Health sleep model monitoring system and method based on artificial intelligence and biomedical sensing technology

Abstract

The invention provides a healthy sleep model monitoring system and method based on artificial intelligence and biomedical sensing technology, and the system comprises a data collection module which is used for collecting the heart rate, breath, blood oxygen, body temperature and in-bed posture data of a patient; the data acquisition module comprises a heart rate and breath monitoring unit which acquires a BCG signal of a patient in a non-contact mode, and the non-contact mode is that an optical fiber microbend sensor embedded into a mattress is adopted and used for indirectly reflecting the heart rate and breath conditions of the patient; and the blood oxygen and body temperature monitoring unit adopts a wireless sensor tag group which is applied to the body surface of the patient. Vital sign data can be collected without direct contact with a patient, pathogen transmission caused by contact is effectively avoided, safety of medical staff and the patient is practically guaranteed, meanwhile, excessive equipment does not need to be pasted or connected to the body of the patient, and constraint and interference to the body of the patient are reduced.

Description

Technical Field

[0001] This invention relates to the field of medical and health monitoring technology, and in particular to a health sleep model monitoring system and method based on artificial intelligence and biomedical sensing technology. Background Technology

[0002] Medical staff face significant infection risks during patient care, especially when monitoring vital signs. Effectively reducing the risk of infection from contact with patients is a major challenge for healthcare workers. Furthermore, with the increasing prevalence of instrument-induced pressure injuries, the question of how to measure vital signs without physical contact is a pressing issue for critically ill patients, those who are bedridden for extended periods, burn victims, newborns, and those with skin allergies.

[0003] In recent years, some research institutions and manufacturers both domestically and internationally have begun developing portable intelligent vital sign monitors and smart mattresses based on accelerometers, fiber optic sensors, and piezoelectric films. For example, Panasonic's accelerometer module, placed at the head of the bed, can monitor the user's heart rate. Scholars from the University of Science and Technology of China and China's aerospace industry have used piezoelectric film technology to collect data on heart rate, respiration, and body movement while the human body is lying in bed, and perform continuous analysis. Artificial intelligence has also been applied to the field of sleep medicine. For instance, the National University of Singapore has developed a monitoring pad based on the principle of fiber optic micro-bending to pick up heart rate and respiration during sleep. A technology company in Taiwan has also begun embedding fiber optics into mattresses to monitor the heart rate and respiration of the elderly and infants during sleep. This demonstrates that there has been some investment in non-contact vital sign and sleep monitoring both domestically and internationally. Currently, most mattresses under development only collect and analyze data on heart rate, respiration, and body movement, and the accuracy is insufficient, especially when using accelerometers, which have weak anti-interference capabilities. They also lack the ability to construct and analyze user health models and monitor body temperature and blood oxygenation.

[0004] However, when using traditional devices (such as fingertip pulse oximeters and wrist blood pressure monitors), these devices can only acquire instantaneous data and cannot capture dynamic indicators such as heart rate variability (HRV) and respiratory rhythm. For example, a nocturnal sleep apnea event may only last for a few seconds, but traditional devices struggle to continuously record such transient abnormalities. Contact sensors (such as electrode pads) may cause skin damage in newborns with prolonged use, and increase the infection rate of burn wounds; contact sensors are also prone to signal noise when patients turn over or move their limbs. For example, a chest lead electrocardiogram (ECG) may be misinterpreted as premature ventricular contractions (PVCs) when a patient coughs. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the defects of the existing technology. The present invention proposes a health sleep model monitoring system and method based on artificial intelligence and biomedical sensing technology.

[0006] To address the technical problems mentioned above—that traditional devices can only acquire instantaneous data and cannot capture dynamic indicators such as heart rate variability and respiratory rhythm; that contact sensors may cause skin damage in newborns with prolonged use of electrode pads, increasing the infection rate of burn patients; and that contact sensors are prone to signal noise when patients turn over or move their limbs—the technical solution adopted by this invention is: A health sleep model monitoring system based on artificial intelligence and biomedical sensing technology includes: The data acquisition module is used to collect data on the patient's heart rate, respiration, blood oxygen, body temperature, and bed posture. The data acquisition module includes: The heart rate and respiration monitoring unit collects the patient's BCG signal in a non-contact manner. The non-contact method uses a fiber optic micro-bend sensor embedded in the mattress to indirectly reflect the patient's heart rate and respiration status. The blood oxygen and body temperature monitoring unit uses a wireless sensor tag group that is applied to the patient's body surface. The wireless sensor tag group is equipped with an LED blood oxygen sensor and a body temperature sensor to collect and monitor real-time data of the patient's blood oxygen saturation and body temperature. The bed posture monitoring unit uses a symmetrically distributed array of piezoresistive pressure sensors to detect real-time data on the patient's bed status and posture changes. The data processing and analysis module processes and analyzes the data collected by the data acquisition module to determine the patient's vital signs and sleep quality. The data processing and analysis module includes: The signal processing unit performs adaptive filtering and peak detection on the BCG signal acquired by the heart rate and respiration monitoring unit; The health assessment and early warning unit performs a comprehensive health assessment on the data processed by the signal processing unit, constructs a patient health model, and outputs a graded early warning signal.

[0007] Preferably, the data processing and analysis module further includes: The blood oxygen and body temperature processing unit analyzes the data monitored by the blood oxygen and body temperature monitoring unit to obtain the patient's blood oxygen saturation and body temperature information; The posture recognition unit determines the patient's posture in bed and whether the patient is in bed based on the detection data from the bed posture monitoring unit. The sleep analysis unit infers and scores the patient's sleep status based on the patient's heart rate, respiration, and bed posture data collected by the data acquisition module.

[0008] Preferably, the fiber optic sensor is laid in a grid pattern in the middle layer of the mattress. The sensitivity of the fiber optic micro-bend sensor is set to a linear pressure measurement range of 0.1-0.7 kPa, and the sampling rate is set to 500Hz-1000Hz to balance accuracy and dynamic response. The wireless sensor tag group consists of a wireless LED light blood oxygen saturation sensor tag and a wireless body temperature sensor tag.

[0009] Preferably, the pressure sensor array is distributed in a 5×6 matrix, and the sensing area of ​​a single pressure sensor is 10cm×10cm, used to identify the patient's supine, lateral, and sitting postures.

[0010] Preferably, the signal processing unit uses a bandpass filter of 10-200Hz to filter out noise, and the signal-to-noise ratio must be ≥20dB when extracting the heart rate signal using the peak detection algorithm.

[0011] Preferably, the blood oxygen and body temperature processing unit is responsible for receiving and parsing the data transmitted by the wireless LED light blood oxygen saturation sensor tag and the wireless body temperature sensor tag, and the preprocessing process includes data verification and outlier removal steps.

[0012] Preferably, the posture recognition unit uses machine learning methods to analyze the patient's posture changes in real time based on the data collected by the pressure sensor, and the posture recognition unit identifies different pressure distribution patterns through training models.

[0013] Preferably, the sleep analysis unit combines the patient's heart rate, respiration, and bed posture data collected by the acquisition module, and uses a sleep analysis algorithm to infer the patient's sleep state and score it.

[0014] Preferably, the health assessment and early warning unit performs a comprehensive health assessment on the data processed by the data processing and analysis module, constructs a patient health model, compares historical data and real-time data, and issues an early warning signal when abnormalities are detected.

[0015] A method for a health sleep model monitoring system based on artificial intelligence and biomedical sensing technology includes the following steps: S1. Upon first use, perform sensor calibration, network connection settings, and patient information entry. Sensor calibration ensures that the output values ​​of each sensor meet the expected range under standard conditions. Network connection settings configure the parameters of the wireless module to ensure that the system is stably connected to the backend server. Patient information entry includes basic information such as name, age, gender, and medical history. S2. After the system starts, the data acquisition module begins to work, collecting the patient's heart rate, respiration, blood oxygen, body temperature and bed posture data in real time. The collected data is transmitted to the data processing and analysis module for processing and analysis via the wireless module. S3. After receiving the data, the data processing and analysis module first performs data verification and preprocessing to remove outliers and noise interference. Then, each processing unit performs in-depth processing and intelligent analysis on the patient's BCG signal, blood oxygen and body temperature data, and pressure sensor data to finally obtain the patient's vital signs indicators and health assessment results. S4. Finally, the patient's vital signs and health assessment results are displayed in real time on the monitoring interface for medical staff to view. At the same time, by comparing historical data and real-time data, abnormal situations can be detected in a timely manner and warning signals can be issued. Monitoring data is recorded long-term, and medical staff can view historical data by accessing the backend server.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention employs non-contact monitoring technology, using accelerometers, fiber optic micro-bending sensors, and wireless sensor tags to collect vital sign data without direct contact with the patient. This effectively avoids the spread of pathogens due to contact, ensuring the safety of medical staff and patients. It also eliminates the need to attach or connect excessive devices to the patient's body, reducing constraints and interference and allowing the patient to maintain a relatively natural and comfortable state during monitoring. This helps improve the patient's sleep quality and promotes physical recovery. This invention integrates multiple sensors to simultaneously monitor various vital signs, including heart rate, respiration, blood oxygen saturation, body temperature, and bed posture, providing comprehensive and rich data. Simultaneously, advanced signal processing and analysis algorithms accurately process and deeply analyze the collected data, effectively removing noise interference and improving data accuracy. This provides medical personnel with more reliable and valuable diagnostic information. Combined with big data and artificial intelligence technologies, the system can perform real-time health assessments of the monitored data, promptly identifying potential health risks and issuing early warnings. This enables medical personnel to take early intervention measures, controlling diseases in their early stages, improving treatment outcomes, reducing the probability of serious complications, and safeguarding patients' health. Attached Figure Description

[0017] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein: Figure 1 This is a schematic diagram of the overall system of the present invention; Figure 2 This is a schematic diagram of the data acquisition module of the system of the present invention; Figure 3 This is a schematic diagram of the data processing and analysis module of the system of the present invention; Figure 4This is a schematic diagram illustrating the use of the wireless LED light-based blood oxygen saturation sensor tag and the wireless body temperature sensor tag in the system of this invention. Figure 5 This is a schematic diagram of the patient body temperature monitoring system of the present invention; Figure 6 This is a schematic diagram of the patient blood oxygen saturation monitoring system of the present invention; Figure 7 This is a schematic diagram of the mattress sensor arrangement in the system of the present invention; Figure 8 This is a waveform data diagram of the patient's heart rate monitoring over 24 hours using the system of the present invention; Figure 9 This is a waveform diagram of respiratory monitoring data of a patient over 24 hours using the system of the present invention; Figure 10 This is a waveform data diagram of 24-hour body movement monitoring of patients using the system of the present invention; Figure 11 This is a graph showing the patient's heart rate and respiratory statistics in the system of the present invention; Figure 12 This is a schematic diagram of the health model of the system of the present invention.

[0018] Figure reference numerals: 1. Data acquisition module; 101. Heart rate and respiration monitoring unit; 102. Blood oxygen and body temperature monitoring unit; 103. Bed posture monitoring unit; 2. Data processing and analysis module; 201. Signal processing unit; 202. Blood oxygen and body temperature processing unit; 203. Posture recognition unit; 204. Sleep analysis unit; 205. Health assessment and early warning unit; 3. Energy supply module. Detailed Implementation

[0019] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0020] Specific embodiments of the present invention are described below with reference to the accompanying drawings.

[0021] Please see Figure 1 In this embodiment, we innovatively propose a highly integrated and intelligent health sleep model monitoring system. This system integrates cutting-edge artificial intelligence technology with biomedical sensing technology, aiming to provide a comprehensive, accurate, and non-invasive monitoring solution for the medical and health field. The system architecture is mainly divided into three core modules: data acquisition module 1, data processing and analysis module 2, and energy supply module 3. Each module works collaboratively to ensure the efficient and stable operation of the system.

[0022] Please see Figures 2-12 Data acquisition module 1 is responsible for collecting patients' vital signs information comprehensively and from multiple dimensions, specifically including: Heart Rate and Respiration Monitoring Unit 101: This unit innovatively employs a high-sensitivity accelerometer (such as a MEMS accelerometer) or a fiber optic micro-bend sensor (based on the principle of micro-bend loss) embedded in the mattress. The sensitivity of the fiber optic micro-bend sensor is set within a linear pressure measurement range of 0.1-0.7 kPa, and the sampling rate is set at 500Hz-1000Hz to balance accuracy and dynamic response. When the patient lies down, each heartbeat causes subtle changes in local pressure on the mattress, leading to a reduction in the bending radius of the fiber optic sensor and resulting in light intensity attenuation. By monitoring light intensity fluctuations in real time (sampling rate up to 1 kHz), the system can accurately capture these changes, forming a BCG signal (Ballistocardiogram), thereby indirectly reflecting heart rate and respiratory status. This design not only avoids the discomfort that may be caused by traditional contact monitoring but also ensures the continuity and accuracy of the data.

[0023] Based on preliminary experiments and literature review, it was found that the BCG signal was weak and severely affected by 50Hz power frequency interference. Therefore, a common-mode amplifier circuit was used for signal amplification, and a 50Hz power frequency notch filter circuit was used for signal processing to eliminate 50Hz power frequency interference. After data smoothing and filtering to remove noise, the basic data of heart rate, respiration, and body movement were obtained from the BCG ECG signal. The 24-hour heart rate monitoring waveform data, 24-hour respiration monitoring waveform data, and 24-hour body movement monitoring waveform data are shown below. Figure 9 , Figure 10 and Figure 11 As shown.

[0024] The fiber optic sensors are laid out in a grid pattern in the middle layer of the mattress. Specifically, along the length and width of the mattress, the optical fibers are bent into U-shapes or S-shapes at certain intervals (such as 10-15cm) to form a regular grid. This arrangement can cover the main pressure areas when the human body is lying down, ensuring that no matter where the body is lying on the mattress, there are sensor nodes to capture signals generated by cardiac output and respiratory movements.

[0025] Blood oxygen and body temperature monitoring unit 102: such as Figure 4As shown, this unit utilizes wireless LED light-based blood oxygen saturation sensor tags and wireless body temperature sensor tags, gently applied to the patient's skin. The wireless blood oxygen sensor tag has a diameter ≤2cm and is applied to the subclavian region, while the body temperature sensor tag has a thickness ≤1.5mm and is installed in the fourth intercostal space along the anterior axillary line to achieve real-time acquisition of blood oxygen saturation and body temperature. The blood oxygen saturation sensor employs reflective photoelectric technology, accurately calculating blood oxygen saturation by measuring the difference in absorption of specific wavelengths of light by oxyhemoglobin and deoxyhemoglobin in the blood. The body temperature sensor uses a high-precision thermistor or infrared thermometer to ensure the accuracy and reliability of the measurement results. The wireless transmission method further enhances patient comfort and reduces the inconvenience of tangled cables.

[0026] Bed posture monitoring unit 103: In order to fully understand the patient's bed status and posture changes, this unit has several pressure sensors (piezoresistive or piezoelectric) symmetrically distributed in sequence inside the mattress. These sensors have high sensitivity and fast response characteristics. When the patient is in different postures, the pressure distribution of different parts of the body on the mattress will change accordingly.

[0027] like Figure 7 As shown, the mattress incorporates highly sensitive pressure sensors installed in a 5x6 matrix at specific locations (this could be further reduced to 18 sensors in a 3x6 matrix for cost and efficiency considerations). Through BCG signals, mechanical energy is converted into digital signals. A series of signal processing and algorithmic analyses yield data that directly reflects sleep patterns, enabling the collection of respiratory and heart rate data during sleep. The mattress's bottom layer uses breathable, antibacterial, and mite-proof high-quality latex, while the middle layer uses flame-retardant polyurethane foam material with good resilience and comfort. The top and bottom layers consist of multiple layers of fleece, providing dust protection. Patients lying on it experience the same comfort as on a regular bed. Sensors and connecting cables are deployed between the fleece layers.

[0028] Because the mattress has multiple layers of fleece, patients feel no discomfort from the sensors and connecting wires. The mattress's texture is no different from a regular bed. Its power primarily comes from non-contact wireless charging or indoor light energy conversion technology, so users can use it with confidence, without worrying about leakage or electric shock. The mattress sensor module receives signals through sensors, processes them, and sends them to a microcontroller, finally transmitting the data to a server via a wireless module. There are two main types of sensors: one for detecting the user's position in bed, and the other for detecting vital signs such as heart rate and respiration. The sensor for detecting the user's sleep breathing and heart rate uses an accelerometer. It employs a non-contact cardiac impulse scan (BCG) detection method. Unlike most clinical cardiac function testing methods, this method does not require sensors to be directly attached to the body, so even those without systematic training can operate it. The BCG-based detection method does not interfere with the user's rest. The BCG signal is essentially a tiny body vibration caused by the physical movement of the heart, so it can be detected using a high-precision accelerometer to achieve the effect of detecting vital signs such as heart rate. Because the BCG signal is weak and easily interfered with, the most obvious interference being the 50Hz power frequency interference, signal processing is required.

[0029] For example, when a patient lies flat, the head, back, buttocks, and legs each exert pressure on specific areas of the mattress, creating a relatively uniform and stable pressure distribution pattern. When the patient lies on their side, one side of the body bears more concentrated pressure, causing a significant change in pressure distribution. If the patient gets up from the bed, the pressure on the mattress will decrease significantly or even disappear. Pressure sensors can detect these pressure changes in real time and convert them into electrical signals, providing raw data for subsequent posture analysis, determining whether the patient is in bed, and providing important evidence for sleep analysis and health assessment.

[0030] Data processing and analysis module 2 is responsible for in-depth processing and intelligent analysis of the collected raw data, specifically including: Signal Processing Unit 201: For the BCG signal, this unit first amplifies and filters it to eliminate noise interference and extract the heart rate and respiratory signal frequency bands. A bandpass filter is used to effectively filter out high-frequency and low-frequency noise, and a peak detection algorithm is used to accurately extract heart rate and respiratory signals. Simultaneously, adaptive filtering technology is introduced to automatically adjust filtering parameters according to signal characteristics, ensuring accurate extraction of vital sign information under various conditions.

[0031] Among these, the heart rate signal has a relatively high frequency, which manifests as a series of periodic peaks in the BCG signal. The basic principle of the peak detection algorithm is to find local maxima in the signal and calculate the heart rate based on the time intervals between these maxima.

[0032] In practice, the filtered signal is first sampled and digitized, converting it into a discrete digital signal. Then, a suitable threshold is set, and the digital signal is iterated through. When the signal amplitude exceeds the threshold, the position of that point is recorded as a peak. To avoid misidentifying noise spikes as heart rate peaks, auxiliary judgment conditions can be used, such as the duration of peak points and the minimum interval between peaks. Finally, the heart rate value is calculated based on the time interval between adjacent peak points: Heart rate (beats / minute) = 60 / time interval between adjacent peak points (seconds).

[0033] Respiratory signals have a low frequency and appear as relatively slow fluctuations in BCG signals. Similar to heart rate signal extraction, peak detection algorithms can also be used to extract respiratory signals, but the threshold and judgment conditions need to be adjusted according to the characteristics of the respiratory signal. Because the amplitude variation of respiratory signals is relatively small, the threshold setting needs to be more precise; a suitable threshold can be determined through statistical analysis of the signal. After detecting a peak point, the respiratory rate is calculated based on the time interval between adjacent peak points. Furthermore, more information about the respiratory state, such as respiratory depth and rhythm, can be obtained by analyzing the waveform characteristics of the respiratory signal, such as the shape and amplitude variations of peaks and troughs.

[0034] Blood oxygen and body temperature processing unit 202: This unit is responsible for receiving data transmitted from wireless sensor tags, parsing and preprocessing it. The preprocessing process includes data verification and outlier removal to ensure data accuracy and reliability. Through rigorous quality control, a solid data foundation is provided for subsequent health assessments.

[0035] Posture Recognition Unit 203: Based on pressure sensor data, this unit uses threshold-based judgment or machine learning methods to analyze changes in the patient's posture in real time. By training a model to identify different pressure distribution patterns, the system can accurately determine the patient's bed posture and whether they are in bed, providing important reference for sleep analysis and health intervention.

[0036] Threshold-based methods are a relatively intuitive and basic approach to posture recognition. The system pre-sets a series of pressure thresholds related to different postures. For example, to determine if a patient is in bed, a minimum pressure threshold is set. When the total pressure value collected by the pressure sensor is below this threshold, the system determines that the patient is not in bed; otherwise, it considers the patient to be in bed. Further subdivisions of in-bed postures, such as lying flat or on one's side, can also be set with corresponding pressure distribution threshold ranges. For instance, when lying on one's side, the pressure on one side of the body increases significantly. The system determines whether the patient is in a side-lying posture by detecting whether the pressure in a specific area exceeds the threshold set for side-lying, combined with the pressure conditions in other areas.

[0037] Machine learning methods offer another solution for posture recognition. First, a large amount of pressure sensor data from different patients in various postures is collected. This data covers common postures such as lying flat, lying on one's side, prone, and sitting up, as well as data samples from patients of different weights and body types, to ensure the generalization ability of the trained model. Then, this data is preprocessed, including noise removal and normalization, to meet the input requirements of machine learning algorithms. Next, appropriate machine learning algorithms, such as Support Vector Machines (SVM), decision trees, and neural networks, are selected to construct a posture recognition model. Taking a neural network as an example, pressure sensor data is used as the input layer, and features are extracted and transformed through multiple hidden layers. Finally, the posture classification result is given at the output layer. During training, the model is repeatedly trained using labeled data, and the model parameters are adjusted so that the model can accurately map the input pressure data to the corresponding posture category.

[0038] Sleep Analysis Unit 204: This unit combines heart rate, respiration, and body movement data, using sleep analysis algorithms to infer and score the patient's sleep state. By incorporating the patient's actual data, it achieves automatic sleep staging and scoring. Through Sleep Analysis Unit 204, targeted intervention suggestions can be provided to healthcare professionals.

[0039] The sleep analysis algorithm references existing sleep staging standards, such as the R&K standard or the AASM standard. These standards divide the sleep process into different stages, each with unique physiological characteristics. For example, the AASM standard divides sleep into wakefulness (W), non-rapid eye movement (N1, N2, N3) sleep, and rapid eye movement (R) sleep. Different sleep stages exhibit different patterns in heart rate, respiration, and body movement. During wakefulness, heart rate and respiration are relatively fast and irregular, and body movement is more frequent; N1 is a transitional stage of sleep, where heart rate and respiration gradually slow down, and body movement decreases; during N2, heart rate and respiration further stabilize, and body movement is even less; N3 is the deep sleep stage, where heart rate and respiration are the slowest and most regular, and body movement is minimal; during R, heart rate and respiration fluctuate to some extent, and there may be rapid eye movements and slight body movements.

[0040] Health Assessment and Early Warning Unit 205: This unit performs a comprehensive health assessment of the processed data. By constructing a patient health model and comparing historical and real-time data, the system can promptly detect anomalies and issue early warning signals. This proactive health management approach helps healthcare professionals take early intervention measures to control diseases in their early stages.

[0041] like Figure 12As shown, a health model is built using a mattress. This model involves collecting sleep data from an individual over a specific number of consecutive time periods, analyzing it using machine learning and artificial neural networks to determine the user's sleep patterns, heart rate, and breathing patterns, and then creating a specific model for that individual based on this data. In our implementation, we attempted to use a backpropagation (BP) neural network to allow the smart mattress to continuously learn and assess the user's health.

[0042] Based on the data processed by the data processing and analysis module, the health assessment and early warning unit uses artificial intelligence technology to construct a patient health model. This model employs a multi-layered architecture, including a data input layer, a feature extraction layer, a model training layer, and a result output layer. The data input layer receives processed multi-source data. The feature extraction layer uses deep learning algorithms, such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and their variants (e.g., Long Short-Term Memory Networks (LSTM) and Gated Recurrent Units (GRUs), to automatically extract representative and discriminative features from the raw data. These features reflect key information about the patient's health status; for example, heart rate variability reflects the functional state of the autonomic nervous system, and breathing patterns during sleep indicate the presence of sleep-disordered breathing. The model training layer trains the model based on the extracted features and known health labels (e.g., healthy, ill, different stages of disease), continuously adjusting the model's parameters to accurately learn the mapping relationship between data and health status. The result output layer presents the health assessment results obtained from model training in an intuitive way, such as health scores and disease risk probabilities.

[0043] The health assessment and early warning unit continuously monitors patients' real-time data and dynamically compares and analyzes it with a constructed patient health model and historical data. On one hand, it compares the real-time collected vital signs data with the model's predicted normal range to determine if the current data exceeds the normal fluctuation range. For example, if the patient's real-time heart rate is consistently higher than the model's predicted upper limit of normal heart rate, and the duration exceeds a certain threshold, it may indicate a heart rate abnormality. On the other hand, it compares real-time data with the patient's historical data longitudinally to observe data trends. For example, by analyzing changes in the patient's sleep efficiency over a period of time, if a gradual decline in sleep efficiency is found, it may suggest that the patient's sleep quality is deteriorating, requiring further attention and intervention.

[0044] When the system detects an anomaly, such as when a patient's five consecutive respiratory intervals exceed 10 seconds, it triggers a sleep apnea warning and immediately generates an alert signal. The warning signal is graded according to the severity and type of the anomaly, such as mild, moderate, and severe warnings. Different levels of warning signals are delivered to medical staff and relevant personnel in different ways. For mild warnings, the system notifies the patient's family or caregivers via SMS or app push notifications, reminding them to pay attention to the patient's condition. For moderate warnings, in addition to notifying family and caregivers, the system also sends a message to the patient's attending physician so that the doctor can understand the situation and make a preliminary assessment. For severe warnings, the system immediately notifies medical staff via telephone or emergency alerts and activates the emergency response mechanism to ensure the patient receives timely treatment. Furthermore, the warning information includes detailed information such as the specific content of the abnormal data, the time of occurrence, possible cause analysis, and suggested intervention measures, providing comprehensive decision support for medical staff.

[0045] The motion monitoring data obtained from the research, acquired through sensors, still needs further extraction and modeling in the background based on actual needs to intuitively and specifically display the user's sleep status. Sleep analysis involves analyzing data such as heart rate, respiration, and body movement during the user's sleep process at night. Based on specific algorithms and logic, it infers the user's awake, sleep, light sleep, and deep sleep states and assigns scores. It analyzes the extreme values ​​and average values ​​of heart rate and respiration each night. The user's health assessment data, obtained through BP neural network calculations, is uniformly transmitted to a cloud server. The application server uniformly filters, transforms, analyzes, and processes the protocol, and finally stores the structured sleep respiration and heart rate data in the data center.

[0046] Energy supply module 3: Energy supply module 3 employs an indoor light energy harvesting device, which uses high-efficiency polycrystalline solar cells to collect indoor light and convert it into electrical energy. After boosting, maximum power point point compensation, and voltage regulation, it provides stable and reliable power support for the data acquisition module and the data processing and analysis module. The light energy harvesting device is installed above or to the side of the mattress to ensure sufficient indoor light reception. Simultaneously, a rechargeable battery is provided as a backup power source to ensure the system can still operate normally in the absence of light, enabling uninterrupted monitoring around the clock.

[0047] Please continue reading. Figures 4-12 The specific steps for using the system in this embodiment are as follows: S1. System Initialization and Configuration: Upon first use, sensor calibration, network connection settings, and patient information entry are performed. Sensor calibration ensures that the output values ​​of each sensor meet the expected range under standard conditions; network connection settings configure the parameters of the wireless module to ensure a stable connection between the system and the backend server; patient information entry includes basic information such as name, age, gender, and medical history for subsequent health assessments and early warnings.

[0048] S2. Data Acquisition and Transmission: After the system starts, the data acquisition module begins working, collecting real-time data on the patient's heart rate, respiration, blood oxygen, body temperature, and bed posture. The collected data is transmitted wirelessly (Wi-Fi or Bluetooth) to the backend server for processing and analysis.

[0049] S3. Data Processing and Analysis: After receiving the data, the backend server first performs data verification and preprocessing to remove outliers and noise interference. Then, each processing unit performs in-depth processing and intelligent analysis on the BCG signal, blood oxygen and body temperature data, and pressure sensor data to finally obtain the patient's vital signs and health assessment results.

[0050] S4. Results Output and Application: The backend server displays the monitoring results in real time on the monitoring interface for medical staff to view. An apnea warning is triggered if the patient's five consecutive respiratory intervals exceed 10 seconds. Simultaneously, by comparing historical and real-time data, abnormalities are promptly detected and warning signals are issued. Long-term recording of monitoring data provides valuable data support for clinical research and disease prevention. Medical staff can access the backend server to view historical data for a better understanding of the patient's health status.

[0051] The healthy sleep model monitoring system in this embodiment demonstrates significant advantages and value in medical settings: 1. Non-contact monitoring reduces infection risk: Traditional contact-based vital sign monitoring methods are prone to cross-infection, especially when dealing with infectious disease patients, where healthcare workers face an extremely high risk of infection. This embodiment employs non-contact monitoring technology, using accelerometers, fiber optic micro-bending sensors, and wireless sensor tags to collect vital sign data without direct contact with patients, effectively avoiding the spread of pathogens through contact and ensuring the safety of both healthcare workers and patients.

[0052] 2. Improve patient comfort and promote recovery: For special groups such as critically ill patients, long-term bedridden patients, burn victims, newborns, and patients with skin allergies, traditional contact monitoring devices may cause skin damage, allergies, and other discomfort, affecting patient rest and recovery. This embodiment of the system eliminates the need to attach or connect excessive devices to the patient's body, reducing constraints and interference. It allows the patient to maintain a relatively natural and comfortable state during monitoring, helping to improve sleep quality and promote physical recovery.

[0053] 3. Comprehensive and accurate data, providing reliable diagnostic basis: This embodiment integrates multiple sensors, capable of simultaneously monitoring multiple vital signs indicators such as heart rate, respiration, blood oxygen, body temperature, and bed posture, providing comprehensive and rich data. Simultaneously, advanced signal processing and analysis algorithms can accurately process and deeply analyze the collected data, effectively removing noise interference and improving data accuracy, providing medical staff with more reliable and valuable diagnostic basis.

[0054] 4. Real-time health assessment and early warning to improve treatment outcomes: By combining big data and artificial intelligence technologies, the system can conduct real-time health assessments of monitored data, promptly identify potential health risks, and issue early warnings. This enables medical staff to take early intervention measures, control diseases in their early stages, improve treatment outcomes, reduce the probability of serious complications, and safeguard patients' health.

[0055] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A health sleep model monitoring system based on artificial intelligence and biomedical sensing technology, characterized in that, include: The data acquisition module is used to collect data on the patient's heart rate, respiration, blood oxygen, body temperature, and bed posture. The data acquisition module includes: The heart rate and respiration monitoring unit collects the patient's BCG signal in a non-contact manner. The non-contact method uses a fiber optic micro-bend sensor embedded in the mattress to indirectly reflect the patient's heart rate and respiration status. The blood oxygen and body temperature monitoring unit uses a wireless sensor tag group that is applied to the patient's body surface. The wireless sensor tag group is equipped with an LED blood oxygen sensor and a body temperature sensor to collect and monitor real-time data of the patient's blood oxygen saturation and body temperature. The bed posture monitoring unit uses a symmetrically distributed array of piezoresistive pressure sensors to detect real-time data on the patient's bed status and posture changes. The data processing and analysis module processes and analyzes the data collected by the data acquisition module to determine the patient's vital signs and sleep quality. The data processing and analysis module includes: The signal processing unit performs adaptive filtering and peak detection on the BCG signal acquired by the heart rate and respiration monitoring unit; The health assessment and early warning unit performs a comprehensive health assessment on the data processed by the signal processing unit, constructs a patient health model, and outputs a graded early warning signal.

2. The health sleep model monitoring system based on artificial intelligence and biomedical sensing technology according to claim 1, characterized in that, The data processing and analysis module also includes: The blood oxygen and body temperature processing unit analyzes the data monitored by the blood oxygen and body temperature monitoring unit to obtain the patient's blood oxygen saturation and body temperature information; The posture recognition unit determines the patient's posture in bed and whether the patient is in bed based on the detection data from the bed posture monitoring unit. The sleep analysis unit infers and scores the patient's sleep status based on the patient's heart rate, respiration, and bed posture data collected by the data acquisition module.

3. The health sleep model monitoring system based on artificial intelligence and biomedical sensing technology according to claim 2, characterized in that, The fiber optic sensors are laid out in a grid pattern in the middle layer of the mattress. The sensitivity of the fiber optic micro-bend sensor is set to a linear pressure measurement range of 0.1-0.7 kPa, and the sampling rate is set to 500Hz-1000Hz to balance accuracy and dynamic response. The wireless sensor tag group consists of wireless LED light blood oxygen saturation sensor tags and wireless body temperature sensor tags.

4. The health sleep model monitoring system based on artificial intelligence and biomedical sensing technology according to claim 3, characterized in that, The pressure sensor array is arranged in a 5×6 matrix, and the sensing area of ​​each pressure sensor is 10cm×10cm, which is used to identify the patient's supine, lateral, and sitting postures.

5. The health sleep model monitoring system based on artificial intelligence and biomedical sensing technology according to claim 4, characterized in that, The signal processing unit uses a bandpass filter to filter out noise in the 10-200Hz range, and the signal-to-noise ratio must be ≥20dB when extracting the heart rate signal using the peak detection algorithm.

6. The health sleep model monitoring system based on artificial intelligence and biomedical sensing technology according to claim 5, characterized in that, The blood oxygen and body temperature processing unit is responsible for receiving and parsing the data transmitted by the wireless LED light blood oxygen saturation sensor tag and the wireless body temperature sensor tag. The preprocessing process includes data verification and outlier removal steps.

7. The health sleep model monitoring system based on artificial intelligence and biomedical sensing technology according to claim 6, characterized in that, The posture recognition unit uses machine learning methods to analyze the patient's posture changes in real time based on the data collected by the pressure sensor, and the posture recognition unit identifies different pressure distribution patterns through training models.

8. The health sleep model monitoring system based on artificial intelligence and biomedical sensing technology according to claim 7, characterized in that, The sleep analysis unit combines the patient's heart rate, respiration, and bed posture data collected by the acquisition module, and uses a sleep analysis algorithm to infer the patient's sleep state and score it.

9. The health sleep model monitoring system based on artificial intelligence and biomedical sensing technology according to claim 8, characterized in that, The health assessment and early warning unit performs a comprehensive health assessment on the data processed by the data processing and analysis module. By constructing a patient health model and comparing historical and real-time data, it issues an early warning signal when abnormalities are detected.

10. The method of the health sleep model monitoring system based on artificial intelligence and biomedical sensing technology according to any one of claims 1-9, characterized in that, Includes the following steps: S1. Upon first use, perform sensor calibration, network connection settings, and patient information entry. Sensor calibration ensures that the output values ​​of each sensor meet the expected range under standard conditions. Network connection settings configure the parameters of the wireless module to ensure that the system is stably connected to the backend server. Patient information entry includes basic information such as name, age, gender, and medical history. S2. After the system starts, the data acquisition module begins to work, collecting the patient's heart rate, respiration, blood oxygen, body temperature and bed posture data in real time. The collected data is transmitted to the data processing and analysis module for processing and analysis via the wireless module. S3. After receiving the data, the data processing and analysis module first performs data verification and preprocessing to remove outliers and noise interference. Then, each processing unit performs in-depth processing and intelligent analysis on the patient's BCG signal, blood oxygen and body temperature data, and pressure sensor data to finally obtain the patient's vital signs indicators and health assessment results. S4. Finally, the patient's vital signs and health assessment results are displayed in real time on the monitoring interface for medical staff to view. At the same time, by comparing historical data and real-time data, abnormal situations can be detected in a timely manner and warning signals can be issued. Monitoring data is recorded long-term, and medical staff can view historical data by accessing the backend server.

Images