A sleep regulation method and system based on dynamic fitting and cooperative rotation control

Abstract

This invention provides a sleep regulation method and system based on dynamic fit and collaborative rotation control, belonging to the field of smart home technology. The invention utilizes a macroscopic rotating skeletal device to slowly and continuously adjust the body's positional angle on the supporting platform, enabling the weight-bearing parts of the body to shift over time. A microscopic adaptive fit device dynamically compensates for local support height, ensuring continuous fit during rotation and reducing shear force and friction. Through multimodal perception and stage discrimination, intervention is performed only during the most suitable sleep stage, reducing the risk of awakening due to control actions. By introducing a graph structure spatiotemporal completion model, specific optimization is performed for scenarios with missing pressure data, restoring the usable pressure field and maintaining the effectiveness of subsequent control links. This solves the technical problems of insufficient static support, significant shear discomfort during rotation, and unstable control in scenarios with missing pressure data in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of smart home technology, specifically to a sleep regulation method and system based on dynamic fit and coordinated rotation control. Background Technology

[0002] During sleep, people typically need to turn over, adjust to their side, or make slight movements to relieve localized circulatory disturbances and discomfort caused by prolonged pressure on the same area. Traditional mattresses mostly improve support by using zoned firmness adjustment, fixed tilt angles, or timed rotation. However, these solutions usually only provide static or semi-static support and cannot simulate the natural positional shifts of the body during sleep, nor can they achieve continuous, smooth, and timely pressure transfer without disrupting the user's sleep continuity. Furthermore, existing static or semi-static adjustment methods cannot fundamentally simulate the natural turning process, limiting their effectiveness in preventing pressure injuries and relieving pressure during deep sleep.

[0003] On the other hand, while some solutions consider bed rotation, they often overlook the issues of relative sliding, shearing forces, and friction between the body and the bed surface during rotation. Furthermore, they fail to adequately consider the impact of differences in user body shape, lack of proper fit in certain areas, and variations in sleep stages on the control effect. In some solutions, while macroscopic skeletal rotation can change the points of force application, the lack of active, microscopic-level conformity and tracking can still cause discomfort or even wakefulness.

[0004] Furthermore, in real-world usage scenarios, mattress covers, thick blankets, localized nursing pads, or abnormal user postures can all obstruct the pressure sensing area, potentially causing localized low response, saturation, or temporary loss of response. If traditional methods of local linear interpolation or simply ignoring missing points are still used, the true body pressure distribution cannot be accurately restored when the missing area is large or the missing location happens to be in a critical pressure zone. This leads to inaccurate judgment of intervention timing, unreasonable selection of rotation direction, and distortion of the fit compensation amplitude. Traditional methods often fail to achieve ideal results in this scenario.

[0005] Therefore, how to establish a sleep regulation system and method that can achieve both macroscopic body position adjustment and dynamic microscopic surface fit during sleep, and maintain high reliability even in special scenarios where pressure data is incomplete, has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] The purpose of this invention is to provide a sleep regulation method based on dynamic fit and coordinated rotation control, so as to at least solve the technical problems in the prior art such as insufficient static support, obvious shear discomfort during rotation, and unstable control in scenarios with missing pressure data.

[0007] To achieve the above objectives, a first aspect of the present invention provides a sleep regulation method based on dynamic fit and coordinated rotation control, the method comprising: Acquire stress perception data, vital sign data, and motion posture data, and perform unified time alignment processing to construct sleep state data frames; Based on the pressure sensing data in the sleep state data frame, pressure calibration, missing data identification, and pressure field completion processing are performed to generate completed pressure field data, and a body surface contour description result and a fitting reference model are generated based on the completed pressure field data. Based on the vital signs data, the completed pressure field data, and the movement posture data, the sleep stage discrimination result and the pressure discomfort risk index are calculated. When the pressure discomfort risk index meets the preset triggering conditions and the sleep stage discrimination result belongs to the preset allowable intervention stage, a coordinated rotation and fit control sequence is generated based on the fit reference model, the completed pressure field data, and the motion posture data; wherein, the coordinated rotation and fit control sequence includes a rotation control parameter sequence and a fit displacement control parameter sequence; The coordinated rotation and fitting control sequence is output, and the user-individualized intervention model is updated based on the post-intervention effect feedback data.

[0008] Optionally, acquire stress-sensing data, vital sign data, and motion posture data, and perform unified time alignment processing to construct sleep state data frames, including: The system collects raw pressure data from the pressure sensor array, raw vital sign data from the non-contact bioradar, and raw motion attitude data from the inertial measurement unit, and adds sampling time markers to each type of raw data. Using a preset time grid as a unified time axis, resampling and interpolation processes are performed on the original pressure data, the original vital signs data, and the original motion posture data to generate time-consistent pressure sequences, vital signs sequences, and posture sequences. Sleep state data frames are constructed based on aligned stress sequences, vital sign sequences, and posture sequences, ensuring that stress information, vital sign information, and posture information are associated simultaneously under the same time index.

[0009] Optionally, pressure calibration, missing data identification, and pressure field completion processing are performed based on the pressure-sensing data in the sleep state data frame to generate completed pressure field data. A body surface contour description result and a fitting reference model are then generated based on the completed pressure field data, including: Zero-point drift compensation, temperature drift compensation, and range normalization are performed on the time-aligned pressure sequence to generate calibrated pressure field data. Based on the calibrated pressure field data, the locations of abnormal distortion, occlusion, and saturation are identified, and an effective sensor mask is generated. When the missing proportion of the effective sensor mask representation is greater than a preset threshold, the graph structure spatiotemporal completion model is invoked, and the missing regions are completed and reconstructed by combining vital sign information, attitude information and historical pressure field information to generate the completed pressure field data. Based on the completed pressure field data, the surface contour description results are extracted, and a bonding reference model is generated to characterize the geometry of the target bonding surface.

[0010] Optionally, based on the vital sign data, the completed pressure field data, and the movement posture data, a sleep stage discrimination result is calculated, including: Based on the vital signs data, heart rate features, respiratory rate features, and heart rate variability features are extracted. Based on the completed pressure field data, body motion features, pressure stability features, and pressure zone change features are extracted. Based on the motion posture data, posture angle features, posture change rate features, and bed motion following features are extracted. The heart rate characteristics, respiratory rate characteristics, heart rate variability characteristics, body movement characteristics, pressure stability characteristics, pressure zone change characteristics, posture angle characteristics, posture change rate characteristics, and bed motion following characteristics are fused to generate a stage discrimination feature sequence. Sleep stage discrimination results are calculated based on stage discrimination feature sequences, and intervention-permitted markers are generated.

[0011] Optionally, based on the vital sign data, the completed pressure field data, and the motion posture data, a pressure discomfort risk index is calculated, including: The completed pressure field data is mapped into multiple partitioned pressure results according to human anatomical regions. Based on the partitioned pressure results, the current supine state represented by the motion posture data, and the autonomous micro-movement state represented by the vital signs data, the cumulative pressure result of each partition within the current time window is determined. The risk value of each zone is calculated based on the cumulative pressure results of each zone, and the risk contribution of different zones is weighted according to the lying position. The risk values ​​of each zone are summarized to generate a stress discomfort risk index. The stress discomfort risk index is then compared with preset trigger conditions to generate a risk trigger flag.

[0012] Optionally, a sequence of rotation control parameters is generated based on the fitting reference model, the completed pressure field data, and the motion attitude data, including: Based on the current lying position, the current bed posture, the completed pressure field distribution, and pressure discomfort risk indicators, a set of candidate rotation directions and a set of candidate rotation amplitudes are constructed. Based on the set of candidate rotation directions, the set of candidate rotation amplitudes, and the pressure prediction results in the future time domain, calculate the pressure averaging benefit, attitude disturbance cost, and motion smoothing cost corresponding to each candidate rotation scheme. Based on the pressure equalization benefit, the attitude perturbation cost, and the motion smoothing cost, the optimal rotation scheme that satisfies the bed motion constraint and the sleep intervention constraint is solved. Based on the optimal rotation scheme, a sequence of rotation control parameters consisting of rotation direction, rotation angular velocity, rotation angular displacement, and pitch compensation is generated.

[0013] Optionally, a fitting displacement control parameter sequence is generated based on the fitting reference model, the completed pressure field data, and the rotation control parameter sequence, including: A target pressure reference distribution is generated based on the completed pressure field data and the fitted benchmark model. The local pressure migration trend during the rotation execution process is predicted based on the rotation control parameter sequence, and the target pressure reference distribution is corrected based on the local pressure migration trend. The corrected target pressure reference distribution is compared with the completed pressure field data to generate the pressure error results for each support unit. Based on the pressure error results and the pressure change results at adjacent times, the displacement increment of each support unit is calculated, and a smoothing constraint process is performed on the displacement increment of each support unit to generate a sequence of fitting displacement control parameters.

[0014] Optionally, before generating the cooperative rotation and bonding control sequence, the following steps are also included: The available confidence level of the stress data is calculated based on the effective sensor mask, and the probability of micro-arousal is calculated based on the stage discrimination feature sequence corresponding to the sleep stage discrimination. The risk trigger marker remains valid as long as the confidence level of the available pressure data is not less than a preset confidence threshold and the probability of micro-awakening is not greater than a preset arousal threshold. The cooperative rotation and bonding control sequence is output only within the time window when the risk trigger flag is valid.

[0015] Optionally, the coordinated rotation and fitting control sequence is output, and the user-individualized intervention model is updated based on the post-intervention effect feedback data, including: The coordinated rotation and fitting control sequence is output to execute the intervention process. Micro-body motion results, micro-awakening results, pressure risk reduction results, and posture stability results are collected within a preset time window after the intervention as effect feedback data. The immediate reward result for this intervention is calculated based on the aforementioned effect feedback data; The parameters of the user-individualized intervention model are updated based on the instant reward results. Based on the updated user-individualized intervention model, the trigger threshold, rotation scheme preference, and fit compensation preference in subsequent sleep cycles are corrected.

[0016] A second aspect of the present invention provides a sleep regulation system based on dynamic fit and coordinated rotation control, the system comprising: Base; A support platform, which is mounted on the base; A macroscopic rotating frame device is connected to the support platform via a transmission; wherein, the macroscopic rotating frame device includes at least a horizontal rotation motor and a pitch adjustment motor, used to drive the support platform to rotate continuously in the horizontal plane and perform pitch adjustment around the transverse axis; A micro-adaptive bonding device is disposed on the support platform; wherein, the micro-adaptive bonding device includes a bonding surface composed of independently adjustable support units arranged in a dense array, each of the independently adjustable support units including at least a flexible contact cap, a micro linear electric cylinder, a pressure-sensitive thin film sensor and a unit housing, for collecting local pressure data and performing local height adjustment; A multimodal sensing device; wherein the multimodal sensing device includes at least a pressure sensor array, a non-contact bio-radar, and an inertial measurement unit, for outputting pressure sensing data, vital sign data, and motion posture data, respectively; The central control module is communicatively connected to the macroscopic rotating skeleton device, the microscopic adaptive bonding device, and the multimodal sensing device, respectively, and is used to receive the pressure sensing data, the vital signs data, and the motion posture data, and generate a rotation control parameter sequence and a bonding displacement control parameter sequence based on the pressure sensing data, the vital signs data, and the motion posture data. The central control module is configured to execute the method described in any of the preceding methods.

[0017] Through the above technical solutions, this invention proposes a sleep regulation method and system based on dynamic fit and collaborative rotation control. It utilizes a macroscopic rotating skeletal device to slowly and continuously adjust the body's positional angle on the support platform, enabling the weight-bearing parts of the body to shift over time, thus achieving proactive pressure timing management. A microscopic adaptive fit device dynamically compensates for local support height, ensuring continuous fit during rotation and reducing shear force and friction. Through multimodal perception and stage discrimination, intervention is performed only during suitable sleep stages, reducing the risk of awakening due to control actions. By introducing a graph-structured spatiotemporal completion model, specific optimizations are performed for scenarios with missing pressure data, allowing the system to recover usable pressure fields and maintain the effectiveness of subsequent control links even in special scenarios. Therefore, this invention solves the technical problems of insufficient static support, significant shear discomfort during rotation, and unstable control in scenarios with missing pressure data in existing technologies.

[0018] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the steps of a sleep regulation method based on dynamic fit and coordinated rotation control according to one embodiment of the present invention. Figure 2 This is a schematic diagram of the central control module of a sleep regulation system based on dynamic fit and coordinated rotation control provided in one embodiment of the present invention. Detailed Implementation

[0020] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] In one embodiment of the present invention, a sleep regulation system based on dynamic fitting and coordinated rotation control is proposed. The system includes, from bottom to top, a base, a macroscopic rotating frame device installed in the base, a microscopic adaptive fitting device covering the macroscopic rotating frame device, a multimodal sensing device arranged in the fitting device and the headboard area, and a central control module.

[0022] In this embodiment of the invention, the macroscopic rotating skeleton module includes a drive mechanism that can drive the support platform to rotate continuously 360 degrees in the horizontal plane and can perform pitch angle adjustment around at least one lateral axis; the microscopic adaptive bonding module is disposed on the support platform and includes a bonding surface composed of independently adjustable support units arranged in a dense array; the multimodal sensing network includes at least a pressure sensor array, a non-contact bio-radar, and an inertial measurement unit; the central control unit is communicatively connected to the above devices and outputs coordinated control commands.

[0023] Furthermore, in one feasible implementation, the macroscopic rotating frame device can consist of a base, a horizontal rotating motor, a pitch adjustment motor, and a support platform that is driven and connected to the aforementioned motors. The horizontal rotating motor is used to output horizontal rotational driving force, causing the support platform to rotate slowly and continuously around the vertical axis; the pitch adjustment motor is used to fine-tune the local pitch attitude of the support platform, thereby meeting the support requirements under different sleeping positions.

[0024] Furthermore, in one feasible implementation, the micro-adaptive bonding device comprises a matrix of thousands of independently adjustable support units. Each independently adjustable support unit includes at least a flexible contact cap, a miniature linear electric cylinder, a pressure-sensitive film sensor, and a unit housing. The flexible contact cap is used to contact the human body surface or the bonding layer, improving contact comfort; the miniature linear electric cylinder is used to drive the unit to extend and retract along the height direction; the pressure-sensitive film sensor is used to detect local pressure changes; and the unit housing is used to structurally fix and protect the above components.

[0025] In the multimodal sensing device, a pressure sensor array is used to provide the contact pressure distribution between the human body and the contact surface; a non-contact bio-radar is used to acquire vital signs such as heart rate and respiratory rate; and an inertial measurement unit is used to acquire information on the current posture, motion changes, and motion state during rotation. It is readily understood that this invention is not limited to specific types of vital sign sensors or inertial devices; any device capable of outputting data related to sleep stage discrimination, posture tracking, and control generation can be included within the scope of this invention.

[0026] The central control module can contain a built-in processor, memory, and program instructions for executing intelligent decision-making algorithms. The central control module receives raw data from multimodal sensing devices, performs data alignment, pressure field modeling, sleep stage discrimination, risk assessment, rotation parameter generation, bonding parameter generation, and user model learning and updating. It then sends the rotation control parameters to the macroscopic rotating skeleton device and the bonding displacement control parameters to the microscopic adaptive bonding device. Simultaneously, it receives status feedback data from these devices, forming feedback control.

[0027] In another embodiment of the invention, such as Figure 1As shown, a sleep regulation method based on dynamic fit and coordinated rotation control is provided. Based on the aforementioned sleep regulation system based on dynamic fit and coordinated rotation control, the method includes: S10: Acquire stress perception data, vital sign data, and motion posture data, and perform unified time alignment processing to construct sleep state data frames.

[0028] Specifically, the system collects raw pressure data from a pressure sensor array, raw vital sign data from a non-contact bioradar, and raw motion posture data from an inertial measurement unit, and adds sampling time stamps to each type of raw data. Using a preset time grid as a unified time axis, the system performs resampling and interpolation processing on the raw pressure data, the raw vital sign data, and the raw motion posture data to generate time-consistent pressure, vital sign, and posture sequences. Based on the aligned pressure, vital sign, and posture sequences, a sleep state data frame is constructed, ensuring that pressure, vital sign, and posture information are associated at least simultaneously under the same time index.

[0029] In this embodiment of the invention, the central control module receives raw pressure data output from the pressure sensor array, raw vital sign data output from the non-contact bioradar, and raw motion attitude data output from the inertial measurement unit. It is easy to understand that, because the sampling rates, reporting cycles, and signal delays of different sensors are not consistent, if the raw data is directly used for subsequent calculations, the pressure state, vital sign state, and attitude state at the same moment may not correspond, easily leading to stage discrimination errors, misaligned risk assessments, and lag in control parameter generation. Therefore, it is necessary to first perform unified time alignment processing.

[0030] In practical applications, a fixed, unified time grid can be set, and resampling and interpolation operations can be performed on the three types of raw data respectively, so that the pressure information, vital sign information, and posture information under the same time index correspond to each other. It is easy to understand that after the above processing, a unified sleep state data frame is formed, which at least contains pressure distribution information, heart rate and breathing related information, and bed posture and motion status information at the corresponding time.

[0031] S20: Perform pressure calibration, missing data identification, and pressure field completion processing based on the pressure sensing data in the sleep state data frame to generate completed pressure field data, and generate body surface contour description results and fitting reference model based on the completed pressure field data.

[0032] Specifically, zero-point drift compensation, temperature drift compensation, and range normalization are performed on the time-aligned pressure sequence to generate calibrated pressure field data. Based on the calibrated pressure field data, abnormal distortion locations, occlusion locations, and saturation locations are identified, and effective sensor masks are generated. When the missing proportion represented by the effective sensor mask is greater than a preset threshold, a graph structure spatiotemporal completion model is invoked, and combined with vital sign information, posture information, and historical pressure field information, the missing regions are completed and reconstructed to generate completed pressure field data. Based on the completed pressure field data, the surface contour description results are extracted, and a bonding reference model is generated to characterize the geometry of the target bonding surface.

[0033] In this embodiment of the invention, the central control module reads the pressure distribution data at the current moment from the sleep state data frame and first performs pressure calibration processing. It is easy to understand that pressure-sensitive film sensors may experience problems such as zero-point offset, inconsistent sensitivity, and local saturation under long-term use, temperature changes, repeated loading, or partial coverage conditions. If calibration is not performed, subsequent judgments of pressure thresholds, accumulation of risks, and adjustment of adhesion height will all be deviated.

[0034] In one executable implementation, the central control module can call the corresponding zero-point drift baseline and sensitivity coefficient for each sensing position, and correct the original pressure value by dividing the difference between the collected pressure value and the zero-point drift baseline value by the sensitivity coefficient.

[0035] Furthermore, in this embodiment of the invention, after pressure calibration is completed, a missing data identification process is performed on the pressure field. Specifically, an effective sensor mask can be generated based on whether the sensing unit exhibits long-term low response, abnormal constant value, over-range saturation, or short-term data loss. This effective sensor mask is used to distinguish which regions in the current pressure field are reliable regions and which regions belong to the missing data regions.

[0036] Furthermore, in this embodiment of the invention, when the detected pressure loss ratio exceeds a preset threshold, or when the loss locations are concentrated in key pressure areas, instead of using traditional simple neighborhood interpolation or ignoring methods, a graph-structured spatiotemporal completion model is invoked to reconstruct the pressure field. This is primarily because traditional methods often only restore local smoothing trends when there are thick mattress covers, folded bedding, localized nursing pads, or partial sensor failures, failing to restore the true pressure center and pressure boundaries, thus leading to distortion in body surface contour recognition.

[0037] To address the limitations of traditional methods in this specific scenario, this invention abstracts the spatial adjacency relationships of the pressure array into a graph structure. Simultaneously, it incorporates vital sign information, attitude information, and historical pressure field information as spatiotemporal constraint inputs, thereby achieving structured reconstruction of missing regions. Its expression is as follows: ; in, This represents the completed pressure field data; This indicates the calibration pressure field data; Indicates the valid sensor mask; This represents element-wise multiplication; Represents a spatiotemporal completion model of graph structures; This represents the parameter set of the spatiotemporal completion model of the graph structure; This represents the feature input formed by the fusion of vital signs information, posture information, and historical pressure field information; Indicates an invalid or missing location mask.

[0038] After generating the completed pressure field data, the central control module further extracts the surface contour description results. Specifically, based on the pressure peak region, pressure isopleths, pressure center of gravity trajectory, and gradient distribution in adjacent regions, the main contact contours of the user's back, waist, hips, and lower limbs can be inferred. Furthermore, in one executable implementation, the surface contour description results can be further converted into a fitting reference model. This fitting reference model defines the geometric support shape that the fitting surface should achieve at each position under ideal conditions and provides a target basis for the subsequent calculation of fitting displacement control parameters.

[0039] S30: Based on the vital signs data, the completed pressure field data, and the movement posture data, calculate the sleep stage discrimination result and the pressure discomfort risk index.

[0040] Specifically, heart rate, respiratory rate, and heart rate variability features are extracted based on the vital signs data; body movement, pressure stability, and pressure zone change features are extracted based on the completed pressure field data; and posture angle, posture change rate, and bed motion following features are extracted based on the movement posture data. These features are then fused to generate a stage discrimination feature sequence. Sleep stage discrimination results are calculated based on the stage discrimination feature sequence, and an intervention permission marker is generated.

[0041] Simultaneously, the completed pressure field data is mapped into multiple partitioned pressure results according to human anatomical regions. Based on the partitioned pressure results, the current supine position represented by the motion posture data, and the voluntary micro-movement state represented by the vital signs data, the cumulative pressure result of each partition within the current time window is determined. Based on the cumulative pressure result of each partition, the partition risk value is calculated, and the risk contribution of different partitions is weighted according to the supine position. The risk values ​​of each partition are summarized to generate a pressure discomfort risk index. The pressure discomfort risk index is compared with preset trigger conditions to generate a risk trigger marker.

[0042] In this embodiment of the invention, the central control module reads vital sign information, completed pressure field information, and motion posture information from the current moment and historical time windows, and performs multimodal feature extraction. Regarding vital signs, indicators reflecting the autonomic nervous system state, such as heart rate, respiratory rate, and heart rate variability, can be extracted. Regarding the pressure field, body movement frequency, local pressure stability, hotspot migration speed, and zoned load change characteristics can be extracted. Regarding posture, the current lying posture type, posture angle change rate, rotational following status, and bed posture disturbance level can be extracted.

[0043] It is easy to understand that relying solely on vital signs makes it difficult to accurately distinguish between deep sleep with a static posture and light sleep with abnormal pressure and no movement; relying solely on pressure distribution also makes it difficult to identify changes in sleep stages when breathing slows down and body movement is sparse. Therefore, this invention adopts a collaborative modeling approach using three types of information: vital signs, pressure distribution, and posture changes, to improve the reliability of stage discrimination.

[0044] In one executable implementation, the aforementioned multimodal features can be arranged in chronological order to form a stage-discriminating feature sequence, and the current sleep stage can be output using a posterior probability discrimination method. ; in, This indicates the sleep stage determination result corresponding to the current time. This indicates the operation corresponding to the stage with the maximum posterior probability; Indicates candidate sleep stages; Represents a set of sleep stages; This indicates that the stage is under the condition of the characteristic sequence. The posterior probability; This represents the stage discrimination feature sequence from the start time to the current time.

[0045] Furthermore, the central control module determines whether the current stage is suitable for micro-intervention based on the posterior probability of each candidate stage, such as the stable period of light sleep, the stable period after natural turning over, or the period of re-entering stable breathing after micro-movement, and generates an intervention permission marker.

[0046] Regarding the calculation of pressure discomfort risk indicators, this embodiment of the invention comprehensively considers the magnitude of pressure, duration, and the differences in importance of various anatomical regions under the current posture. Specifically, the central control module first maps the completed pressure field into multiple anatomical zones, and then calculates the cumulative pressure result of each zone within a time window. The expression is as follows: ; in, Indicates the first The cumulative compression results of each anatomical region; This indicates the backtracking step size relative to the current time. Indicates the cumulative time window length; Represents the natural exponential function; Indicates the time decay coefficient; Indicates the first The average pressure of each anatomical region at the time of regression; Indicates the first Pressure threshold of each anatomical region; Indicates the anatomical partition index; This means that only the portion exceeding the threshold will be retained.

[0047] S40: When the pressure discomfort risk index meets the preset triggering condition and the sleep stage discrimination result belongs to the preset allowable intervention stage, a coordinated rotation and fit control sequence is generated based on the fit reference model, the completed pressure field data, and the motion posture data; wherein, the coordinated rotation and fit control sequence includes a rotation control parameter sequence and a fit displacement control parameter sequence.

[0048] Specifically, based on the current lying posture, the current bed posture, the completed pressure field distribution, and pressure discomfort risk indicators, a set of candidate rotation directions and a set of candidate rotation amplitudes are constructed. Based on the set of candidate rotation directions, the set of candidate rotation amplitudes, and the pressure prediction results in the future time domain, the pressure averaging benefit, posture disturbance cost, and motion smoothing cost corresponding to each candidate rotation scheme are calculated. Based on the pressure averaging benefit, posture disturbance cost, and motion smoothing cost, the optimal rotation scheme that satisfies the bed motion constraint and sleep intervention constraint is solved. Based on the optimal rotation scheme, a rotation control parameter sequence consisting of rotation direction, rotation angular velocity, rotation angular displacement, and pitch compensation is generated.

[0049] Following this, a target pressure reference distribution is generated based on the completed pressure field data and the fitting reference model; the local pressure migration trend during the rotation execution process is predicted based on the rotation control parameter sequence, and the target pressure reference distribution is corrected based on the local pressure migration trend; the corrected target pressure reference distribution is differentially analyzed with the completed pressure field data to generate pressure error results for each support unit; based on the pressure error results and the pressure change results at adjacent times, the displacement increment of each support unit is calculated, and smoothing constraint processing is performed on the displacement increment of each support unit to generate a fitting displacement control parameter sequence.

[0050] In this embodiment of the invention, the central control module will only initiate the control generation process when the stress discomfort risk index reaches a preset level and the current sleep stage is within the permissible intervention stage. Therefore, this invention further subdivides control generation into two coupled processes: rotational control parameter sequence generation and conformal displacement control parameter sequence generation. (1) Generation of rotation control parameter sequence: First, the central control module generates a set of candidate rotation directions and a set of candidate rotation amplitudes based on the current lying position, the current posture of the bed, the completed pressure field distribution, and the distribution of risk hotspots.

[0051] It's easy to understand that the direction of rotation is not fixed, but should be determined comprehensively based on which side of the body the pressure point is located on, the user's current lying position, and the current posture of the bed. For example, when the user is lying on their back and there is continuous high pressure in the left back and waist area, it is advisable to slowly rotate to the right first; if the user is lying on their right side and there is high pressure in the right shoulder, it is more appropriate to release the pressure through pitch compensation and local contact, rather than continuing to rotate to the same side.

[0052] Furthermore, based on the candidate rotation direction, candidate rotation amplitude, and pressure prediction results in the future time domain, the central control module calculates the pressure averaging benefit, attitude disturbance cost, and motion smoothing cost under different candidate schemes, thereby selecting the optimal scheme from among multiple schemes. Its objective function is as follows: in, This represents the sequence of rotational control parameters to be solved; Indicates the length of the prediction time domain; , , , Indicates the weighting coefficient; Indicates the first The variance of the predicted pressure field at each prediction step size is used to characterize the degree of pressure uniformity. Indicates the first Predicted pressure field under a prediction step size; Indicates the first The horizontal rotation angle increment under each prediction step size; Indicates the first Pitch angle increment per prediction step; Indicates the first The attitude perturbation increment under a predicted step size; This represents the L2 norm.

[0053] In one specific executable implementation, an upper limit can be set for the rotational angular velocity to keep it within a perceptible or slightly perceptible range, such as a slow horizontal rotational speed of about 0.8 degrees per minute; at the same time, a variation constraint can be set for the pitch compensation amount so that the bed will not produce abrupt attitude changes in a short period of time.

[0054] (2) Generation of the fit displacement control parameter sequence: After the rotation control parameter sequence is determined, the central control module does not immediately drive the bed to complete the rotation. Instead, it further calculates the fitting displacement control parameter sequence by combining the fitting reference model with the current completed pressure field. It is easy to understand that if there is only macroscopic rotation without fitting compensation, the user's body will tend to slip during the rotation due to changes in the gravitational component and the untimely deformation of the contact surface. This is one of the reasons why traditional rotation devices are prone to causing discomfort.

[0055] Specifically, the central control module first generates a target pressure reference distribution based on the fitting reference model. Then, combining the local pressure migration trend predicted by the rotation control parameters, it performs a forward-looking correction to the target pressure reference distribution. Subsequently, the corrected target pressure reference distribution is compared with the current completed pressure field to obtain the pressure error of each support element. Based on the pressure error and the pressure change results at adjacent time points, the displacement increment of each element is calculated. The expression is as follows: ; in, Indicates the first Displacement increment of each support unit; Indicates the proportionality coefficient; Indicates the difference coefficient; Indicates the first Target pressure reference value at each support unit location; Indicates the current time. The completion pressure value at each support unit location; Indicates the previous moment. The completion pressure value at each support unit location; This indicates the support unit index.

[0056] In practical applications, the central control module also applies smoothing constraints to the displacement increments of each support unit to prevent excessive displacement jumps between adjacent time steps, thus avoiding a noticeable foreign body sensation or lifting feeling for the user. In one executable implementation, the fitting displacement control parameters are not only used to establish the initial fitting surface before rotation, but are also continuously updated during rotation, thereby forming an execution logic of first fitting modeling, then rotation linkage, and continuous compensation during rotation.

[0057] It should be noted that before generating the cooperative rotation and fitting control sequence, the process includes: calculating the available confidence level of pressure data based on the effective sensor mask; calculating the micro-arousal probability based on the stage discrimination feature sequence corresponding to the sleep stage discrimination; keeping the risk triggering flag valid when the available confidence level of pressure data is not less than a preset confidence threshold and the micro-arousal probability is not greater than a preset arousal threshold; and outputting the cooperative rotation and fitting control sequence only within the time window when the risk triggering flag is valid.

[0058] In this embodiment of the invention, to avoid accidental control triggering when pressure data is unreliable or the user is in a highly aroused and sensitive state, confidence gating and micro-arousal gating are set before outputting the cooperative rotation and fit control sequence. Specifically, the central control module calculates the available confidence of the current pressure field based on the effective sensor mask, and calculates the micro-arousal probability of the current time window based on the stage discrimination feature sequence: ; in, Indicates the probability of micro-arousal; Represents the natural exponential function; This represents the transpose of the classification weight vector; This represents the stage-specific feature vector corresponding to the current time window; This indicates the bias term.

[0059] Therefore, in this embodiment of the invention, the central control module maintains the risk trigger flag and continues to output the control sequence only when the available confidence level of the pressure field reaches a preset level and the micro-awakening probability is not higher than a preset threshold. It should be noted that this gating process, together with the aforementioned graph structure spatiotemporal completion model, constitutes the core technical route of this invention: first, data availability is improved through a completion algorithm, and then confidence gating prevents excessive reliance on low-confidence results, thereby avoiding the chain reaction of mis-completion, misjudgment, and miscontrol in traditional solutions.

[0060] S50: Output the coordinated rotation and fitting control sequence, and update the user's individualized intervention model based on the post-intervention effect feedback data.

[0061] Specifically, the coordinated rotation and fit control sequence is output to execute the intervention process. Micro-body movement results, micro-arousal results, pressure risk reduction results, and posture stability results are collected within a preset time window after the intervention as effect feedback data. The immediate reward result of this intervention is calculated based on the effect feedback data. The parameters of the user-individualized intervention model are updated based on the immediate reward result. Based on the updated user-individualized intervention model, the trigger threshold, rotation scheme preference, and fit compensation preference in subsequent sleep cycles are corrected.

[0062] In this embodiment of the invention, the central control module sends the solved rotation control parameter sequence to the macroscopic rotation skeleton device and the fitting displacement control parameter sequence to the microscopic adaptive fitting device. Based on this, the macroscopic rotation skeleton device drives the support platform to complete slow rotation and necessary pitch compensation, while the microscopic adaptive fitting device drives each independent adjustable support unit to complete corresponding height adjustments, ensuring continuous fit and smooth pressure transfer for the human body throughout the rotation process.

[0063] It is easy to understand that this invention does not output fixed control commands once and then stop updating them. Instead, it continuously receives pressure feedback, vital sign feedback, and posture feedback during the control execution process, and dynamically corrects the execution state. In other words, there is constant data interaction between the central control module, the macroscopic rotating skeleton device, and the microscopic adaptive bonding device.

[0064] Furthermore, after each intervention, the system also records post-intervention feedback data. For example, it can statistically analyze the number of micro-movements, micro-arousals, the decline in risk indicators, and the degree of postural stability within a preset time window after the intervention, and calculate the immediate reward for this intervention based on these results. ; in, Indicates the immediate reward result; This indicates the increase in micromotor activity or microarousal after intervention; Indicators representing the risk of stress and discomfort after intervention; This represents the cost of attitude perturbation after intervention; and This represents the tradeoff coefficient.

[0065] Subsequently, the central control module updates the user's individualized intervention model based on the immediate reward result. It is easy to understand that the updated objects can be trigger thresholds, rotation direction preferences, rotation amplitude preferences, rotation speed preferences, fit compensation gains, or weights of various anatomical regions, etc. Anything that can make subsequent interventions more adaptable to the user's individual sleeping posture habits and arousal sensitivity can be included within the scope of this invention.

[0066] In one specific implementation, after the user falls asleep, the initial fitting phase begins. The central control module controls all independently adjustable support units in the micro-adaptive fitting device to collect the initial pressure distribution and complete the initial shaping of the fitting surface within 10-30 seconds, forming corresponding fitting support surfaces for the back, waist, and legs. During this phase, the central control module establishes a fitting reference model based on the initial pressure field and the body surface contour description results.

[0067] The system then enters the dynamic monitoring and decision-making phase. Non-contact bio-radar continuously outputs heart rate and respiratory information, a pressure sensor array continuously monitors pressure distribution, and an inertial measurement unit continuously provides feedback on the bed's attitude and motion. The central control module aligns these three types of data to form continuous sleep state data frames and performs joint analysis on the sleep stage, pressure discomfort risk, and current posture within each time window. If the duration of pressure on a certain area exceeds a preset threshold, and the current stage is within a stable light sleep period, and the data confidence level meets the requirements, then the control generation phase begins.

[0068] The next stage is the coordinated rotation execution phase. The central control module first calculates the candidate rotation direction and amplitude, then solves for the rotation control parameter sequence based on future pressure predictions and attitude disturbance constraints, and simultaneously generates the fitting displacement control parameter sequence. The macroscopic rotating skeleton device drives the horizontal rotation motor and pitch adjustment motor according to the calculated rotation direction and angular velocity, while the microscopic adaptive fitting device drives each independent adjustable support unit to perform millimeter-level height compensation based on local pressure errors. Through this coordinated execution method, which changes the overall load-bearing position through macroscopic rotation and maintains contact continuity through microscopic fitting, the user's body surface and the fitting surface can remain approximately synchronized during rotation, thereby significantly reducing the relative sliding tendency caused by traditional bed rotation.

[0069] Finally, the system enters the learning and optimization phase. After each intervention, the system records whether micro-awakening has decreased, risk has declined, and posture has become more stable. These results are then input into the user's individualized intervention model to update parameters, in order to obtain a more suitable triggering and control strategy for the next intervention.

[0070] It should be noted that the graph structure spatiotemporal completion model, sleep stage discrimination model, rotation scheme optimization model, and user individualized intervention model described in this invention are not limited to any specific algorithm implementation. As long as the corresponding functions can be accomplished, they can be implemented using neural networks, probabilistic graphical models, state-space models, reinforcement learning models, constrained optimization models, or any combination of the above models.

[0071] It should also be noted that although the inertial measurement unit is set as the main source of motion attitude data in the preferred embodiment of the present invention, angle encoders, gyroscope assemblies, or bed attitude feedback modules can also be introduced in other implementations to provide the same or similar attitude information. As long as the relevant information can participate in sleep stage discrimination, risk assessment, rotation parameter generation, and execution status feedback, it will not affect the core technical concept of the present invention.

[0072] like Figure 2 As shown, in a sleep regulation system based on dynamic fit and coordinated rotation control provided by an embodiment of the present invention, the central control module is configured to execute the method described in any of the preceding claims, the central control module comprising: The acquisition unit is used to acquire pressure perception data, vital sign data, and motion posture data, and perform unified time alignment processing to construct sleep state data frames. The execution unit is used to perform pressure calibration, missing data identification and pressure field completion processing based on the pressure sensing data in the sleep state data frame, so as to generate the completed pressure field data, and generate the body surface contour description result and the fitting reference model based on the completed pressure field data. The calculation unit is used to calculate the sleep stage discrimination result and the stress discomfort risk index based on the vital signs data, the completed pressure field data and the movement posture data. The generation unit is used to generate a coordinated rotation and fit control sequence based on the fit reference model, the completed pressure field data, and the motion posture data when the pressure discomfort risk index meets the preset triggering conditions and the sleep stage discrimination result belongs to the preset allowable intervention stage; wherein, the coordinated rotation and fit control sequence includes a rotation control parameter sequence and a fit displacement control parameter sequence; The output unit is used to output the coordinated rotation and fitting control sequence and update the user's individualized intervention model based on the effect feedback data after the intervention.

[0073] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a microcontroller, chip, or processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0074] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details described above. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention. It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not further describe the various possible combinations.

[0075] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the embodiments of the present invention, they should also be regarded as the content disclosed by the embodiments of the present invention.

Claims

1. A sleep regulation method based on dynamic fitting and cooperative rotation control, characterized in that, The method includes: Acquire stress perception data, vital sign data, and motion posture data, and perform unified time alignment processing to construct sleep state data frames; Based on the pressure sensing data in the sleep state data frame, pressure calibration, missing data identification, and pressure field completion processing are performed to generate completed pressure field data, and a body surface contour description result and a fitting reference model are generated based on the completed pressure field data. Based on the vital signs data, the completed pressure field data, and the movement posture data, the sleep stage discrimination result and the pressure discomfort risk index are calculated. When the pressure discomfort risk index meets the preset triggering conditions and the sleep stage discrimination result belongs to the preset allowable intervention stage, a coordinated rotation and fit control sequence is generated based on the fit reference model, the completed pressure field data, and the motion posture data; wherein, the coordinated rotation and fit control sequence includes a rotation control parameter sequence and a fit displacement control parameter sequence; The coordinated rotation and fitting control sequence is output, and the user-individualized intervention model is updated based on the post-intervention effect feedback data.

2. The sleep regulation method based on dynamic fit and coordinated rotation control according to claim 1, characterized in that, Acquire stress-sensing data, vital sign data, and movement posture data, and perform unified time alignment processing to construct sleep state data frames, including: The system collects raw pressure data from the pressure sensor array, raw vital sign data from the non-contact bioradar, and raw motion attitude data from the inertial measurement unit, and adds sampling time markers to each type of raw data. Using a preset time grid as a unified time axis, resampling and interpolation processes are performed on the original pressure data, the original vital signs data, and the original motion posture data to generate time-consistent pressure sequences, vital signs sequences, and posture sequences. Sleep state data frames are constructed based on aligned stress sequences, vital sign sequences, and posture sequences, ensuring that stress information, vital sign information, and posture information are associated simultaneously under the same time index.

3. The sleep regulation method based on dynamic fit and coordinated rotation control according to claim 1, characterized in that, Based on the pressure-sensing data in the sleep state data frames, pressure calibration, missing data identification, and pressure field completion processing are performed to generate completed pressure field data. Then, based on the completed pressure field data, a body surface contour description result and a fitting reference model are generated, including: Zero-point drift compensation, temperature drift compensation, and range normalization are performed on the time-aligned pressure sequence to generate calibrated pressure field data. Based on the calibrated pressure field data, the locations of abnormal distortion, occlusion, and saturation are identified, and an effective sensor mask is generated. When the missing proportion of the effective sensor mask representation is greater than a preset threshold, the graph structure spatiotemporal completion model is invoked, and the missing regions are completed and reconstructed by combining vital sign information, attitude information and historical pressure field information to generate the completed pressure field data. Based on the completed pressure field data, the surface contour description results are extracted, and a bonding reference model is generated to characterize the geometry of the target bonding surface.

4. The sleep regulation method based on dynamic fit and coordinated rotation control according to claim 1, characterized in that, Based on the vital signs data, the completed pressure field data, and the movement posture data, the sleep stage discrimination result is calculated, including: Based on the vital signs data, heart rate features, respiratory rate features, and heart rate variability features are extracted. Based on the completed pressure field data, body motion features, pressure stability features, and pressure zone change features are extracted. Based on the motion posture data, posture angle features, posture change rate features, and bed motion following features are extracted. The heart rate characteristics, respiratory rate characteristics, heart rate variability characteristics, body movement characteristics, pressure stability characteristics, pressure zone change characteristics, posture angle characteristics, posture change rate characteristics, and bed motion following characteristics are fused to generate a stage discrimination feature sequence. Sleep stage discrimination results are calculated based on stage discrimination feature sequences, and intervention-permitted markers are generated.

5. The sleep regulation method based on dynamic fit and coordinated rotation control according to claim 1, characterized in that, Based on the vital sign data, the completed pressure field data, and the movement posture data, a pressure discomfort risk index is calculated, including: The completed pressure field data is mapped into multiple partitioned pressure results according to human anatomical regions. Based on the partitioned pressure results, the current supine state represented by the motion posture data, and the autonomous micro-movement state represented by the vital signs data, the cumulative pressure result of each partition within the current time window is determined. The risk value of each zone is calculated based on the cumulative pressure results of each zone, and the risk contribution of different zones is weighted according to the lying position. The risk values ​​of each zone are summarized to generate a stress discomfort risk index. The stress discomfort risk index is then compared with preset trigger conditions to generate a risk trigger flag.

6. The sleep regulation method based on dynamic fit and coordinated rotation control according to claim 1, characterized in that, Based on the fitting reference model, the completed pressure field data, and the motion attitude data, a rotation control parameter sequence is generated, including: Based on the current lying position, the current bed posture, the completed pressure field distribution, and pressure discomfort risk indicators, a set of candidate rotation directions and a set of candidate rotation amplitudes are constructed. Based on the set of candidate rotation directions, the set of candidate rotation amplitudes, and the pressure prediction results in the future time domain, calculate the pressure averaging benefit, attitude disturbance cost, and motion smoothing cost corresponding to each candidate rotation scheme. Based on the pressure equalization benefit, the attitude perturbation cost, and the motion smoothing cost, the optimal rotation scheme that satisfies the bed motion constraint and the sleep intervention constraint is solved. Based on the optimal rotation scheme, a sequence of rotation control parameters consisting of rotation direction, rotation angular velocity, rotation angular displacement, and pitch compensation is generated.

7. The sleep regulation method based on dynamic fit and coordinated rotation control according to claim 1, characterized in that, Based on the aforementioned fitting reference model, the completed pressure field data, and the rotation control parameter sequence, a fitting displacement control parameter sequence is generated, including: A target pressure reference distribution is generated based on the completed pressure field data and the fitted benchmark model. The local pressure migration trend during the rotation execution process is predicted based on the rotation control parameter sequence, and the target pressure reference distribution is corrected based on the local pressure migration trend. The corrected target pressure reference distribution is compared with the completed pressure field data to generate the pressure error results for each support unit. Based on the pressure error results and the pressure change results at adjacent times, the displacement increment of each support unit is calculated, and a smoothing constraint process is performed on the displacement increment of each support unit to generate a sequence of fitting displacement control parameters.

8. The sleep regulation method based on dynamic fit and coordinated rotation control according to claim 3, characterized in that, Before generating the coordinated rotation and bonding control sequence, the following steps are also included: The available confidence level of the stress data is calculated based on the effective sensor mask, and the probability of micro-arousal is calculated based on the stage discrimination feature sequence corresponding to the sleep stage discrimination. The risk trigger marker remains valid as long as the confidence level of the available pressure data is not less than a preset confidence threshold and the probability of micro-awakening is not greater than a preset arousal threshold. The cooperative rotation and bonding control sequence is output only within the time window when the risk trigger flag is valid.

9. The sleep regulation method based on dynamic fit and coordinated rotation control according to claim 1, characterized in that, Output the coordinated rotation and fitting control sequence, and update the user-individualized intervention model based on the post-intervention effect feedback data, including: The coordinated rotation and fitting control sequence is output to execute the intervention process. Micro-body motion results, micro-awakening results, pressure risk reduction results, and posture stability results are collected within a preset time window after the intervention as effect feedback data. The immediate reward result for this intervention is calculated based on the aforementioned effect feedback data; The parameters of the user-individualized intervention model are updated based on the instant reward results. Based on the updated user-individualized intervention model, the trigger threshold, rotation scheme preference, and fit compensation preference in subsequent sleep cycles are corrected.

10. A sleep regulation system based on dynamic fit and coordinated rotation control, characterized in that, The system includes: Base; A support platform, which is mounted on the base; A macroscopic rotating frame device is connected to the support platform via a transmission; wherein, the macroscopic rotating frame device includes at least a horizontal rotation motor and a pitch adjustment motor, used to drive the support platform to rotate continuously in the horizontal plane and perform pitch adjustment around the transverse axis; A micro-adaptive bonding device is disposed on the support platform; wherein, the micro-adaptive bonding device includes a bonding surface composed of independently adjustable support units arranged in a dense array, each of the independently adjustable support units including at least a flexible contact cap, a micro linear electric cylinder, a pressure-sensitive thin film sensor and a unit housing, for collecting local pressure data and performing local height adjustment; A multimodal sensing device; wherein the multimodal sensing device includes at least a pressure sensor array, a non-contact bio-radar, and an inertial measurement unit, for outputting pressure sensing data, vital sign data, and motion posture data, respectively; The central control module is communicatively connected to the macroscopic rotating skeleton device, the microscopic adaptive bonding device, and the multimodal sensing device, respectively, and is used to receive the pressure sensing data, the vital signs data, and the motion posture data, and generate a rotation control parameter sequence and a bonding displacement control parameter sequence based on the pressure sensing data, the vital signs data, and the motion posture data. The central control module is configured to execute the method described in any one of claims 1 to 9.

Images