System for supervising sleep health of old people, identifying nursing workers and recording nursing information

Abstract

The invention provides a system for supervising sleep health of old people, identifying nursing workers and recording nursing information, and relates to the field of nursing monitoring, comprising a user management module comprising a plurality of intelligent beds and a data management unit, the intelligent bed comprises a bed body and a basic information recording assembly, a sleep monitoring assembly, an off-bed judgment assembly and an alarm assembly which are arranged on the bed body, the sleep monitoring assembly is used for obtaining sleep monitoring data of a user and predicting the falling risk of the user, and the off-bed judgment assembly is used for determining the off-bed duration of the user and sending the off-bed duration to the alarm assembly. The alarm component is used for giving an alarm based on the falling risk of the user and / or the off-bed duration of the user; the personnel attendance module comprises a plurality of personnel punch-in units and a plurality of service identification units, each bed body is provided with one personnel punch-in unit and one service identification unit, the personnel punch-in units are used for identifying service personnel, the service identification units are used for recording service contents of the service personnel, and the system has the advantage of improving the old-age care quality.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of old-age care monitoring, in particular to a system for monitoring the sleep health of the elderly and identifying caregivers and recording nursing information. BACKGROUND

[0002] With the trend of population aging, the problem of old-age care is becoming more and more concerned, and with the rapid development of Internet technology and intelligent hardware technology, the use of intelligent Internet of Things devices combined with portable human data monitoring terminals can realize remote monitoring of user status.

[0003] The existing old-age care monitoring system generally sets up a monitoring camera device at the user's residence, obtains image information of the user's residence and analyzes it, and then judges the user's state. However, this monitoring method has high reliability, but has many problems in actual operation. First, the monitoring device can easily leak the user's privacy, and there is a risk of information leakage due to device attacks. Second, the analysis of image information will generate a large amount of data transmission and data processing, which will obviously increase the cost of the monitoring process.

[0004] Therefore, it is necessary to provide a system for monitoring the sleep health of the elderly and identifying caregivers and recording nursing information to improve the quality of old-age care monitoring. SUMMARY

[0005] The present application provides a system for monitoring the sleep health of the elderly and identifying caregivers and recording nursing information, comprising: a user management module comprising a plurality of intelligent beds and a data management unit, wherein the intelligent bed comprises a bed body and a basic information recording assembly, a sleep monitoring assembly, a bed leaving determination assembly and an alarm assembly arranged on the bed body, the sleep monitoring assembly is used to obtain sleep monitoring data of the user and predict the fall risk of the user, the bed leaving determination assembly is used to determine the bed leaving duration of the user, and the alarm assembly is used to alarm based on the fall risk of the user and / or the bed leaving duration of the user; a personnel attendance module comprising a plurality of personnel clock-in units and a plurality of service identification units, wherein each bed body is provided with a personnel clock-in unit and a service identification unit, the personnel clock-in unit is used to identify service personnel, and the service identification unit is used to record the service content of the service personnel.

[0006] Further, the personnel clocking unit comprises a display screen, a face recognition component and a voiceprint recognition component, wherein the display screen is configured to display characters for assisting personnel recognition, the face recognition component is configured to capture an image after the display screen displays the characters for assisting personnel recognition, and extract face features and mouth shape features from the image, the voiceprint recognition component is configured to capture first voice information synchronously with the face recognition component after the display screen displays the characters for assisting personnel recognition, and extract sound features from the first voice information, and the personnel recognition component is configured to recognize the service personnel based on the face features, the mouth shape features and the sound features.

[0007] Further, determining the characters for assisting personnel recognition comprises: determining a plurality of characters; for each character, obtaining mouth shape features and sound features of a plurality of reference personnel on the candidate character, determining a comprehensive difference value of the candidate character according to the mouth shape features and the sound features of each reference personnel on the candidate character; determining a sampling weight of each character according to the comprehensive difference value of each character; and determining the characters for assisting personnel recognition according to the sampling weight of each character.

[0008] Further, recognizing the service personnel based on the face features, the mouth shape features and the sound features comprises: judging whether the personnel is a service personnel with authority based on the face features; if yes, recognizing the character read by the personnel according to the mouth shape features and the sound features, determining whether the current recognition is valid according to the character read by the personnel and the characters for assisting personnel recognition displayed on the display screen, and if yes, the recognition is passed.

[0009] Further, the service recognition unit comprises a voice capturing component and a content recognition component, wherein the voice capturing component is configured to, and the content recognition component is configured to capture second voice information after the personnel clocking unit completes the service personnel recognition, and the content recognition component is configured to recognize and record the service content from the second voice information.

[0010] Further, the off-bed determination component comprises an off-bed sensing device and an off-bed determination device, wherein the off-bed sensing device is configured to capture state data of a bed body, the off-bed determination device is configured to determine whether the user has left the bed according to the state data of the bed body, if yes, record an off-bed start time, and determine an off-bed duration according to the off-bed start time, and the alarm component is configured to alarm based on the fall risk of the user and / or the off-bed duration of the user, comprising: judging whether to alarm based on the off-bed duration and the off-bed duration threshold value corresponding to the user, and if yes, alarming.

[0011] Further, determining the off-bed duration threshold value corresponding to the user comprises: determining the off-bed duration threshold value corresponding to the user according to the fall risk of the user and the health information of the user.

[0012] Further, the basic information recording component is configured to record bed information, bed building time, bed address information, bed cost information, room information, bed subsidy records, user information, and bed history information.

[0013] Further, the sleep monitoring component is configured to predict a fall risk of a user: obtain a big data sample, determine a plurality of user types, a sleep quality factor associated with each user type, and a weight of the sleep quality factor; for each user type, establish a fall risk prediction model corresponding to the user type; determine a user type of the user; obtain sleep monitoring data of the user by using the sleep monitoring device; extract a sleep quality feature of the user from the sleep monitoring data of the user based on the sleep quality factor associated with the user type of the user; and predict a fall risk of the user by using the fall risk prediction model corresponding to the user type of the user based on the sleep quality feature of the user.

[0014] Further, the sleep monitoring component is configured to predict a fall risk of a user: obtain a big data sample, determine a plurality of user types, a sleep quality factor associated with each user type, and a weight of the sleep quality factor; for each user type, establish a fall risk prediction model corresponding to the user type; determine a user type of the user; obtain sleep monitoring data of the user by using the sleep monitoring device; extract a sleep quality feature of the user from the sleep monitoring data of the user based on the sleep quality factor associated with the user type of the user; and predict a fall risk of the user by using the fall risk prediction model corresponding to the user type of the user based on the sleep quality feature of the user.

[0015] Compared with the prior art, the system for monitoring sleep health of the old people, identifying caregivers, and recording nursing information has at least the following beneficial effects: 1、By analyzing sleep monitoring data and other related factors (such as the age of the elderly, medical history, etc.), the system can predict the risk of falls for the elderly. For example, poor sleep quality and frequent turning over may indicate that the elderly are unwell or have decreased balance, increasing the likelihood of falls. Predicting falls in advance allows institutions to take preventive measures in a timely manner, such as adjusting bed height, increasing handrails, etc., to reduce the incidence of falls among the elderly. The off-bed determination component can accurately determine the duration of the elderly leaving the bed. When the elderly leave the bed for more than a set threshold, the alarm component will issue an alarm in a timely manner. This is particularly important for some elderly people with limited mobility or cognitive impairment, as it can prevent accidents such as falls and getting lost when they are away from the bed for a long time. The personnel check-in unit can accurately identify the identity of service personnel (nurses), avoiding the occurrence of irregular behaviors such as proxy check-in. Through attendance records, institutions can understand the work attendance of nurses, ensuring that each bed has enough nurses to provide services, improving the quality and efficiency of nursing services. The service recognition unit can record the service content of nurses, such as assisting the elderly in eating, washing, turning over, and rehabilitation training. These records not only provide detailed information for the elderly's care files, but also help institutions supervise and evaluate the quality of service provided by nurses. For example, by reviewing service records, institutions can understand whether nurses provide services in accordance with the prescribed process and standards, and whether there are cases of inadequate service.

[0016] 2、The personnel check-in unit combines facial features, mouth features, and voice features to identify service personnel. Single facial recognition may be affected by factors such as light, obstruction (such as masks, hats), etc., leading to recognition errors. By combining mouth features and voice features, it is equivalent to adding additional recognition dimensions. For example, when facial recognition cannot accurately determine due to dark light, mouth features (such as lip movements when reading characters) and voice features (such as tone and intonation when reading characters) can further assist in identification, greatly improving the accuracy of identification. When determining the characters for auxiliary personnel identification, the comprehensive difference value of each character is considered and the sampling weight is determined, and then the characters for auxiliary identification are determined. This means that the selected characters have good discrimination in mouth features and voice features. For example, some characters with obvious mouth changes and unique voice features are more easily and accurately identified. By using these characters as auxiliary identification content, the facial recognition component and voiceprint recognition component can more specifically collect and analyze features, improving the accuracy of identification. The mouth and voice features of different service personnel differ, and by considering the features of multiple reference personnel on candidate characters, the characters for auxiliary identification are determined, which can better adapt to the features of different personnel. For example, for some service personnel with low or high-pitched voices, the selected characters can more accurately reflect their voice features, thereby improving the identification effect.

[0017] 3. By distinguishing different user types and establishing corresponding fall risk prediction models for each user type, personalized fall warning is achieved, improving the accuracy and relevance of the warning. By including sleep quality factors in fall risk prediction, the potential impact of sleep on fall risk is considered, making the warning system more comprehensive and scientific. Based on large data samples, user types, sleep quality factors and their weights are determined, making the warning have a solid data foundation and improving the reliability and effectiveness of the warning. By real-time acquisition of user sleep monitoring data through sleep monitoring devices and extraction of sleep quality characteristics based on these data, real-time fall risk prediction is realized, which helps to discover and prevent fall events in time. Through effective fall warning, users (especially the elderly or people with poor health) can take preventive measures to reduce the harm caused by falls, thereby improving their quality of life. BRIEF DESCRIPTION OF DRAWINGS

[0018] The present specification will be further illustrated in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, the same numbers represent the same structures, wherein: Figure 1 is a module schematic diagram of a system for monitoring the sleep health of the elderly and identifying caregivers and recording care information according to some embodiments of the present specification; Figure 2 is a flowchart of a process for predicting the fall risk of a user according to some embodiments of the present specification. DETAILED DESCRIPTION

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present specification, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some examples or embodiments of the present specification, and for those skilled in the art, the present specification can be applied to other similar scenarios without creating creative labor. Unless it is obvious from the language environment or otherwise stated, the same reference numbers in the figures represent the same structure or operation.

[0020] Figure 1 is a module schematic diagram of a system for monitoring the sleep health of the elderly and identifying caregivers and recording care information according to some embodiments of the present specification, as shown in Figure 1 The system for monitoring the sleep health of the elderly and identifying caregivers and recording care information includes a user management module and a personnel attendance module.

[0021] The user management module includes a plurality of smart beds and a data management unit, wherein the smart bed includes a bed body and a basic information recording assembly, a sleep monitoring assembly, a bed leaving determination assembly, and an alarm assembly arranged on the bed body, the sleep monitoring assembly is configured to obtain sleep monitoring data of a user and predict a fall risk of the user, the bed leaving determination assembly is configured to determine a bed leaving duration of the user, and the alarm assembly is configured to alarm based on the fall risk of the user and / or the bed leaving duration of the user.

[0022] The personnel attendance module includes a plurality of personnel clock-in units and a plurality of service identification units, wherein one personnel clock-in unit and one service identification unit are arranged on each bed body, the personnel clock-in unit is configured to identify a service personnel, and the service identification unit is configured to record service content of the service personnel.

[0023] In some embodiments, the personnel clock-in unit includes a display screen, a face recognition assembly, and a voiceprint recognition assembly, wherein the display screen is configured to display characters for assisting personnel identification, the face recognition assembly is configured to capture an image after the display screen displays the characters for assisting personnel identification, and extract face features and mouth shape features from the image, the voiceprint recognition assembly is configured to capture first voice information synchronously with the face recognition assembly after the display screen displays the characters for assisting personnel identification, and extract sound features from the first voice information, and the personnel identification assembly is configured to identify the service personnel based on the face features, the mouth shape features, and the sound features.

[0024] In some embodiments, determining the characters for assisting personnel identification includes: determining a plurality of characters; for each character, obtaining mouth shape features and sound features of a plurality of reference personnel for a candidate character, and determining a comprehensive difference value of the candidate character according to the mouth shape features and the sound features of each reference personnel for the candidate character; determining a sampling weight of each character according to the comprehensive difference value of each character; determining the characters for assisting personnel identification according to the sampling weight of each character.

[0025] Specifically, a character set is determined in advance, and the character set can include Chinese characters, English letters, Chinese characters, or combinations thereof.

[0026] For each candidate character in the character set, a plurality of reference personnel are invited to read or pronounce. In the process of reading or pronouncing by the reference personnel, the system uses the face recognition assembly to capture the mouth shape features, and uses the voiceprint recognition assembly to capture the sound features. These features will serve as basic data for subsequent analysis.

[0027] For each candidate character, the system calculates its comprehensive difference value based on the collected mouth shape features and sound features of multiple reference persons. The comprehensive difference value reflects the degree of difference in mouth shape features and sound features when different reference persons read or pronounce the character. The larger the difference value, the more obvious the pronunciation difference between the character and different people, which may be more conducive to the system to accurately identify.

[0028] By comparing the mouth shape features of different reference persons (such as the shape of the lips, the degree of opening, etc.), the difference degree of the mouth shape features is calculated. By comparing the sound features of different reference persons (such as tone, timbre, speed, etc.), the difference degree of the sound features is calculated. The mouth shape feature difference and the sound feature difference are weighted and summed to obtain the comprehensive difference value of the candidate character.

[0029] According to the comprehensive difference value of each candidate character, the system can determine the sampling weight of each character. The sampling weight reflects the probability of the character being selected in the subsequent identification process. The larger the comprehensive difference value of the character, the higher the sampling weight, and the more likely it is to be selected as the character to be identified by the auxiliary person.

[0030] The character needs to be displayed in high contrast and large font to ensure that the service personnel can still clearly identify it within a certain range from the screen. The character display time needs to be sufficient for the service personnel to complete reading or pronunciation, while avoiding too long to cause recognition efficiency to decline.

[0031] When the display screen displays the character, the face recognition component immediately starts image acquisition. The acquired image needs to be transmitted to the processing unit in real time to avoid delay causing inaccurate feature extraction. Through a deep learning model (such as FaceNet, ArcFace), facial key points are extracted for identity verification. Combined with lip region segmentation and geometric feature analysis (such as lip opening degree, shape change), dynamic features related to pronunciation are extracted. The face features and mouth shape features are fused to form a more comprehensive biometric feature representation, improving the recognition robustness.

[0032] The voiceprint recognition component needs to be strictly synchronized with the face recognition component to ensure that the voice information matches the character display time, and to extract sound features such as tone, timbre, speed, and spectral features.

[0033] In some embodiments, based on the face features, mouth shape features, and sound features, the service personnel identification is performed, including: Based on the face features, it is judged whether the person is a service personnel with authority; If so, according to the mouth shape features and sound features, the character read by the person is identified, and according to the character read by the person and the character displayed on the display screen to be identified by the auxiliary person, it is determined whether the current identification is valid, and if so, the identification is passed.

[0034] Specifically, the facial recognition component uses deep learning algorithms to extract key facial features from the image, including the facial contour, the location and shape of key points such as the eyes, nose, and mouth, as well as facial texture information. The extracted facial features are then compared with facial feature templates of authorized service personnel pre-stored in the system. Feature similarity (such as cosine similarity or Euclidean distance) is calculated to determine whether the currently captured facial features match any of the templates. If the similarity exceeds a preset threshold, the individual is deemed an authorized service personnel. Image processing techniques (such as edge detection and region segmentation) are used to extract mouth shape features, such as mouth shape and degree of opening. These features can reflect the changes in the service personnel's mouth shape when reading aloud. Combining these mouth shape features with vocal characteristics, pattern recognition or machine learning algorithms (such as support vector machines and neural networks) are used to recognize the characters read aloud by the service personnel. The recognized characters are then compared with the characters recognized by the assistant on the display screen. If the two match, it indicates that the service personnel read the correct characters as instructed. If the service personnel reads the correct characters as required, the current recognition is considered valid and the system will consider that the service personnel has successfully completed the clock-in operation. The system can record relevant data such as clock-in time and personnel information.

[0035] In some embodiments, the service identification unit includes a sound collection component and a content recognition component, wherein the sound collection component is used to obtain second voice information after the personnel punching unit completes the service personnel identification, and the content recognition component is used to identify the service content from the second voice information and record it.

[0036] Specifically, the sound collection component is the hardware device in the service recognition unit responsible for capturing the service personnel's voice information. It can be a high-sensitivity microphone array that can adapt to the voice collection needs of different environments. After the service personnel completes the clocking in, the sound collection component enters a continuous monitoring state to ensure that no service-related voice information is missed. The voice information can be the voice information stated by the service personnel. The denoised voice information is input into the ASR (Automatic Speech Recognition) model, and the speech transcription result in text form is output. The service keywords in the transcribed text are identified through regular expressions or NLP models to generate service content. The extracted service content is stored in the database in formats such as JSON and XML, and associated with information such as the service personnel ID and timestamp. A complete log containing the original voice, transcribed text, and recognition results is generated to support subsequent tracing and review.

[0037] The bed exit determination component may be enabled at a specific time, for example, during a time period when the user is accustomed to sleeping.

[0038] In some embodiments, the bed leaving determination component includes a bed leaving sensing device and a bed leaving determination device, wherein the bed leaving sensing device is used to collect status data of the bed body, and the bed leaving determination device is used to determine whether the user has left the bed based on the status data of the bed body. If so, record the start time of leaving the bed, and determine the duration of leaving the bed based on the start time of leaving the bed.

[0039] Specifically, the bed leaving sensing device may include multiple pressure sensors embedded inside the mattress, and the multiple pressure sensors monitor the pressure distribution of different areas of the bed surface in real time. When the user lies on the bed, the sensor will detect the change in the pressure value of a specific area; if the user leaves the bed, the pressure value of these areas will drop significantly. Set a pressure threshold, and when the pressure sensed by multiple pressure sensors is less than the pressure threshold, it is determined that the user has left the bed. The pressure threshold can be obtained based on human experience or experimental data. When it is determined that the user has left the bed, the bed leaving determination device will immediately record the current bed leaving start time, and calculate the bed leaving duration by continuously monitoring whether the user returns to the bed. If the user does not return to the bed within a certain period of time (such as the set threshold time), it is determined to be a complete bed leaving event, and the bed leaving duration is recorded.

[0040] In some embodiments, the alarm component is used to issue an alarm based on the user's fall risk and / or the user's bed exit duration, including: Based on the duration of leaving the bed and the duration threshold of leaving the bed corresponding to the user, it is determined whether to issue an alarm. If so, an alarm is issued.

[0041] Specifically, when the duration of time spent out of bed exceeds a threshold, an alarm can be triggered. When the user's fall risk exceeds a threshold, an alarm can be triggered, notifying relevant personnel (such as family members, medical staff, and service personnel) to address the risk. The fall risk threshold can be derived from experience or experimental data.

[0042] In some embodiments, determining a user's corresponding out-of-bed duration threshold includes: Determine the user's corresponding out-of-bed duration threshold based on the user's fall risk and the user's health information.

[0043] Specifically, the user's health information may include: Basic health data: including height, weight, blood pressure, heart rate, blood sugar and other routine physiological indicators. These data can reflect the user's overall health status and physical function level.

[0044] Chronic disease information: Understand whether the user has chronic diseases such as diabetes, hypertension, heart disease, etc., as well as the disease control status and treatment plan. Chronic diseases may affect the user's physical endurance and mobility, thereby affecting safety after leaving bed.

[0045] Medication: Record the types of medication the user is taking, the dosage, and the time of administration. Some medications can have side effects such as dizziness, fatigue, drowsiness, etc.

[0046] The off-bed duration threshold can be periodically re-evaluated (e.g., every 3-6 months) by human or machine learning models based on the user's fall risk and health information, and adjusted according to the evaluation results. As the user's age, physical condition, or disease progresses, their fall risk and health needs may change, so it is necessary to regularly update the threshold to ensure its reasonableness and effectiveness. For users with a high fall risk, as they are more likely to fall after leaving the bed, a shorter off-bed duration threshold should be set. For example, set the threshold to 15-30 minutes, and once the user's off-bed time exceeds this threshold, the system should promptly send a reminder to notify relevant personnel to pay attention to the user's situation. The off-bed duration threshold for medium-risk users can be appropriately relaxed, such as setting it to 30-60 minutes. However, close attention should still be paid to the user's off-bed situation to ensure their safety. For users with a lower fall risk, the off-bed duration threshold can be relatively longer, such as 60-120 minutes. However, it cannot be completely ignored, and regular safety reminders and checks should still be conducted. If the user's basic health data shows abnormalities, such as a sudden increase or decrease in blood pressure, a rapid or slow heart rate, etc., it may indicate that the user's physical condition is unstable, in which case the off-bed duration threshold should be appropriately shortened to ensure user safety. When the user's chronic disease has an acute attack, such as a diabetic patient's blood sugar rising or falling sharply, a heart attack patient's sudden angina, etc., the off-bed duration threshold should be immediately shortened and monitoring should be strengthened. If the user experiences significant side effects from medication, such as dizziness, fatigue, etc., which affect their mobility, the off-bed duration threshold should be adjusted accordingly based on the severity of the side effects In some embodiments, the basic information recording component is used to record bed information, including bed information, bed information, bed information, bed information, room information, bed subsidy records, user information, and bed history information.

[0047] Specifically, the institution name: clearly record the specific name of the institution to which the bed belongs, such as "XX nursing home" "XX hospital rehabilitation department" etc. This helps to distinguish between bed resources in different institutions, facilitating unified management and resource allocation.

[0048] Institution type: mark the nature of the institution, such as a nursing institution, a medical institution, a rehabilitation institution, etc.

[0049] Institution contact information: including the contact information of the institution, such as the contact number, email, official website, etc.

[0050] Institution address: Record the institution’s geographical location in detail, including province, city, district, street, house number, etc.

[0051] Bed establishment date: This accurately records the specific year, month, and day the bed was established. Bed establishment date is an important starting point in the bed's usage history and can be used to analyze bed usage duration, maintenance cycles, and more.

[0052] Bed address information: the specific location of the bed within the institution, such as room number, floor, bed number, etc.

[0053] Bed fee information: clearly record the bed charging standards, including daily fees, monthly fees or fees calculated based on other periods.

[0054] Room information: Mark the type of room, such as single room, double room, multi-person room, etc.

[0055] Bed subsidy records: record the types of bed subsidies, such as government subsidies, charity funding, corporate sponsorship, etc.

[0056] User information: including the user's name, gender, age, ID number, contact information, etc. It also records the user's health status, such as whether they have chronic diseases, allergies, disabilities, etc., as well as the user's emergency contact name, relationship, contact number and other information.

[0057] Bed historical occupancy information: records the user's name, check-in time, check-out time and other information for each check-in.

[0058] The following experiments demonstrate the association between sleep quality and fall risk.

[0059] Sleep monitoring and fall data were collected from 300 elderly people aged 65-90 years across three nursing homes between January 2024 and March 2025. Hospital A housed single rooms in a quiet environment, Hospital B housed double rooms in a standard environment, and Hospital C housed a natural sound environment, with 100 residents each. All residents had no severe cognitive impairment, could walk independently, and volunteered to participate in the monitoring.

[0060] Sleep monitoring: Use millimeter-wave radar to continuously record total sleep time, deep sleep percentage, and number of awakenings at night.

[0061] Fall records: Nurses record fall events (including time, location, and degree of injury) daily to exclude involuntary falls.

[0062] Logistic regression was used to analyze the association between sleep quality indicators and fall risk, and odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. The chi-square test was used to compare fall rates among different nursing homes.

[0063] The results of the logistic regression analysis of sleep quality indicators and fall risk are shown in Table 1.

[0064] Table 1 Sleep quality indicators Fall incidence OR value (95% CI) P value Total sleep duration < 6 hours 32.6% 3.12（1.89-5.14） <0.001 Deep sleep percentage < 15% 28.4% 2.76（1.67-4.55） 0.002 Nightly awakenings ≥ 3 times / night 37.1% 4.05（2.33-7.02） <0.001 Table 1 shows that the incidence of falls in elderly people with total sleep duration <6 hours, deep sleep percentage <15%, and nighttime awakenings ≥3 times / night was significantly increased, with OR values of 3.12 (95% CI: 1.89-5.14), 2.76 (95% CI: 1.67-4.55), and 4.05 (95% CI: 2.33-7.02), respectively, with P values all <0.05. The number of nighttime awakenings had the greatest impact on fall risk (OR=4.05).

[0065] The fall rates in different nursing homes are shown in Table 2.

[0066] Table 2 Institution Mean deep sleep duration Falls per 100 person years Percentage of severe injuries A hospital 1.8 ± 0.4 hours 18.2 12.1% B hospital 1.2 ± 0.6 hours 42.7 27.6% C hospital 2.1 ± 0.3 hours 9.5 5.3% As shown in Table 2, the average deep sleep duration in C (natural sound environment) was the longest (2.1±0.3 hours), the number of falls per hundred people per year was the lowest (9.5 times), and the percentage of severe injuries was the lowest (5.3%). Significance test showed that the fall rate in C was significantly lower than that in A (18.2 times per hundred people per year) and B (42.7 times per hundred people per year) (χ²=15.32, P<0.001).

[0067] Mechanism analysis: 1. Physiological pathway: Insufficient deep sleep leads to decreased vestibular function (23.7% reduction in balance ability) and increased daytime fatigue (41.2%).

[0068] Nighttime awakenings cause orthostatic hypotension (systolic blood pressure drop >20 mmHg, risk increased 2.3 times).

[0069] 2. Environmental interaction: Noise >50 decibels increases the number of awakenings by 1.8 times (B vs C, P=0.008).

[0070] Single rooms reduce nighttime toilet path obstacles (A's fall rate is 57.4% lower than B's).

[0071] Experiments have confirmed that sleep quality is an important factor affecting the fall risk of the elderly. The number of nighttime awakenings has the greatest impact on fall risk (OR=4.05), which may be related to the decrease in balance ability and delayed reaction caused by nighttime awakenings. Insufficient deep sleep indirectly increases the risk of falls by affecting vestibular function and daytime fatigue, which is consistent with previous studies.

[0072] Figure 2is a flowchart of predicting a fall risk of a user according to some embodiments of the present specification, as shown in Figure 2 As shown, predicting a fall risk of a user can include the following steps.

[0073] In step 210, a big data sample is obtained, and a plurality of user types, sleep quality factors associated with each user type, and weights of the sleep quality factors are determined.

[0074] The big data sample includes medical history information, sleep monitoring data, and fall risk assessment data of a plurality of sample users. Specifically, the medical history information can include the user's age, gender, medical history (such as chronic diseases, surgical history, etc.), medication, etc. The sleep monitoring data is obtained by wearable devices, mattress sensors, or millimeter wave radars, etc. The sleep monitoring data continuously records the total sleep duration, deep sleep proportion, number of night awakenings, and other sleep quality indicators of the user. The fall risk assessment data can include a plurality of fall risk assessment values, wherein the plurality of fall risk assessment indicators at least include vestibular nerve signal conduction efficiency, creatine kinase level, melatonin secretion cycle, and systolic blood pressure.

[0075] In some embodiments, the plurality of user types are determined, including: A plurality of diseases to be screened and a plurality of drugs to be screened are determined, wherein the diseases to be screened are diseases closely related to the fall risk, such as Parkinson's disease, sequelae of stroke, osteoporosis, cardiovascular disease, etc., and the drugs to be screened are drugs that can affect sleep quality or increase the risk of falling, such as sedatives, antidepressants, antihypertensive drugs, etc. According to the medical history information and fall data of the plurality of sample users, an influence coefficient of each disease to be screened on the fall risk is determined, and an influence disease is screened from the plurality of diseases to be screened according to the influence coefficient of each disease to be screened on the fall risk. Specifically, statistical methods (such as Logistic regression, Cox proportional hazards model, etc.) are used to analyze the influence coefficient of each disease to be screened on the fall risk, and according to the size of the influence coefficient and the significance level (such as P value <0.05), the influence disease that has a significant influence on the fall risk is screened from the plurality of diseases to be screened. According to the medical history information and fall data of the plurality of sample users, an influence coefficient of each drug to be screened on the fall risk is determined, and an influence drug is screened from the plurality of drugs to be screened according to the influence coefficient of each drug to be screened on the fall risk. Specifically, statistical methods (such as Logistic regression, Cox proportional hazards model, etc.) are used to analyze the influence coefficient of each drug to be screened on the fall risk, and according to the size of the influence coefficient and the significance level (such as P value <0.05), the influence disease that has a significant influence on the fall risk is screened from the plurality of diseases to be screened. Based on the influence disease, the influence drug, and the medical history information of the plurality of sample users, a plurality of user types are determined.

[0076] In some embodiments, based on the impact diseases, the impact drugs and the medical record information of the plurality of sample users, a plurality of user types are determined, comprising: For each sample user, an impact disease identification vector of the sample user is determined according to the medical record information of the sample user and the impact diseases, and an impact drug identification vector of the sample user is determined according to the medical record information of the sample user and the impact drugs; The identification vector similarity of any two sample users is determined according to the impact disease identification vector and the impact drug identification vector of the two sample users. The plurality of sample users are divided into a plurality of user clusters according to the identification vector similarity of any two sample users by a clustering algorithm, and each user cluster corresponds to a user type.

[0077] Specifically, for each sample user, it is checked whether the impact diseases are contained in the medical record information of the sample user, and a binary vector (identification vector) is constructed, the length of the vector being equal to the total number of impact diseases. If the user has a certain impact disease, the corresponding position is marked as 1; otherwise, it is marked as 0. Assuming that there are 3 impact diseases (A, B, C), a user has diseases A and C, and the impact disease identification vector of the user is [1, 0, 1].

[0078] For each sample user, it is checked whether the impact drugs are being used in the medical record information of the sample user, and a binary vector (identification vector) is constructed, the length of the vector being equal to the total number of impact drugs. If the user is using a certain impact drug, the corresponding position is marked as 1; otherwise, it is marked as 0. Assuming that there are 2 impact drugs (X, Y), a user is using drug X, and the impact drug identification vector of the user is [1, 0].

[0079] For any two sample users, the cosine similarity of the impact disease identification vectors and the cosine similarity of the impact drug identification vectors are calculated respectively, and the cosine similarity of the impact disease identification vectors and the cosine similarity of the impact drug identification vectors are weighted and summed to obtain the identification vector similarity of the two sample users.

[0080] The plurality of sample users can be divided into a plurality of user clusters according to the identification vector similarity of any two sample users by a K-means clustering algorithm, specifically comprising the following steps: S11, initializing cluster centers: Randomly selecting K sample users as initial cluster centers.

[0081] S12, assigning samples to clusters: For each sample user, the distance (such as Euclidean distance) between its comprehensive identification vector and all cluster centers is calculated.

[0082] The sample is assigned to the nearest cluster.

[0083] S13, update cluster center: Recalculate the center of each cluster (i.e., the mean of all sample identification vectors within the cluster).

[0084] S14, iterative optimization: Repeat steps S12 and S13 until the cluster centers no longer change significantly or the maximum number of iterations is reached.

[0085] For each user cluster, count the distribution of users within the user cluster who have an impact on the disease and an impact on the drug.

[0086] Example: User cluster 1: 80% of users have Parkinson's disease, and 60% of users are using antidepressants.

[0087] User cluster 2: 70% of users have osteoporosis, and there is no significant impact on the drug.

[0088] According to the common characteristics of users within the user cluster, name each user cluster.

[0089] Example: User cluster 1 is named "Parkinson's disease combined with depression patient type".

[0090] User cluster 2 is named "osteoporosis patient type".

[0091] In some embodiments, a sleep quality factor associated with each user type is determined, including: determining a plurality of sleep quality factors to be screened and a plurality of fall risk assessment indicators, wherein the plurality of sleep quality factors to be screened at least includes the number of awakenings, total sleep duration, first stage sleep duration, and second stage sleep duration; the plurality of fall risk assessment indicators at least includes vestibular nerve signal conduction efficiency, creatine kinase level, melatonin secretion cycle, and systolic blood pressure; For each user type, based on the sleep monitoring data and fall risk assessment data of the plurality of sample users included in the user cluster, determine the correlation coefficient of each sleep quality factor to be screened and each fall risk assessment indicator, and determine the sleep quality factor associated with the user type according to the correlation coefficient of each sleep quality factor to be screened and each fall risk assessment indicator.

[0092] Specifically, the number of awakenings: the number of awakenings during night sleep, reflecting the continuity of sleep, defined as body movement > 3 times / hour or heart rate > 100 bpm for 5 minutes, which is considered as awakening.

[0093] Total sleep duration: total time of night sleep, reflecting the adequacy of sleep.

[0094] First stage sleep duration: Time spent in light sleep, typically less than 50% of total sleep duration, defined by dominant brain waves theta (4-7 Hz) and respiratory rate ≤ 12 breaths per minute.

[0095] Second stage sleep duration: Time spent in deep sleep, crucial for physical recovery and repair, typically 15-25% of total sleep duration, defined by dominant brain waves delta (0.5-4 Hz) and body movement frequency ≤ 1 movement per hour.

[0096] Vestibular nerve signaling efficiency: Reflects the function of the vestibular system, which is related to balance and spatial orientation. A decline in function may increase the risk of falls.

[0097] Creatine kinase level: Reflects muscle damage or metabolic status. A decline in muscle function may affect walking stability.

[0098] Melatonin secretion cycle: Reflects the biological clock and sleep rhythm. Disruption may lead to sleep disorders and increased risk of falls.

[0099] Systolic blood pressure: Blood pressure fluctuations may affect brain blood supply, leading to dizziness or balance disorders, increasing the risk of falls.

[0100] From each user cluster corresponding to each user type, extract the sleep monitoring data and fall risk assessment data of all sample users. Each sample user's data should include: the measurement value of the sleep quality factor to be screened (such as the number of awakenings, total sleep duration, etc.). The measurement value of the fall risk assessment index (such as vestibular nerve signaling efficiency, creatine kinase level, etc.). The measurement value of the sleep quality factor to be screened and the measurement value of the fall risk assessment index of the user cluster corresponding to the user type are substituted into the Spearman rank correlation coefficient, and the correlation coefficient of the sleep quality factor to be screened and the fall risk assessment index is calculated.

[0101] According to the significance and size of the correlation coefficient, screen out the sleep quality factors that are significantly correlated with the fall risk assessment index as the sleep quality factors associated with this user type.

[0102] Significance: The P-value of the correlation coefficient is less than the set significance level (such as 0.05).

[0103] Correlation strength: According to research needs, set the absolute value threshold of the correlation coefficient (such as |r|>0.3).

[0104] Screening process: For each user type, iterate through all correlation coefficients of the sleep quality factors to be screened and the fall risk assessment index.

[0105] Screen out the sleep quality factors corresponding to the correlation coefficients that meet the screening criteria.

[0106] Example: In the "Parkinson's disease combined with depression patient type", it was found that the "number of awakenings" was significantly negatively correlated with the "vestibular nerve signal conduction efficiency" (r = -0.45, P<0.01), so the "number of awakenings" is the sleep quality factor associated with this user type.

[0107] In the "osteoporosis patient type", it was found that "total sleep duration" was significantly positively correlated with "creatine kinase level" (r = 0.38, P<0.05), so "total sleep duration" is the sleep quality factor associated with this user type.

[0108] In some embodiments, determining the weight of the sleep quality factor associated with the user type includes: The weight of the sleep quality factor associated with the user type is determined according to the correlation coefficient between each sleep quality factor associated with the user type and each fall risk assessment indicator.

[0109] Specifically, for each sleep quality factor associated with the user type, the correlation coefficients between the sleep quality factor and each fall risk assessment indicator are averaged to obtain the correlation coefficient mean.

[0110] The correlation coefficient means of each sleep quality factor associated with the user type are summed to obtain the sum of the correlation coefficient means, and the ratio of the correlation coefficient of the sleep quality factor associated with the user type to the sum of the correlation coefficient means is used as the weight of the sleep quality factor associated with the user type.

[0111] Step 220: For each user type, establish a fall risk prediction model corresponding to the user type.

[0112] Specifically, the fall risk prediction model corresponding to the user type can be a deep learning model.

[0113] There are significant differences in the fall risk mechanisms of different user types, and it is necessary to build a personalized fall risk prediction model based on their unique sleep quality factors and risk characteristics.

[0114] Deep learning models (such as long short-term memory networks, Transformer, convolutional neural networks, etc.) are suitable for processing high-dimensional, nonlinear, and time series data, and can capture the complex relationship between sleep quality factors and fall risk.

[0115] Input and output design of the fall risk prediction model: 1. Input layer: Time series data: sleep quality factors for multiple consecutive days (e.g., 7 days × 5 factors = 35-dimensional features) and the weights of sleep quality factors associated with user types.

[0116] 2. Output layer: Binary classification: Fall risk (0 = low risk, 1 = high risk).

[0117] Regression: Probability of fall risk (0~1) or prediction of future fall counts.

[0118] 3. Model training and optimization Data preprocessing: Normalization: Z-score standardization on sleep quality factors.

[0119] Sliding window: Splitting consecutive multi-day data into fixed-length sequences (e.g., 7 days per window).

[0120] Data augmentation: Adding Gaussian noise or random occlusion to time series data to improve generalization.

[0121] 4. Model training: Loss function: Binary cross-entropy (classification) or mean squared error (regression).

[0122] Optimizer: Adam (adaptive learning rate).

[0123] Regularization: Dropout (to prevent overfitting), L2 regularization.

[0124] 5. Hyperparameter tuning: Use grid search or Bayesian optimization to adjust the following parameters: Number of LSTM layers (1~3 layers).

[0125] Number of hidden units (32~256).

[0126] Learning rate (1e-3~1e-5).

[0127] Evaluation metrics: Classification tasks: Accuracy, recall, F1 score, ROC-AUC.

[0128] Regression tasks: Mean squared error (MSE), mean absolute error (MAE).

[0129] Step 230, determine the user type of the current user.

[0130] Specifically, the user type of the current user can be determined according to the medical record information of the user.

[0131] Step 240, obtain sleep monitoring data of the current user through a sleep monitoring device.

[0132] The sleep monitoring device includes a millimeter wave radar component and a heart rate monitoring component.

[0133] Millimeter wave radar component Working Principle: Millimeter wave radar emits high-frequency electromagnetic waves (usually 24 GHz or 77 GHz) and receives reflected signals to analyze the distance, speed, and angle of target objects.

[0134] In sleep monitoring, radar can penetrate lightweight fabrics such as bed sheets and obtain physiological and movement information without direct contact with the user.

[0135] Function and Data Collection: Respiratory Monitoring: Detects respiratory rate and depth through chest micro-movement.

[0136] Body Movement Monitoring: Captures large movements such as turning over and sitting up to assess sleep fragmentation.

[0137] Micro-movement Analysis: Identifies subtle limb twitches or convulsions to assist in determining rapid eye movement sleep (REM) or abnormal sleep behaviors (such as RBD, Rapid Eye Movement Sleep Behavior Disorder).

[0138] Spatial Positioning: Monitors body position changes during sleep (e.g., supine, lateral).

[0139] Heart Rate Monitoring Component Working Principle: Through photoplethysmography (PPG) or electrocardiogram (ECG) technology, measures user heart rate and heart rate variability (HRV).

[0140] PPG sensors are usually integrated into wearable devices (such as wristbands, chest straps) or mattress sensors, detecting blood volume changes through LED light and photodiodes.

[0141] Function and Data Collection: Heart Rate Monitoring: Records resting heart rate and heart rate changes during sleep in real time.

[0142] Step 250: Based on the sleep quality factors associated with the user type of the current user, extract the sleep quality features of the current user from the sleep monitoring data of the current user.

[0143] In some embodiments, step 250 specifically includes: For each user type, based on the sleep quality factors associated with the user type, establish a feature extraction model corresponding to the user type; Through the feature extraction model corresponding to the user type of the current user, extract the sleep quality features of the current user from the sleep monitoring data of the current user.

[0144] Specifically, in the fall risk prediction, the influence mechanism of sleep quality factors of different user types on fall risk is significantly different. Therefore, it is necessary to customize the feature extraction model for each user type to extract features highly related to fall risk from the original sleep monitoring data. The feature extraction model can be used for the deep learning model.

[0145] Input data preprocessing: Standardize, normalize, or sliding window segmentation on the current user's sleep monitoring data.

[0146] Example: Segment the micro-motion signal of the millimeter wave radar into 1-minute windows and calculate the micro-motion energy of each window.

[0147] Feature extraction: Input the preprocessed data into the feature extraction model corresponding to the user type to output sleep quality features.

[0148] Step 260, based on the sleep quality features of the current user, predict the fall risk of the current user through the fall risk prediction model corresponding to the user type of the current user.

[0149] Specifically, the sleep quality features of the current user can be input into the fall risk prediction model corresponding to the user type of the current user to predict the fall risk of the current user.

[0150] In some embodiments, the method further comprises step 270, if the fall risk of the current user is greater than the preset fall risk threshold, determining the environment control strategy and exercise recommendation based on the sleep environment information of the current user and the sleep quality features of the current user.

[0151] Specifically, through correlation analysis or causal inference model, the association between environmental factors and sleep quality features is established, and key risk factors are identified.

[0152] Based on the correlation analysis results, the following environmental control strategies are developed: 1. Sound environment optimization Applicable scenario: User sleep quality features show multiple awakenings or sleep fragmentation.

[0153] Specific measures: Noise detection: Monitor the noise level in the bedroom at night through wearable devices or environmental sensors.

[0154] Hierarchical intervention: If the noise is > 40 decibels: Install soundproof curtains, seal door and window gaps, or use active noise-cancelling headphones.

[0155] If there is a sudden noise (such as door closing): Introduce a white noise machine (such as a fan sound, rain sound) to mask the sudden noise.

[0156] Effect verification: Monitor whether the number of awakenings is reduced after the intervention (target reduction ≥30%).

[0157] 2. Lighting control Applicable scenarios: Users who frequently get up at night or feel dizzy.

[0158] Specific measures: Lighting modification: Install induction floor lights (brightness <10 lux, color temperature <2700K) to avoid strong light stimulation at night.

[0159] Remove the main bedroom light and use a dimmable bedside reading lamp instead.

[0160] Adjust furniture layout: Ensure that walking paths are clear at night to avoid tripping over debris.

[0161] Effect verification: After the intervention, the incidence of dizziness when getting up at night was counted (target reduction ≥50%).

[0162] 3. Temperature and humidity control Applicable scenario: The user's sleep quality characteristics show short deep sleep duration or frequent jitters during REM sleep.

[0163] Specific measures: Temperature Control: Summer: Set the air conditioner temperature to 24-26℃ to avoid sweating due to overheating or curling up due to overcooling.

[0164] Winter: Use an electric blanket (low setting) or thick bedding to maintain body surface temperature at 32-34°C.

[0165] Humidity regulation: If humidity is <30%: Use a humidifier to maintain humidity at 40-60% (to reduce respiratory dryness).

[0166] If humidity > 70%: Use a dehumidifier to prevent mold growth.

[0167] Effect verification: Monitor whether the deep sleep duration increases after the intervention (target increase ≥15%).

[0168] Based on sleep quality characteristics, the following exercise program is recommended to improve sleep and reduce the risk of falls: 1. Tai Chi training Applicable scenarios: Users with poor balance or frequent jitters during REM sleep.

[0169] Specific measures: Action selection: Give priority to movements that emphasize the transfer of center of gravity, such as "Cloud Hands", "Single Whip", and "Wild Horse's Mane".

[0170] Avoid single-leg standing movements like "Golden Rooster Standing" (if the user has very poor balance).

[0171] Training Plan: 3-5 times per week, 30-45 minutes each time, completed in 2-3 sets.

[0172] Combine with breathing exercises (such as "rise-inhale-fall-exhale"), to enhance autonomic nervous regulation ability.

[0173] Effect Verification: After intervention, assess balance improvement through balance tests (such as Berg Scale).

[0174] 2. Balance and Strength Training Applicable Scenario: Users experience dizziness when getting up at night or lower limb muscle weakness.

[0175] Specific Measures: Single-Leg Standing: 3 sets per day, 30 seconds each (with wall support), gradually transition to no wall support.

[0176] Increase difficulty: Close your eyes or stand on a soft mat.

[0177] Dynamic Balance Exercises: "heel-toe walking": Walk along a straight line, with the front heel touching the back toe, 10 meters each time.

[0178] "Sit-to-Stand Test": Stand up from a chair without support and sit down, record the time (target <10 seconds).

[0179] Lower Limb Strength Training: Wall Squats: 3 sets per day, 30 seconds each, knees not exceeding the toes.

[0180] Resistance Band Training: Perform hip abduction, knee flexion and extension movements.

[0181] Effect Verification: After intervention, assess fall risk through the Timed Up and Go Test (TUGT).

[0182] 3. Flexibility Training Applicable Scenario: Users experience muscle stiffness at night or limited body position changes.

[0183] Specific Measures: Pre-Sleep Stretching: Calf Triceps: Stand facing the wall, hands on the wall, back foot straight, heel on the ground, feel the stretch on the back of the calf.

[0184] Quadriceps: Stand on one leg, grasp the same side ankle with the other hand, stretch towards the hips.

[0185] Back muscles: Lie on your back, bend your legs, and wrap your hands around your knees to gently pull them towards your chest.

[0186] Training plan: Hold each movement for 15-30 seconds, repeat 2-3 sets.

[0187] Combine deep breathing (relax during inhalation, deepen stretching during exhalation).

[0188] Effect verification: After intervention, assess flexibility improvement through joint range of motion test.

[0189] Personalized intervention plan generation process: 1. Data input: Sleep environment information (noise, lighting, temperature and humidity, slip resistance).

[0190] Sleep quality characteristics (number of awakenings, REM period tremor, AHI, body position change frequency).

[0191] 2. Risk factor identification: Identify key risk factors through correlation analysis (such as Pearson correlation coefficient) or decision tree model.

[0192] Example: If the correlation coefficient between the number of awakenings and noise level is >0.5, then noise is a key factor.

[0193] 3. Strategy matching: Match environmental control strategies and exercise recommendations based on risk factors.

[0194] For example: Scenario Example 1: Parkinson's disease patient Sleep quality characteristics: Frequent REM period limb tremor (>2 times / minute), multiple night awakenings (>8 times / night).

[0195] Environmental information: Bedroom background noise >45 decibels (window traffic noise), no floor light at night, need to turn on main light to use the bathroom.

[0196] Intervention plan: Environment: Install soundproof windows + white noise machine, install inductive floor light beside the bed.

[0197] Exercise: Daily Tai Chi training, calf stretching before sleep.

[0198] Scenario Example 2: Diabetic patient Sleep quality characteristics: Sleep apnea hypopnea index (AHI) >15 times / hour, large sleep period blood glucose fluctuation (>10%).

[0199] Environmental information: Bright bedroom lighting at night (color temperature 4000K), cluttered items beside the bed, narrow walking space.

[0200] Intervention plan: Environment: Replace warm light bulbs, clean up clutter around bed.

[0201] Exercise: Daily balance training (single leg standing) + flexibility training (full body stretching).

[0202] In the end, it should be understood that the embodiments described in this specification are only to illustrate the principles of the embodiments described in this specification. Other variations can also belong to the scope of this specification. Therefore, as an example but not limitation, alternative configurations of the embodiments of this specification can be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. A system for monitoring the sleep health of the elderly, identifying caregivers, and recording nursing information, characterized by: include: The user management module includes multiple smart beds and a data management unit. The smart bed includes a bed body and a basic information recording component, a sleep monitoring component, a bed exit determination component, and an alarm component provided on the bed body. The sleep monitoring component is used to obtain the user's sleep monitoring data and predict the user's fall risk. The bed exit determination component is used to determine the user's bed exit duration. The alarm component is used to issue an alarm based on the user's fall risk and / or the user's bed exit duration. The personnel attendance module includes multiple personnel punching units and multiple service identification units, wherein each bed body is provided with a personnel punching unit and a service identification unit, the personnel punching unit is used to identify the service personnel, and the service identification unit is used to record the service content of the service personnel.

2. The system for monitoring the sleep health of the elderly, identifying caregivers and recording nursing information according to claim 1 is characterized in that: The personnel punching unit includes a display screen, a face recognition component and a voiceprint recognition component, wherein the display screen is used to display characters for auxiliary personnel identification, the face recognition component is used to capture an image after the display screen displays the characters for auxiliary personnel identification, and extract facial features and mouth shape features from the image, the voiceprint recognition component is used to collect first voice information synchronously with the face recognition component after the display screen displays the characters for auxiliary personnel identification, and extract sound features from the first voice information, and the personnel identification component is used to identify service personnel based on facial features, mouth shape features and sound features.

3. The system for monitoring the sleep health of the elderly, identifying caregivers and recording nursing information according to claim 2 is characterized in that: Determine the characters recognized by the auxiliary personnel, including: Identify multiple characters; For each character, obtain the mouth shape features and voice features of the candidate character from multiple reference persons, and determine the comprehensive difference value of the candidate character based on the mouth shape features and voice features of the candidate character from each reference person; Determine the sampling weight of each character based on the comprehensive difference value of each character; Based on the sampling weight of each character, the characters recognized by the auxiliary personnel are determined.

4. The system for monitoring the sleep health of the elderly, identifying caregivers and recording nursing information according to claim 2 is characterized in that: Identify service personnel based on facial features, mouth shape features, and voice features, including: Based on facial features, determine whether the person is an authorized service personnel; If so, the characters read by the recognition personnel are identified based on the mouth shape features and voice features, and the validity of the current recognition is determined based on the characters read by the personnel and the characters recognized by the auxiliary personnel displayed on the display screen. If so, the recognition is passed.

5. The system for monitoring the sleep health of the elderly, identifying caregivers and recording nursing information according to claim 1 is characterized in that: The service identification unit includes a sound collection component and a content recognition component, wherein the sound collection component is used to obtain second voice information after the personnel punching unit completes the service personnel identification, and the content recognition component is used to identify the service content from the second voice information and record it.

6. The system for monitoring the sleep health of the elderly, identifying caregivers and recording nursing information according to any one of claims 1 to 5, characterized in that: The bed leaving determination component includes a bed leaving sensing device and a bed leaving determination device, wherein the bed leaving sensing device is used to collect bed body status data, and the bed leaving determination device is used to determine whether the user has left the bed based on the bed body status data, and if so, record the bed leaving start time and determine the bed leaving duration based on the bed leaving start time; The alarm component is used to issue an alarm based on the user's fall risk and / or the user's bed leaving duration, including: Based on the duration of leaving the bed and the duration threshold of leaving the bed corresponding to the user, it is determined whether to issue an alarm. If so, an alarm is issued.

7. The system for monitoring the sleep health of the elderly, identifying caregivers and recording nursing information according to claim 6 is characterized in that: Determine the user's corresponding bed leaving duration threshold, including: Determine the user's corresponding out-of-bed duration threshold based on the user's fall risk and the user's health information.

8. The system for monitoring the sleep health of the elderly, identifying caregivers, and recording care information according to any one of claims 1 to 5, characterized in that: The basic information recording component is used to record the bed's institution information, bed establishment time, bed address information, bed fee information, room information, bed subsidy records, user information and bed occupancy history information.

9. The system for monitoring the sleep health of the elderly, identifying caregivers, and recording care information according to any one of claims 1 to 5, characterized in that: The sleep monitoring component predicts the user's fall risk: Obtaining a large data sample, determining multiple user types, sleep quality factors associated with each user type, and weights of the sleep quality factors; For each user type, a fall risk prediction model corresponding to the user type is established; Determine the user type of the user; Obtain the current user's sleep monitoring data through the sleep monitoring device; Extracting the sleep quality characteristics of the current user from the user's sleep monitoring data based on the sleep quality factor associated with the user's user type; The fall risk prediction model corresponding to the user's user type is used to predict the current user's fall risk based on the user's sleep quality characteristics.

10. The system for monitoring the sleep health of the elderly, identifying caregivers and recording nursing information according to claim 8, characterized in that: Determine the sleep quality factors associated with each user type, including: Identify multiple sleep quality factors and fall risk assessment indicators to be screened; For each user type, based on the sleep monitoring data and fall risk assessment data of multiple sample users included in the user cluster, the correlation coefficient between each sleep quality factor to be screened and each fall risk assessment indicator is determined. According to the correlation coefficient between each sleep quality factor to be screened and each fall risk assessment indicator, the sleep quality factor associated with the user type is determined.

Images