Mattress type sleep monitoring data real-time processing method based on edge calculation

Abstract

The invention provides a mattress type sleep monitoring data real-time processing method based on edge calculation, relates to the technical field of medical health data processing, and aims to perform macroscopic state layer calculation, pathological event layer calculation and stability layer calculation at the same time in a layering manner based on a constructed dynamic sensing matrix, generate a structured sleep health report, and realize real-time sleep monitoring through one-time sleep monitoring. The comprehensive information about the sleep macrostructure, the respiratory pathology event burden and the sleep stability can be synchronously output; through a sleep health assessment report, total sleep duration and sleep efficiency calculated by the macroscopic state layer, apnea and type and severity calculated by the pathological event layer, and a sleep micro-awakening state and sleep fragmentation degree calculated by the stability layer can be clearly known, and a multi-dimensional and integrated comprehensive assessment result with high medical value is realized. The information amount of the sleep health portrait is enriched, and a comprehensive and reliable data support is provided for accurately positioning the root of the sleep problem and formulating a personalized intervention or treatment scheme.

Description

Technical Field

[0001] This invention relates to the field of medical and health data processing technology, specifically to a real-time processing method for mattress-based sleep monitoring data based on edge computing. Background Technology

[0002] In the current technology for sleep monitoring and data collection, polysomnography, smart wearable devices, and early smart mattress systems are generally used. These systems acquire various data points during sleep to assess sleep quality. For example, the "Health Monitoring Method, Device, Equipment, and Storage Medium Based on Smart Mattress" disclosed in application number CN202510875006.1 uses cluster analysis to perform sleep monitoring and early warning. However, the following problems exist in achieving widespread and accurate sleep health monitoring: Problem 1: Existing sleep monitoring devices such as mattresses and wristbands focus on macroscopic analyses such as deep sleep and light sleep, failing to identify and classify specific sleep events. They only provide a single score for sleep monitoring without providing specific reasons, resulting in limited value for auxiliary diagnosis of sleep-disordered breathing and insomnia. Furthermore, the monitoring results emphasize single functions, such as recording blood oxygen, respiration, and heart rate separately, failing to form a unified analytical framework for sleep assessment. The monitoring data has low medical value density, is singular, and lacks clinical diagnostic value. The second issue is that existing smart mattresses generally use two methods for data processing. One is to upload all the raw waveform data with a high sampling rate to the cloud, which leads to high network bandwidth pressure, high power consumption, and a significant increase in the risk of user privacy leakage. Moreover, the latency of cloud feedback makes it impossible to support millisecond-level real-time intervention. The other is to perform only extremely simple feature extraction (such as calculating the average heart rate) at the edge. Although this ensures low latency, it sacrifices the dimension and depth of analysis, resulting in a rigid data processing architecture that cannot balance real-time performance and analytical depth. Summary of the Invention

[0003] To achieve the above objectives, the present invention provides the following technical solution: a real-time processing method for mattress-based sleep monitoring data based on edge computing, the method comprising: The system acquires raw physiological signals from a multimodal sensor array embedded in the mattress in real time, performs validity screening on the raw physiological signals to generate valid signal sources, and constructs a dynamic sensing matrix based on the valid signal sources. Based on the dynamic perception matrix, three levels of computational analysis are performed in parallel, including macroscopic state layer calculation, pathological event layer calculation, and stability layer calculation, generating three levels of analysis results. The analysis results from the three levels are integrated to generate a sleep health assessment report.

[0004] Furthermore, the construction of the dynamic sensing matrix based on effective signal sources includes: The multimodal sensor array includes a pressure sensor array, a bioradar sensor, and a temperature sensor. It performs validity screening on each signal source of the acquired raw physiological signals to generate valid signal sources. For each valid signal source, a standardized physiological feature vector is calculated, which includes time-domain features and frequency-domain features; The standardized physiological feature vectors corresponding to all effective signal sources are used as column vectors, and the effective signal sources are used as row vectors to construct a dynamic sensing matrix. The number of rows in the dynamic sensing matrix is ​​dynamically adjusted according to the number of effective signal sources passing through within a preset time window.

[0005] Furthermore, the macroscopic state layer calculation is used to assess the macroscopic structure of sleep, and the macroscopic state layer calculation includes: The dynamic perception matrix is ​​dimensionality reduced to extract the main feature vector representing the overall physiological pattern of the current time window, which is then used as the physiological state vector. Over Q consecutive time windows, the changes in the physiological state vector are tracked to form a physiological state trajectory; The dynamic characteristics of physiological state trajectories are analyzed, including moving variance and changing trend. Based on these dynamic characteristics, sleep is divided into different macroscopic states. Based on the division of macro-states, macro-structure summary data is generated.

[0006] Furthermore, the macroscopic states include stable core sleep, active brain activity, wakefulness, and unstable sleep, and the macroscopic structure summary data includes sleep efficiency, the percentage of each macroscopic state in the total sleep time, and the number of awakenings.

[0007] Furthermore, the pathological event layer calculation is used to identify and classify specific sleep breathing events, and the pathological event layer calculation includes: Row vectors related to respiratory monitoring are dynamically selected from the dynamic sensing matrix to form a respiratory monitoring submatrix; Based on the respiratory monitoring sub-matrix, the respiratory effort signal is separated from the pressure sensor array, and the respiratory airflow proxy signal is extracted from the bioradar sensor; Real-time analysis of the synergistic relationship between respiratory effort signals and respiratory airflow proxy signals, and identification and classification of respiratory events based on the synergistic relationship.

[0008] Furthermore, the respiratory events include obstructive sleep apnea events, central sleep apnea events, and airflow limitation events, and the synergistic relationships include the amplitude relationship, morphological relationship, and cross-correlation coefficient of the signals.

[0009] Furthermore, the pathological event layer calculation also includes: All identified respiratory events are statistically analyzed to generate pathological event summary data. The pathological event summary data includes the apnea-hypopnea index, obstructive apnea index, central apnea index, percentage of airflow-limited events in total sleep time, and a detailed list of various respiratory events, including the start time, duration, and type of the respiratory event.

[0010] Furthermore, the stability layer calculation is used to evaluate the microstructure and stability of sleep, and the stability layer calculation includes: Over Q consecutive time windows, the difference between the dynamic perception matrices of adjacent time windows is calculated to generate a matrix perturbation index. Set the matrix perturbation threshold V5. When the peak value of the matrix perturbation index is within the fourth preset event threshold T6, and continues to exceed V5, and the duration of the peak value exceeding V5 is less than the fifth preset time threshold T7, then the event is marked as a micro-awakening event. When the duration of the matrix perturbation index exceeding the V5 peak exceeds T7, it is determined to be a full awakening event and output to the macrostructure summary data; Based on the monitored micro-awakening events, stability summary data is calculated, which includes the micro-awakening index and the sleep fragmentation index.

[0011] Furthermore, the analysis results from the three levels are fused to generate a sleep health assessment report, including: The acquired macroscopic structural summary data, pathological event summary data, and stability summary data are comprehensively scored to generate an overall risk assessment level; The report integrates macroscopic structural summary data, pathological event summary data, stability summary data, comprehensive score, and overall risk assessment level into a unified report to form a sleep health assessment report.

[0012] Furthermore, the sleep health assessment reports generated in each monitoring cycle are stored as personalized user profiles, which serve as a benchmark for comparison between macroscopic state layer calculations, stability layer calculations, and sleep health assessment reports.

[0013] This invention provides a real-time processing method for mattress-based sleep monitoring data based on edge computing. It offers the following advantages: 1. This invention employs a three-layer computing architecture based on a dynamic perception matrix, simultaneously performing macroscopic state layer calculations, pathological event layer calculations, and stability layer calculations to generate a structured sleep health report. Through a single sleep monitoring session, comprehensive information on the macroscopic structure of sleep, respiratory pathological event burden, and sleep stability can be output simultaneously. The sleep health assessment report clearly shows the total sleep duration and sleep efficiency calculated at the macroscopic state layer, the sleep apnea and its type and severity calculated at the pathological event layer, and the sleep micro-awake state and sleep fragmentation degree calculated at the stability layer. This achieves a multi-dimensional, integrated, and highly medically valuable comprehensive assessment result, enriching the information content of the sleep health profile and providing comprehensive and reliable data support for accurately locating the root cause of sleep problems and developing personalized intervention or treatment plans.

[0014] 2. This invention collects raw physiological signals by embedding a multimodal sensor array inside the mattress to construct a dynamic sensing matrix. Based on the dynamic sensing matrix, it performs hierarchical processing calculations. The dimensionality reduction and trajectory analysis of the macroscopic state layer, the signal reconstruction and logical judgment of the pathological event layer, and the matrix difference calculation of the stability layer all adopt computationally lightweight and interpretable mathematical methods, such as principal component analysis, matrix operations, and threshold logic. Compared with the traditional method of uploading all data to the cloud for processing, the lightweight distributed architecture reduces the computational pressure, resulting in shorter computation time, lower cost, and higher efficiency. It can complete the entire process from data acquisition to event recognition within a few hundred milliseconds, while ensuring the depth and breadth of analysis. It improves the data processing capability under edge resource constraints, ensuring computational efficiency while taking into account the real-time nature and depth of data analysis. Attached Figure Description

[0015] Figure 1 This is a flowchart of the real-time processing method for mattress-type sleep monitoring data based on edge computing according to the present invention. Figure 2 This is a data transmission diagram of the real-time processing method for mattress-type sleep monitoring data based on edge computing according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example 1:

[0017] like Figures 1 to 2 As shown, a real-time processing method for mattress-based sleep monitoring data based on edge computing is described, and the method includes: Step S100: Real-time acquisition of raw physiological signals collected by the multimodal sensor array embedded in the mattress; after screening the effectiveness of the raw physiological signals, generating effective signal sources; and constructing a dynamic sensing matrix based on the effective signal sources. Step S101: The multimodal sensor array includes a pressure sensor array, a bio-radar sensor, and a temperature sensor. The validity of each signal source of the acquired raw physiological signals is screened to generate valid signal sources; wherein: In addition to pressure sensor arrays, bioradar sensors, and temperature sensors, the multimodal sensor array also includes one or more of heart rate sensors, blood oxygen sensors, humidity sensors, and sound sensors. The pressure sensor array is typically evenly distributed across the torso and limb areas of the mattress to collect pressure distribution signals for extracting respiratory effort signals. The bioradar sensor is installed at the corresponding chest position on the mattress, collecting chest movement signals by emitting and receiving electromagnetic waves for extracting respiratory airflow proxy signals. The temperature sensor is embedded in the mattress surface to collect surface temperature signals, aiding in the assessment of thermoregulation changes during sleep. The heart rate sensor is integrated into the back or chest area of ​​the mattress, collecting heart rate signals through photoplethysmography for assessing cardiovascular activity. The blood oxygen sensor is placed as a patch at a designated location on the mattress to collect blood oxygen saturation signals for detecting hypoxia events. The humidity sensor is installed on the mattress surface to collect ambient humidity signals for correcting temperature signals and assessing the sleep environment. The raw physiological signals include raw voltage or digital signals acquired from the various sensors in the multimodal sensor array, specifically including pressure signals, radar signals, temperature signals, heart rate signals, blood oxygenation signals, and humidity signals, etc. The effectiveness screening includes signal quality assessment, noise filtering, and physiological rationality checks. First, each signal source is baseline-calibrated to remove DC offset and environmental interference. Then, the signal-to-noise ratio (SNR) is calculated. If it is lower than a preset SNR threshold (e.g., 20 dB), it is marked as invalid. Next, the signal amplitude of the signal source is checked to see if it is within the physiologically reasonable range. For example, the pressure signal amplitude should be between 0.5-10V, and the radar signal frequency should be between 0.1-2Hz, corresponding to the respiratory range. If it is outside the range, it is discarded. Finally, by comparing the synchronicity of multiple sensors within the same time window, if a signal source is inconsistent with most signal sources, it is considered invalid. This completes the effectiveness screening of the signal sources, obtaining valid signal sources. Effective signal sources refer to high-quality original physiological signals retained after effectiveness screening, including pressure signals, bioradar signals, temperature signals, etc. Each signal source is calibrated and verified to ensure its reliability and consistency. Effective signal sources serve as input data for constructing the dynamic sensing matrix, avoiding interference from noise or abnormal data on sleep monitoring results and improving the accuracy and real-time performance of the overall analysis.

[0018] Step S102: Calculate a standardized physiological feature vector for each valid signal source. The standardized physiological feature vector includes predefined, lightweight time-domain and frequency-domain features; wherein: Standardized physiological feature vectors are used to quantify the physiological characteristics of each effective signal source; the calculation of standardized physiological feature vectors for effective data sources includes: First, each valid signal source is preprocessed, including denoising and normalization, scaling the signal amplitude of the valid signal source to a standard range, such as [0, 1]. Then, within a preset time window, time-domain characteristics are calculated. These characteristics include the signal mean, variance, and zero-crossing rate. The signal mean is the arithmetic mean of all sampling points, used to represent the overall level of the signal. The variance is the average of the squares of the differences between each sampling point and the signal mean, used to measure the volatility of the signal. The zero-crossing rate is the number of times the signal crosses zero, used to indicate the frequency components and activity of the signal. Next, the signal is converted to the frequency domain using a Fast Fourier Transform to calculate frequency domain features. Frequency domain features include signal energy, dominant frequency, and spectral entropy. Signal energy is the sum of the squares of all frequency components and is used to represent the signal strength. Dominant frequency is the frequency component with the highest energy identified and is used to reflect the dominant period of the signal. Spectral entropy is the probability entropy of the frequency distribution and is used to measure the complexity and randomness of the signal. Finally, the calculated time-domain and frequency-domain features are combined to form a standardized physiological feature vector, which facilitates unified comparison and analysis in the dynamic perception matrix, thereby supporting multi-level assessment of sleep state.

[0019] Step S103: Use the standardized physiological feature vectors corresponding to all effective signal sources as column vectors and the effective signal sources as row vectors to construct a dynamic sensing matrix. The number of rows in the dynamic sensing matrix is ​​dynamically adjusted according to the number of effective signal sources passing through the preset time window to ensure that the dynamic sensing matrix can still use the remaining effective signal sources for calculation when some sensor signals temporarily fail. The number of rows in the dynamic sensing matrix is ​​dynamically adjusted based on the number of valid signal sources within a preset time window, including: At the beginning of each time window, count the number of valid signal sources that have passed the validity screening, and denote it as M as a row vector; Then, the standardized physiological feature vectors corresponding to each effective signal source are used as column vectors and arranged in rows to construct an M-row N-column matrix, where N is the fixed dimension (i.e. the number of features) of the standardized physiological feature vectors. If a sensor signal fails, M is reduced and the number of rows is reduced accordingly, but the number of columns N remains unchanged, ensuring that the dynamic sensing matrix can still be calculated using the remaining effective signal sources. For example, suppose there are initially 5 valid signal sources: pressure, radar, temperature, heart rate, and blood oxygen. The dynamic sensing matrix is ​​5 rows and N columns. If the temperature sensor temporarily fails, the number of valid signal sources is reduced to 4. The matrix is ​​then adjusted to 4 rows and N columns. The feature vectors of the remaining 4 signal sources are used to perform calculations for the macroscopic state layer, pathological event layer, and stability layer. This maintains the continuity of monitoring when some sensors fail and avoids analysis interruptions caused by data loss.

[0020] Step S200: Based on the dynamic perception matrix, three levels of computational analysis are executed in parallel. These computational analyses include macroscopic state layer calculation, pathological event layer calculation, and stability layer calculation, generating analysis results at all three levels. The macroscopic state layer calculation is mainly used to assess the macroscopic structure of sleep. By analyzing the dynamic characteristics of physiological state trajectories, sleep stages are divided, generating macroscopic structure summary data to help users understand the overall quality and cycle distribution of sleep. The pathological event layer calculation is used to identify and classify specific sleep breathing events. By analyzing the synergistic relationship between respiratory effort signals and respiratory airflow proxy signals, pathological event summary data is generated for diagnosing sleep-disordered breathing and assessing respiratory health risks. The stability layer calculation is used to assess the microstructure and stability of sleep. By calculating the matrix perturbation index, micro-awakening and full awakening events are detected, generating stability summary data to reveal the continuity and depth of sleep and help identify sleep interruption and fragmentation issues. Through the parallel execution of these three levels of computation, comprehensive sleep health insights are provided. Step S201: Macroscopic state layer calculation includes: Step S2011: Perform dimensionality reduction on the dynamic perception matrix and extract the main feature vector representing the overall physiological pattern of the current time window as the physiological state vector; Dimensionality reduction is used to reduce the dimensionality of data to extract the principal feature vector representing the current time window. Dimensionality reduction reduces the amount of computation while ensuring that the principal feature vector is not lost, thereby improving computation speed and efficiency, and saving computational costs and time. First, the covariance matrix of the dynamic sensing matrix (an M-row N-column matrix, where M is the number of effective signal sources and N is the dimension of the standardized physiological feature vector) is calculated. Then, the eigenvalues ​​and eigenvectors of the covariance matrix are solved. The eigenvector with the largest eigenvalue is selected as the principal feature vector. The principal feature vector captures the largest variance in the covariance matrix, thus representing the overall physiological pattern of the current time window. The physiological state vector is the main feature vector extracted after dimensionality reduction. It is a low-dimensional vector that contains the main physiological features after compression. It is used to characterize the overall physiological pattern of the current time window, such as physical activity level, breathing pattern and heart rate variability, and provides a unified basis for subsequent tracking of physiological state trajectory and sleep stage division. Step S2012: Track the changes in physiological state vectors over Q consecutive time windows to form a physiological state trajectory. The physiological state trajectory is a sequence formed by connecting the physiological state vectors of each time window in chronological order over Q consecutive time windows. It contains the physiological state vector at each time point and reflects the changes in physiological state over time. By analyzing the dynamic characteristics of the physiological state trajectory, the transition and stability of the macroscopic sleep state can be identified. Step S2013: Analyze the dynamic characteristics of the physiological state trajectory. The dynamic characteristics include moving variance and trend of change. Based on the dynamic characteristics, sleep is divided into different macro states. The moving variance and trend of change of the dynamic characteristics are used to quantify the volatility and directionality of the physiological state trajectory. The moving variance is calculated by taking the variance of the physiological state vector within a sliding window (e.g., 5 consecutive time points). The variance of each feature component is calculated and then the average value is taken. This is used to measure the stability of the physiological state trajectory. During the calculation, a moving variance threshold is set to distinguish between high and low moving variance. Low moving variance indicates a stable state. The trend of change is obtained by fitting the points of the physiological state trajectory within the sliding window through linear regression to obtain the slope value. This is used to determine whether the state tends to be stable or changing. With a horizontal 0° straight line as the benchmark, a positive slope indicates an upward state and a negative slope indicates a downward state. The moving variance and trend of change serve as quantitative indicators for the division of sleep stages and help distinguish different macro states. The macroscopic sleep state includes stable core sleep, active brain activity, wakefulness, and unstable sleep. Stable core sleep represents the deep sleep stage, where the physiological state is highly stable. The length of the data reflects the duration of deep sleep, and a longer duration indicates high sleep quality. Active brain activity represents the rapid eye movement (REM) sleep stage, where brain activity is active and dreams are frequent. The length of the data reflects the proportion of REM sleep, which is about 20%-25% in normal adults. Wakefulness represents the fully awake state, where the physiological state changes abruptly. The length of the data reflects the frequency and duration of sleep interruptions, and a longer duration indicates low sleep efficiency. Unstable sleep represents the light sleep or transitional stage, where the physiological state is between stable core sleep and wakefulness. The length of the data reflects the degree of sleep fragmentation, and a longer duration indicates poor sleep quality. When the variance of the physiological state trajectory within a continuous first time period T1 is lower than the first threshold V1, and the absolute value of the slope of the trend is less than the second threshold S1, it is determined to be a stable core sleep period. When the frequency of the periodic peak of the moving variance of the physiological state trajectory is higher than the third threshold F1 and the amplitude of the peak is lower than the fourth threshold V2 within the second consecutive time period T2, it is determined to be an active brain activity period. When the instantaneous rate of change of the variance of the physiological state trajectory exceeds the fifth threshold V3, and the Euclidean distance of the physiological state vector relative to the previous time window exceeds the sixth threshold D1, it is determined to be the awakening period. When the variance of the movement of the physiological state trajectory is between the first threshold V1 and the seventh threshold V4, and the absolute value of the slope of its trend is between the second threshold S1 and the eighth threshold S2, but does not meet the criteria for the awakening period, it is judged as an unstable sleep period. Among them, the first time period T1, the second time period T2, the first threshold V1, the second threshold S1, the third threshold F1, the fourth threshold V2, the fifth threshold V3, the sixth threshold D1, the seventh threshold V4 and the eighth threshold S2 are predefined thresholds obtained based on user group data and through self-learning from user personalized profiles. For example, the first time period T1 is 10 minutes, the second time period T2 is 5 minutes, the first threshold V1 is 0.1 (moving variance units), the second threshold S1 is 0.01 (slope units), the third threshold F1 is 0.5 Hz (peak frequency), the fourth threshold V2 is 0.5 (peak amplitude units), the fifth threshold V3 is 0.3 (instantaneous rate of change units), the sixth threshold D1 is 0.5 (Euclidean distance units), the seventh threshold V4 is 0.3, and the eighth threshold S2 is 0.05; if the physiological trajectory moves within 10 consecutive minutes... If the moving variance is consistently below 0.1 and the absolute value of the slope of the trend is less than 0.01, it is considered a stable core sleep period; if the frequency of periodic peaks in the moving variance is higher than 0.5 Hz and the peak amplitude is lower than 0.5 within 5 minutes, it is considered an active brain activity period; if the instantaneous rate of change of the moving variance exceeds 0.3 and the Euclidean distance of the physiological state vector exceeds 0.5, it is considered a wakefulness period; if the moving variance is between 0.1 and 0.3 and the absolute value of the slope is between 0.01 and 0.05, but does not meet the wakefulness period criteria, it is considered an unstable sleep period. Step S2014: Based on the distribution and transformation of macro-states throughout the sleep process, generate macro-structure summary data. This data includes sleep efficiency, the percentage of each macro-state in total sleep time, and the number of awakenings. The macro-structure summary data reflects the overall structure and quality of sleep, providing users with a quantitative sleep assessment to identify sleep problems and support health decisions. Sleep efficiency is the ratio of total sleep time to total time in bed, reflecting sleep effectiveness. For example, a sleep efficiency of 90% means that 90% of time in bed is actually spent sleeping; low efficiency may indicate insomnia or environmental disturbances. The percentage of each macro-state in total sleep time represents each sleep stage (such as stable core sleep, active brain activity, etc.). The percentage of time spent in core sleep (such as sleep periods) reflects the health of the sleep structure. For example, a 40% percentage of stable core sleep indicates sufficient deep sleep, while a percentage below 20% may indicate insufficient sleep. The number of awakenings is the total number of fully awake events during sleep, reflecting the frequency of sleep interruptions. For example, 5 awakenings indicate relatively continuous sleep, while more than 15 awakenings may indicate severe sleep fragmentation. For example, if a sleep health assessment report shows a sleep efficiency of 85%, a 30% percentage of stable core sleep, a 20% percentage of active brain activity, and 8 awakenings, this indicates that the user has moderate sleep quality, good deep sleep but a high number of awakenings, and may need to optimize the sleep environment or investigate pathological events. Step S202: Calculation of the pathological event layer, including: Step S2021: Dynamically filter row vectors related to respiratory monitoring from the dynamic sensing matrix to form a respiratory monitoring sub-matrix; the row vectors related to respiratory monitoring refer to the rows corresponding to effective signal sources directly related to respiratory activity selected from the dynamic sensing matrix. The row vectors mainly come from the pressure sensor array and bio-radar sensor; the respiratory monitoring sub-matrix is ​​a matrix composed of respiratory-related row vectors. The number of rows is the number of effective signal sources related to respiratory monitoring, and the number of columns is consistent with the dynamic sensing matrix, that is, the dimension of the standardized physiological feature vector. The respiratory monitoring sub-matrix concentrates the scattered respiratory-related signals into a unified data structure, providing a dedicated and efficient data foundation for the subsequent extraction and analysis of the synergistic relationship of respiratory effort signals and respiratory airflow proxy signals, thereby ensuring the accuracy and real-time performance of pathological event identification; Step S2022: Based on the respiratory monitoring sub-matrix, the respiratory effort signal is separated from the pressure sensor array using a blind source separation algorithm, and the respiratory airflow proxy signal representing thoracic cavity movement is extracted from the bio-radar sensor using a spectrum focusing algorithm; wherein: The respiratory effort signal is a signal reflecting the intensity of respiratory muscle activity in the thoracic or abdominal cavity and is used to distinguish the type of sleep apnea. When acquiring the respiratory effort signal, firstly, all column vectors of the pressure sensor array, i.e., its standardized physiological feature vectors, are extracted from the respiratory monitoring submatrix. Then, the column vectors are reconstructed into the original or preprocessed multi-channel pressure signal. Finally, blind source separation technology is applied to the multi-channel pressure signal. By calculating the statistical independence of the signal, the independent components that are synchronized with the rhythm of respiratory muscle movement and whose amplitude is proportional to the respiratory effort are separated, which are the respiratory effort signal. The respiratory airflow proxy signal is a signal that simulates airflow through the mouth and nose and reflects the changes in the strength of respiratory airflow. It is used to directly determine whether breathing exists and the degree of airflow unobstructedness. When acquiring the respiratory airflow proxy signal, firstly, the row vector corresponding to the bioradar sensor is extracted from the respiratory monitoring sub-matrix. Then, the row vector is back-mapped to its frequency domain characteristics, mainly the narrowband spectrum near the main frequency. Finally, through the spectrum focusing algorithm, the signal energy is integrated in the narrowband spectrum near the main frequency and its change curve over time is tracked. This change curve is the respiratory airflow proxy signal representing the chest cavity undulation movement. Step S2023: Analyze the synergistic relationship between the respiratory effort signal and the respiratory airflow proxy signal in real time. The synergistic relationship includes the amplitude relationship, morphological relationship, and cross-correlation coefficient of the signals. The amplitude relationship refers to the relative change in amplitude between the respiratory effort signal and the respiratory airflow proxy signal, which is obtained by comparing the percentage change in amplitude of the two signals relative to their respective baseline levels during the respiratory event determination period. The morphological relationship refers to the similarity or difference in waveform between the respiratory effort signal and the respiratory airflow proxy signal, which is evaluated by dynamic time warping or by directly observing the phase consistency of the waveforms. The cross-correlation coefficient is an indicator that quantifies the degree of linear correlation between the respiratory effort signal and the respiratory airflow proxy signal in the time domain, which is obtained by dividing the covariance of the two signals within a sliding time window (e.g., one respiratory cycle) by the product of their standard deviations. Based on a pre-defined effort-airflow state judgment matrix, respiratory events are identified and classified according to synergistic relationships. Respiratory events include obstructive apnea events, central apnea events, and airflow limitation events. Obstructive apnea events represent complete collapse of the upper airway leading to cessation of airflow, but respiratory drive still exists, reflecting problems with the anatomy or muscle function of the upper airway. Central apnea events represent a temporary disappearance of the respiratory center drive in the brain, causing respiratory effort and airflow to stop simultaneously, reflecting abnormal regulation of the central nervous system. Airflow limitation events represent partial obstruction of the upper airway leading to a decrease in airflow but not complete cessation, while respiratory effort is compensated for and enhanced, reflecting a pathological state of increased upper airway resistance. When a respiratory effort signal is detected or its amplitude is increased, and the amplitude of the respiratory airflow proxy signal decreases by more than a first preset percentage threshold P1 compared to the baseline level before the event, and the duration of this state exceeds a first preset time threshold T3, it is determined to be an obstructive sleep apnea event. When the amplitudes of the respiratory effort signal and the respiratory flow proxy signal are detected to have decreased by more than the second preset percentage threshold P2 compared with the baseline level before the event, and the duration of this state exceeds the second preset time threshold T4, it is determined to be a central sleep apnea event. When the amplitude of the breathing effort signal shows a progressively increasing trend within a continuous third preset time threshold T5, while the amplitude of the breathing airflow proxy signal shows a progressively decreasing trend within the same time period, and the overall amplitude decreases by more than the third preset percentage threshold P3, it is determined to be an airflow limitation event. Among them, the first preset percentage threshold P1, the second preset percentage threshold P2, the third preset percentage threshold P3, the first preset time threshold T3, the second preset time threshold T4, and the third preset time threshold T5 are fixed values ​​preset based on clinical medical standards, or adaptive values ​​dynamically adjusted based on the user's personalized profile; for example: the first preset percentage threshold P1 is 90% (airflow decreases by more than 90% from the baseline), the second preset percentage threshold P2 is 80% (both effort and airflow decrease by more than 80%), the third preset percentage threshold P3 is 50% (airflow gradually decreases by 50%), the first preset time threshold T3 is 10 seconds, the second preset time threshold T4 is 10 seconds, and the third preset time threshold T5 is 30 seconds; for example: if it is detected that the respiratory effort signal amplitude remains constant within 12 consecutive seconds, but the respiratory airflow proxy signal amplitude decreases by 95% from the baseline level, based on the thresholds P1=90% and T3=10 seconds, it is determined to be an obstructive sleep apnea event; The preset effort-airflow state judgment matrix is ​​a two-dimensional lookup table. Its rows represent the state of respiratory effort signals (such as "present / enhanced", "disappeared / weakened"), and its columns represent the state of respiratory airflow proxy signals (such as "normal", "weakened", "disappeared"). Each cell defines the respiratory event type or no event determined by the corresponding state combination. The settings are based on the definition of respiratory events in clinical medical standards (such as the American Academy of Sleep Medicine AASM guidelines), that is, based on the specific combination pattern of effort and airflow, to provide a standardized and quickly queryable rule set, which can automatically and accurately classify respiratory events according to the real-time calculated synergistic relationship. The baseline level before the event refers to the average amplitude of the respiratory effort signal and the respiratory flow proxy signal during a stable respiratory period (usually the most recent 2 minutes) before the suspected respiratory event occurs. It is obtained by calculating the moving average of the signal amplitude during this stable period and serves as a reference benchmark for judging the percentage decrease in signal amplitude during the event. Step S2024: Statistical analysis is performed on all identified respiratory events to generate pathological event summary data. This data includes the apnea-hypopnea index, obstructive apnea index, central apnea index, percentage of airflow-limited events in total sleep time, and a detailed list of each type of respiratory event, including the start time, duration, and type. The pathological event summary data is used to quantify and summarize respiratory events occurring throughout the monitoring period, providing clinicians or users with a report of key indicators of sleep-related respiratory health risks. This data is used to assist in the diagnosis of sleep-disordered breathing (such as sleep apnea-hypopnea syndrome, SAHS) and to assess the effectiveness of treatment interventions. The Apnea-Hypopnea Index (AHI) is the total number of apnea (airflow decrease ≥90%) and hypopnea (airflow decrease ≥30% accompanied by decreased blood oxygenation or awakening) events per hour of sleep. It is calculated as: (number of apnea events + number of hypopnea events) / total sleep hours. It is used to assess the overall severity of sleep-disordered breathing. For example, an AHI ≥5 times / hour can be diagnosed as sleep apnea-hypopnea syndrome (SAHS). The Obstructive Sleep Apnea Index (OSAI) is the average number of obstructive sleep apnea events occurring per hour of sleep, calculated as the total number of obstructive sleep apnea events divided by the total number of sleep hours. It is used to assess the burden of upper airway obstructive events separately. The Central Apnea Index is the number of central apnea events that occur per hour of sleep on average. It is calculated as the total number of central apnea events divided by the total number of sleep hours and is used to assess the frequency of central respiratory events. The percentage of airflow-limited events in total sleep time refers to the proportion of the total duration of airflow-limited events during the entire sleep process. It is calculated as (sum of the durations of all airflow-limited events / total sleep time) × 100%, reflecting the overall impact of increased upper airway resistance during sleep. A detailed list of various respiratory events is a record of events arranged in chronological order, including the start time (timestamp), duration (seconds), and specific type (e.g., obstructive, central) for each respiratory event. It provides raw respiratory event data for in-depth analysis and review, such as determining whether respiratory events are concentrated in a specific sleep stage or position.

[0021] Step S203: Stability layer calculation, including: Step S2031: Calculate the difference between the dynamic perception matrices of adjacent time windows over Q consecutive time windows to generate a matrix perturbation index; the difference is quantified by calculating the matrix norm or matrix cosine similarity, and the difference is defined as the matrix perturbation index. The matrix perturbation index is a scalar value used to quantify the overall change in physiological state between adjacent time windows. It contains amplitude information extracted from the differences in the dynamic perception matrix and can capture transient physiological fluctuations ignored during macroscopic state layer calculations. The matrix perturbation index is generated as follows: First, the dynamic perception matrix of the current time window (denoted as matrix A) and the dynamic perception matrix of the previous time window (denoted as matrix B) are cached to ensure that the two matrices have the same dimensions, i.e., the same number of rows M and columns N. Next, the difference matrix between matrix A and matrix B is calculated, denoted as matrix C, i.e., C=AB. Then, the overall difference is quantified by calculating the Frobenius norm of the difference matrix C. The Frobenius norm is the square root of the sum of the squares of all elements in matrix C. Specific formulas can be found in publicly available techniques and will not be elaborated here. Finally, the norm value of the Frobenius norm is directly defined as the matrix perturbation index. The larger the value of the matrix perturbation index, the more drastic the change in physiological patterns between adjacent time windows, indicating the occurrence of micro-awakening or other sleep disruption events. Step S2032: Set the matrix perturbation threshold V5. When the peak value of the matrix perturbation index is within the fourth preset event threshold T6, and continuously exceeds V5, and the duration of the peak value exceeding V5 is less than the fifth preset time threshold T7, then the event is marked as a micro-awakening event. When the duration of the matrix perturbation index exceeding the V5 peak exceeds T7, it is determined to be a full awakening event and output to the macrostructure summary data; The matrix perturbation threshold V5 is set based on the statistical distribution of historical data of the user group. It is usually set as the 95th percentile of the matrix perturbation index value during stable sleep for a specific user group. For example, through the analysis of a large amount of healthy sleep data, it was found that the general setting value of V5 is 0.85 (based on the normalized matrix norm), which means that only changes in the perturbation index that exceed this threshold are considered to have physiological significance and correspond to sleep interruption events. The fourth preset time threshold T6 and the fifth preset time threshold T7 are set based on clinical experience and the typical duration of physiological events. T6 is usually set to 10 seconds, based on the fact that the cortical EEG changes of a meaningful micro-arousal event usually need to last for more than 3 seconds. Here, a more conservative 10 seconds is used to improve specificity. T7 is usually set to 30 seconds, based on the fact that a fully aroused event is usually accompanied by a relatively long period of body movement and consciousness, and a duration of more than half a minute is a common judgment criterion. Example: A peak of 0.92 (exceeding V5=0.85) was detected in the matrix perturbation index within 12 consecutive seconds, but the peak only lasted for 8 seconds before falling back. Since the duration of 8 seconds is less than T7=30 seconds, it is judged as a micro-arousal event. If the peak state lasts for 35 seconds, it is judged as a fully aroused event.

[0022] Step S2033: Based on the monitored micro-arousal events, calculate stability summary data. Stability summary data is a set of indicators that quantify the continuity of sleep and the stability of microstructure. Stability summary data includes micro-arousal index and sleep fragmentation index. The microarousal index refers to the number of microarous events that occur on average per hour of sleep, reflecting the frequency of brief interruptions to sleep. It is calculated by dividing the total number of microarous events identified throughout the entire sleep monitoring cycle by the total sleep time (in hours). For example, if a total sleep time of 6 hours is 30 microarous events detected, the microarousal index is 5 times / hour. The sleep fragmentation index reflects the degree of interruption (including micro-awakening and full awakening) in the sleep process. To calculate it, first, the total number of awakenings during the entire sleep process is counted, including micro-awakening and full awakening; then, the total number of awakenings is divided by the total sleep time (in hours). For example, if a person sleeps for 7 hours and experiences 21 awakenings of various types, the sleep fragmentation index is 3 times / hour. The higher the sleep fragmentation index value, the more discontinuous the sleep and the more difficult it is to maintain deep sleep. Step S300: Integrate the analysis results from the three levels to generate a sleep health assessment report.

[0023] Step S301: Compare and analyze the acquired macroscopic structure summary data, pathological event summary data, and stability summary data with the user's historical baseline data. Based on a preset weighting model, comprehensively score macroscopic sleep structure, pathological event burden, and sleep stability to generate an overall risk assessment level; wherein: Historical baseline data is typical or average data extracted from multiple sleep health assessment reports stored in a user's personal profile. It includes typical macroscopic structure summary data, pathological event summary data, and stability summary data for the user, providing a personalized comparison benchmark for current sleep data and thus identifying abnormalities that deviate from the norm. The pre-defined weighting model is a mathematical framework that defines the importance of different sleep dimensions in the overall score, including three core weights: macro-sleep structure weight. Pathological event burden weight Sleep stability weight For example, 0.2, and The weighting is based on clinical importance; for example, pathological events have the greatest impact on health risks and are therefore given a higher weight. In the comprehensive scoring process, firstly, the macroscopic sleep structure summary data, pathological event summary data, and stability summary data are compared with the user's historical baseline data to calculate a score of 0-100 for each dimension of macroscopic sleep structure, pathological event burden, and sleep stability; then, the scores of the three dimensions of macroscopic sleep structure, pathological event burden, and sleep stability are multiplied by their corresponding weights. , , Finally, the scores are summed to obtain the overall score, with a total score of 100 points; an overall score of ≥85 points is "Excellent", 70-84 points is "Good", and below 70 points is "Needs Attention". The overall risk assessment level is a qualitative classification based on a composite score and key pathological indicators (such as AHI). For example, a low-risk level is a composite score ≥80 and AHI <5; a medium-risk level is a composite score 60-79 or 5≤AHI<15; and a high-risk level is a composite score <60 or AHI≥15. Each level includes corresponding health recommendations and warning levels. Key pathological indicators can be any one or more pathology-related indices from the pathological event summary data. Step S302: Integrate the macroscopic structural summary data, pathological event summary data, stability summary data, comprehensive score, and overall risk assessment level into a unified report to form a structured sleep health assessment report. The sleep health assessment report is a structured data document that integrates macroscopic structural summary data, pathological event summary data, stability summary data, comprehensive score, and overall risk assessment level. It provides users or medical professionals with a comprehensive and quantitative summary of their sleep quality and health status. It not only helps users intuitively understand their own sleep patterns and discover potential problems, such as sleep apnea and sleep fragmentation, but also allows them to evaluate the effectiveness of lifestyle adjustments or treatment interventions by tracking changes in each report. It is a key output for achieving personalized sleep health management. Step S303: Store the sleep health assessment report generated in each monitoring cycle as a user personalized profile. The user personalized profile serves as a benchmark for comparison between the macro state layer calculation, the stability layer calculation, and the sleep health assessment report.

[0024] In this embodiment, the edge computing device and the cloud server are used together to realize this method, and the edge computing device and the cloud server cooperate with each other. Edge computing devices periodically encrypt and upload structured sleep health assessment reports to a cloud server, storing them as personalized user profiles. These personalized user profiles are database records stored in the cloud or locally, containing the user's historical sleep health assessment reports, various computational thresholds optimized through learning (such as macroscopic state judgment thresholds and matrix perturbation thresholds), and feature extraction parameters. This allows the sleep monitoring system to evolve from a general mode to a personalized mode, making state judgment and risk assessment more accurate by using the user's own historical data as a benchmark. The cloud server performs aggregation analysis and model optimization based on user personalized profile data, generating optimized parameters or adjusted thresholds for updating edge computing logic. These optimized parameters or adjusted thresholds include: new thresholds for the macro-state layer (such as updated V1, S1), adaptive thresholds for the pathological event layer (such as dynamically adjusted P1, T3), new thresholds for the stability layer (such as V5), and feature extraction parameters (such as normalization coefficients in standardized calculations). The specific logic and steps for optimization and adjustment are as follows: Anonymized data from a large number of users is aggregated through the cloud server; machine learning models are used to discover better group parameters; simultaneously, for individual users, long-term data trends are analyzed, and personalized thresholds that deviate from the norm are fine-tuned. For example, if it is found that a user's normal respiratory airflow amplitude baseline is generally lower than the group average, the percentage threshold P1 for their individual respiratory events will be lowered accordingly. The cloud server packages the optimized parameters or adjusted thresholds into an update package and sends it to the edge computing device. The edge computing device receives the update package during idle periods and performs integrity and security verification. After successful verification, the new values ​​in the update package are written to a designated configuration area of ​​non-volatile memory. The next time the device starts to perform analysis and calculation, it will automatically call these new and optimized parameters, thereby enabling the continuous evolution of the monitoring algorithm, and performing internal preset judgment rules, weight models, feature extraction parameters, and various thresholds to achieve continuous evolution.

[0025] In this embodiment, a three-layer computing architecture based on a dynamic perception matrix is ​​adopted, which simultaneously performs macroscopic state layer calculation, pathological event layer calculation, and stability layer calculation to generate a structured sleep health report. Through a single sleep monitoring, comprehensive information on the macroscopic structure of sleep, respiratory pathological event burden, and sleep stability can be output simultaneously. The sleep health assessment report clearly shows the total sleep duration and sleep efficiency calculated at the macroscopic state layer, the sleep apnea and its type and severity calculated at the pathological event layer, and the sleep micro-awake state and sleep fragmentation degree calculated at the stability layer. This achieves a multi-dimensional, integrated, and highly medically valuable comprehensive assessment result, enriching the information content of the sleep health profile and providing comprehensive and reliable data support for accurately locating the root cause of sleep problems and formulating personalized intervention or treatment plans. By acquiring raw physiological signals through a multimodal sensor array embedded inside the mattress to construct a dynamic sensing matrix, and performing hierarchical processing calculations based on the dynamic sensing matrix, the dimensionality reduction and trajectory analysis of the macroscopic state layer, the signal reconstruction and logical judgment of the pathological event layer, and the matrix difference calculation of the stability layer all adopt computationally lightweight and interpretable mathematical methods, such as principal component analysis, matrix operations, and threshold logic. Compared with the traditional method of uploading all data to the cloud for processing, the lightweight distributed architecture reduces the computational pressure, resulting in shorter computation time, lower cost, and higher efficiency. It can complete the entire process from data acquisition to event recognition within hundreds of milliseconds, while ensuring the depth and breadth of analysis. It improves the data processing capability under edge resource constraints, ensuring computational efficiency while taking into account the real-time nature and depth of data analysis. Example 2:

[0026] This application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories store computer-readable code that, when executed by the one or more processors, can perform the real-time processing method for mattress-based sleep monitoring data based on edge computing as described above.

[0027] The method or system according to the embodiments of this application can also be implemented using the architecture of the electronic device shown in this application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as a ROM or hard disk, may store the real-time processing method for mattress-type sleep monitoring data based on edge computing provided in this application. Furthermore, the electronic device also includes a user interface. Of course, the architecture shown in this application is merely exemplary; when implementing different devices, one or more components of the electronic device shown in this application may be omitted according to actual needs. Example 3:

[0028] One embodiment of this application discloses a computer-readable storage medium. The computer-readable storage medium stores computer-readable instructions. When the computer-readable instructions are executed by a processor, a real-time processing method for mattress-type sleep monitoring data based on edge computing, according to an embodiment of this application and referring to the above-described figures, can be performed. The storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0029] Furthermore, according to embodiments of this application, the processes described in the above-referenced flowcharts can be implemented as computer software programs. For example, this application provides a non-transitory machine-readable storage medium storing machine-readable instructions that can be executed by a processor to perform instructions corresponding to the method steps provided in this application, such as a real-time processing method for mattress-based sleep monitoring data based on edge computing. When this computer program is executed by a central processing unit (CPU), it performs the functions defined in the method of this application.

[0030] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0031] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for real-time processing of mattress sleep monitoring data based on edge computing, characterized in that, The method includes: The system acquires raw physiological signals from a multimodal sensor array embedded in the mattress in real time, performs validity screening on the raw physiological signals to generate valid signal sources, and constructs a dynamic sensing matrix based on the valid signal sources. Based on the dynamic perception matrix, three levels of computational analysis are performed in parallel, including macroscopic state layer calculation, pathological event layer calculation, and stability layer calculation, generating three levels of analysis results. The analysis results from the three levels are integrated to generate a sleep health assessment report.

2. The edge computing based mattress type sleep monitoring data real-time processing method according to claim 1, characterized in that, The construction of the dynamic sensing matrix based on effective signal sources includes: The multimodal sensor array includes a pressure sensor array, a bioradar sensor, and a temperature sensor. It performs validity screening on each signal source of the acquired raw physiological signals to generate valid signal sources. For each valid signal source, a standardized physiological feature vector is calculated, which includes time-domain features and frequency-domain features; The standardized physiological feature vectors corresponding to all valid signal sources are used as column vectors, and the valid signal sources are used as row vectors to construct a dynamic sensing matrix. The number of rows in the dynamic sensing matrix is ​​dynamically adjusted according to the number of valid signal sources within a preset time window.

3. The edge computing based mattress type sleep monitoring data real-time processing method according to claim 1, characterized in that, The macroscopic state layer calculation is used to assess the macroscopic structure of sleep, and the macroscopic state layer calculation includes: The dynamic perception matrix is ​​dimensionality reduced to extract the main feature vector representing the overall physiological pattern of the current time window, which is then used as the physiological state vector. Over Q consecutive time windows, the changes in the physiological state vector are tracked to form a physiological state trajectory; The dynamic characteristics of physiological state trajectories are analyzed, including moving variance and changing trend. Based on these dynamic characteristics, sleep is divided into different macroscopic states. Based on the division of macro-states, macro-structure summary data is generated.

4. The edge computing based mattress type sleep monitoring data real-time processing method according to claim 3, characterized in that, The macroscopic states include stable core sleep, active brain activity, wakefulness, and unstable sleep. The macroscopic structure summary data includes sleep efficiency, the percentage of each macroscopic state in the total sleep time, and the number of awakenings.

5. The edge computing based mattress type sleep monitoring data real-time processing method according to claim 1, characterized in that, The pathological event layer calculation is used to identify and classify specific sleep breathing events, and the pathological event layer calculation includes: Row vectors related to respiratory monitoring are dynamically selected from the dynamic sensing matrix to form a respiratory monitoring submatrix; Based on the respiratory monitoring sub-matrix, the respiratory effort signal is separated from the pressure sensor array, and the respiratory airflow proxy signal is extracted from the bioradar sensor; Real-time analysis of the synergistic relationship between respiratory effort signals and respiratory airflow proxy signals, and identification and classification of respiratory events based on the synergistic relationship.

6. The real-time processing method for mattress-type sleep monitoring data based on edge computing according to claim 5, characterized in that, The respiratory events include obstructive sleep apnea events, central sleep apnea events, and airflow limitation events, and the synergistic relationships include the amplitude relationship, morphological relationship, and cross-correlation coefficient of the signals.

7. The real-time processing method for mattress-type sleep monitoring data based on edge computing according to claim 6, characterized in that, The pathological event layer calculation also includes: All identified respiratory events are statistically analyzed to generate pathological event summary data, which includes the apnea-hypopnea index, obstructive apnea index, central apnea index, percentage of airflow-limited events in total sleep time, and a detailed list of various respiratory events.

8. The real-time processing method for mattress-type sleep monitoring data based on edge computing according to claim 1, characterized in that, The stability layer calculation is used to assess the microstructure and stability of sleep, and the stability layer calculation includes: Over Q consecutive time windows, the difference between the dynamic perception matrices of adjacent time windows is calculated to generate a matrix perturbation index. Set the matrix perturbation threshold V5. When the peak value of the matrix perturbation index is within the fourth preset event threshold T6, and continues to exceed V5, and the duration of the peak value exceeding V5 is less than the fifth preset time threshold T7, then the event is marked as a micro-awakening event. When the duration of the matrix perturbation index exceeding the V5 peak exceeds T7, it is determined to be a full awakening event and output to the macrostructure summary data; Based on the monitored micro-awakening events, stability summary data is calculated, which includes the micro-awakening index and the sleep fragmentation index.

9. The real-time processing method for mattress-type sleep monitoring data based on edge computing according to claim 1, characterized in that, The analysis results from the three levels are integrated to generate a sleep health assessment report, including: The acquired macroscopic structural summary data, pathological event summary data, and stability summary data are comprehensively scored to generate an overall risk assessment level; The report integrates macroscopic structural summary data, pathological event summary data, stability summary data, comprehensive score, and overall risk assessment level into a unified report to form a sleep health assessment report.

10. The real-time processing method for mattress-type sleep monitoring data based on edge computing according to claim 9, characterized in that, The sleep health assessment report generated in each monitoring cycle is stored as a user's personalized profile, which serves as a benchmark for comparison between the macro-state layer calculation, the stability layer calculation, and the sleep health assessment report.

Images