Home health risk assessment method and system based on night sleep activity

Abstract

The present application relates to the technical field of health monitoring, and discloses a home health risk assessment method and system based on night sleep activity. The present application collects night sleep physiological data and night activity trajectory data of a user through a sign monitoring module under a mattress and a night behavior trajectory module in a residence, and uploads the data after time stamp alignment by a gateway. Abnormal physiological values are cleaned out, sleep-behavior correlation features are constructed, and composite features such as difficulty falling asleep, heart rate sudden change and apnea are extracted. The risk of heart, respiration and cognitive decline is assessed in different dimensions, and long-term risk aggravation warning can be generated through 30-day personal dynamic baseline and linear regression significance analysis. The present application solves the problems of data isolation, rough index and cognitive assessment blind area in the prior art, realizes non-invasive and accurate early warning of home health, and is suitable for home care monitoring of the elderly.

Description

Technical Field

[0001] This invention belongs to the field of data processing and medical health monitoring technology, specifically relating to a home health risk assessment method and system based on nighttime sleep activity, which can be implemented through multimodal Internet of Things sensing methods. Background Technology

[0002] With the accelerating aging of the global population, home-based elder care has become the mainstream model of my country's elder care service system. While various monitoring solutions have been developed to address the health monitoring needs of the elderly living at home, significant technological limitations still exist. First, existing monitoring schemes generally suffer from data silos. Most systems only analyze sleep physiological data or collect home behavior data in isolation, failing to achieve deep correlation and integration between nighttime sleep physiological data and concurrent activity trajectory data. This makes it difficult to comprehensively reflect the overall health status of the elderly during their nighttime sleep cycle, resulting in a high false alarm rate. Second, current technologies utilize monitoring data in a rather coarse manner. Traditional schemes only focus on macro-level indicators such as total sleep duration and the percentage of baseline stages, lacking the ability to mine fine-grained indicators such as duration of difficulty falling asleep, sudden changes in heart rate, and sleep apnea, leading to a significant lag in health risk warnings. Furthermore, existing technologies have blind spots in cognitive function assessment. Current systems only focus on immediate physiological abnormalities, lacking long-term trend modeling, and cannot achieve early quantitative screening for neurodegenerative diseases such as Alzheimer's disease, causing early-stage patients to miss the opportunity for intervention. In summary, existing technologies cannot meet the needs of elderly people living at home for seamless, continuous, and accurate nighttime health monitoring. There is an urgent need for a new home-based health risk assessment solution that can integrate nighttime sleep and activity data, mine fine-grained health indicators, and achieve real-time risk assessment and long-term trend early warning. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies by providing a home-based health risk assessment method and system based on nighttime sleep activity. By expanding the fine-grained indicators of raw sleep data and performing spatiotemporal correlation and deep fusion with concurrent activity trajectories, this invention achieves a leap from "single data monitoring" to "sleep-behavioral multidimensional health assessment," accurately identifying and assessing users' sleep disorders, cardiac and respiratory risks, and cognitive decline risks.

[0004] The objective of this invention is achieved through the following technical solution: This invention provides a home-based health risk assessment method based on nighttime sleep activity, comprising the following steps: S1 collects users' sleep physiological data through a vital sign monitoring module deployed under the mattress, and collects users' nighttime behavioral trajectory data through behavioral trajectory modules deployed in multiple locations within the residence; S2 performs data cleaning and spatiotemporal alignment on the collected data, aligning the time series of nighttime behavioral trajectory data with the time axis of sleep physiological data; S3 constructs a "sleep-behavior" correlation feature vector, which associates sleep physiological data with nighttime behavioral trajectory data at corresponding time points to generate composite features including sleep difficulty features, nighttime toilet behavior features, heart rate mutation events, sleep apnea events, and circadian rhythm features; S4 performs a risk assessment based on the aforementioned composite characteristics, generating assessment results including cardiac risk, respiratory risk, and cognitive decline risk. S5 outputs the assessment results and risk warnings.

[0005] Furthermore, this method also includes long-term trend modeling and risk aggravation warning steps: based on the sleep physiological data, nighttime behavioral trajectory data and composite feature data of the past 30 days, an individual dynamic health baseline is established, and the moving average and standard deviation of core health indicators are calculated; the linear regression slope of multiple key health indicators is calculated, and when the slope changes significantly and the significance p<0.05, a corresponding risk aggravation warning is generated.

[0006] Furthermore, for multi-person home scenarios, this method also includes user differentiation and data matching steps: through a vital sign monitoring module that is bound one-to-one with each user, the nighttime sleep physiological data of different users is differentiated to obtain each user's in-bed / out-of-bed status; based on a unified timestamp, the nighttime activity trajectory data is correlated and matched with the nighttime sleep physiological data of users who are out of bed at the corresponding time period, and the behavior trajectory module does not need to perform individual identification; there is no need to perform precise individual matching on a single nighttime activity trajectory data, but the health status of the user is determined by the long-term trend changes of the core health indicators of the corresponding bound user, and a personalized data report and risk warning are generated for the corresponding individual.

[0007] Correspondingly, the present invention also provides a home health risk assessment system based on nighttime sleep activity for performing the above assessment method. The system includes a data acquisition layer, a data processing layer, and an application layer. The data acquisition layer includes: The vital signs monitoring module is a piezoelectric or fiber optic sleep monitoring belt deployed under the mattress and bound one-to-one with each user. One sleep monitoring belt corresponds to one user and is used to collect the corresponding user's nighttime sleep physiological data, naturally realizing the differentiation of individual users in a multi-person home setting. The behavior trajectory module consists of infrared sensors or millimeter-wave radar sensors deployed at multiple key movement locations within the residence to collect users' nighttime behavior trajectory data without requiring individual identification. The gateway module is used to receive the data collected by the vital signs monitoring module and the behavior trajectory module, and upload the data to the data processing layer after adding a precise timestamp. The data processing layer is used to clean and spatiotemporally align the sleep physiological data and nighttime behavioral trajectory data, construct a sleep-behavior correlation feature vector, assess the user's cardiac risk, respiratory risk and cognitive function risk in different dimensions based on the correlation feature vector, and establish a personal dynamic baseline for long-term trend modeling and risk aggravation warning. The application layer is used to output the assessment results, risk warnings, individual-specific data reports, and family multi-user health summary reports.

[0008] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects: (1) Improve the sensitivity and accuracy of early warning: This invention constructs a “sleep-behavior” correlation feature vector and cross-validates nighttime sleep physiological data with concurrent activity trajectory data. For example, it eliminates the interference of body movement on heart rate mutation events through nighttime activity data and accurately defines the physiological impact of nighttime toileting behavior through sleep staging, effectively solving the problems of isolated data and high false alarm rate in existing technologies.

[0009] (2) Achieve early screening of cognitive function: By analyzing sleep regularity (work-rest standard deviation) and nighttime behavior patterns (wandering path entropy), the quantitative assessment of cognitive function decline is incorporated into the home health monitoring system for the first time, providing a non-intrusive and quantitative screening tool for early intervention of neurodegenerative diseases such as Alzheimer's disease.

[0010] (3) More comprehensive assessment dimensions: This invention deeply integrates multimodal data of "physiology-behavior-environment", realizing the leap from traditional single "sleep monitoring" to "multidimensional health risk assessment", which can simultaneously assess the user's sleep health, cardiac and respiratory risks and cognitive function status, and is more in line with the comprehensive care needs of home-based elderly care.

[0011] (4) Non-contact monitoring improves user compliance: The present invention uses a non-contact sleep monitoring belt and infrared sensor, which eliminates the need for users to wear wearable devices and achieves non-contact and continuous health monitoring, which is especially suitable for elderly people living alone. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the overall architecture of the system of the present invention; Figure 2 This is a schematic diagram of the network layer-platform layer architecture of the system architecture of this invention. Figure 3 This is a schematic diagram of the data processing flow of the present invention; Figure 4 This is a schematic diagram of the "sleep-behavior" association feature extraction in this invention. Detailed Implementation

[0013] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Example 1

[0014] This embodiment uses the home health monitoring of Mr. Zhang, an elderly person living alone, as an example to illustrate the implementation process of the present invention: I. System Deployment and Overall Architecture The overall system architecture of this invention is as follows: Figure 1 and Figure 2 As shown, the system is divided into a perception layer, a network layer, and an application layer from bottom to top. Based on this architecture, it can be adapted to two core scenarios: single-person living and multi-person living, and complete the system deployment.

[0015] Perception Layer: For each monitored user in the home environment, a vital sign monitoring module is deployed under their corresponding mattress. Specifically, this is a piezoelectric sleep monitoring belt or a fiber optic sleep monitoring belt. One sleep belt corresponds to one user, with a one-to-one binding, which naturally achieves the differentiation of individual data for multiple users and is used to collect the corresponding user's nighttime sleep physiological data. Behavioral trajectory modules are deployed along key movement lines in the bedroom, bathroom, living room, and kitchen. Specifically, these are infrared human body sensors or millimeter-wave radar sensors, used to collect nighttime activity trajectory data for the entire residence without the need for individual identification by sensors.

[0016] Sleep physiological data includes sleep initiation and wake-up times, sleep stages, duration of the first awake period in bed, heart rate data, respiratory data, blood oxygen data, and snoring data; nighttime behavioral trajectory data includes trigger time and spatial location encoding, and sensor-collected location time-series data, used to generate personnel movement trajectories. The heart rate data collected by the vital signs monitoring module includes real-time heart rate, nighttime average heart rate, and heart rate variability, used to identify heart rate instability and sudden heart rate events; the respiratory data includes real-time respiratory rate, respiratory waveform, and combined with blood oxygen data, used to detect sleep apnea and hypoventilation events.

[0017] Network Layer: Deploys IoT gateways to receive data from two types of sensors, assigns a unified millisecond-level timestamp to all data, and synchronously uploads it to the cloud platform layer to ensure the temporal consistency of multi-source data. The platform layer performs data preprocessing, sleep-behavioral data association matching, feature extraction, risk assessment, and long-term trend modeling. In multi-person home scenarios, by using the sleep tracker's in-bed / out-of-bed status and a unified timestamp, it associates and matches infrared-collected nighttime activity trajectory data with the corresponding user's sleep physiological data, achieving the correspondence between different users' behavioral and physiological data without requiring individual identification at the sensor end.

[0018] Application layer: This is the client-side app, used for subsequent result display. The client-side app receives and displays: Sleep Daily Report: Displays sleep onset time, wake-up time, total sleep duration, percentage of each sleep stage, with special note on the first stage of awake time in bed, and provides a "difficulty falling asleep" alert.

[0019] Cardiac and respiratory risk dashboard: Displays nighttime heart rate curves, heart rate mutation event markers, apnea event distribution, apnea-hypopnea index (AHI) and trend chart.

[0020] Trend warnings are divided into immediate risks (such as "a sudden change in heart rate was detected, with an amplitude of 30 beats / minute, please pay attention") and trend risks (such as "difficulty falling asleep has worsened in the past week, it is recommended to improve bedtime habits").

[0021] This embodiment is adapted to the scenario of living alone, and relevant sensing devices were deployed in the residence of Zhang, an elderly person living alone.

[0022] II. Raw Data Collection The system ran for one night and collected the following raw data: At 23:10, Mr. Zhang lay down in bed, and the sleep monitoring tape began recording data. During the periods of 23:10-23:45 and 02:45-03:25, the user was awake in bed. The first period of awake time was 35 minutes, and the second period of awake time was 40 minutes. The bedroom infrared sensor was not triggered during both periods, indicating that the user did not leave the bed.

[0023] Sleep stages: Light sleep: 23:45-00:40, 02:05-02:30, 03:25-04:05, 04:55-05:35, 06:15-06:33; Deep sleep: 00:40-02:05, 05:35-06:15; REM sleep (04:05-04:55); Time out of bed: 02:30-02:45; Actual effective sleep duration that night was 6 hours.

[0024] At 02:32, the bedroom infrared sensor is triggered, marking the user as leaving the bed; at 02:34, the bathroom infrared sensor is triggered, marking the user as entering the bathroom; at 02:42, the bathroom infrared triggering ends; at 02:43, the user returns to bed.

[0025] Heart rate monitoring during sleep showed that between 02:35 and 02:38, the user's heart rate suddenly dropped from 62 beats / min to 48 beats / min, and recovered after 3 minutes. During this period, the sleep monitoring belt did not detect any body movement, and the user was in the bathroom.

[0026] Physiological data such as heart rate, respiration, and snoring are collected synchronously during sleep.

[0027] Multiple apnea and hypopnea events occurred during sleep, with a total of 108 apnea and hypopnea events occurring at night.

[0028] Snoring monitoring showed that between 03:10 and 03:15, there were continuous interruptions in snoring, accompanied by apnea events, with an apnea duration of 25 seconds, and a 4% drop in blood oxygen was detected simultaneously.

[0029] There is a momentary interference noise of 250 times / minute in the raw data, which is an outlier value collected by the device.

[0030] III. Cloud Data Processing The data processing flow is as follows: Figure 3 As shown, the cleaning and alignment, feature extraction, risk assessment, and trend modeling are completed sequentially: 1. Data cleaning and spatiotemporal alignment First, the raw data was cleaned to remove an abnormal heart rate of 250 beats per minute at a moment. Then, the nighttime behavioral trajectory data from the infrared sensor was aligned with the physiological data from the sleep monitoring belt to ensure that the time sequence of all events was completely matched, laying the foundation for subsequent correlation analysis.

[0031] The abnormal heart rate of 250 beats / min exceeded the normal physiological range of 30-200 beats / min. During this cleaning process, abnormal physiological values ​​with respiratory rates of <4 breaths / min or >40 breaths / min, as well as environmental noise data, were removed.

[0032] 2. Sleep-behavior association feature extraction The correlation feature extraction process described here can be referenced. Figure 4 This figure synchronously displays the correspondence between sleep stages, infrared triggering events, and physiological abnormalities on a unified timeline. Based on this synchronous temporal sequence, we extracted five core composite features: Characteristics of difficulty falling asleep: The first period of wakefulness in bed was 35 minutes, which has exceeded the threshold of 30 minutes for 5 consecutive days. Combined with the data of not getting out of bed, it was determined to be a characteristic of psychogenic insomnia. Nighttime toileting characteristics: The delay from the end of light sleep to getting out of bed is 2 minutes, and the duration of stay in the toilet is 11 minutes. Combined with synchronized heart rate data, the physiological burden characteristics of toileting are generated. Heart rate mutation characteristics: The heart rate variability rate is calculated using a sliding window with a window length of 30 seconds and a step length of 5 seconds. If the heart rate variability rate reaches 14 beats / min / second during toilet use, exceeding the threshold of 6 beats / min / second, and there is no body movement interference, it is marked as a heart rate mutation event. The heart rate variability rate threshold can be adjusted according to the actual situation and individual physical condition.

[0033] Sleep apnea characteristics: An event with a 92% decrease in respiratory amplitude lasting for 25 seconds was detected. Combined with the interruption of snoring, it was determined to be obstructive sleep apnea, and the AHI was calculated to be 18. Apnea events and AHI calculation: AHI is the apnea-hypopnea index, which refers to the total number of apneas and hypopneas per hour of sleep. This calculation method is a clinically accepted standard in this field. In this embodiment, the total number of events was 108, and the actual sleep time was 6 hours, resulting in an AHI of 18.

[0034] Circadian rhythm characteristics: Based on sleep initiation time and wake-up time over multiple consecutive days, the time standard deviation is calculated. The dispersion of the statistical time series characterizes the regularity of the user's sleep patterns. If the standard deviation continues to increase (e.g., exceeding 1.5 hours), the infrared trigger location sequence from 2:00 AM to 4:00 AM is extracted. The path entropy value is obtained by calculating the path disorder using the information entropy formula. This entropy value is used to quantify the orderliness of nighttime behavior. A lower entropy value indicates that the behavior is purposeful and there is no nighttime wandering. Based on this, the circadian rhythm characteristics are extracted. The standard deviation calculation and information entropy formula are existing technologies known in the field and will not be described in detail here.

[0035] 3. Multi-dimensional risk assessment Based on the extracted composite features, risk assessment is completed across multiple dimensions: Cardiac Risk Assessment: Five consecutive minutes of heart rate interval data during deep sleep (excluding body movement interference) are extracted, and the LF / HF ratio (a frequency domain index) is calculated. This ratio is calculated using internationally accepted techniques in heart rate variability analysis to reflect the autonomic nervous system's balance. An elevated ratio indicates increased sympathetic nerve tone and increased cardiac load. If the LF / HF ratio during deep sleep has been consistently elevated over the past 30 days, it suggests autonomic nervous system dysfunction. The frequency of nocturnal heart rate abrupt changes is statistically analyzed, combined with the magnitude of the abrupt changes (e.g., a sudden drop from 60 bpm to 40 bpm) and their duration. If the abrupt change is accompanied by apnea or body movement, it is considered a cardiac compensatory response. If an isolated abrupt change occurs repeatedly, a "cardiac arrhythmia risk" warning is generated. When the algorithm detects a prolonged interval of more than 2 minutes with a nighttime heart rate <40 bpm, and synchronous infrared data shows the user is in a resting state (without body movement), it is considered a high-risk cardiac event, and an emergency warning is immediately sent. All preset detection thresholds can be adjusted as needed.

[0036] This incident did not trigger any warnings for autonomic nervous system dysfunction or high-risk cardiac events. However, the sudden change in heart rate during nighttime urination, which occurred at rest, suggests possible vagal nerve over-excitation and requires monitoring of cardiac conduction function. Respiratory risk assessment: Sleep apnea risk levels are generated based on the hourly apnea-hypopnea index (AHI) classification, combined with the nocturnal decrease in blood oxygen saturation in respiratory data. Clinically, the AHI classification standard is: AHI < 5 is normal, 5 ≤ AHI < 15 is mild sleep apnea, 15 ≤ AHI < 30 is moderate sleep apnea, and AHI ≥ 30 is severe sleep apnea. This invention uses this standard to determine the respiratory risk level. In this case, AHI = 18 combined with a 4% decrease in blood oxygen saturation indicates a moderate risk of sleep apnea. Cognitive risk assessment: Based on the changing trend of the standard deviation between the sleep initiation time and the wake-up time, and the entropy value of the infrared trigger path, early signals of cognitive dysfunction are generated.

[0037] No significant abnormalities were observed in the rhythm and entropy values, and no cognitive warning was generated.

[0038] 4. Long-term trend modeling A personal dynamic baseline is established using data from the past 30 days. The moving average and standard deviation of indicators such as daily sleep patterns (standard deviation of sleep onset / wake-up time), mean duration of the first awake period in bed, percentage of deep sleep, AHI index, and frequency of heart rate variability are calculated. The linear regression slopes of several key indicators are also calculated. When the slope increases significantly and the significance p < 0.05, a corresponding risk aggravation warning is generated.

[0039] For difficulty falling asleep: Calculate the linear regression slope of the first period of awake time in bed. If it continues to rise and the significance p<0.05, generate an "increased risk of insomnia" warning.

[0040] Regarding cardiac risk: Calculate the regression slope of the frequency of heart rate mutations. If it continues to rise and is statistically significant (p<0.05), it indicates an increased risk of cardiovascular events.

[0041] Regarding cognitive function: Based on existing technology, the linear regression slope of the sleep initiation time phase shift and path entropy value is calculated. If the slope continues to increase and the significance p<0.05, it indicates that the monitor's circadian rhythm disorder is aggravated and the orderliness of nighttime behavior is continuously reduced, which increases the risk of developing neurodegenerative diseases. This can be used for early screening of neurodegenerative diseases.

[0042] The term "continuous rise" refers to a statistical period of weeks or months, in which the statistical value of the corresponding indicator shows a monotonically increasing trend over multiple consecutive statistical periods, rather than random fluctuations.

[0043] IV. Results Output Finally, the system pushed the assessment results and early warning information to Zhang's caregiver's mobile phone. The push message was as follows: Immediate warning: A nighttime breathing apnea event was detected, lasting 25 seconds, with a sudden change in heart rate during nighttime urination, and the event occurred at rest. This may indicate excessive vagal nerve excitation, requiring attention to cardiac conduction function. Medical examination is recommended.

[0044] Trend report: Your difficulty falling asleep has worsened in the past week, and the number of sudden changes in your heart rate at night has increased. It is recommended that you adjust your bedtime habits and consult your doctor to adjust your medication plan. Example 2

[0045] This embodiment uses a family scenario where two elderly people live together as an example to illustrate the multi-user adaptation solution of the present invention: A sleep monitoring belt is installed in the master bedroom and the second bedroom of the residence. The sleep monitoring belt in the master bedroom is bound to user Zhang (male, 72 years old), and the sleep monitoring belt in the second bedroom is bound to user Li (female, 70 years old). Infrared sensors are installed at the bedroom door, bathroom and living room, without the need to configure individual identification functions.

[0046] The core data collected and the processing logic during the system's operation cycle are as follows: Data Matching and Differentiation: The gateway assigns a unified timestamp to all data. The cloud uses the in-bed / out-of-bed status of the two sleep tracking zones to match and correlate nighttime behavioral data. For example, if the master bedroom sleep tracking zone shows that Mr. Zhang got out of bed at 02:12, and the infrared sensors at the master bedroom door and bathroom are triggered sequentially, this behavioral data is associated with Mr. Zhang. However, if the second bedroom sleep tracking zone shows that Mr. Li was in bed and asleep throughout the entire time, this behavioral data does not match Mr. Li.

[0047] Individual feature extraction and risk assessment: Based on physiological data from two sleep cycles, exclusive "sleep-behavior" correlation feature vectors were constructed for the two users, and risk assessments were completed in different dimensions to generate their respective daily sleep reports and risk assessment results; among them, Zhang's AHI index was consistently in the 15-20 range, which was judged as a moderate risk of sleep apnea; Li's first awake time in bed exceeded 40 minutes for several consecutive days, which was judged as a risk of difficulty falling asleep.

[0048] Long-term trends and family health management: Based on data from the past 30 days, personal dynamic health baselines are established for two users, and the linear regression slope and significance p-value of core indicators are calculated. When the slope of Zhang's heart rate mutation frequency increases significantly (p<0.05), a personalized warning of increased cardiovascular event risk is generated for him. At the same time, the client can generate a family health summary report, displaying the core health indicators and risk warnings of the two users, realizing health monitoring and management of all family members.

[0049] This embodiment achieves individual differentiation through a "sleep belt one-to-one binding user" method, eliminating the need for complex individual identification using infrared sensors. It weakens the need for precise individual matching of single behavioral data. The core is to determine the health status of corresponding family members through long-term trend changes. The solution is simple and easy to implement, and perfectly suited for home-based elderly care scenarios such as multi-child families and elderly couples living together.

Claims

1. A home-based health risk assessment method based on nighttime sleep activity, characterized in that, Includes the following steps: S1 collects users' sleep physiological data through a vital sign monitoring module deployed under the mattress, and collects users' nighttime behavioral trajectory data through behavioral trajectory modules deployed in multiple locations within the residence; S2 performs data cleaning and spatiotemporal alignment on the collected data, aligning the time series of nighttime behavioral trajectory data with the time axis of sleep physiological data; S3 constructs a "sleep-behavior" related feature vector, which associates sleep physiological data with nighttime behavioral trajectory data at corresponding time points to generate composite features including sleep difficulty features, nighttime toilet behavior features, heart rate mutation events, sleep apnea events, and circadian rhythm features; S4 performs a risk assessment based on the aforementioned composite characteristics, generating assessment results including cardiac risk, respiratory risk, and cognitive decline risk; S5 outputs the assessment results and risk warnings.

2. The home health risk assessment method based on nighttime sleep activity according to claim 1, characterized in that, The sleep physiological data mentioned in S1 includes sleep initiation and wake-up times, sleep stages, duration of the first awake period in bed, heart rate data, respiratory data, blood oxygen data, and snoring data; the vital signs monitoring module is a sleep monitoring strip deployed one-to-one with a user, with one sleep monitoring strip corresponding to one user, thereby enabling the differentiation of sleep physiological data of different users in a multi-person home setting; the nighttime activity trajectory data includes trigger time and spatial location encoding. In a multi-person home setting, the nighttime sleep physiological data also includes the user target ID bound to the vital signs monitoring module, used to distinguish individual data of different users.

3. The home health risk assessment method based on nighttime sleep activity according to claim 1, characterized in that, The data cleaning described in S2 includes removing abnormal physiological values ​​such as heart rate <30 beats / min or >200 beats / min, respiratory rate <4 breaths / min or >40 breaths / min, and environmental noise.

4. The home health risk assessment method based on nighttime sleep activity according to claim 2, characterized in that, The construction of the "sleep-behavior" correlation feature vector described in S3 specifically includes: Extract the first segment of awake time in bed. When the awake time exceeds the preset threshold of 30 minutes for several consecutive days, combine it with nighttime behavior trajectory data to determine whether there is repeated behavior of getting out of bed before sleep, so as to generate characteristics of difficulty falling asleep. Identify the delay time from the end of deep sleep or light sleep to the triggering of the infrared sensor in the bedroom, as well as the duration of continuous infrared triggering in the bathroom, and combine this with synchronized heart rate data to generate nighttime toilet behavior characteristics. Heart rate data sequences during sleep are extracted, and the heart rate variability is calculated using a sliding window with a window length of 30 seconds and a step size of 5 seconds. When the variability exceeds the preset heart rate variability threshold and lasts for a certain period of time, it is marked as a heart rate mutation event. Combined with the sleep stage and nighttime behavior trajectory data at the same time to exclude body movement interference and generate heart rate mutation event features. By combining snoring data, breathing data, and heart rate data, apnea and hypopnea events are detected, obstructive apnea and central apnea are distinguished, the apnea-hypopnea index (AHI) is calculated every hour, and apnea characteristics are generated. The standard deviation between sleep initiation time and wake-up time is calculated, and the infrared trigger path entropy value within the preset time period of 2:00-4:00 am is analyzed to generate circadian rhythm characteristics.

5. The home health risk assessment method based on nighttime sleep activity according to claim 4, characterized in that, The risk assessment based on the composite characteristics described in S4 specifically includes: Cardiac risk assessment: Extract heart rate interval data during deep sleep excluding body movement interference, calculate the LF / HF ratio of heart rate variability. If this ratio continues to increase in the long-term trend, an autonomic nervous system dysfunction risk is generated. Statistical analysis of the frequency of nocturnal heart rate mutation events, combined with the mutation amplitude and duration, generates arrhythmia risk. When a preset low-value, long-interval state of nighttime heart rate is detected, and synchronous nighttime behavioral trajectory data shows that the user is in a resting state, a high-risk cardiac event warning is generated. Respiratory risk assessment: Based on the characteristics of apnea, the apnea-hypopnea index (AHI) is graded by hour, and combined with the nighttime blood oxygen saturation level, a sleep apnea risk level is generated. Cognitive decline risk assessment: Based on circadian rhythm characteristics, the standard deviation of sleep initiation time and wake-up time is calculated to determine the trend, as well as the entropy value of the infrared trigger path, to generate early signals of cognitive impairment.

6. The home health risk assessment method based on nighttime sleep activity according to claim 1, characterized in that, It also includes long-term trend modeling and risk aggravation early warning steps, specifically: Based on the sleep physiological data, nighttime behavioral trajectory data and composite characteristic data of the past 30 days, an individual dynamic health baseline was established, and the standard deviation of sleep onset / wake time, mean first awake time in bed, proportion of deep sleep, AHI index, and moving average and standard deviation of heart rate mutation frequency were calculated. Calculate the linear regression slope of multiple key health indicators. When the slope changes significantly and the significance p < 0.05, generate a corresponding risk aggravation warning. Specifically, for difficulty falling asleep: the linear regression slope of the first period of awake time in bed is calculated, and if the slope continues to rise, an "increased risk of insomnia" warning is generated; for cardiac risk: the regression slope of the frequency of heart rate mutations is calculated, and if the slope continues to rise, it indicates an increased risk of cardiovascular events; for cognitive function: the slope of the phase shift of sleep initiation time and the path entropy value are calculated, and if the slope continues to increase, it indicates an increased risk of developing neurodegenerative diseases.

7. The home health risk assessment method based on nighttime sleep activity according to claim 2, characterized in that, For multi-person home scenarios, the process also includes user differentiation and data matching steps, specifically: By using a vital sign monitoring module that is linked to each user on a one-to-one basis, the system can distinguish the nighttime sleep physiological data of different users and obtain each user's in-bed / out-of-bed status. Based on a unified timestamp, the nighttime activity trajectory data is associated and matched with the nighttime sleep physiological data of users who are out of bed during the corresponding time period. The behavior trajectory module does not require individual identity recognition. There is no need for precise individual matching of single nighttime activity trajectory data. By analyzing the long-term trend changes of the core health indicators of the corresponding bound user, the health status of the user can be determined, and a personalized data report and risk warning can be generated for the individual.

8. A home-based health risk assessment system based on nighttime sleep activity, characterized in that, Used to perform the method according to any one of claims 1 to 7; It includes a data acquisition layer, a data processing layer, and an application layer: The data acquisition layer includes: The vital signs monitoring module is a piezoelectric or fiber optic sleep monitoring belt deployed under the mattress and bound one-to-one with each user. One sleep monitoring belt corresponds to one user and is used to collect the corresponding user's nighttime sleep physiological data, naturally realizing the differentiation of individual users in a multi-person home setting. The behavior trajectory module consists of infrared sensors or millimeter-wave radar sensors deployed at multiple key movement locations within the residence to collect users' nighttime behavior trajectory data without requiring individual identification. The gateway module is used to receive the data collected by the vital signs monitoring module and the behavior trajectory module, and upload the data to the data processing layer after adding a precise timestamp. The data processing layer is used to clean and spatiotemporally align the sleep physiological data and nighttime behavioral trajectory data, construct a sleep-behavior correlation feature vector, assess the user's cardiac risk, respiratory risk and cognitive function risk in different dimensions based on the correlation feature vector, and establish a personal dynamic baseline for long-term trend modeling and risk aggravation warning. The application layer is used to output the assessment results, risk warnings, individual-specific data reports, and family multi-user health summary reports.

Images