Environment adjusting method, system and equipment based on sleep state monitoring and medium

Abstract

The invention relates to the technical field of data processing, and particularly provides an environment adjusting method, system and device based on sleep state monitoring and a medium, the method comprises the steps that state data are collected through a multi-mode non-inductive sensing network, and the state data comprise bedding pressure data, vital sign data and environment parameters; extracting key features from the state data, wherein the key features comprise an off-bed tendency feature, a sleep stability feature and an early awakening feature; analyzing the key features through a machine learning model, and predicting sleep behavior intentions of the user, including an off-bed tendency, an awakening precursor and an early awakening risk; triggering an environment actuator according to the predicted intention, and adjusting at least one parameter of illumination, temperature and humidity; recording user state data before and after intervention, and optimizing a decision threshold and intervention parameters through a reinforcement learning algorithm; the invention provides an unprecedented safe and comfortable sleep health management solution for sensitive people with cognitive impairment and the like.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and specifically relates to an environmental regulation method, system, device and medium based on sleep state monitoring. Background Technology

[0002] With the development of smart healthcare technologies, sleep health management is receiving increasing attention. Existing technologies largely rely on wearable devices for sleep monitoring, which suffers from issues such as discomfort and disruption to natural sleep. Furthermore, these systems are mostly limited to passively recording and alerting about already occurring sleep events, such as issuing an alarm after detecting a user leaving the bed, failing to provide early warning and intervention before risks materialize. Regarding environmental regulation, existing solutions often rely on simple environmental parameter thresholds or fixed sleep stages for single adjustments, lacking the ability to predict users' personalized sleep behaviors and intentions, resulting in rigid interventions and limited effectiveness. Therefore, existing technologies struggle to achieve the goal of "prevention before the event and seamless resolution" in sleep health management, especially failing to meet the urgent needs of sensitive groups such as those with cognitive impairments for undisturbed, personalized deep sleep promotion and safety protection. Summary of the Invention

[0003] In view of the above-mentioned shortcomings of the prior art, the present invention provides an environmental regulation method, system, device and medium based on sleep state monitoring to solve the above-mentioned technical problems.

[0004] In a first aspect, the present invention provides an environmental regulation method based on sleep state monitoring, comprising: Status data is collected through a multimodal non-sensor network, including mattress pressure data, vital signs data, and environmental parameters. Key features are extracted from the state data, including the tendency to get out of bed, sleep stability, and early awakening. By analyzing key features through machine learning models, users' sleep behavior intentions can be predicted, including the tendency to get out of bed, pre-awakening symptoms, and risk of early awakening. The environmental actuator is triggered according to the predicted intent to adjust at least one parameter among light, temperature, and humidity. Record user status data before and after intervention, and optimize decision thresholds and intervention parameters through reinforcement learning algorithms; Among them, the characteristics of getting out of bed include at least the longitudinal displacement velocity of the center of gravity of the pressure center trajectory and the pressure release area ratio; the characteristics of sleep stability include at least the respiratory disturbance index and the low-frequency / high-frequency power ratio of heart rate variability; and the characteristics of early awakening include at least the rest-activity ratio before dawn.

[0005] In one optional implementation, state data is acquired via a multimodal non-sensing network, including: By deploying a distributed pressure sensor array under the mattress or integrating it into the mattress, the user's body movement and pressure distribution data are collected. The data includes the pressure center trajectory, the rate of change of pressure distribution, and the pressure pattern of specific postures. The bio-radar installed on the bedside table or ceiling collects the user's vital signs signals, including micro-movement breathing waveforms, heartbeat vibration signals, and macroscopic body movements. Environmental parameters, including ambient temperature, relative humidity, and ambient light intensity, are collected by temperature and humidity sensors and light sensors installed in the bedroom space.

[0006] In one optional implementation, key features are extracted from the state data, including: The collected mattress pressure data and vital sign radar signals were subjected to noise reduction and filtering to remove high-frequency noise and low-frequency drift. The vital signs radar signal is subjected to signal separation to separate the respiratory, heartbeat and body movement components. Extracting bed-leaking tendency features from pressure data includes at least calculating the ratio of the longitudinal displacement velocity of the center of gravity of the pressure center to the pressure release area. Extracting sleep stability features from vital signs includes at least calculating the low-frequency / high-frequency power ratio of the respiratory disturbance index and heart rate variability. Early awakening characteristics are extracted from body movement data and environmental parameters, including at least calculating the pre-dawn rest-activity ratio and performing a correlation analysis between this ratio and ambient light intensity.

[0007] In one optional implementation, early awakening features are extracted from body movement data and environmental parameters, including: Within a predetermined time window before the user's expected wake-up time, the user's pre-dawn rest-activity ratio is calculated; the pre-dawn rest-activity ratio is the ratio of the user's body movement frequency within the predetermined time window to the body movement frequency within a historical reference period after the user enters deep sleep. Simultaneously, the ambient light intensity collected synchronously during the predetermined time window when the user's body movement frequency significantly increases is recorded and analyzed. The value of the pre-dawn rest-activity ratio was correlated with the ambient light intensity to determine the risk of early awakening for users and its causal relationship with premature morning light.

[0008] In one optional implementation, key features are analyzed using a machine learning model to predict the user's sleep behavior intent, including: During the initial 3 to 7 nights of system operation, the system will operate in a monitoring-only, non-intervention mode to learn and establish the user's personal sleep baseline, which includes at least typical sleep onset time, deep sleep period distribution time, average heart rate, respiratory rate, and typical number of nighttime awakenings. Based on the individual's sleep baseline, a specific machine learning model is used for real-time state classification and prediction, including: For the prediction of the tendency to get out of bed, a support vector machine or a lightweight neural network is used as a classifier, with the longitudinal displacement velocity of the center of gravity and the pressure release area ratio as input features; when the feature value exceeds a dynamic threshold set based on the individual's sleep baseline, it is determined that the user has a high probability of intending to get out of bed. For the prediction of sleep disruption, a hidden Markov model or recurrent neural network is used to analyze the respiratory disturbance index sequence and the low-frequency / high-frequency power ratio sequence of heart rate variability to predict the probability of awakening within a specific time period in the future; when the predicted probability is consistently higher than a preset threshold and accompanied by an increase in microbody movements, it is determined to be a premonitory symptom of awakening. To assess the risk of early awakening, if the pre-dawn rest-activity ratio calculated within a specific time window from the expected wake-up time continuously rises and remains at a set high threshold for more than 5 minutes compared to the user's baseline ratio during deep sleep, then the risk of early awakening is determined.

[0009] In one optional implementation, an environmental actuator is triggered based on the predicted intent to adjust at least one parameter among light, temperature, and humidity, including: Based on the predicted different sleep behavior intentions, pre-set, differentiated, and collaborative intervention strategies are triggered, including: When the predicted intention is a high probability of getting out of bed, the triggered intervention strategy is to activate the LED light strips placed by the bedside or on the ground, so that they emit a gradually brightening warm light with a brightness of less than 5 lumens, forming a guiding light path towards the bathroom. When the predicted intention is a premonition of awakening, the triggered intervention strategy is to control the air conditioning, fresh air or humidifier system to raise the ambient temperature by 0.8-1.2℃ and adjust the ambient relative humidity to about 60% RH. When the predicted intention is the risk of early awakening, the triggered intervention strategy is: to simultaneously activate a controllable light source, a micro-vibration device, and an aromatherapy diffuser; wherein, the controllable light source emits an amber light curtain with a wavelength of 590nm, the micro-vibration device generates gentle periodic vibrations, and the aromatherapy diffuser releases a soothing aroma with cedarwood as the main scent.

[0010] In an optional implementation, user state data before and after the intervention is recorded, and the decision threshold and intervention parameters are optimized using a reinforcement learning algorithm, including: The system continuously records and builds a feedback dataset, which includes: Pre-intervention state: The predictive intent and corresponding key feature values ​​upon which the intervention was triggered; Intervention actions: the specific intervention strategies implemented and their parameters; Post-intervention status: The user's physiological and behavioral responses within a predetermined period of time after the intervention, including at least whether they successfully continued to sleep, whether they safely got out of bed, or whether they were fully awake. The post-intervention state is used as the reward signal for the reinforcement learning algorithm to dynamically optimize the decision threshold and intervention parameters; The decision thresholds include a dynamic threshold for longitudinal displacement velocity of the center of gravity used to predict the tendency to get out of bed, a threshold for the probability of arousal used to predict arousal precursors, and a threshold for the pre-dawn rest-activity ratio used to determine the risk of early arousal. The intervention parameters include the brightness and gradual brightening speed of the light guidance, the amplitude of temperature adjustment, and the intensity and period of micro-vibration.

[0011] Secondly, the present invention provides an environmental regulation system based on sleep state monitoring, comprising: The data acquisition module is used to collect status data through a multimodal non-sensor network. The status data includes mattress pressure data, vital sign data, and environmental parameters. The feature extraction module is used to extract key features from the state data, including features of tendency to get out of bed, sleep stability, and early awakening. The intent recognition module is used to analyze key features through machine learning models to predict users' sleep behavior intent, including the tendency to get out of bed, pre-awakening aura, and risk of early awakening. An environmental control module is used to trigger an environmental actuator based on a predicted intent to adjust at least one parameter among light, temperature, and humidity. The decision optimization module records user status data before and after intervention and optimizes decision thresholds and intervention parameters through reinforcement learning algorithms. Among them, the characteristics of getting out of bed include at least the longitudinal displacement velocity of the center of gravity of the pressure center trajectory and the pressure release area ratio; the characteristics of sleep stability include at least the respiratory disturbance index and the low-frequency / high-frequency power ratio of heart rate variability; and the characteristics of early awakening include at least the rest-activity ratio before dawn.

[0012] Thirdly, a device is provided, comprising: Memory for storing environmental adjustment programs based on sleep state monitoring; A processor is configured to implement the steps of the sleep state monitoring-based environmental regulation method provided in the first aspect when executing the sleep state monitoring-based environmental regulation program.

[0013] Fourthly, a computer-readable medium is provided, on which an environmental regulation program based on sleep state monitoring is stored, wherein when the environmental regulation program based on sleep state monitoring is executed by a processor, the steps of the environmental regulation method based on sleep state monitoring provided in the first aspect are implemented.

[0014] The beneficial effects of this invention lie in the fact that the environmental regulation method, system, device, and medium based on sleep state monitoring provided by this invention achieve a paradigm shift from "passive alarm" to "proactive resolution" through a non-intrusive sensor network and machine learning prediction model. Its beneficial effects are: the system can accurately predict risks such as users getting out of bed or waking up in a zero-disruption manner, and trigger dynamically coordinated environmental interventions, thereby significantly reducing the risk of falling out of bed and effectively extending the duration of deep sleep and effective morning sleep; at the same time, the system, through a reinforcement learning closed loop, continuously adaptively optimizes, achieving a high degree of personalization, and providing an unprecedented safe and comfortable sleep health management solution for sensitive groups such as those with cognitive impairments. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic flowchart of a method according to an embodiment of the present invention.

[0017] Figure 2 This is a schematic block diagram of a system according to an embodiment of the present invention.

[0018] Figure 3 This is a schematic diagram of the structure of a device provided in an embodiment of the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0021] The environmental regulation method based on sleep state monitoring provided in this embodiment of the invention is executed by a computer device, and correspondingly, the environmental regulation system based on sleep state monitoring runs in the computer device.

[0022] Figure 1 This is a schematic flowchart illustrating a method according to an embodiment of the present invention. Wherein, Figure 1 The implementing entity can be an environmental regulation system based on sleep state monitoring. Depending on different needs, the order of the steps in this flowchart can be changed, and some can be omitted.

[0023] like Figure 1 As shown, the method includes: S1. Collect status data through a multimodal non-sensor network, the status data including mattress pressure data, vital signs data and environmental parameters; S2. Extract key features from the state data, including features of bed-avoidance tendency, sleep stability, and early awakening; S3. Analyze key features through machine learning models to predict users' sleep behavior intentions, including the tendency to get out of bed, pre-awakening symptoms, and risk of early awakening; S4. Trigger the environmental actuator according to the predicted intent to adjust at least one parameter among light, temperature, and humidity; S5. Record user status data before and after intervention, and optimize decision thresholds and intervention parameters through reinforcement learning algorithms; Among them, the characteristics of getting out of bed include at least the longitudinal displacement velocity of the center of gravity of the pressure center trajectory and the pressure release area ratio; the characteristics of sleep stability include at least the respiratory disturbance index and the low-frequency / high-frequency power ratio of heart rate variability; and the characteristics of early awakening include at least the rest-activity ratio before dawn.

[0024] In one embodiment of the present invention, based on step S1, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner.

[0025] S101. Specific implementation of distributed pressure sensor array: The distributed pressure sensor array is preferably a fiber Bragg grating sensor array or a flexible thin-film pressure sensor array. Taking a fiber Bragg grating sensor array as an example, its specific implementation is as follows: Installation method: Multiple sensing optical fibers, each embedded with multiple fiber optic grating sensing points, are sewn or pressed into the mattress lining in a grid structure with a spacing of 5-10 cm, or laid flat between the mattress and the bed board.

[0026] Data Acquisition: Each sensor point monitors minute pressure changes at its location in real time at a sampling frequency of no less than 10Hz. Data from all sensor points is aggregated through a hub and then transmitted to the central processing unit via CAN bus or Wi-Fi protocol.

[0027] Data output: The system generates the user's pressure center trajectory (represented in two-dimensional coordinates) in real time by solving the pressure data of the entire grid; it obtains the pressure distribution change rate by calculating the standard deviation of the pressure value per unit time; and it identifies the user's specific posture pressure pattern through a pre-stored pressure distribution pattern library (containing typical images of lying flat, side lying, sitting up, etc.) and a pattern matching algorithm.

[0028] S102. Specific implementation of bio-radar: The bio-radar is preferably a continuous-wave Doppler bio-radar or an ultra-wideband pulse bio-radar. Its specific implementation is as follows: Installation and Calibration: Install the radar module on the bedside table, aligning its beam center with the user's torso, or embed it in the ceiling, ensuring its beam covers the entire bed surface. During initial system installation, baseline calibration must be performed with the user in a lying position to eliminate ambient static clutter.

[0029] Signal acquisition: The radar continuously transmits and receives echoes using electromagnetic waves in the 24GHz band, with a sampling rate of no less than 100Hz. By demodulating the phase information, the micro-Doppler signals caused by chest wall micro-movements and body movements are extracted.

[0030] Signal separation and output: A blind source separation algorithm based on empirical mode decomposition is used to separate respiratory signals with a frequency of 0.1-0.3Hz (used to generate respiratory waveforms and calculate frequencies), heartbeat signals with a frequency of 0.8-2.0Hz (used to extract heart rate and heart rate variability), and macroscopic body movement signals with higher frequency and larger amplitude (used to identify large movements such as turning over) from the mixed micro-Doppler signals.

[0031] S103. Specific implementation of environmental sensors: Deployment and Communication: A digital temperature and humidity sensor (such as the SHT30 series) and a wide-range photosensitizer (range 0-1000 Lux) are integrated into a miniature node, which connects to the central processing unit via ZigBee or LoRa wireless communication protocols. This node should be deployed in the bedroom away from air conditioning vents and direct sunlight from windows to reflect the actual parameters of the user's environment.

[0032] Data Acquisition: The temperature and humidity sensor collects ambient temperature (accuracy ±0.3℃) and relative humidity (accuracy ±3% RH) at a frequency of 1Hz. The photosensor simultaneously collects ambient light intensity, and its data is used to determine the diurnal rhythm and identify abnormal light interference.

[0033] In one embodiment of the present invention, based on step S2, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner.

[0034] S201. Data Preprocessing: Noise Reduction and Filtering For mattress pressure data: A fourth-order Butterworth bandpass filter is used for processing, with a passband frequency range of 0.1Hz to 3Hz. This operation can effectively filter out the 50Hz high-frequency noise introduced by electrical equipment, as well as the low-frequency drift caused by the user's slow body movement or the temperature drift of the sensor itself, while retaining the effective pressure change signals caused by actions such as turning over and sitting up.

[0035] For vital signs radar signals: First, a Chebyshev Type I low-pass filter (cutoff frequency 10Hz) is used to initially smooth the raw signal. Then, for the separated respiratory signal (typically 0.1-0.3Hz) and heartbeat signal (typically 0.8-2.0Hz), secondary fine filtering is performed using bandpass filters of 0.05-0.5Hz and 0.5-3Hz, respectively, to further improve the signal-to-noise ratio.

[0036] S202. Separation of vital signs signals The present invention preferably employs the Empirical Mode Decomposition (EMD) algorithm to perform blind source separation on the preprocessed radar signal.

[0037] The specific steps are as follows: The algorithm adaptively decomposes the mixed radar signal into several intrinsic mode functions (IMFs), and then selects the corresponding IMF components for reconstruction based on prior knowledge (such as respiratory rate 0.1-0.3Hz and heart rate 0.8-2.0Hz) to obtain pure respiratory and heart rate signals. Residual high-frequency IMF components not included in the above frequency ranges are identified as body motion components.

[0038] S203. Extraction of Bed-Leaving Tendency Features longitudinal displacement velocity of the center of gravity: First, based on the pressure sensor array data, the coordinates of the pressure center on the longitudinal (Y-axis) of the bed at the current moment are calculated at a frequency of 10 times per second. .

[0039] Then, calculate the rate of change of this coordinate over the past 0.5 seconds: .

[0040] when A sitting-up acceleration signal is considered valid if it lasts for more than 2 seconds beyond a threshold set according to an individual's baseline (e.g., 0.15 m / s).

[0041] Pressure relief area ratio: The system defines the main pressure-bearing area of ​​the torso (an area that occupies approximately 40% of the total bed surface area).

[0042] Calculate the area R of sensor points within this region where the pressure value is 30% lower than the average pressure. release .

[0043] When R release The jump from less than 10% to more than 35% within 1 second, combined with the increased longitudinal displacement velocity, constitutes a strong characteristic of "bed-off tendency".

[0044] S204. Extraction of Sleep Stability Features The Respiratory Disorder Index (RDI) is determined by peak detection of continuous respiratory waveforms. Apnea is defined as a respiratory amplitude that is 20% lower than baseline and lasts for more than 10 seconds; hypopnea is defined as a respiratory amplitude that is 50% lower and lasts for more than 10 seconds. The total number of apneas and hypopneas per hour of sleep is calculated as the RDI. A peak in this index within a short period (e.g., 5 minutes) (e.g., >10 times / hour) is considered a sign of sleep instability.

[0045] Heart rate variability low-frequency / high-frequency power ratio: RR interval sequences were extracted from the isolated heartbeat signals, and spectral estimation was performed using Lomb-Scargle spectral analysis. The power spectral density of the low-frequency band (0.04-0.15 Hz) and the high-frequency band (0.15-0.4 Hz) was calculated. The ratio LF / HF was calculated. When the mean LF / HF ratio increases by more than 50% compared to the sleep baseline within 5 minutes, it indicates enhanced sympathetic nerve activity and is a potential precursor to wakefulness.

[0046] S205. Early Wake-Up Feature Reminder Calculation of the pre-dawn rest-activity ratio: Define time window: The “pre-set time window” is set to 1.5 hours before the user’s expected wake-up time.

[0047] Define reference period: The “historical reference period” is a stable 1-hour period after the user enters N3 stage deep sleep (usually 1-2 hours after falling asleep).

[0048] Calculate the ratio: Within the predetermined time window, calculate the user's body movement frequency (unit: times / minute), and obtain the ratio R of this frequency to the body movement frequency of the reference time period. early .

[0049] Risk assessment: If R early If the score remains above 3.0 for 10 consecutive minutes, the system determines that the user is in an "early awakening state with a low probability of naturally falling back asleep".

[0050] Correlation analysis with ambient light: Within the same time window that is identified as an early awakening state, the system synchronously reads the ambient light intensity L collected by the photosensor. env .

[0051] Perform association analysis: If L env If the light level is >50 lux (equivalent to the dawn's first light), the system determines that there is a strong causal relationship between this early awakening and premature sunlight, and prioritizes shading measures in subsequent interventions; if L env If the value is less than 10 lux, it is considered endogenous premature awakening, and the intervention strategy will focus on rhythm guidance.

[0052] In one embodiment of the present invention, based on step S3, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner.

[0053] S301. Establish a user's personal sleep baseline model After initial system deployment, a baseline learning mode of 5 nights is set (which can be adaptively adjusted within 3-7 nights based on the user's sleep patterns). In this mode, the system only collects and analyzes data without performing any intervention.

[0054] Data aggregation and calculation: The system aggregates the data collected each night and calculates the mean and standard deviation of the following core baseline parameters: Sleep onset time: The average time from turning off the lights and intending to fall asleep to first entering N2 stage sleep.

[0055] Distribution of deep sleep period: The total duration of N3 stage (deep sleep) sleep each night and its main distribution period in the whole night's sleep (such as the first half of the night).

[0056] Resting vital signs: Calculate the mean heart rate and mean respiratory rate during stable N2 and N3 sleep stages.

[0057] Typical nighttime urination frequency: The average number of times each night is triggered by getting out of bed to use the toilet.

[0058] Ultimately, the system establishes a dynamic range (e.g., mean ± 1.5 standard deviations) for each parameter, serving as a personalized benchmark for subsequent real-time predictions.

[0059] S302. Specific Implementation of Real-Time State Classification and Prediction (1) Prediction of bed-leaving tendency Model selection and deployment: A lightweight convolutional neural network is preferred, whose input layer is a two-dimensional feature matrix composed of the longitudinal displacement velocity sequence of the center of gravity and the pressure release area ratio sequence within the past 10-second time window.

[0060] Personalized threshold setting: The classification probability threshold output by the model is not a fixed value, but is dynamically set based on the activity level reflected by the "typical number of nighttime urinations" in the individual's baseline. For example, for users who frequently urinate at night in the baseline, the system will appropriately increase the trigger threshold to reduce false alarms.

[0061] Predictive Judgment: When the probability of “leaving the bed” output by the model exceeds the dynamic threshold of 65% for 2 consecutive seconds, the system determines that the user has a high probability of leaving the bed and immediately triggers intervention.

[0062] (2) Prediction of sleep interruption (wakefulness) Model selection and deployment: A long short-term memory network is used as the specific implementation of the recurrent neural network to process time series data. Its input consists of a respiratory disturbance index sequence and a heart rate variability low-frequency / high-frequency power ratio sequence calculated in 30-second time windows over the past 5 minutes.

[0063] Temporal prediction: This LSTM model was trained to predict the probability of awakening (defined as the transition from sleep stage N1, N2 or N3 to wakefulness) occurring within the next 3 minutes.

[0064] Comprehensive judgment: When the model predicts that the probability of arousal exceeds the threshold of 70% for 1 minute, and during this period the frequency of micro-body movements detected by the bio-radar increases by more than 50% of the baseline level, the system will comprehensively judge it as a pre-awakening sign.

[0065] (3) Assessment of the risk of early awakening Logical judgment implementation: This risk judgment is mainly based on a pre-set, quantitative decision-making logic, rather than a simple machine learning classifier.

[0066] Specific steps: Time window setting: The system defines the period starting 1.5 hours before the user's expected wake-up time as the "early wake-up risk assessment window".

[0067] Baseline value acquisition: The body movement frequency during the deep sleep phase (N3) in the user's personal sleep baseline is used as the baseline value F. baseline .

[0068] Real-time monitoring and calculation: Within the decision window, the current body movement frequency F is calculated in real time. current With F baseline The ratio R early .

[0069] Risk Trigger: If R early If a user's body movement frequency reaches ≥ 2.5 times that of deep sleep for 5 consecutive minutes, the system will determine that the user is at risk of waking up early and that the possibility of falling back asleep naturally is low.

[0070] In one embodiment of the present invention, based on step S4, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner.

[0071] All actuators are connected to the central decision-making system via IoT communication protocols (such as Wi-Fi and Zigbee) and are subject to unified scheduling.

[0072] S401. Implementation of interventions targeting "high probability of leaving the bed" Actuator configuration: The LED light strip is preferably an RGBW four-color light strip, and its controller has a preset "warm light" formula that matches the target spectrum. The light strip is specifically installed on both sides of the bed foot, 10-15 cm above the ground, and extends a light path to the bathroom door.

[0073] Intervention process: Trigger: After the system determines that there is a "high probability of leaving the bed", it sends a command to the intelligent controller of the LED light strip.

[0074] Start-up: The controller drives the light strip, starting from 0 lumens, and gradually brightens to a stable brightness of 3 lumens within 3 seconds in a smooth S-shaped curve.

[0075] Spectral control: The color temperature of the emitted light is strictly controlled below 2200K, which is warm yellow light. The intensity of the blue light band (460-480nm) in its spectrum is significantly suppressed to ensure that it does not interfere with the user's melatonin secretion.

[0076] Termination: Once the pressure sensor array confirms that the user has safely returned to bed or entered the bathroom, the light strip smoothly turns off within 1 second.

[0077] S402. Implementation of interventions targeting "pre-awakening auras" Actuator configuration: The system achieves environmental control by establishing communication connections with smart air conditioners and independent humidifiers.

[0078] Intervention process: Parameter acquisition: The system first reads the real-time ambient temperature (T) of the current bedroom. current ) and relative humidity (RH) current ).

[0079] Target calculation: Calculate the target temperature T based on fine-tuning of the individual's comfort baseline. target =T current +1.0℃, target humidity RH target =60%.

[0080] Collaborative execution: Send a command to the smart air conditioner to set its operating mode to "heating" or "fan" (depending on the season), and set the target temperature to T. target .

[0081] Send a command to the standalone humidifier to start and run until the ambient humidity reaches RH. target .

[0082] Maintenance of effect: This intervention strategy lasts for 30 minutes, or until the system determines that the user has re-entered a stable sleep state (such as N2 or N3).

[0083] S403. Implementation of interventions targeting "early awakening risk" Actuator configuration: Controllable light source: It adopts a full-spectrum LED light panel with precise dimming and is installed on the ceiling to avoid direct light shining into people's eyes.

[0084] Micro-vibration device: It adopts a linear resonant actuator, whose vibration frequency and amplitude are programmable and controlled, and is installed inside the mattress or bed frame.

[0085] Aroma diffuser: It uses an ultrasonic atomizing aroma diffuser and a special aromatherapy liquid with cedarwood essential oil as its main ingredient.

[0086] Intervention process: Synchronous Trigger: After the system determines the "risk of early wake-up", it sends a synchronous start command to the three actuators.

[0087] Multimodal intervention: Light intervention: A controllable light source emits amber light with a center wavelength of 590nm, forming a uniform light curtain with an average illuminance of <10 Lux above the bed surface for 20 minutes.

[0088] Vibration intervention: The linear resonant actuator generates gentle periodic vibrations at a frequency of 0.8 Hz (simulating resting heart rate), with each vibration cycle lasting 2 seconds and an interval of 1 second.

[0089] Olfactory intervention: The aroma diffuser operates in intermittent mode (working for 10 seconds and stopping for 50 seconds) to ensure that the aroma concentration is maintained at a perceptible but not strong level.

[0090] Termination conditions: The intervention lasts for 20 minutes, or the system determines that the user has successfully returned to sleep through vital sign monitoring during the intervention period.

[0091] In one embodiment of the present invention, based on step S5, a possible embodiment will be given below, and its specific implementation will be described in a non-limiting manner.

[0092] S501. Construction of the Feedback Dataset The system creates a complete record for each intervention event, constructing a time-series feedback dataset. Each record contains the following triplet information: State (s) t ): This refers to the state before intervention. A multidimensional vector, including: Predictive intentions that trigger intervention (such as tendency to get out of bed, pre-awakening symptoms, risk of early awakening).

[0093] All key characteristic values ​​corresponding to the trigger (such as longitudinal displacement velocity of the center of gravity, respiratory disturbance index, rest-activity ratio before dawn, etc.).

[0094] Action (a) t ): This refers to the intervention action. A multi-dimensional vector that specifically describes the executed strategy and its parameters, for example: For light guidance: [Brightness = 3 lumens, Gradient speed = 3 seconds] For temperature regulation: [Temperature change range = +1.0℃] For micro-vibrations: [Intensity = 0.3G, Period = 0.8Hz] New Status and Rewards (s) {t+1} , r t ): That is, the reward signal of the state after intervention and its transformation.

[0095] The system monitors the user's physiological and behavioral responses within a predetermined 15-minute time window after the intervention, and calculates the reward value r accordingly. t .

[0096] S502. Design of the Reward Function reward function r t It is guided by reinforcement learning optimization, and its design is as follows: For interventions outside the bed: r t =+5: If the user safely gets out of bed and returns within the next 5 minutes.

[0097] r t =-5: If the user does not get out of bed within the next 5 minutes, but the system triggers an intervention (considered a false alarm).

[0098] r t =-20: If a user falls out of bed or is suspected of falling (judged by the abnormal impact pattern of the pressure array).

[0099] Intervention for Awakening: r t =+3: If the user successfully maintains or returns to N2 / N3 stage sleep within 15 minutes after intervention.

[0100] r t =0: If the user's sleep stage neither worsens nor improves.

[0101] r t =-3: If the user is fully awake within 15 minutes after the intervention.

[0102] Intervention for early awakening: r t =+5: If the user successfully returns to sleep after intervention until the expected wake-up time.

[0103] r t =+2: If the user is not fully asleep, but their body movement level is significantly reduced, they are in a light sleep state.

[0104] r t =-2: Several pre-invalid, users fully awakened.

[0105] S503. Specific Implementation of Reinforcement Learning Algorithms Algorithm selection: The proximal strategy optimization algorithm is adopted because it has advantages in handling continuous action space and ensuring training stability.

[0106] Network structure: Actor Network: Input is state s t The output is a multidimensional Gaussian distribution defined on the action parameters, with the mean used to generate new actions (such as new thresholds or new brightness) and the variance used for exploration.

[0107] Critics Network: Input is state s t The output is the value estimate V(s) for that state. t (), used to evaluate the merits of the current strategy.

[0108] Optimization process: Data collection: The system runs continuously in the background, collecting a large number of interaction experience tuples (s t , a t ,r t ,s {t+1} And store it in the experience replay pool.

[0109] Periodic training: The system starts a training cycle every 1,000 new experiences accumulated.

[0110] Policy Update: The actor network updates its policy based on the advantage function computed by the critic network, aiming to maximize the expected cumulative reward. The PPO algorithm maintains stability by pruning the proxy objective function to ensure that each policy update does not deviate too far from the old policy.

[0111] Parameter Output: After training convergence, the system decodes the optimal motion parameters output by the actor network into specific, executable system configurations: Decision threshold: For example, the dynamic threshold for the longitudinal displacement velocity of the center of gravity of the bed-off tendency is optimized from 0.15 m / s to 0.18 m / s.

[0112] Intervention parameters: such as optimizing the micro-vibration intensity of early awakening intervention from 0.3G to 0.25G, or adjusting the gradual brightening speed of light guidance from 3 seconds to 4 seconds.

[0113] In some embodiments, the sleep state monitoring-based environmental regulation system may include multiple functional modules composed of computer program segments. The computer programs for each program segment in the sleep state monitoring-based environmental regulation system may be stored in the memory of a computer device and executed by at least one processor to perform (see details). Figure 1 (Description) Functions for environmental regulation based on sleep state monitoring.

[0114] In this embodiment, the sleep state monitoring-based environmental regulation system can be divided into multiple functional modules according to its functions, such as... Figure 2 As shown. The module referred to in this invention is a series of computer program segments that can be executed by at least one processor and perform a fixed function, and is stored in memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0115] The data acquisition module is used to collect status data through a multimodal non-sensor network. The status data includes mattress pressure data, vital sign data, and environmental parameters. The feature extraction module is used to extract key features from the state data, including features of tendency to get out of bed, sleep stability, and early awakening. The intent recognition module is used to analyze key features through machine learning models to predict users' sleep behavior intent, including the tendency to get out of bed, pre-awakening aura, and risk of early awakening. An environmental control module is used to trigger an environmental actuator based on a predicted intent to adjust at least one parameter among light, temperature, and humidity. The decision optimization module records user status data before and after intervention and optimizes decision thresholds and intervention parameters through reinforcement learning algorithms. Among them, the characteristics of getting out of bed include at least the longitudinal displacement velocity of the center of gravity of the pressure center trajectory and the pressure release area ratio; the characteristics of sleep stability include at least the respiratory disturbance index and the low-frequency / high-frequency power ratio of heart rate variability; and the characteristics of early awakening include at least the rest-activity ratio before dawn.

[0116] Figure 3The environmental adjustment method based on sleep state monitoring provided in the embodiments of this application can be applied to devices. Those skilled in the art will understand that the device structure involved in the embodiments of this invention does not constitute a limitation on the device. A device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. In the embodiments of this invention, the device includes, but is not limited to, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of this application described and / or claimed herein.

[0117] The device 300 may include a processor 310, a memory 320, and a communication unit 330. These components communicate via one or more buses. Those skilled in the art will understand that the server structure shown in the figure does not constitute a limitation of the present invention. It may be a bus topology or a star topology, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0118] The memory 320 can be used to store execution instructions of the processor 310. The memory 320 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. When the execution instructions in the memory 320 are executed by the processor 310, the device 300 is able to perform some or all of the steps in the above method embodiments.

[0119] The processor 310 serves as the control center of the storage device, connecting various parts of the electronic device via various interfaces and lines. It executes software programs and / or modules stored in the memory 320, and calls data stored in the memory to perform various functions of the electronic device and / or process data. The processor can be composed of integrated circuits (ICs), such as a single packaged IC or multiple packaged ICs with the same or different functions connected together. For example, the processor 310 may consist only of a central processing unit (CPU). In this embodiment of the invention, the CPU may have a single processing core or include multiple processing cores.

[0120] The communication unit 330 is used to establish a communication channel, enabling the storage device to communicate with other devices. It can receive user data sent by other devices or send user data to other devices.

[0121] The present invention also provides a computer medium, wherein the computer medium may store a program, which, when executed, may include some or all of the steps provided in the embodiments of the present invention. The medium may be a magnetic disk, an optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0122] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a medium 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, or any other medium capable of storing program code. It includes several instructions to cause a computer device (which may be a personal computer, a server, or a second device, network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0123] The same or similar parts between the various embodiments in this specification can be referred to mutually. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple, and the relevant parts can be referred to the description in the method embodiments.

[0124] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or modules may be electrical, mechanical, or other forms.

[0125] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0126] In addition, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0127] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the present invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the present invention by those skilled in the art without departing from the spirit and essence of the invention, and such modifications or substitutions should all be within the scope of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should also be covered within the protection scope of the present invention.

Claims

1. An environmental regulation method based on sleep state monitoring, characterized in that, include: Status data is collected through a multimodal non-sensor network, including mattress pressure data, vital signs data, and environmental parameters. Key features are extracted from the state data, including the tendency to get out of bed, sleep stability, and early awakening. By analyzing key features through machine learning models, users' sleep behavior intentions can be predicted, including the tendency to get out of bed, pre-awakening symptoms, and risk of early awakening. The environmental actuator is triggered according to the predicted intent to adjust at least one parameter among light, temperature, and humidity. Record user status data before and after intervention, and optimize decision thresholds and intervention parameters through reinforcement learning algorithms; Among them, the characteristics of getting out of bed include at least the longitudinal displacement velocity of the center of gravity of the pressure center trajectory and the pressure release area ratio; the characteristics of sleep stability include at least the respiratory disturbance index and the low-frequency / high-frequency power ratio of heart rate variability; and the characteristics of early awakening include at least the rest-activity ratio before dawn.

2. The method according to claim 1, characterized in that, Status data is collected through a multimodal, non-sensor network, including: By deploying a distributed pressure sensor array under the mattress or integrating it into the mattress, the user's body movement and pressure distribution data are collected. The data includes the pressure center trajectory, the rate of change of pressure distribution, and the pressure pattern of specific postures. The bio-radar installed on the bedside table or ceiling collects the user's vital signs signals, including micro-movement breathing waveforms, heartbeat vibration signals, and macroscopic body movements. Environmental parameters, including ambient temperature, relative humidity, and ambient light intensity, are collected by temperature and humidity sensors and light sensors installed in the bedroom space.

3. The method according to claim 1, characterized in that, Key features are extracted from the state data, including: The collected mattress pressure data and vital sign radar signals were subjected to noise reduction and filtering to remove high-frequency noise and low-frequency drift. The vital signs radar signal is subjected to signal separation to separate the respiratory, heartbeat and body movement components. Extracting bed-leaking tendency features from pressure data includes at least calculating the ratio of the longitudinal displacement velocity of the center of gravity of the pressure center to the pressure release area. Extracting sleep stability features from vital signs includes at least calculating the low-frequency / high-frequency power ratio of the respiratory disturbance index and heart rate variability. Early awakening characteristics are extracted from body movement data and environmental parameters, including at least calculating the pre-dawn rest-activity ratio and performing a correlation analysis between this ratio and ambient light intensity.

4. The method according to claim 3, characterized in that, Early awakening features were extracted from body movement data and environmental parameters, including: Within a predetermined time window before the user's expected wake-up time, the user's pre-dawn rest-activity ratio is calculated; the pre-dawn rest-activity ratio is the ratio of the user's body movement frequency within the predetermined time window to the body movement frequency within a historical reference period after the user enters deep sleep. Simultaneously, the ambient light intensity collected synchronously during the predetermined time window when the user's body movement frequency significantly increases is recorded and analyzed. The value of the pre-dawn rest-activity ratio was correlated with the ambient light intensity to determine the risk of early awakening for users and its causal relationship with premature morning light.

5. The method according to claim 1, characterized in that, By analyzing key features using machine learning models, we can predict users' sleep behavior intentions, including: During the initial 3 to 7 nights of system operation, the system will operate in a monitoring-only, non-intervention mode to learn and establish the user's personal sleep baseline, which includes at least typical sleep onset time, deep sleep period distribution time, average heart rate, respiratory rate, and typical number of nighttime awakenings. Based on the individual's sleep baseline, a specific machine learning model is used for real-time state classification and prediction, including: For the prediction of the tendency to get out of bed, a support vector machine or a lightweight neural network is used as a classifier, with the longitudinal displacement velocity of the center of gravity and the pressure release area ratio as input features; when the feature value exceeds a dynamic threshold set based on the individual's sleep baseline, it is determined that the user has a high probability of intending to get out of bed. For the prediction of sleep disruption, a hidden Markov model or recurrent neural network is used to analyze the respiratory disturbance index sequence and the low-frequency / high-frequency power ratio sequence of heart rate variability to predict the probability of awakening within a specific time period in the future; when the predicted probability is consistently higher than a preset threshold and accompanied by an increase in microbody movements, it is determined to be a premonitory symptom of awakening. To assess the risk of early awakening, if the pre-dawn rest-activity ratio calculated within a specific time window from the expected wake-up time continuously rises and remains at a set high threshold for more than 5 minutes compared to the user's baseline ratio during deep sleep, then the risk of early awakening is determined.

6. The method according to claim 1, characterized in that, Based on the predicted intent, an environmental actuator is triggered to adjust at least one parameter among light, temperature, and humidity, including: Based on the predicted different sleep behavior intentions, pre-set, differentiated, and collaborative intervention strategies are triggered, including: When the predicted intention is a high probability of getting out of bed, the triggered intervention strategy is to activate the LED light strips placed by the bedside or on the ground, so that they emit a gradually brightening warm light with a brightness of less than 5 lumens, forming a guiding light path towards the bathroom. When the predicted intention is a premonition of awakening, the triggered intervention strategy is to control the air conditioning, fresh air or humidifier system to raise the ambient temperature by 0.8-1.2℃ and adjust the ambient relative humidity to about 60% RH. When the predicted intention is the risk of early awakening, the triggered intervention strategy is: to simultaneously activate a controllable light source, a micro-vibration device, and an aromatherapy diffuser; wherein, the controllable light source emits an amber light curtain with a wavelength of 590nm, the micro-vibration device generates gentle periodic vibrations, and the aromatherapy diffuser releases a soothing aroma with cedarwood as the main scent.

7. The method according to claim 1, characterized in that, Record user status data before and after intervention, and optimize decision thresholds and intervention parameters through reinforcement learning algorithms, including: The system continuously records and builds a feedback dataset, which includes: Pre-intervention state: The predictive intent and corresponding key feature values ​​upon which the intervention was triggered; Intervention actions: the specific intervention strategies implemented and their parameters; Post-intervention status: The user's physiological and behavioral responses within a predetermined period of time after the intervention, including at least whether they successfully continued to sleep, whether they safely got out of bed, or whether they were fully awake. The post-intervention state is used as the reward signal for the reinforcement learning algorithm to dynamically optimize the decision threshold and intervention parameters; The decision thresholds include a dynamic threshold for longitudinal displacement velocity of the center of gravity used to predict the tendency to get out of bed, a threshold for the probability of arousal used to predict arousal precursors, and a threshold for the pre-dawn rest-activity ratio used to determine the risk of early arousal. The intervention parameters include the brightness and gradual brightening speed of the light guidance, the amplitude of temperature adjustment, and the intensity and period of micro-vibration.

8. An environmental regulation system based on sleep state monitoring, characterized in that, include: The data acquisition module is used to collect status data through a multimodal non-sensor network. The status data includes mattress pressure data, vital sign data, and environmental parameters. The feature extraction module is used to extract key features from the state data, including features of tendency to get out of bed, sleep stability, and early awakening. The intent recognition module is used to analyze key features through machine learning models to predict users' sleep behavior intent, including the tendency to get out of bed, pre-awakening aura, and risk of early awakening. An environmental control module is used to trigger an environmental actuator based on a predicted intent to adjust at least one parameter among light, temperature, and humidity. The decision optimization module records user status data before and after intervention and optimizes decision thresholds and intervention parameters through reinforcement learning algorithms. Among them, the characteristics of getting out of bed include at least the longitudinal displacement velocity of the center of gravity of the pressure center trajectory and the pressure release area ratio; the characteristics of sleep stability include at least the respiratory disturbance index and the low-frequency / high-frequency power ratio of heart rate variability; and the characteristics of early awakening include at least the rest-activity ratio before dawn.

9. An environmental regulation device based on sleep state monitoring, characterized in that, include: Memory for storing environmental adjustment programs based on sleep state monitoring; A processor, configured to implement the steps of the sleep state monitoring-based environmental regulation method as described in any one of claims 1-7 when executing the sleep state monitoring-based environmental regulation program.

10. A computer-readable medium storing a computer program, characterized in that, The readable medium stores an environmental regulation program based on sleep state monitoring, which, when executed by a processor, implements the steps of the environmental regulation method based on sleep state monitoring as described in any one of claims 1-7.

Images