4155 Pa / duty and 4299 Pa / duty represent the magnetic pressure response sensitivity of each zone, and off-diagonal elements -471 Pa / duty and -552 Pa / duty represent the cross-interference magnitude between zones.

[0147] The coupling ratios |C_21 / C_11| = 0.133 and |C_12 / C_22| = 0.111 confirm the magnitude of coupling between zones. These figures are actual measurement examples and may vary depending on the structure and usage conditions of the air cell mattress system.

[0148] In summary, the combination matrix calculation module (230) collects pressure responses of multiple zones to an identification signal applied by the identification signal application module (220), and calculates elements of the combination matrix as the ratio of the pressure change amount of the output zone to the size of the identification signal of the input zone at the time of response extraction, thereby providing information necessary for the combination compensation module (240) to calculate a compensation input that reduces pressure interference between zones.

[0149] The combination compensation module (240) calculates a compensation input that reduces pressure interference between zones using the combination matrix calculated by the combination matrix calculation module (230), and performs the role of controlling the driving unit (120) with the calculated compensation input.

[0150] In order for the combined compensation module (240) to calculate the compensation input, it must first calculate a virtual input vector v based on the zone-specific pressure error (e_i = P*_i - P_i), which is defined as the difference between the target pressure and the actual pressure of each zone.

[0151] At this time, virtual input refers to the amount of pressure change required in a zone when it is assumed that each zone operates independently before the combined compensation is applied, and the virtual input vector v = [v_1, v_2, ..., v_n]^T consists of the virtual inputs of each of the multiple zones.

[0152] To calculate a virtual input vector, the controller (200) of the present invention may include a PI controller or a PID controller.

[0153] The PI controller (proportional-integral controller) is a controller that produces a virtual input v_i as the sum of a proportional term obtained by multiplying the zone-specific pressure error e_i by a proportional gain K_p and an integral term obtained by multiplying the value obtained by integrating the zone-specific pressure error over time by an integral gain K_i, and has the characteristic of converging the steady-state pressure error to zero.

[0154] A PID controller (Proportional-Integral-Derivative controller) is a controller that improves transient response characteristics by adding a derivative term to the PI controller, which is the derivative of the time derivative of the pressure error by zone multiplied by the derivative gain K_d.

[0155] The controller (200) independently operates a PI controller or a PID controller corresponding to each zone and constructs a virtual input vector v with the virtual input v_i calculated by each controller as a component.

[0156] An example in which the combined compensation module (240) calculates a compensation input from a virtual input vector v is described in two examples as follows.

[0157] The first embodiment is a structure that derives a decoupling matrix from a coupling matrix, applies the decoupling matrix to a virtual input vector, and then performs a saturation operation to calculate a compensation input. This can be expressed mathematically as follows.

[0158] formula.

[0159]

[0160] Here, Δu is a compensation input vector applied to the driving unit (120), D is a decoupling matrix derived from the inverse or pseudo-inverse of the coupling matrix C, v is a virtual input vector, and sat(·) represents a saturation function considering the physical driving range of the driving unit (120).

[0161] The saturation function sat(·) is an element-wise clipping operation that limits the control command value of each zone, calculated by multiplying the decoupling matrix D and the virtual input vector v, to within the range [u_min, u_max] that the driving unit (120) can realize.

[0162] Since a warp may occur in the integral term of the PI controller or PID controller when saturation occurs, the combined compensation module (240) may apply a warp prevention algorithm that freezes the integral term when saturation occurs or feeds back a correction value proportional to the saturation error to the integrator.

[0163] The second embodiment is a structure that applies feedforward cross-compensation without inverse matrix calculation, and calculates the compensation input for each zone using the following formula.

[0164] formula.

[0165]

[0166] Here, Δu_i represents the reward input of the i-th zone, v_i represents the virtual input of the i-th zone, C_ij represents the element in the i-th row and j-th column of the combination matrix, and C_ii represents the diagonal element in the i-th row and i-th column of the combination matrix.

[0167] Applying the second implementation form as a numerical example to the previously calculated 2×2 combined matrix yields the following.

[0168] Under the condition that the pressure error in the first zone is 200 Pa and the pressure error in the second zone is 0 Pa, the virtual input vector is v = [200 Pa, 0 Pa]^T.

[0169] When the second embodiment is applied to the combined compensation module (240), the compensation input of the first zone is Δu_1 = v_1 - (C_12 / C_11)·v_2 = 200 - (-471 / 4155)×0 = 200 Pa, and the compensation input of the second zone is Δu_2 = v_2 - (C_21 / C_22)·v_1 = 0 - (-552 / 4299)×200 = 25.7 Pa. In this way, by applying a compensation input corresponding to 25.7 Pa to the second zone, the cross-interference caused by the control input of the first zone to the second zone can be offset in advance.

[0170] This type of implementation is easy to implement in low-spec microcontroller environments as it does not require inverse matrix calculation, but the compensation accuracy is high when the combination matrix satisfies diagonal dominance (|C_ii| ≥ α·Σ_{j≠i}|C_ij|, α ≥ 1.2).

[0171] The combined compensation module (240) can reduce inter-zone pressure interference that occurs while the pressure of each of the multiple zones follows the target value by controlling the driving unit (120) with the calculated compensation input vector Δu.

[0172] However, for the combined compensation module (240) to calculate an accurate compensation input from a virtual input vector v, the operation of deriving a decoupling matrix from a combined matrix and the operation of applying the decoupling matrix to the virtual input vector to calculate a compensation input must be performed separately according to their roles.

[0173] In other words, since the method for deriving the decoupling matrix can vary depending on the characteristics of the coupling matrix, and the quality of the derived decoupling matrix directly affects the accuracy of the compensation input calculation, separating the two operations to clearly define the roles and responsibilities of each is advantageous for ensuring system stability.

[0174] To this end, the coupling compensation module (240) may include a decoupling matrix derivation unit (241) and a compensation input calculation unit (242).

[0175] FIG. 2 is a control block diagram of the coupling compensation module of the present invention.

[0176] The decoupling matrix derivation unit (241) derives a decoupling matrix D that offsets the inter-zone coupling effect from the coupling matrix C calculated by the coupling matrix calculation module (230).

[0177] The decoupling matrix derivation unit (241) derives the inverse matrix (C^{-1}) of the combination matrix as the decoupling matrix when the absolute value of the determinant det(C) is greater than a predetermined threshold, and derives the pseudo-inverse matrix (C^+) ​​as the decoupling matrix otherwise.

[0178] In an air cell mattress system composed of two zones, the decoupling matrix derivation unit (241) can calculate the inverse matrix in real time using the analytical solution of the following formula.

[0179] formula.

[0180]

[0181] Here, det(C) represents the determinant of the combined matrix C.

[0182] Based on this, if the previously calculated coupling matrix is ​​applied as a numerical example, det(C) = 4155×4299 - (-471)×(-552) = 17,602,153, and the decoupling matrix derivation unit (241) derives the decoupling matrix as follows.

[0183]

[0184] In an air cell mattress system composed of three or more zones, the decoupling matrix derivation unit (241) can calculate the inverse matrix by applying an LU decomposition or QR decomposition algorithm.

[0185] The compensation input calculation unit (242) calculates a compensation input by applying the decoupling matrix D derived by the decoupling matrix derivation unit (241) to a virtual input vector v composed of virtual inputs calculated by a PI controller or PID controller based on zone-specific pressure errors, and performing a saturation operation.

[0186] Based on the numerical example presented earlier, when a virtual input vector v = [200 Pa, 0 Pa]^T is input, the compensation input calculation unit (242) calculates Δu = sat(D·v) = sat([2.44×10^{-4}×200 + 2.68×10^{-5}×0, 3.14×10^{-5}×200 + 2.36×10^{-4}×0]^T) = sat([0.049, 0.006]^T) = [0.049 duty, 0.006 duty]^T and outputs it to the driving unit (120).

[0187] Since both values ​​are within the driving range of the valve (121), which is 0 to 1, they are output as is without clipping by saturation calculation, and through this compensation input, the pressure in the first zone approaches the target value of 8.2 kPa and the pressure in the second zone maintains the target value of 8.0 kPa, thereby reducing pressure interference between zones.

[0188] In summary, since the combination compensation module (240) is configured separately into a decoupling matrix derivation unit (241) and a compensation input calculation unit (242), it is possible to independently perform an operation to selectively derive a suitable decoupling matrix among an inverse matrix or a pseudo-inverse matrix according to the characteristics of the combination matrix, and an operation to calculate a compensation input by applying the derived decoupling matrix to a virtual input vector.

[0189] Through this, numerical stability in the decoupling matrix derivation stage and saturation calculation processing in the compensation input calculation stage can be performed according to the role of each stage, thereby simultaneously ensuring the performance of reducing inter-zone pressure interference and system stability of the air cell mattress system.

[0190] The operation flow of the system of the present invention is described below.

[0191] FIG. 3 is a flowchart illustrating the operation flow of the controller of the present invention.

[0192] The identification execution condition determination module (210) of the controller (200) continuously monitors the pressure change rate and the user's body movement indicator for each of the plurality of zones, and determines whether the identification execution condition is satisfied in which the two indicators are maintained at a state smaller than their respective stability thresholds for a stability threshold time.

[0193] When the identification execution condition is satisfied, the identification signal application module (220) starts operation and sequentially applies an identification signal with an amplitude limited to a threshold smaller than the user's perceived threshold to the input zone, and during the period of application of the identification signal and the response stabilization period after the end of application, the pressure control of the driving unit (120) for the other zone is fixed.

[0194] The combination matrix calculation module (230) measures the pressure change amount of each output zone at the time of response extraction when the pressure change rate of the output zone after the end of application of the identification signal becomes smaller than the response stability threshold, and calculates a combination matrix having elements defined as the ratio of the pressure change amount of the output zone to the magnitude of the identification signal applied to the input zone.

[0195] The coupling compensation module (240) derives a decoupling matrix from the calculated coupling matrix, applies the decoupling matrix to a virtual input vector calculated by each zone's PI controller or PID controller based on the zone-specific pressure error, performs a saturation operation to calculate a compensation input, and reduces pressure interference between zones by controlling the driving unit (120) with the calculated compensation input.

[0196] In summary, the system of the present invention automatically identifies the inter-zone coupling matrix using an identification signal that does not cause user perception and compensates for it in real time, thereby reducing inter-zone pressure interference that inevitably occurs in conventional independent pressure control and providing the characteristic of precisely tracking the target pressure of each of multiple zones.

[0197] Meanwhile, in order for the system of the present invention to accurately limit the amplitude of the identification signal to below the user's perceived threshold, the perceived threshold must be calculated in advance for each user.

[0198] Since the perceived threshold varies significantly by individual based on the user's weight, physical sensitivity, and sleep status, applying a fixed value uniformly may result in an overly conservative effect for some users, leading to a decrease in the signal-to-noise ratio of the identification signal, or conversely, causing a perceived problem for some users.

[0199] To solve this, the identification signal application module (220) may further include a calibration unit (221) that calculates a user-specific comfort threshold (A_comfort).

[0200] The calibration unit (221) applies a test signal with a stepwise increasing amplitude to a user in an awake state and calculates a perceived threshold from the amplitude of the test signal at the point in time when at least one of the user's feedback and the user's body movement indicator is detected.

[0201] Here, the term "awakened state" means a state in which the user is fully awake and capable of recognizing and responding to external stimuli.

[0202] The calibration unit (221) can determine the awakened state based on the body movement index (J_move) calculated by the identification performance condition determination module (210).

[0203] For example, the state of awake can be determined when the body movement index is greater than a predetermined awake determination threshold and the user's breathing component is clearly detected in the pressure signal measured by the pressure sensor unit (130). As another example, the state of awake can be determined when the user directly requests to enter the calibration mode through a separate control unit (e.g., a remote control button or a touch pad).

[0204] Specifically, the calibration unit (221) sequentially applies the amplitude of the test signal to each zone while gradually increasing it from a predetermined initial value. The initial value of the amplitude of the test signal can be set to 0.5% or less relative to the base pressure, and the stepwise increase amount of the amplitude of the test signal can be set to a range of 0.2% to 0.5% relative to the base pressure.

[0205] For example, if the base pressure is 10 kPa, the initial amplitude value can be set to 50 Pa and the test signal can be applied while increasing it step by step by 50 Pa.

[0206] The calibration unit (221) continuously monitors user feedback and body movement indicators while applying a test signal.

[0207] User feedback is an input signal generated by the user recognizing pressure changes while in an awake state, and can be received in the form of, for example, remote control button input, touchpad contact, or voice input.

[0208] The body movement index is the previously defined J_move, and sensory events can be indirectly detected by utilizing the characteristic that the body movement index increases when a user moves their body in response to pressure changes while sleeping or awake.

[0209] The calibration unit (221) determines the test signal amplitude at the time when at least one of the user's feedback and body movement indicators is detected as the initial perceived detection amplitude (A_th).

[0210] Next, the calibration unit (221) calculates the perceived threshold by applying a safety margin factor β to A_th. This can be expressed as a formula as follows.

[0211] formula.

[0212]

[0213] Here, A_comfort represents the comfort threshold, A_th represents the initial comfort detection amplitude, and β represents the safety margin factor.

[0214] The safety margin factor β can be set in the range of 0.5 to 0.9. For example, when user feedback is first detected at an amplitude of 300 Pa while the test signal amplitude is increased stepwise by 50 Pa under a base pressure of 10 kPa, A_th = 300 Pa, and if β = 0.7 is applied, the comfort threshold is calculated as A_comfort = 0.7 × 300 = 210 Pa.

[0215] Afterwards, the identification signal application module (220) can limit the amplitude of the identification signal to 210 Pa or less to prevent user perception while securing the signal-to-noise ratio required for estimating the combination matrix.

[0216] Meanwhile, since direct feedback from the user is not possible during sleep unlike in the waking state, the calibration unit (221) can update the perceived threshold online by utilizing the change in the body movement index (J_move) observed during or immediately after the application of the identification signal as an indirect indicator of the perceived event.

[0217] Specifically, when a perceived event occurs in which the body movement index exceeds the body movement threshold (V_abort) while the identification signal is applied, the calibration unit (221) conservatively updates the perceived threshold to A_comfort ← min(A_comfort, A_probe).

[0218] The amplitude of the identification signal during sleep can be limited to A_probe ≤ β_sleep · A_comfort by additionally applying a compensation coefficient β_sleep during sleep to the perceived threshold calculated during wakefulness, and β_sleep can be set in the range of 0.3 to 0.8.

[0219] That is, the calibration unit (221) directly measures the initial perceived detection amplitude for each user in the waking state and applies a safety margin coefficient to personalize the perceived threshold, thereby maximizing the signal-to-noise ratio required for estimating the combination matrix within a range where the amplitude of the identification signal does not induce the user's perception, and during sleep, the perceived threshold is conservatively maintained through online updates based on the body movement index, thereby stably securing identification performance conditions without sleep disturbance.

[0220] The system of the present invention must be able to protect the user's sense of touch and safety even when sudden changes in posture, body movements, or unexpected pressure deviations occur during the application period of the identification signal.

[0221] Even if an identification signal is applied at the point when the identification execution conditions are satisfied, if the user changes their posture or the pressure changes rapidly due to external factors during the period of application of the identification signal, the amount of pressure change or the rate of pressure change in the input zone may exceed the allowable range, causing the user to feel the pressure change or hindering the stable operation of the air cell mattress system.

[0222] Furthermore, the identification signal response data collected under these circumstances is contaminated data mixed with pressure changes caused by user movement or pressure deviation, which carries the potential to cause errors in the estimation of the combined matrix.

[0223] To resolve this, the identification signal application module (220) may further include a safety interruption unit (222) that monitors abnormal conditions in real time and responds immediately during the application period of the identification signal.

[0224] The safety interruption unit (222) monitors the amount of pressure change, the rate of pressure change, and the user's body movement indicator in real time during the period of application of the identification signal, and if at least one of these is detected to be greater than the respective interruption threshold, it immediately stops the application of the identification signal and re-applies the identification signal with reduced amplitude after a predetermined waiting time has elapsed.

[0225] The three indicators monitored by the safety interruption unit (222) and their respective interruption thresholds are described in detail as follows.

[0226] The first indicator is the amount of pressure change in the input zone.

[0227] The pressure change amount in the input zone is defined as the absolute value of the difference between the base pressure immediately before the application of the identification signal and the current pressure measured during the application of the identification signal.

[0228] Normal pressure changes detected by identification signals are limited to below the perceived threshold, but this is an indicator to detect cases where the pressure in the input zone changes significantly and unexpectedly due to sudden changes in the user's posture or load transfer.

[0229] The interruption threshold (P_abort) for pressure change amount can be set to 10% of the base pressure or in the range of 0.3 kPa to 1.0 kPa.

[0230] For example, if the base pressure is 10 kPa and P_abort = 1.0 kPa is set, the safety interruption unit (222) immediately stops the application of the identification signal when the pressure in the input zone changes by more than 1.0 kPa relative to the base pressure.

[0231] The second indicator is the pressure change rate, which is the absolute value of the derivative of the pressure measurement of the input zone with respect to time, and is an indicator for detecting cases where the pressure changes rapidly due to the user's minute movement or external impact. The interruption threshold (R_abort) for the pressure change rate is set to a value greater than the stability threshold (ε_p) of the identification execution condition determination module (210), and can be set in the range of 0.5 kPa / s to 5.0 kPa / s.

[0232] For example, if R_abort = 1.0 kPa / s is set, the safety interruption unit (222) operates when a state is detected where the pressure change rate exceeds 1.0 kPa / s while the identification signal is applied.

[0233] The third indicator is the user's body movement indicator (J_move). As previously defined, the body movement indicator is defined as the sum of the variances of the components to which a high-pass filter is applied to the pressure signal of each air cell (110), and reflects the user's body movement.

[0234] The suspension threshold (V_abort) for the body movement indicator is set separately from the body movement threshold (V_th) of the identification execution condition determination module (210) and represents the maximum body movement level allowed during the application of the identification signal. V_abort may be set based on long-term statistics or the calibration results of the calibration unit (221).

[0235] The three indicators are monitored independently, and as soon as a state is detected where any one exceeds its respective suspension threshold, the safety suspension unit (222) operates in the following order.

[0236] First, the safety interruption unit (222) immediately stops the application of the identification signal and releases the pressure control fixation of the drive unit (120) to return to normal pressure tracking control.

[0237] At the same time, the input and response data collected from the corresponding identification signal application are determined to be contaminated data and discarded.

[0238] Next, the safety interruption unit (222) instructs the coupling compensation module (240) to set the decoupling matrix to the identity matrix, and the air cell mattress system maintains an independent pressure-following control state without inter-zone coupling compensation.

[0239] After a predetermined waiting time has elapsed following the occurrence of a safety interruption, the safety interruption unit (222) re-applies an identification signal with reduced amplitude. The amplitude of the re-applied identification signal is set to a value lower than the amplitude of the identification signal immediately before the interruption.

[0240] At this time, the decrease in re-approval amplitude can be expressed by the following formula.

[0241] formula.

[0242]

[0243] Here, A_probe,new is the amplitude of the re-applied identification signal, A_probe,prev is the amplitude of the identification signal immediately before interruption, and η is an amplitude reduction factor that can be set to a range of 0.5 to 0.9.

[0244] In the event that safety interruption is repeated, the safety interruption unit (222) can reduce the amplitude stepwise by repeatedly applying the above formula, or increase the normalization coefficient λ used in the normalized least squares estimation of the combined matrix calculation module (230) to enhance the stability of the estimated numerical value.

[0245] The entire operation of the above-described safety suspension part (222) is explained as follows.

[0246] Assume that while applying an identification signal to the first zone under conditions of a base pressure of 10 kPa and an identification signal amplitude of 200 Pa (2% relative to the base pressure), at 0.3 seconds after application, the rate of change of pressure in the first zone is measured to be 1.2 kPa / s, and a state exceeding R_abort = 1.0 kPa / s is detected.

[0247] In this case, the safety interruption unit (222) immediately stops the application of the identification signal, discards the collected data, and returns to normal pressure-following control.

[0248] Next, after a predetermined waiting time (e.g., 10 seconds) has elapsed, the safety interruption section (222) reapplies the identification signal with a reduced amplitude of A_probe,new = 0.7 × 200 = 140 Pa (1.4% relative to the base pressure) by applying η = 0.7.

[0249] When identification is successfully completed by the re-authorized identification signal, the combination matrix is ​​calculated and the combination compensation module (240) resumes the calculation of the compensation input.

[0250] According to such safety interruption unit (222), even if unexpected user movement or pressure deviation occurs during the application period of the identification signal, the user's sense of safety can be protected in real time, and the accuracy of the combined matrix estimation can be maintained through a conservative retry algorithm that immediately discards contaminated data and reapplies it with reduced amplitude.

[0251] Figure 4 is a timing diagram illustrating the application of the identification signal and the extraction of the response.

[0252] Furthermore, the system of the present invention can simplify system costs and configuration by adopting a structure that sequentially measures the pressure of a plurality of air cells (110) by combining a pressure sensor (131) and a scanning valve (132), instead of an independent sensor arrangement structure in which the number of pressure sensors and wiring increases in proportion to the number of air cells (110).

[0253] Specifically, the pressure sensor unit (130) includes a pressure sensor (131) and a scanning valve (132) that sequentially connects a plurality of air cells (110) to the pressure sensor (131).

[0254] The pressure sensor (131) is a device that converts the internal pressure of the air cell (110) into an electrical signal, and, for example, a capacitive pressure sensor or a piezoresistive pressure sensor may be used, and the measurement range may be set to 0 kPa to 20 kPa and the resolution to 5 Pa to 20 Pa.

[0255] The pressure sensor (131) can be placed near a control box or manifold (123) outside the mattress body (100) and is fluidly connected sequentially to a plurality of air cells (110) through a scanning valve (132).

[0256] The scanning valve (132) is a switching valve that connects one of the plurality of air cells (110) to the pressure sensor (131) and blocks the connection with the remaining air cells (110) according to a control signal from the controller (200).

[0257] The implementation form of the scanning valve (132) can be implemented, for example, by arranging a 2-port solenoid valve for each air cell (110) and opening and closing it sequentially, or as another example, by implementing a pneumatic switching valve of the multiplexer type.

[0258] The time during which the scanning valve (132) remains connected to a specific air cell (110) is called the dwell time (T_dwell), and is set as the minimum time for the pressure sensor (131) to stably measure the pressure of the air cell (110).

[0259] The dwell time (T_dwell) is set to be greater than the sum of the response time of the scanning valve (132) and the stabilization time of the pressure sensor (131), and can be set to, for example, a range of 50 ms to 200 ms.

[0260] The controller (200) can measure the pressure of each air cell (110) during the dwell time (T_dwell) while sequentially switching the scanning valve (132).

[0261] For example, in an air cell mattress system composed of n air cells (110), when the scanning valve (132) sequentially switches the air cells (110), the pressure sensor (131) measures the pressure of each air cell (110) in the order of the first air cell and the second air cell for a dwell time (T_dwell) and completes the measurement up to the nth air cell (110).

[0262] At this time, the time required for the scanning valve (132) to circulate through all of the multiple air cells (110) once is defined as the scanning period (T_scan).

[0263] The scanning period (T_scan) is calculated as the product of the quantity of air cells (110) and the dwell time (T_dwell), and can be expressed as a formula as follows.

[0264] formula.

[0265]

[0266] Here, T_scan represents the scanning period, n represents the quantity of air cells (110), and T_dwell represents the dwell time during which the scanning valve (132) remains connected to each air cell (110).

[0267] Since the scanning period (T_scan) is the time required for the pressure sensor unit (130) to update the pressure of each of the plurality of air cells (110) once, the pressure measurement values ​​of some air cells (110) may not yet be updated before the scanning period has elapsed.

[0268] Accordingly, the scanning period (T_scan) is used as a criterion for the combined matrix calculation module (230) to determine the time for extracting the response.

[0269] For example, in an air cell mattress system composed of 6 air cells (110), if the dwell time (T_dwell) is set to 50 ms, the scanning cycle is calculated as T_scan = 6 × 50 ms = 300 ms.

[0270] In this case, the pressure sensor unit (130) updates the pressure of all six air cells (110) once every 300 ms, and the combination matrix calculation module (230) determines the response extraction time after at least 300 ms have elapsed since the end of the application of the identification signal.

[0271] In response to this, the combination matrix calculation module (230) can determine the first time point that simultaneously satisfies two conditions as the response extraction time point when determining the response extraction time point after the end of the application of the identification signal.

[0272] The first condition is that at least one scanning cycle must elapse after the termination of the authorization of the identification signal.

[0273] In a structure where the pressure sensor unit (130) combines the pressure sensor (131) and the scanning valve (132), the pressure of a plurality of air cells (110) is measured sequentially, so immediately after the application of the identification signal ends, the pressure measurement values ​​of some air cells (110) may not yet be updated.

[0274] When the response is extracted in this state, underestimation occurs due to scanning delay, where the pressure change amount in some output zones is measured to be smaller than the actual steady-state value.

[0275] For example, in an air cell mattress system composed of six air cells (110), if the scanning cycle is 300 ms, immediately after the application of the identification signal ends, the pressure measurement of the last updated air cell (110) may be a value from up to 300 ms ago, so a response extraction error may occur.

[0276] In actual experiments, it was confirmed that when a response is extracted immediately after the end of the application of the identification signal, the estimation error of the combination matrix elements increases to 40% or more, whereas when a response is extracted after at least one scanning cycle has elapsed, the estimation error decreases to 10% or less.

[0277] Accordingly, the combination matrix calculation module (230) starts searching for the response extraction time from the point in time when at least one scanning cycle has elapsed after the end of the application of the identification signal.

[0278] The second condition is that it must be the point at which the rate of change of pressure in the output zone becomes smaller than the response stability threshold (ε_resp).

[0279] Immediately after the application of the identification signal ends, the pneumatic transient response of the air cell (110) has not yet stabilized, so the pressure is in a state of continuous change. In this state, the pressure change amount measured is a transient value that has not reached a steady state value, and if this is used to calculate the elements of the combination matrix, an underestimation error occurs in which it is estimated to be smaller than the actual steady state combination matrix.

[0280] When the two conditions are combined, the condition for the response extraction time determined by the combination matrix calculation module (230) can be expressed by the following formula.

[0281] formula.

[0282]

[0283] Here, t_extract is the response extraction time, t_end is the end time of application of the identification signal, T_scan is the scanning period, dP_i(t) / dt is the rate of change of pressure in the i-th output zone, and ε_resp is the response stability threshold.

[0284] The above condition means that the first time point at which the rate of change of pressure in all output zones becomes smaller than the response stability threshold after at least one scanning cycle has elapsed following the termination of the application of the identification signal is determined as the response extraction time point.

[0285] The response stability threshold (ε_resp) is a reference value for determining whether the pneumatic transient response of the air cell (110) has been sufficiently stabilized, and can be set in the range of 0.05 kPa / s to 0.2 kPa / s. If ε_resp is excessively large, the response is extracted before the transient response is sufficiently stabilized, which increases the estimation error, and if it is excessively small, the timing of the response extraction is excessively delayed, which may cause a problem of increased waiting time until the next identification signal is applied.

[0286] Based on the numerical examples presented earlier, the process for determining the response extraction time is explained as follows.

[0287] It is assumed that under the conditions of a scanning period T_scan = 300 ms and a response stability threshold ε_resp = 0.1 kPa / s, an identification signal is applied for 1.0 second with the first zone as the input zone, and the application ends at t_end = 1.0 second.

[0288] The combination matrix calculation module (230) starts monitoring the pressure change rate of the output zones from the time point t_end + T_scan = 1.0 + 0.3 = 1.3 seconds, and if it is confirmed that the pressure change rate of all output zones decreases to less than 0.1 kPa / s at the time point 1.5 seconds, it determines t_extract = 1.5 seconds as the response extraction time point, and measures the pressure change amount of each output zone at that time point to calculate the elements of the combination matrix.

[0289] In summary, the combination matrix calculation module (230) determines the response extraction time by simultaneously applying the scanning cycle elapsed condition and the pressure change rate stability condition, thereby having the advantage of minimizing the combination matrix estimation error caused by scanning delay and pneumatic transient response and calculating an accurate combination matrix even in a low-cost structure combining a single pressure sensor (131) and a scanning valve (132).

[0290] Meanwhile, if the combination matrix calculation module (230) described above directly calculates the elements of the combination matrix based only on the result of applying a single identification signal to a single zone, errors may be mixed into the measurement values ​​due to measurement noise of the pressure sensor unit (130), scanning delay, and fine movements of the user, thereby reducing the accuracy of the combination matrix estimation.

[0291] In addition, as the number of zones (n) increases, the number of elements in the combination matrix increases to n², which may lead to a problem where it is difficult to ensure the reliability of the estimation through simple direct calculation alone.

[0292] To solve this, the combination matrix calculation module (230) may further include a normalization estimation unit (231) that estimates the combination matrix by applying least squares including a normalization term to the input matrix and response matrix formed from the results of multiple identification signal applications for different input zones.

[0293] The operation of the normalization estimation unit (231) is explained in two stages as follows.

[0294] The first step is to construct the input matrix and the response matrix.

[0295] Specifically, the normalization estimation unit (231) constructs an input matrix U and a response matrix P using data collected by the identification signal application module (220) applying an identification signal m times to different input zones.

[0296] If we define Δu(k) as the magnitude vector of the identification signal applied to the input zone at the k-th (k = 1, 2, ..., m) identification signal application, and ΔP(k) as the vector of the pressure change amount of each zone measured at the time of response extraction, then the input matrix U and the response matrix P are constructed as follows.

[0297] formula.

[0298]

[0299] Here, U is an m×n input matrix consisting of the magnitude of the identification signal applied to each zone in m identification signal applications, P is an m×n response matrix consisting of the pressure change amount of each zone in m identification signal applications, and n represents the number of zones.

[0300] For example, in an air cell mattress system composed of two zones, when an identification signal is applied once to the first zone and the second zone respectively (m = 2), when applied to the first zone, Δu(1) = [80 Pa, 0 Pa]^T and the response ΔP(1) = [72.3 Pa, -9.6 Pa]^T, and when applied to the second zone, Δu(2) = [0 Pa, 80 Pa]^T and the response ΔP(2) = [-8.2 Pa, 74.8 Pa]^T, the input matrix and the response matrix are configured as follows according to the above formula.

[0301]

[0302] The second step is to estimate the normalized least squares.

[0303] The normalization estimation unit (231) can calculate an estimated combination matrix from the constructed input matrix U and response matrix P using the following mathematical formula 1.

[0304] Mathematical formula 1.

[0305]

[0306] Here, U is the input matrix, P is the response matrix, λ is the normalization coefficient, I is the identity matrix, represents the estimated joint matrix.

[0307] Equation 1 assumes a linear model P = U·C^T + E (E: error matrix) between the input matrix U and the response matrix P, and the sum of squared residuals (||P - U· The normalization term (λ||) in ^T||_F²) It is derived as the solution that minimizes the cost function obtained by adding ||_F²).

[0308] The unnormalized least squares (OLS) without a normalization term is It is expressed in the form ^T = (U^TU)^{-1} U^TP, and is applicable only when U^TU is invertible.

[0309] On the other hand, the normalized least squares of Equation 1 can produce numerically stable estimates even when U^TU is irreversible or the number of conditions is large, as the number of conditions of (U^TU + λI) is reduced by the normalization coefficient λ.

[0310] The normalization factor λ is a key parameter that determines the balance between numerical stability and estimation accuracy.

[0311] If λ is too small, the condition number of (U^TU + λI) becomes large, making the inverse matrix calculation unstable, and if it is too large, the influence of the normalization term becomes excessive, resulting in underestimation where the estimated elements of the combined matrix are estimated to be smaller than their actual values.

[0312] The normalization coefficient λ can be automatically selected as the minimum value that satisfies the condition number (cond(U^TU + λI)) of (U^TU + λI) being less than or equal to a predetermined upper limit (κ_max).

[0313] For example, it can be implemented in the form of selecting the minimum λ that satisfies cond(U^TU + λI) ≤ 100 from the candidate set Λ = {0.001, 0.01, 0.05, 0.1, 0.2}.

[0314] The effects of normalized least squares are specifically explained through examples and comparative examples as follows.

[0315] Based on the input matrix U and response matrix P constructed earlier, it is assumed that five repeated measurements are performed in an environment mixed with measurement noise (standard deviation 15 Pa).

[0316] As a comparative example, when applying the normalization factor λ = 0 (OLS), the condition number is very high with cond(U^TU) = 1850, making the inverse matrix calculation unstable, and the estimation error of the elements of the joint matrix estimated in a noisy environment increases to more than 35%, and the deviation of the estimates is large, making it difficult to produce a reliable joint matrix.

[0317] As an example, applying a normalization factor λ = 0.01 significantly reduces the number of conditions to cond(U^TU + λI) = 115, thereby stabilizing the inverse matrix calculation, and the estimated combination matrix is ​​calculated as follows by Equation 1.

[0318]

[0319] In this case, the estimation error is reduced to the level of 8% to 9%, and the condition number cond(┴) of the decoupling matrix calculated by the decoupling matrix derivation unit (241) is 1.28, so the reversibility is good and the coupling compensation module (240) can calculate an accurate compensation input.

[0320] When compared with the comparative example (λ = 0, OLS), it can be seen that while the inverse matrix calculation is unstable in the comparative example with an estimation error of 35% or more and a condition number of 1850, the numerical stability is significantly improved in the example (λ = 0.01) with the estimation error reduced to 8% to 9% and the condition number of 115.

[0321] In addition, when combined compensation is applied, the pressure interference index (ΔP_RMS,Z2) of the second zone is at the 48 Pa level in the comparative example, but there is a risk of overcompensation due to inverse matrix instability, whereas in the example, it is maintained at the 55 Pa level, providing stable compensation performance without overcompensation.

[0322] In summary, the normalization estimation unit (231) applies normalization least squares to produce a numerically stable combination matrix estimate even in a measurement noise environment, thereby significantly reducing the estimation error compared to direct calculation using only the result of applying a single identification signal, and simultaneously improving the compensation accuracy of the combination compensation module (240) and the stability of the air cell mattress system.

[0323] If the coupling matrix calculated by the coupling matrix calculation module (230) or the decoupling matrix derived by the decoupling matrix derivation unit (241) is numerically unstable, and if it is used as is for coupling compensation, the compensation input may be excessively calculated, which may cause a problem in which the pressure control of the system of the present invention becomes unstable.

[0324] Numerical instability can occur due to various causes, such as measurement noise, abrupt changes in coupling characteristics caused by changes in user posture, or a decrease in the signal-to-noise ratio due to the excessively small amplitude of the identification signal.

[0325] For example, if the size of a non-diagonal element in the row direction of the coupling matrix approaches or exceeds a diagonal element, a small measurement error is amplified during the inverse matrix calculation process, so that the elements of the decoupling matrix differ significantly from the actual values, and overcompensation may occur in which the compensation input calculated by the coupling compensation module (240) deviates significantly from the physical driving range of the driving unit (120).

[0326] To prevent this, the decoupling matrix derivation unit (241) verifies numerical stability for at least one of the coupling matrix and the decoupling matrix, and if it does not satisfy a predetermined criterion, the decoupling matrix is ​​set as the identity matrix.

[0327] The criteria for verifying numerical stability by the decoupling matrix derivation unit (241) are three: condition number, determinant, and diagonal dominance, and numerical stability is determined based on at least one of these.

[0328] The first verification criterion is the condition number. The condition number is an indicator that quantifies the numerical stability of a matrix, and for an n×n matrix A, the condition number is defined as the ratio of A's maximum singular value (σ_max) to its minimum singular value (σ_min).

[0329] Here, the singular value is defined as the positive square root of the eigenvalues ​​of A^TA for matrix A, and serves as a measure of how much the matrix stretches or compresses the input vector. This can be expressed mathematically as follows.

[0330] formula.

[0331]

[0332] Here, cond(A) represents the condition number of matrix A, σ_max(A) represents the maximum singular value of matrix A, and σ_min(A) represents the minimum singular value of matrix A. The closer the condition number is to 1, the more numerically stable the matrix is; the larger the condition number, the greater the influence of measurement error on the estimation result when calculating the inverse matrix.

[0333] The decoupling matrix derivation unit (241) determines that numerical stability is not satisfied if the number of conditions of the combination matrix or decoupling matrix is ​​greater than a predetermined condition number threshold (κ_th). The condition number threshold (κ_th) can be set in the range of 50 to 200.

[0334] For example, if κ_th ​​is set to 100, if the number of conditions of the combination matrix is ​​greater than 100, it is determined that numerical stability is not satisfied.

[0335] The second verification criterion is the determinant. The determinant is an indicator used to determine the invertibility of a matrix; the closer the determinant of an n×n square matrix A is to zero, the more likely the matrix is ​​effectively irreversible or has very low invertibility, which increases the likelihood of numerical instability when calculating the inverse matrix.

[0336] The decoupling matrix derivation unit (241) determines that numerical stability is not satisfied if the absolute value of the determinant of the combined matrix is ​​smaller than a predetermined determinant reference value (ε_det).

[0337] The determinant value (ε_det) can be set to an appropriate value depending on the element size and number of zones of the combined matrix, for example, in an air cell mattress system composed of two zones, it can be set to a level of 1% to 5% of the absolute value of the determinant of a normal combined matrix.

[0338] The third verification criterion is diagonal dominance. Diagonal dominance refers to the property where the absolute value of the diagonal elements in each row of a matrix is ​​greater than the sum of the absolute values ​​of the non-diagonal elements in that row, and this can be expressed mathematically as follows.

[0339] formula.

[0340]

[0341] Here, C_ii is the i-th row and i-th column diagonal element of the combined matrix, C_ij is the i-th row and j-th column non-diagonal element of the combined matrix, and α is the diagonal dominance slack factor, which is set to a value greater than 1.0, for example, 1.2. If α = 1.0, it is a strict diagonal dominance condition, and if α > 1.0, it is a strong diagonal dominance condition including slack.

[0342] If the diagonal dominance condition is not satisfied, the non-diagonal elements representing cross-interference in the coupling matrix are relatively larger than the diagonal elements representing self-response, so there is a high risk of overcompensation occurring when applying the decoupling matrix. The decoupling matrix derivation unit (241) determines that numerical stability is not satisfied if the coupling matrix does not satisfy the diagonal dominance condition.

[0343] The decoupling matrix derivation unit (241) sets the decoupling matrix to the identity matrix (I) if it is determined that at least one of the three verification criteria above is not satisfied.

[0344] When the decoupling matrix is ​​set to an identity matrix, the coupling compensation module (240) operates in an independent pressure-following control state in which the virtual input calculated by the PI controller or PID controller of each zone is applied directly to the driving unit (120) without coupling compensation, thereby protecting the system of the present invention from overcompensation caused by the numerically unstable decoupling matrix.

[0345] Examples and comparative examples regarding this are as follows.

[0346] As an example, the case of the normal combination matrix calculated earlier is applied. In this combination matrix, cond(C) = 1.28, which is smaller than the condition number criterion (κ_th = 100), and det(C) = 17,602,153, which is sufficiently large for the absolute determinant. When examining the first row for the diagonal dominance condition, |C_11| = 4155 and α·|C_12| = 1.2×471 = 565.2, satisfying 4155 > 565.2, and when examining the second row, |C_22| = 4299 and α·|C_21| = 1.2×552 = 662.4, satisfying 4299 > 662.4, thus passing all three verification criteria.

[0347] In this case, the decoupling matrix derivation unit (241) derives the inverse matrix as the decoupling matrix, and the coupling compensation module (240) produces a normal compensation input.

[0348] As a comparative example, it is assumed that in an environment where measurement noise is excessively mixed and the amplitude of the identification signal is too small, resulting in a reduced signal-to-noise ratio, the estimated combination matrix is ​​calculated as follows.

[0349]

[0350] Here, C_bad represents the comparison combination matrix.

[0351] The fact that the absolute value of the off-diagonal element (-3980, -4020) in this combination matrix is ​​estimated to be close to the absolute value of the diagonal element (4155, 4299) is an example of a case where noise is excessively mixed into the off-diagonal element estimate value in a situation where the pressure response becomes equivalent to the measurement noise level of the pressure sensor (131) by applying an identification signal with an under-amplitude (about 16 Pa) of 0.2% or less relative to the base pressure.

[0352] In this case, cond(C_bad) increases significantly and exceeds the condition number threshold (κ_th = 100), and det(C_bad) = 4155×4299 - 3980×4020 = 17,862,345 - 15,999,600 = 1,862,745, which is about 10.6% of the determinant of the normal combination matrix (17,602,153), and the diagonal dominance condition is also not satisfied in the first row, as |C_11| = 4155 and α·|C_12| = 1.2×3980 = 4776, which is 4155 < 4776.

[0353] Since two or more of the three numerical stability verification criteria are determined to be unsatisfactory, the decoupling matrix derivation unit (241) sets the decoupling matrix as the identity matrix.

[0354] In the comparative example, if the inverse matrix of C_bad is applied as a decoupling matrix without numerical stability verification, the inverse matrix elements are calculated to be tens of times larger than the actual decoupling matrix elements, resulting in saturation where the compensation input significantly exceeds the driving range of the valve (121) and pressure interference between zones is amplified. On the other hand, in the embodiment where numerical stability verification is applied, the independent pressure tracking control state is maintained through identity matrix return, thereby preventing pressure instability caused by overcompensation.

[0355] In summary, the decoupling matrix derivation unit (241) performs numerical stability verification based on condition number, determinant, and diagonal dominance, thereby preventing numerically unstable decoupling matrices from being used for coupling compensation due to measurement noise or coupling matrix estimation error, and safely returns to independent pressure tracking control in an abnormal state, thereby providing a characteristic that can simultaneously protect the pressure control stability of the air cell mattress system and the user's perceived quality.

[0356] Furthermore, even if the decoupling matrix derivation unit (241) passes the numerical stability verification and derives the decoupling matrix from the combination matrix, there may be cases where the derived decoupling matrix does not accurately reflect the combination characteristics of the actual air cell mattress system due to errors introduced during the estimation process of the combination matrix or the non-linear characteristics of the air cell (110).

[0357] In this case, the compensation input calculated by the combination compensation module (240) does not correspond to the actual combination characteristics, so there is a possibility that the pressure interference between zones may actually increase.

[0358] To prevent this, the decoupling matrix derivation unit (241) applies a verification input to the input zone after deriving the decoupling matrix, and if the error between the predicted pressure change amount calculated based on the coupling matrix and the actual pressure change amount measured by the verification input is greater than the verification threshold, the decoupling matrix can be set as an identity matrix.

[0359] At this time, the term "verification input" refers to a small pressure control input applied to an input zone to verify whether the derived decoupling matrix accurately reflects the coupling characteristics of the actual air cell mattress system after the decoupling matrix is ​​derived.

[0360] The amplitude of the verification input is set to a value that is smaller than the amplitude of the identification signal used to calculate the coupling matrix but sufficiently larger than the measurement noise level of the pressure sensor unit (130), and is applied at a level that generates a minimum pressure change necessary to verify the coupling characteristics while minimizing the user's sensation. The waveform of the verification input can be implemented in the form of a pulse or a step, and the duration can be set to a range of 0.2 seconds to 1.0 seconds.

[0361] First, the decoupling matrix derivation unit (241) calculates the theoretical pressure change amount, i.e., the predicted pressure change amount, that must be observed in each output zone when a verification input is applied, in order to verify whether the derived decoupling matrix accurately reflects the coupling characteristics of the current air cell mattress system.

[0362] The predicted pressure change is a value calculated by multiplying the verification input by the coupling matrix under the assumption that the coupling matrix perfectly reflects the coupling characteristics of the actual air cell mattress system, and serves as a criterion for retrospectively verifying the accuracy of the coupling matrix through comparison with the actual pressure change.

[0363] For example, if the coupling matrix is ​​accurate, the actual pressure change when a verification input is applied should deviate from the predicted pressure change by only the error level of the measurement noise, and if the coupling characteristics change and the coupling matrix does not reflect the current state, the actual pressure change will differ significantly from the predicted pressure change.

[0364] The decoupling matrix derivation unit (241) can calculate the predicted pressure change amount, for example, using the following formula.

[0365] formula.

[0366]

[0367] As another example, it can also be implemented by performing independent linear regression for each output zone to calculate the predicted pressure change.

[0368] Next, the decoupling matrix derivation unit (241) actually applies the verification input to the input zone and collects the actual pressure change amount (ΔP_meas) of each output zone measured by the pressure sensor unit (130) at the time of response extraction.

[0369] As previously defined, the response extraction point is determined as the point at which the rate of change of pressure in the output zone becomes smaller than the response stability threshold.

[0370] Next, the decoupling matrix derivation unit (241) calculates an error that combines the difference between the calculated predicted pressure change amount and the measured actual pressure change amount into a single scalar value. At this time, the larger the error, the more the coupling matrix fails to reflect the coupling characteristics of the current air cell mattress system.

[0371] In order to comprehensively reflect the prediction errors occurring in multiple output zones, the decoupling matrix derivation unit (241) preferably calculates the error as the Euclidean norm of the error vector, that is, the square root of the sum of squares of the prediction errors in each output zone.

[0372] The reason for using the Euclidean norm is that it allows for verification judgment with a single scalar value while equally reflecting prediction errors occurring simultaneously in multiple output zones.

[0373] As an example of this, the decoupling matrix derivation unit (241) can calculate the error using the following formula.

[0374] formula.

[0375]

[0376] Here, e_verify represents the error between the predicted pressure change and the actual pressure change, ΔP_meas represents the actual pressure change vector, ΔP_pred represents the predicted pressure change vector, and ||·||_2 represents the Euclidean norm.

[0377] As another example, it can also be implemented in a form using the L1 norm (||ΔP_meas - ΔP_pred||1), which is the sum of the absolute values ​​of the errors of each output zone, or the maximum value norm (||ΔP_meas - ΔP_pred||∞), which is the absolute value of the maximum component of the error vector, as the error.

[0378] The decoupling matrix derivation unit (241) determines that the coupling matrix does not accurately reflect the coupling characteristics of the current air cell mattress system when the calculated error (e_verify) is greater than the verification threshold (ε_verify), and sets the decoupling matrix as the identity matrix.

[0379] The verification threshold (ε_verify) is set based on the measurement noise standard deviation (σ_sensor) of the pressure sensor unit (130) and can be set to a range of 2 to 3 times σ_sensor.

[0380] The above explanation is illustrated with an example where numerical values ​​are substituted as follows.

[0381] It is assumed that under the condition of the measurement noise standard deviation of the pressure sensor (131) σ_sensor = 15 Pa, the verification threshold is set to ε_verify = σ_sensor × 2 = 30 Pa, and the verification input Δu_verify = [0.005 duty, 0 duty]^T is applied to the first zone.

[0382] The predicted pressure change amount calculated by the decoupling matrix derivation unit (241) based on the previously calculated coupling matrix is ​​ΔP_pred = [4155×0.005, -552×0.005]^T = [20.8 Pa, -2.8 Pa]^T.

[0383] As an example, when the coupling matrix accurately reflects the coupling characteristics of the current air cell mattress system, if the actual pressure change amount measured by the pressure sensor unit (130) is ΔP_meas = [19.5 Pa, -2.5 Pa]^T, the error calculated by the decoupling matrix derivation unit (241) is e_verify = ||[-1.3, 0.3]||_2 = √(1.69+0.09) = approximately 1.33 Pa.

[0384] At this time, since 1.33 Pa < 30 Pa, the verification is passed and the decoupling matrix derivation unit (241) retains the derived decoupling matrix as is.

[0385] As a comparative example, when the magnetic gain of the first zone decreases and the cross gain of the second zone increases significantly due to a change in the user's posture, if the actual pressure change amount measured by the pressure sensor unit (130) is ΔP_meas = [2.5 Pa, -30.0 Pa]^T, the error calculated by the decoupling matrix derivation unit (241) is e_verify = ||[2.5-20.8, -30.0-(-2.8)]||_2 = ||[-18.3, -27.2]||_2 = √(334.89+739.84) = √1074.73 = approximately 32.78 Pa.

[0386] Here, since 32.78 Pa > 30 Pa, the decoupling matrix derivation unit (241) determines that the coupling matrix does not reflect the current coupling characteristics and sets the decoupling matrix to an identity matrix so that the coupling compensation module (240) operates in an independent pressure-following control state.

[0387] If the decoupling matrix derived without verification is applied as is, the compensation input is inconsistent with the changed coupling characteristics, which actually amplifies pressure interference between zones, whereas unnecessary overcompensation can be prevented by returning to the identity matrix through verification.

[0388] In summary, the decoupling matrix derivation unit (241) performs post-verification through a verification input after deriving the decoupling matrix, thereby securing a dual safety mechanism that additionally detects cases where the decoupling matrix that has passed numerical stability verification is inconsistent with the coupling characteristics of the actual air cell mattress system and returns to the identity matrix, thereby protecting the pressure control stability of the air cell mattress system and the user's perceived quality without overcompensation due to coupling matrix estimation error or changes in user posture.

[0389] The aforementioned coupling matrix has time-varying characteristics that vary depending on the user's weight, posture, and the material deformation characteristics of the air cell (110).

[0390] If the coupling matrix calculated through initial identification is used as is even after a change in the user's posture or load distribution, the compensation input calculated by the coupling compensation module (240) may not be consistent with the current coupling characteristics, and thus may not sufficiently reduce or may instead amplify the pressure interference between zones.

[0391] To resolve this, the combination matrix calculation module (230) further includes a re-identification trigger unit (232) that monitors changes in the user's load distribution and the elapsed time since the last identification and determines the need to recalculate the combination matrix based on this.

[0392] FIG. 5 is a flowchart illustrating the re-identification trigger and decoupling matrix update operation of the present invention.

[0393] The first monitoring indicator of the re-identification trigger unit (232) is the amount of change in load distribution.

[0394] The re-identification trigger unit (232) continuously monitors the amount of change in load distribution and the elapsed time since the last identification, and if at least one of these is greater than each re-identification threshold, it generates a re-identification trigger, and the re-identification trigger unit (232) causes the combination matrix calculation module (230) to recalculate the combination matrix in response to the re-identification trigger.

[0395] The load distribution is a vector composed of the long-term average pressure of each of the multiple zones as a component, which quantitatively represents the state in which the user's body weight is distributed to each zone on the mattress body (100).

[0396] Long-term average pressure is defined as the output value of a pressure signal to which a low-pass filter is applied to smooth short-term pressure fluctuations, and the time constant of the low-pass filter can be set to a range of several seconds to tens of seconds.

[0397] As an example, the re-identification trigger unit (232) can define the load distribution vector using the following formula.

[0398] formula.

[0399]

[0400] Here, L is a load distribution vector consisting of n zones, and P_bar_i represents the long-term average pressure of the i-th zone.

[0401] As another example, it can also be implemented by using characteristic values, such as pressure distribution in each zone or the relative pressure ratio between sensors, as proxy indicators of the load distribution.

[0402] The re-identification trigger unit (232) calculates the amount of change between the load distribution vector (L_t) at the current time and the load distribution vector (L_ref) at the time when the last combination matrix was calculated.

[0403] The amount of change in load distribution is an indicator that combines the difference between two load distribution vectors into a single scalar value, and must be able to equally reflect load changes occurring simultaneously in multiple zones.

[0404] In this regard, the re-identification trigger unit (232) can calculate the amount of change in load distribution using the following formula.

[0405] formula.

[0406]

[0407] Here, |ΔL| represents the change in load distribution, L_t is the load distribution vector at the current time point, L_ref is the load distribution vector at the time point of calculating the last combination matrix, and ||·||_2 is the Euclidean norm representing the square root of the sum of the squares of each component of the vector.

[0408] The reason for using the Euclidean norm is that it allows for trigger determination using a single scalar value while comprehensively reflecting load changes occurring in multiple zones regardless of direction.

[0409] As another example, it may also be implemented in a form where the maximum value norm (||L_t - L_ref||_∞), which is the maximum value of the change in each component of the load distribution vector, or the L1 norm (||L_t - L_ref||_1), which is the sum of the absolute values ​​of the change in each component, is used as the change in the load distribution.

[0410] The re-identification trigger unit (232) determines that if the amount of change in load distribution |ΔL| is greater than the load distribution re-identification threshold (L_th), it is highly likely that the combination matrix has changed due to a change in the user's posture or load movement, and generates a re-identification trigger.

[0411] The load distribution re-identification threshold (L_th) can be set to a value corresponding to 5% to 15% of the base pressure of each zone, and when the base pressure is 10 kPa, L_th can be set to a range of 500 Pa to 1500 Pa.

[0412] This is explained as follows using an example where numerical values ​​are substituted.

[0413] In an air cell mattress system composed of two zones, when L_th is set to 800 Pa, when the load distribution vector at the time of the last combined matrix calculation is L_ref = [8000 Pa, 8000 Pa]^T and the load distribution vector at the current time is L_t = [8700 Pa, 7400 Pa]^T, the amount of change in load distribution calculated by the re-identification trigger unit (232) is |ΔL| = ||[8700-8000, 7400-8000]||_2 = ||[700, -600]||_2 = √(700²+600²) = √(490000+360000) = √850000 = approximately 922 Pa.

[0414] At this time, since 922 Pa > 800 Pa, the re-identification trigger unit (232) generates a re-identification trigger.

[0415] The second monitoring indicator of the re-identification trigger unit (232) is the elapsed time since the last time the combination matrix was calculated.

[0416] Even if a change in the user's posture is not detected, the bonding characteristics may gradually change due to viscoelastic deformation of the air cell (110) material, temperature change, or long-term load transfer, so it is desirable to recalculate the bonding matrix after a certain period of time has elapsed since the last identification.

[0417] The re-identification trigger unit (232) generates a re-identification trigger when the elapsed time from the time of the last combined matrix calculation is greater than the time-based re-identification threshold (T_ID).

[0418] The time-based re-identification threshold (T_ID) can be set in the range of minutes to tens of minutes. For example, if T_ID is set to 30 minutes, a re-identification trigger occurs regardless of the amount of change in load distribution after 30 minutes have elapsed since the last combination matrix calculation.

[0419] When a re-identification trigger occurs, the re-identification trigger unit (232) instructs the combination compensation module (240) to set the decoupling matrix to the identity matrix to temporarily suspend combination compensation, and when the identification execution condition determination module (210) determines the identification execution condition and it is satisfied, the identification signal application module (220) applies the identification signal so that the combination matrix calculation module (230) recalculates the combination matrix.

[0420] When the recalculation is completed, the decoupling matrix derivation unit (241) derives a new decoupling matrix from the updated combination matrix, and the combination compensation module (240) resumes the calculation of the compensation input.

[0421] In summary, the re-identification trigger unit (232) actively responds to the time-varying nature of the coupling matrix by double-monitoring the amount of change in load distribution and the elapsed time, thereby continuously maintaining the compensation accuracy of the coupling compensation module (240) by recalculating the coupling matrix in a timely manner even if a change in the user's posture or long-term load movement occurs, and provides the advantage of stably responding to undetectable gradual changes in coupling characteristics by using a time-based trigger in parallel.

[0422] When the system of the present invention operates in an environment where multiple users use it simultaneously, cross-user pressure interference may occur, in which the adjustment of one user's air cell mattress system affects the pressure in the zone occupied by another user.

[0423] The basic coupling compensation performed by the coupling compensation module (240) described above compensates for the coupling characteristics between zones equally, so it does not distinguish between interference between zones within the same user and interference between zones between different users.

[0424] However, in a multi-user environment, since cross-user pressure interference directly affects the user's sleep quality, it is desirable to preferentially suppress cross-user pressure interference by relatively increasing the compensation gain for the cross-user component.

[0425] To this end, the controller (200) further includes a user separation module (250) that estimates the occupancy area of ​​each of a plurality of users and calculates a zone-specific membership weight, and the combined compensation module (240) applies a weighting factor greater than 1 to the compensation input based on a cross-user mask.

[0426] FIG. 6 is an example of a mattress top view and zone division in a multi-user environment according to the present invention.

[0427] The user separation module (250) estimates the occupied area of ​​each of the multiple users based on the coordinates of each of the multiple zones and the long-term average pressure, and calculates the membership weight of each user for each of the multiple zones.

[0428] The coordinates of each zone are values ​​representing the position of the corresponding zone on the upper surface of the mattress body (100), and can be defined, for example, as (x_i, y_i) coordinates based on the length direction and width direction of the mattress body (100).

[0429] The long-term average pressure is the pressure signal of each zone to which a low-pass filter has been applied, as previously defined, and reflects the magnitude of the load exerted by the user's body weight on the corresponding zone.

[0430] The user separation module (250) estimates the center point of the user-specific occupied area using the coordinates of each zone and the long-term average pressure.

[0431] For example, the user separation module (250) can estimate the center points μ_a and μ_b of the occupied areas of the two users by applying a k-means clustering (k=2) algorithm using long-term average pressure as a weight, and can also be implemented by applying a Gaussian Mixture Model (GMM) for estimation.

[0432] The user separation module (250) calculates user-specific membership weights for each zone based on the estimated center point of the occupied area.

[0433] At this time, the membership weight w_a(i) is an indicator that represents the degree to which the i-th zone belongs to the a-th user's occupied area as a continuous value between 0 and 1.

[0434] The reason for defining membership weights as continuous values ​​is that it is difficult to determine dichotomously which user a zone belongs to in boundary zones or zones where the occupied areas of two users overlap.

[0435] The user separation module (250) can calculate membership weights by applying a soft assignment algorithm based on the distance between the coordinates of each zone and the center point of the user's occupied area, for example.

[0436] In this case, the membership weight of the i-th zone for the a-th user is calculated as a value closer to 1 as the zone coordinates are closer to the center point of the a-th user's occupied area and closer to 0 as the zone coordinates are closer to the center point of the b-th user's occupied area, and the sum of the membership weights of the a-th user and the b-th user is normalized so that w_a(i) + w_b(i) = 1.

[0437] Another example is that the membership weight of each zone can be defined as a hard assignment, such that w_a(i) ∈ {0, 1} and w_b(i) = 1 - w_a(i), but in this case, boundary zone processing may be inaccurate.

[0438] The combination compensation module (240) calculates a cross-user mask using the membership weights calculated by the user separation module (250).

[0439] The cross-user mask M_ab(i,j) is an indicator representing the degree to which the i-th zone and the j-th zone belong to different users, reflecting both the component where the i-th zone belongs to the a-th user and the j-th zone belongs to the b-th user, and the opposite component.

[0440] As the cross-user mask is larger, the likelihood that the combination between the i-th zone and the j-th zone corresponds to cross-user interference is higher, so the combination compensation module (240) relatively increases the compensation gain of the corresponding pair to preferentially suppress cross-user pressure interference.

[0441] Preferably, the combined compensation module (240) can calculate a cross-user mask by the following mathematical formula 2.

[0442] Mathematical formula 2.

[0443]

[0444] Here, M_ab(i,j) represents the cross-user mask between the i-th zone and the j-th zone, w_a(i) represents the membership weight of user a for the i-th zone, and w_b(j) represents the membership weight of user b for the j-th zone.

[0445] The first term w_a(i)·w_b(j) of mathematical formula 2 represents the degree to which the i-th zone belongs to the a-th user and the j-th zone belongs to the b-th user, and the second term w_b(i)·w_a(j) represents the opposite direction, that is, the component in which the i-th zone belongs to the b-th user and the j-th zone belongs to the a-th user.

[0446] By summing the two terms, the cross-user mask symmetrically reflects all cases where the i-th zone and the j-th zone belong to different users, and since M_ab(i,j) = M_ab(j,i) holds, the same cross-user mask value is calculated even when the roles of the input zone and the output zone are swapped.

[0447] For example, if the i-th zone completely belongs to the a-th user (w_a(i) = 1, w_b(i) = 0) and the j-th zone completely belongs to the b-th user (w_a(j) = 0, w_b(j) = 1), Equation 2 is calculated as M_ab(i,j) = 1×1 + 0×0 = 1, which has the maximum cross-user mask value.

[0448] On the other hand, when the i-th zone and the j-th zone both belong completely to the same user (w_a(i) = 1, w_b(i) = 0, w_a(j) = 1, w_b(j) = 0), Equation 2 is calculated as M_ab(i,j) = 1×0 + 0×1 = 0, so the cross-user mask value becomes 0.

[0449] In cases where the affiliations of two users are mixed, such as in a boundary zone, it is calculated as an intermediate value between 0 and 1.

[0450] The combined compensation module (240) applies a weighting factor greater than 1 to the compensation input based on the calculated cross-user mask (M_ab(i,j)).

[0451] The weighting factor is set larger as the cross-user mask value increases, and the combined compensation module (240) can define the weighting factor using, for example, the following formula.

[0452] formula.

[0453]

[0454] Here, G_ab(i,j) is a weighting coefficient between the i-th zone and the j-th zone, and ρ is a weighting strength coefficient that controls the cross-user compensation strength and is set to a value greater than 0. If ρ = 0, cross-user weighting is disabled and operates identically to the basic combined compensation of claim 1, and if ρ > 0, the compensation gain for the cross-user component increases to 1 + ρ·M_ab(i,j) > 1 so that cross-user pressure interference is compensated preferentially.

[0455] As another example, it can also be implemented by defining a non-linear function for M_ab(i,j), such as a sigmoid function or threshold-based binary weights as weighting coefficients.

[0456] An example of this is described as follows.

[0457] It is assumed that in an air cell mattress system composed of 6 zones, zones 1 through 3 are mainly occupied by user A and zones 4 through 6 are mainly occupied by user B.

[0458] If the membership weights calculated by the user separation module (250) using a soft assignment algorithm are w_a(3) = 0.7 and w_b(3) = 0.3 for the third zone, and w_a(4) = 0.3 and w_b(4) = 0.7 for the fourth zone, then the cross-user mask between the third zone (i=3) and the fourth zone (j=4) is calculated by Equation 2 as M_ab(3,4) = w_a(3)·w_b(4) + w_b(3)·w_a(4) = 0.7×0.7 + 0.3×0.3 = 0.49 + 0.09 = 0.58.

[0459] When ρ is set to 1.0, the weighting coefficient between the 3rd zone and the 4th zone is calculated as G_ab(3,4) = 1 + 1.0×0.58 = 1.58.

[0460] On the other hand, the cross-user mask between the first zone (i=1) and the second zone (j=2) (both occupied by user A, w_a(1) = approximately 1, w_b(1) = 0, w_a(2) = approximately 1, w_b(2) = approximately 0) is calculated as M_ab(1,2) = w_a(1)·w_b(2) + w_b(1)·w_a(2) = 1×0 + 0×1 = 0, so the weighting coefficient is G_ab(1,2) = 1, with no additional weighting.

[0461] Through this, the compensation gain for cross-user pressure interference between boundary zones occupied by different users is set to be 1.58 times greater than the compensation gain between zones within the same user, thereby preferentially suppressing cross-user pressure interference.

[0462] Meanwhile, the user separation module (250) includes a safety mechanism that calculates user separation reliability and disables cross-user weighting if user separation is uncertain.

[0463] User separation confidence (Conf) is the average of the absolute values ​​of the difference in membership weights between two users across multiple zones; the closer it is to 1, the clearer the user separation, and the closer it is to 0, the more the occupied areas of the two users are mixed, indicating an uncertain separation.

[0464] As an example, the user separation module (250) can calculate the user separation reliability using the following formula.

[0465] formula.

[0466]

[0467] Here, Conf is the user separation confidence, n is the number of zones, w_a(i) is the membership weight of user a for the i-th zone, and w_b(i) is the membership weight of user b for the i-th zone.

[0468] The user separation module (250) determines that the occupied areas of the two users are not sufficiently separated when the calculated user separation confidence (Conf) is smaller than a predetermined confidence threshold (δ_conf), disables cross-user weighting (ρ = 0), and returns to the basic combination reward.

[0469] In summary, the user separation module (250) continuously calculates user-specific membership weights based on the coordinates of each zone and long-term average pressure, and the combined compensation module (240) relatively increases the compensation gain for cross-user components based on the cross-user mask of Equation 2, thereby providing the advantage of preferentially suppressing cross-user pressure interference in which one user's pressure adjustment interferes with another user's sleep in an environment where multiple users use the air cell mattress system simultaneously.

[0470] In addition, through the continuous processing of membership weights, stable compensation is performed even in boundary zones where the occupancy areas of two users overlap, while a user separation reliability-based deactivation safety mechanism ensures stable operation of the system even in situations where occupancy separation is uncertain.

[0471] As explained above, the configuration and operation of an air cell mattress system having an air cell coupling compensation control function according to the present invention have been described and drawn in the above description and drawings. However, this is merely an example, and the concept of the present invention is not limited to the above description and drawings. It is understood that various changes and modifications are possible within the scope of the technical concept of the present invention. Explanation of the symbols

[0472] 100: Mattress body 110: Air cell 120: Actuator 121: Valve 122: Pump 123: Manifold 130: Pressure sensor unit 131: Pressure sensor 132: Scanning valve 200: Controller 210: Identification Execution Condition Determination Module 220: Identification Signal Application Module 221: Calibration section 222: Safety suspension section 230: Combined Matrix Calculation Module 231: Normalization Estimator 232: Re-identification Trigger Section 240: Combined Compensation Module 241: Decoupling Matrix Derivation Unit 242: Compensation Input Calculation Unit 250: User Separation Module

Claims

Claim 1 A mattress body equipped with multiple air cells, divided into multiple zones; a driving unit comprising multiple valves connected to the air cells to perform filling and exhausting, and a pump that supplies air to the air cells through the valves; and a pressure sensor unit that measures the pressure of each air cell. An air cell mattress system characterized by comprising: a controller including: an identification execution condition determination module for each of the above zones, which determines an identification execution condition in which at least one of the pressure change rate and the user's body movement index is maintained in a state smaller than a stable threshold for a stable threshold time; an identification signal application module that, when the identification execution condition is satisfied, applies an identification signal having an amplitude limited to a user's perceived threshold to an input zone among the plurality of zones, and fixes the pressure control of the driving unit for other zones during the application period of the identification signal and the response stabilization period after the application ends; a combination matrix calculation module that calculates a combination matrix having an element defined as the ratio of the pressure change amount of the output zone to the magnitude of the identification signal applied to the input zone at a response extraction point where the pressure change rate of the output zone after the application of the identification signal ends is smaller than a response stable threshold; and a combination compensation module that calculates a compensation input to reduce pressure interference between zones using the combination matrix and controls the driving unit with the compensation input. Claim 2 An air cell mattress system according to claim 1, wherein the coupling compensation module comprises a decoupling matrix derivation unit that derives a decoupling matrix from the coupling matrix, and a compensation input calculation unit that calculates the compensation input by applying the decoupling matrix to a virtual input vector calculated based on the zone-specific pressure error and performing a saturation operation. Claim 3 An air cell mattress system according to claim 1, wherein the identification signal application module further comprises a calibration unit that applies a test signal with a stepwise increase in amplitude to a user in an awake state, and calculates the perceived threshold from the amplitude of the test signal at a point in time when at least one of the user's feedback and the user's body movement indicator is detected. Claim 4 An air cell mattress system according to claim 1, wherein the identification signal application module further comprises a safety interruption unit that stops the application of the identification signal when a state is detected in which at least one of the pressure change amount, pressure change rate of the input zone, and the user's body movement index is greater than a interruption threshold during the application period of the identification signal, and re-applies an identification signal with reduced amplitude after a predetermined waiting time has elapsed following the suspension of the application of the identification signal. Claim 5 An air cell mattress system according to claim 1, wherein the pressure sensor unit comprises a pressure sensor and a scanning valve sequentially connected to a plurality of air cells to allow the pressure sensor to sequentially measure the pressure of the air cells, wherein the time for the scanning valve to circulate through the plurality of air cells once is defined as a scanning period, and wherein the coupling matrix calculation module determines the point in time at which the pressure change rate of the output zone becomes smaller than the response stability threshold after at least one scanning period has elapsed following the termination of the application of the identification signal as the response extraction point. Claim 6 An air cell mattress system according to claim 1, wherein the combination matrix calculation module further comprises a normalization estimation unit that constructs an input matrix and a response matrix by applying the identification signal to different input zones multiple times, and estimates the combination matrix from the input matrix and the response matrix by the following mathematical formula 1. (Here, U is the input matrix, P is the response matrix, λ is the normalization coefficient, I is the identity matrix, is the estimated combination matrix) Claim 7 An air cell mattress system according to claim 2, wherein the decoupling matrix derivation unit verifies numerical stability based on at least one of a condition number, a determinant, and diagonal dominance for at least one of the coupling matrix and the decoupling matrix, and if the numerical stability does not satisfy a predetermined criterion, the decoupling matrix is ​​set as an identity matrix. Claim 8 An air cell mattress system according to claim 2, wherein the decoupling matrix derivation unit applies a verification input to the input zone after updating the decoupling matrix, and sets the decoupling matrix as an identity matrix when the error between the predicted pressure change amount calculated based on the coupling matrix and the actual pressure change amount measured by the verification input is greater than a verification threshold. Claim 9 An air cell mattress system according to claim 1, wherein the coupling matrix calculation module further comprises a re-identification trigger unit that generates a re-identification trigger when at least one of the amount of change in load distribution calculated from the long-term average pressure of the plurality of zones and the elapsed time since the last time the coupling matrix was calculated is greater than a re-identification threshold, and recalculates the coupling matrix in response to the re-identification trigger. Claim 10 In claim 1, the controller further comprises a user separation module that estimates the occupied area of ​​each of the plurality of users based on the coordinates of each of the plurality of zones and the long-term average pressure, and calculates the membership weight of the a-th user for the i-th zone for each of the plurality of zones, and the combined compensation module calculates a cross-user mask for pairs of the i-th zone and the j-th zone for different a-th user and b-th user by the following Equation 2, and applies a weighting coefficient greater than 1 to the compensation input based on the cross-user mask, characterized in that the air cell mattress system. Equation 2. (Here, M_ab(i,j) is the cross-user mask, w_a(i) is the membership weight of the a-th user for the i-th zone, and w_b(j) is the membership weight of the b-th user for the j-th zone)

Images