A non-contact vital sign intelligent monitoring system and a monitoring method

Abstract

The present application relates to the field of medical system, especially to a kind of non-contact vital signs intelligent monitoring system and monitoring method.The system includes: non-contact sensor, gathers the original micro-motion signal caused by breathing heartbeat;Edge computing processing module, deployed in local ward, after the pre-processing of original signal, separate out breathing and heart rate waveform, calculate the breathing frequency and heart rate value, and compare with preset threshold to generate judgment result;Central monitoring display module, deployed in nurse station or cloud, receive judgment result and display state identifier, trigger alarm and push to mobile terminal when abnormal.The present application realizes non-contact, continuous automatic monitoring, solves the problem that existing monitoring mode intrudes patient, cannot continuous monitoring, early warning is not timely.

Description

Technical Field

[0001] This invention relates to the field of medical systems, and more particularly to a non-contact intelligent monitoring system and method for vital signs. Background Technology

[0002] In clinical nursing care in hospital wards, the monitoring of vital signs, especially key indicators such as respiration and heart rate at night, has long relied on nurses' regular manual rounds. This method has many insurmountable technical shortcomings, becoming a core pain point in clinical nursing. Manual rounds require medical staff to enter the ward and turn on the lighting, inevitably disturbing patients' sleep and disrupting their rest rhythm, which is detrimental to postoperative recovery and disease progression. Furthermore, regular rounds are a discrete monitoring method, with long monitoring gaps between rounds, making continuous monitoring of vital signs impossible. If a patient experiences respiratory depression, apnea, or abnormal heart rate during these gaps, it may be difficult to detect in time, potentially leading to serious medical accidents. In addition, the monitoring results of manual rounds are highly dependent on the nurse's personal experience and attention. When medical staff are fatigued or overworked, misjudgments and omissions of vital signs are more likely, compromising the objectivity and accuracy of the monitoring results. Moreover, the large amount of repetitive rounds occupies nurses' core working time, preventing them from focusing their energy on patients requiring urgent or complex care, significantly reducing the efficiency of nursing work. While existing contact-based vital signs monitors can provide continuous monitoring, they suffer from high costs, numerous wiring connections, and significant discomfort for patients. Furthermore, these devices are only suitable for intensive care units such as ICUs and cannot meet the needs of patients in general wards who require a quiet, undisturbed rest environment, hindering their widespread adoption in ordinary wards. Therefore, the clinical nursing field urgently needs a monitoring system that can continuously, automatically, and reliably monitor patients' vital signs such as respiration and heart rate without disturbing their rest, and can promptly issue warning signals when abnormalities occur. Summary of the Invention

[0003] This invention provides a non-contact intelligent monitoring system and method for vital signs, aiming to solve the problems of existing monitoring methods that cannot simultaneously achieve non-invasiveness and continuous automatic monitoring, and that the monitoring results lack objectivity and accuracy and are not timely in warning.

[0004] To achieve the above objectives, the following technical solution is adopted.

[0005] A non-contact intelligent vital sign monitoring system includes: At least one non-contact sensor, said at least one non-contact sensor being used to acquire raw micro-motion signals caused by the patient's breathing and heartbeat; An edge computing processing module connected to the at least one non-contact sensor, the edge computing processing module being deployed locally in the ward, the edge computing processing module receiving the original micro-motion signal and preprocessing the original micro-motion signal to obtain a purified micro-motion signal, the edge computing processing module performing signal separation on the purified micro-motion signal to obtain respiratory waveform data and heart rate waveform data, the edge computing processing module calculating a respiratory rate value based on the respiratory waveform data and a heart rate value based on the heart rate waveform data, the edge computing processing module comparing the respiratory rate value and the heart rate value with a preset safety threshold range to generate a local judgment result; The central monitoring and display module is communicatively connected to the edge computing processing module. The central monitoring and display module is deployed at the nurse station or cloud server. The central monitoring and display module receives the local judgment result and generates a corresponding status identifier on the display interface. When the local judgment result is an abnormal status identifier, the central monitoring and display module triggers the corresponding alarm signal and pushes the alarm signal to the designated mobile terminal.

[0006] Optionally, the at least one non-contact sensor is a millimeter-wave radar sensor, which is installed on the ceiling or headboard above the patient's bed and the detection area of ​​the millimeter-wave radar sensor is aligned with the patient's torso area. The millimeter-wave radar sensor emits electromagnetic waves with a wavelength of 4mm to 12mm and receives echo signals reflected from the human body surface. The edge computing processing module performs phase demodulation on the echo signals to extract the chest cavity micro-displacement information caused by breathing and heartbeat as the original micro-displacement signal. Alternatively, the at least one non-contact sensor is a high-sensitivity piezoelectric thin-film sensor, which is laid under the mattress with its sensing surface in contact with the underside of the mattress. The high-sensitivity piezoelectric thin-film sensor senses minute pressure changes caused by the patient's breathing and heartbeat and generates an electrical charge signal. The edge computing processing module amplifies and converts the electrical charge signal to obtain the original micro-motion signal.

[0007] Optionally, when the edge computing processing module performs signal separation on the pure micro-motion signal, the edge computing processing module first performs empirical mode decomposition on the pure micro-motion signal to obtain multiple intrinsic mode function components. The edge computing processing module calculates the sample entropy value of each intrinsic mode function component. The edge computing processing module reconstructs the respiratory waveform data into the intrinsic mode function components whose sample entropy value is less than a first threshold, and reconstructs the heart rate waveform data into the intrinsic mode function components whose sample entropy value is greater than a second threshold, wherein the first threshold is less than the second threshold.

[0008] Optionally, when the edge computing processing module calculates the respiratory frequency value based on the respiratory waveform data, the edge computing processing module first performs wavelet transform on the respiratory waveform data to obtain a respiratory signal time-frequency graph. The edge computing processing module extracts the main frequency band variation curve over time from the respiratory signal time-frequency graph. The edge computing processing module performs peak detection on the variation curve and calculates the time interval between adjacent peaks. The edge computing processing module generates the respiratory frequency value based on the reciprocal of the time interval. When the edge computing processing module calculates the heart rate value based on the heart rate waveform data, the edge computing processing module first performs autocorrelation analysis on the heart rate waveform data to obtain an autocorrelation function. The edge computing processing module extracts the time delay corresponding to the first maximum point other than zero from the autocorrelation function. The edge computing processing module generates the heart rate value based on the reciprocal of the time delay.

[0009] Optionally, in the step where the edge computing processing module compares the respiratory rate value and the heart rate value with a preset safety threshold range, the edge computing processing module first obtains the current patient's historical average respiratory rate and historical average heart rate. Based on the historical average respiratory rate, the edge computing processing module dynamically adjusts the upper and lower limits of the safety threshold range to generate a personalized safety threshold range. The edge computing processing module compares the real-time calculated respiratory rate value and heart rate value with the personalized safety threshold range one by one. When the results of three consecutive comparisons all exceed the personalized safety threshold range, the edge computing processing module generates the abnormal state identifier.

[0010] A non-contact intelligent monitoring method for vital signs includes the following steps: Raw micro-motion signals caused by the patient's breathing and heartbeat are collected using non-contact sensors; The edge computing processing module receives the original micro-motion signal and preprocesses it to obtain a clean micro-motion signal. The edge computing processing module performs signal separation on the pure micro-motion signal to obtain respiratory waveform data and heart rate waveform data. The edge computing processing module calculates the respiratory rate value based on the respiratory waveform data and the heart rate value based on the heart rate waveform data. The edge computing processing module compares the respiratory rate value and the heart rate value with a preset safety threshold range to generate a local judgment result containing a normal state identifier or an abnormal state identifier. The edge computing processing module sends the local judgment result to the central monitoring and display module. The central monitoring and display module receives the local judgment result and generates the corresponding status identifier on the display interface. When the local judgment result is an abnormal status identifier, the corresponding alarm signal is triggered through the central monitoring and display module and the alarm signal is pushed to the designated mobile terminal.

[0011] Optionally, in the step of acquiring the original micro-motion signal caused by the patient's breathing and heartbeat through a non-contact sensor, when a millimeter-wave radar sensor is used as the non-contact sensor, it includes emitting electromagnetic waves with a wavelength of 4mm to 12mm through the millimeter-wave radar sensor and receiving the echo signal reflected from the human body surface, and performing phase demodulation on the echo signal through the edge computing processing module to extract the chest cavity micro-motion displacement information caused by breathing and heartbeat as the original micro-motion signal; when a high-sensitivity piezoelectric thin-film sensor is used as the non-contact sensor, it includes sensing the minute pressure changes caused by the patient's breathing and heartbeat through the high-sensitivity piezoelectric thin-film sensor and generating a charge signal, and performing charge amplification and voltage conversion on the charge signal through the edge computing processing module to obtain the original micro-motion signal.

[0012] Optionally, the step of performing signal separation on the pure micro-motion signal through the edge computing processing module to obtain respiratory waveform data and heart rate waveform data includes performing empirical mode decomposition on the pure micro-motion signal through the edge computing processing module to obtain multiple intrinsic mode function components, calculating the sample entropy value of each intrinsic mode function component through the edge computing processing module, reconstructing the respiratory waveform data by the intrinsic mode function components whose sample entropy value is less than a first threshold through the edge computing processing module, and reconstructing the heart rate waveform data by the intrinsic mode function components whose sample entropy value is greater than a second threshold through the edge computing processing module, wherein the first threshold is less than the second threshold.

[0013] Optionally, the step of calculating the respiratory frequency value based on the respiratory waveform data and the heart rate value based on the heart rate waveform data by the edge computing processing module includes: performing wavelet transform on the respiratory waveform data to obtain a respiratory signal time-frequency graph by the edge computing processing module; extracting the main frequency band change curve over time from the respiratory signal time-frequency graph by the edge computing processing module; performing peak detection on the change curve and calculating the time interval between adjacent peaks by the edge computing processing module; generating the respiratory frequency value based on the reciprocal of the time interval by the edge computing processing module; performing autocorrelation analysis on the heart rate waveform data to obtain an autocorrelation function by the edge computing processing module; extracting the time delay corresponding to the first maximum point other than zero from the autocorrelation function by the edge computing processing module; and generating the heart rate value based on the reciprocal of the time delay by the edge computing processing module.

[0014] Optionally, the step of comparing the respiratory rate value and the heart rate value with a preset safety threshold range through the edge computing processing module to generate a local judgment result containing a normal state identifier or an abnormal state identifier includes obtaining the current patient's historical average respiratory rate and historical average heart rate through the edge computing processing module; dynamically adjusting the upper and lower limits of the safety threshold range based on the historical average respiratory rate through the edge computing processing module to generate a personalized safety threshold range; and successively comparing the real-time calculated respiratory rate value and heart rate value with the personalized safety threshold range through the edge computing processing module. When three consecutive comparisons are made... When all comparison results exceed the personalized security threshold range, the edge computing processing module generates the abnormal status identifier. The step of triggering the corresponding alarm signal through the central monitoring display module and pushing the alarm signal to the designated mobile terminal includes identifying whether the current system time is in a preset nighttime silent period when the abnormal status identifier is generated. If the current system time is in the nighttime silent period, the central monitoring display module only triggers the flashing of the alarm light without triggering the sound alarm of the audible and visual alarm, and pushes the alarm information corresponding to the abnormal status identifier to the vibrating call device or smart bracelet carried by the nurse.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This monitoring system collects raw micro-motion signals caused by the patient's breathing and heartbeat through non-contact sensors. Combined with a locally deployed edge computing module in the ward, it performs signal preprocessing, separation, and vital sign calculation. This forms a collaborative monitoring architecture with the central monitoring display module, enabling continuous automatic monitoring of the patient's vital signs without contact or intrusion. This completely solves the problems of existing monitoring methods that cannot simultaneously achieve non-invasiveness and continuous automatic monitoring, and that lack objective accuracy and timely warnings. It eliminates the disturbance to the patient's rest caused by manual rounds and fills the monitoring gaps during manual rounds through 24 / 7 signal acquisition and analysis, allowing abnormalities in the patient's vital signs to be detected promptly. The targeted application and signal processing methods of millimeter-wave radar sensors and piezoelectric film sensors further adapt to the actual use scenario in the ward. The radar sensor can penetrate thin blankets and is not dependent on light, while the piezoelectric film sensor is laid under the mattress without any wearing sensation. Both sensing methods can accurately extract effective raw signals, ensuring the stability and adaptability of signal acquisition. Based on empirical mode decomposition and sample entropy value-based signal separation, precise separation of respiratory and heart rate waveform data is achieved, avoiding interference from signal mixing in vital sign calculations. The separate application of wavelet transform and autocorrelation analysis further enhances the accuracy and real-time performance of respiratory rate and heart rate calculations. Combining personalized safety threshold ranges based on patients' historical vital sign averages with anomaly judgment rules based on three consecutive comparisons makes the judgment of abnormal vital signs more closely aligned with individual patient characteristics, effectively reducing the probability of false alarms and missed alarms and improving the scientific rigor of the judgment results. Simultaneously, the system can identify quiet nighttime periods and execute corresponding silent alarm strategies, providing early warnings only through flashing alarm lights and vibration push notifications from personal devices. This ensures timely transmission of abnormal information while avoiding disturbance to patients in the ward at night. Targeted push notifications of alarm signals to designated mobile terminals allow medical staff to quickly locate abnormal beds and handle them promptly, significantly improving the response efficiency of nursing work. This frees medical staff from mechanical, timed rounds, allowing them to proactively address abnormal warnings and significantly improve the safety and efficiency of clinical nursing. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the module structure of an embodiment of a non-contact intelligent vital signs monitoring system of the present invention.

[0017] Figure 2 This is a schematic diagram of the steps of a non-contact intelligent monitoring method for vital signs according to the present invention. Detailed Implementation

[0018] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0019] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0020] like Figure 1 As shown, the non-contact intelligent vital sign monitoring system of this embodiment is suitable for ward scenarios such as general wards, geriatric departments, and psychiatric departments in hospitals that require low-interference continuous vital sign monitoring. It can achieve non-contact, highly accurate, continuous real-time monitoring of patients' respiratory rate and heart rate, and simultaneously complete auxiliary monitoring of patients' physical activity intensity and bed-off status. When patients' vital signs are abnormal, it can provide graded, targeted, and intelligent early warning prompts. The system consists of non-contact sensors, edge computing processing modules, and central monitoring and display modules. The modules establish a stable two-way data interaction connection through a dedicated encrypted communication link for the hospital, forming a collaborative monitoring architecture of local sensing and acquisition, edge intelligent processing, and centralized display and early warning. The system has undergone targeted innovation and optimization in hardware selection, deployment methods, and algorithm design. It completely avoids the inherent defects of traditional manual inspection and contact monitoring equipment. Through multi-dimensional anti-interference design, personalized monitoring design, and intelligent early warning design, it improves the accuracy of monitoring and the timeliness of early warning. The hardware composition, deployment logic, functional implementation, and innovative processing methods of each module are all designed around the actual monitoring needs of the ward. It can achieve full automation of vital sign monitoring without disturbing the patient's rest. At the same time, all technical designs are based on the actual needs of clinical monitoring, with no redundant design, which facilitates large-scale deployment and application in various wards.

[0021] As the core signal acquisition unit at the front end of the system, the number of non-contact sensors is independently configured in a one-to-one correspondence with the number of beds in the ward. Each bed area is equipped with a separate non-contact sensor to collect the raw micro-motion signals generated by the patient's breathing and heartbeat. The selection of sensors can be flexibly chosen based on the ward's layout, the patient's physical characteristics, and the specific needs of clinical monitoring, either millimeter-wave radar sensors or high-sensitivity piezoelectric film sensors. Both types of sensors adopt a non-contact, non-sensory signal acquisition design, requiring no additional equipment worn by the patient and causing no interference to the patient's sleep, turning over, or limb movements. Furthermore, through innovative hardware structure design and optimized signal acquisition logic, the accuracy and anti-interference capability of the raw micro-motion signal acquisition are effectively improved, ensuring the effectiveness of subsequent data processing from the source of signal acquisition.

[0022] If a millimeter-wave radar sensor is selected, it employs a miniaturized, low-power, frequency-modulated continuous wave hardware design with an integrated, sealed structure. This allows it to adapt to the temperature and humidity environment of hospital wards. The installation location is chosen to be the ceiling above the patient's bed or the wall beside the headboard, maintaining a reasonable vertical distance from the bed surface. During installation, the sensor's built-in angle adjustment structure allows for precise calibration of the detection angle, ensuring the sensor's detection beam is accurately aimed at the patient's torso area. Furthermore, the narrow beamforming design ensures precise detection only of the patient's torso area, effectively avoiding signal interference from other objects and personnel in the ward. This millimeter-wave radar sensor emits electromagnetic waves with wavelengths from 4mm to 12mm. These waves can easily penetrate thin blankets, sheets, cotton clothing, and other flexible coverings, enabling the acquisition of micro-motion signals without direct contact with the human body. The acquisition process is completely independent of ambient light, maintaining stable acquisition results even in low-light or no-light environments such as when lights are off or curtains are drawn. When the electromagnetic waves emitted by the sensor come into contact with the human body surface, they are reflected by the patient's breathing and heartbeat. The echo signal carries core information such as the displacement and frequency of the micro-movements in the chest and abdominal cavities. The sensor transmits the real-time collected echo signal to the edge computing processing module via encrypted wireless communication. The edge computing processing module performs precise phase demodulation processing on the echo signal and extracts the chest cavity micro-movement displacement information caused only by the patient's breathing and heartbeat through demodulation algorithms. This displacement information is the original micro-movement signal required for subsequent system processing. During the demodulation process, stray phase information caused by electromagnetic wave reflection is automatically filtered out to improve the purity of the original micro-movement signal.

[0023] If a high-sensitivity piezoelectric thin-film sensor is selected, it employs a flexible array hardware structure design, consisting of multiple high-sensitivity piezoelectric sensing units arranged in a matrix. The overall design offers excellent flexibility and conformability, perfectly adapting to mattresses of varying thicknesses and curvatures. The sensor is directly placed under the patient's mattress, with its sensing surface seamlessly integrated with the mattress's underside. Its placement precisely corresponds to the area below the patient's chest and abdomen when lying flat, ensuring comprehensive sensing of minute pressure changes caused by the patient's breathing and heartbeat. Each piezoelectric sensing unit has independent signal acquisition capabilities and is equipped with a temperature and humidity drift compensation circuit. This automatically offsets the signal drift caused by temperature and humidity changes in the hospital ward, ensuring the stability of pressure signal acquisition. This sensor acquires signals based on the piezoelectric effect. When a patient breathes and their heart beats, the minute fluctuations in the chest and abdominal cavities create periodic, minute pressure changes on the mattress. These pressure changes are transmitted to the sensing surface of the piezoelectric thin-film sensor, causing the piezoelectric material inside the sensor to polarize. This generates a charge signal that matches the patient's breathing and heartbeat patterns. The amplitude and frequency changes of this charge signal are completely synchronized with the pressure changes in the patient's chest and abdominal cavities. The sensor transmits the real-time generated charge signal to the edge computing processing module via a shielded wired link to avoid signal interference from wireless transmission. The edge computing processing module first performs low-noise charge amplification on the charge signal, amplifying the weak charge signal to a recognizable range. A low-noise operational amplifier is used during amplification to effectively suppress noise introduction. Subsequently, the amplified charge signal undergoes precise voltage conversion, converting the charge signal into a continuous voltage signal. This voltage signal is the original micro-motion signal required for subsequent system processing.

[0024] As the core data processing unit of the system, the edge computing processing module adopts a high-performance embedded AI hardware architecture, integrating a dedicated signal processing chip, an embedded AI computing chip, a local data storage chip, a multi-protocol communication module, and a power management module. The overall design is a miniaturized, wall-mounted processing box structure, deployed locally in the ward. The specific installation location can be selected in areas that do not affect the activities of medical staff and patients, such as the ward entrance or a corner of the ward. One edge computing processing module is configured for each ward. This module establishes a two-way communication connection with all non-contact sensors in the ward through a dedicated encrypted communication link in the hospital, and at the same time establishes a stable encrypted communication connection with the central monitoring and display module through the hospital intranet. The core function of the edge computing processing module is to perform intelligent, high-precision processing of the raw micro-motion signals transmitted by non-contact sensors throughout the entire process, and to complete the real-time calculation of vital signs and the local judgment of abnormal states. All data processing is completed locally in the ward, effectively reducing the latency caused by data transmission and improving the real-time performance of monitoring and early warning. The specific processing process is divided into four consecutive and interconnected steps: signal preprocessing, signal separation, vital sign calculation, and threshold comparison and judgment. Each step adopts innovative algorithm design and processing logic, which effectively improves the accuracy and anti-interference ability of data processing. At the same time, auxiliary analysis of patient movement and bed-off status is completed simultaneously during the data processing.

[0025] In the signal preprocessing stage, the edge computing processing module receives the raw micro-motion signals transmitted by all non-contact sensors in the corresponding ward in real time through the communication module. During the acquisition and transmission of the raw micro-motion signals, electromagnetic interference from the environment, slight vibrations of the equipment, and noise signals caused by slight movement of people in the ward are inevitably mixed in. At the same time, the raw micro-motion signals may also have baseline drift problems. These factors will directly affect the accuracy of subsequent signal separation and vital sign calculation. Therefore, the edge computing processing module performs multi-dimensional fusion preprocessing on the received raw micro-motion signals. The preprocessing process includes three innovative sub-steps: baseline drift correction, adaptive filtering, and multi-scale wavelet denoising. The three sub-steps are executed continuously to optimize the raw micro-motion signals layer by layer. First, a baseline drift correction sub-step is executed. An adaptive polynomial fitting algorithm is used to perform baseline fitting on the original micro-motion signal, extracting the baseline drift trend and performing reverse compensation to completely eliminate the baseline drift problem in the original micro-motion signal, allowing the signal waveform to return to the standard baseline. Next, an adaptive filtering sub-step is executed. An adaptive filtering algorithm based on minimum mean square error is used to filter the corrected signal. This algorithm dynamically adjusts the filtering parameters according to the frequency characteristics and amplitude changes of the original micro-motion signal, accurately identifying and removing noise signals unrelated to respiration and heartbeat, retaining only the effective signal components related to the patient's respiration and heartbeat. Finally, a multi-scale wavelet denoising sub-step is executed. A multi-scale wavelet decomposition algorithm is used to denoise the filtered signal, decomposing the signal into wavelet spaces of different scales. Noise coefficients are threshold-quantized in each scale space, and then reconstructed using wavelets to obtain the denoised signal, effectively suppressing random noise and making the signal waveform smoother. Through the fusion preprocessing of the above three sub-steps, a clean micro-motion signal with complete removal of baseline drift, clutter signals and random noise is finally obtained, providing a high-quality signal foundation for subsequent signal separation steps.

[0026] In the signal separation stage, the edge computing processing module performs high-precision signal separation on the pure micro-motion signal, extracting the respiratory waveform data and heart rate waveform data mixed in the same signal completely independently to avoid mutual interference. This stage adopts an algorithm framework that combines empirical mode decomposition with sample entropy analysis, and on this framework, two innovative optimizations, endpoint effect suppression and component correlation verification, have been carried out, which greatly improves the accuracy of signal separation. First, the edge computing processing module performs Empirical Mode Decomposition (EMD) on the pure micro-motion signal. Before decomposition, the two ends of the pure micro-motion signal are extended using the mirror extension method to effectively suppress the endpoint effect that is prone to occur during EMD and avoid distortion of the decomposition results. Then, EMD is performed on the extended signal to decompose the continuous pure micro-motion signal into multiple intrinsic mode function (IMF) components with different time scales and frequency characteristics. Each IMF component corresponds to an independent frequency component in the signal. The micro-motion signal corresponding to the patient's respiration has a lower frequency and a smoother waveform change, corresponding to the low-frequency IMF component. The micro-motion signal corresponding to the patient's heartbeat has a higher frequency and a more rapid waveform change, corresponding to the high-frequency IMF component. Next, the edge computing processing module calculates the sample entropy value of each IMF component. The sample entropy value can accurately reflect the complexity and irregularity of the signal. The component corresponding to the respiration waveform has a smooth waveform change and strong regularity, with a smaller sample entropy value. The component corresponding to the heart rate waveform has a rapid waveform change and relatively weak regularity, with a larger sample entropy value. The system incorporates two distinct sample entropy thresholds: a first threshold for determining the components corresponding to the respiratory waveform and a second threshold for determining the components corresponding to the heart rate waveform. The first threshold is less than the second. The edge computing module performs preliminary reconstruction on the intrinsic mode function (IMF) components with sample entropy values ​​less than the first threshold to obtain preliminary respiratory waveform data, and on the same module, it performs preliminary reconstruction on the IMF components with sample entropy values ​​greater than the second threshold to obtain preliminary heart rate waveform data. Finally, the system verifies the component correlation of the pre-reconstructed respiratory and heart rate waveform data by calculating the correlation coefficients between the two preliminary waveform data and the original pure micro-motion signal. If the correlation coefficients do not meet the preset correlation standard, the sample entropy thresholds are readjusted and reconstruction is performed until the correlation coefficients reach the standard, ultimately yielding completely independent and highly accurate respiratory and heart rate waveform data. While separating respiratory and heartbeat waveform data, the edge computing processing module also performs auxiliary analysis on the amplitude changes and signal continuity of pure micro-motion signals. By monitoring the changes in signal amplitude in real time, it judges the intensity of the patient's physical activity and realizes graded monitoring of body movement status. By monitoring whether effective micro-motion signals have not been collected for a long time, it judges the patient's out-of-bed status. The analysis results of body movement and out-of-bed status are transmitted to the central monitoring and display module along with respiratory and heart rate data.

[0027] In the vital sign value calculation stage, the edge computing processing module calculates the real-time respiratory rate and heart rate values ​​based on the separated respiratory waveform data and heart rate waveform data. This stage adopts an adapted algorithm framework and has been innovatively optimized for the different characteristics of respiratory and heart rate waveform data. The respiratory rate value calculation adopts an algorithm combining wavelet transform and peak detection, while the heart rate value calculation adopts an algorithm combining autocorrelation analysis and extreme point extraction. In addition, an innovative sub-step for body motion interference removal has been added to the calculation of both vital sign values ​​to ensure the accuracy and real-time performance of the calculation results. The calculation of respiratory rate and heart rate values ​​is executed synchronously and independently, and the operation cycle is controlled within a short time to ensure that subtle changes in the patient's vital signs can be captured in real time. For calculating the respiratory rate, the separated respiratory waveform data is first subjected to body motion interference removal. When the detected body motion amplitude exceeds a preset body motion threshold, the calculation of the respiratory rate is temporarily paused until the body motion amplitude recovers below the threshold and the signal stabilizes, thus avoiding calculation errors caused by body motion. Subsequently, wavelet transform processing is performed on the respiratory waveform data without body motion interference to convert the one-dimensional time-domain respiratory waveform data into a two-dimensional respiratory signal time-frequency graph. The time-frequency graph can clearly reflect the frequency change of the respiratory signal over time, avoiding the limitations of single time-domain analysis. Next, the curve of the main frequency band changing over time is extracted from the respiratory signal time-frequency graph. The extraction process uses an adaptive main frequency band identification algorithm, which can automatically identify and extract the core main frequency band of the respiratory signal at different time periods, avoiding the main frequency band extraction error caused by slight changes in the patient's respiratory rhythm. Then, peak detection is performed on the curve. An adaptive peak detection algorithm is used to identify all characteristic peaks in the curve and automatically remove false peaks caused by small signal fluctuations. Finally, the time interval between two adjacent characteristic peaks is calculated. This time interval is one respiratory cycle of the patient. The reciprocal of this time interval is used to obtain the real-time respiratory rate value. For heart rate calculation, body movement interference is first eliminated to avoid calculation errors caused by body movement. Then, autocorrelation analysis is performed on the heart rate waveform data without body movement interference. The autocorrelation operation eliminates random interference in the heart rate signal, highlights the periodic characteristics of the heart rate signal, and obtains the autocorrelation function of the heart rate waveform data. During the operation, a noise threshold filtering algorithm is used to automatically filter out noise components in the autocorrelation function. Next, the first maximum point other than the zero point is extracted from the autocorrelation function. The extraction process uses an extreme point precise positioning algorithm to ensure accurate identification of the maximum point. The time delay corresponding to this maximum point is one heartbeat cycle of the patient. Calculating the reciprocal of this time delay and completing the unit conversion yields the real-time heart rate value.

[0028] In the threshold comparison and judgment stage, the edge computing processing module accurately compares the real-time calculated respiratory rate and heart rate values ​​with the preset safety threshold range, generating a local judgment result. This local judgment result is divided into two types: normal state indicator and abnormal state indicator. To make the threshold comparison more closely reflect the individual physiological characteristics of patients and avoid misjudgments and missed judgments caused by uniform thresholds, this stage incorporates three innovative optimizations based on the personalized safety threshold range design: sliding window mean statistics, dynamic threshold updates, and continuous comparison verification. These improvements make the judgment of abnormal states more scientific and accurate. First, the local data storage chip of the edge computing processing module continuously and in real-time records the respiratory rate and heart rate monitoring data of each patient. A sliding window statistical method is used to process historical monitoring data. The window duration can be set according to clinical needs. The sliding window statistics obtain the historical average respiratory rate and historical average heart rate of the current patient under normal physiological conditions. This average reflects the patient's baseline vital signs in real time, avoiding mean deviations caused by single historical data anomalies. Subsequently, the edge computing processing module dynamically adjusts the upper and lower limits of the preset general safety threshold range based on the historical average respiratory rate and historical average heart rate, combined with the safety standards of clinical monitoring. This generates a personalized safety threshold range that adapts to the individual physiological characteristics of the current patient. For patients with weaker constitutions and lower baseline heart rates, the lower limit of the heart rate threshold will be adjusted appropriately. For patients with poor lung function and slower baseline respiratory rates, the lower limit of the respiratory rate threshold will be adjusted appropriately. This personalized safety threshold range is dynamically updated. After each statistical cycle, the system recalculates the average and adjusts the threshold range based on new historical monitoring data to ensure that the threshold range always matches the patient's actual physiological condition. After setting the personalized safety threshold range, the edge computing processing module compares the real-time calculated respiratory rate and heart rate values ​​with this personalized safety threshold range in real time. Each comparison result is marked and stored. If a single comparison result exceeds the threshold range, the system does not generate an abnormal status indicator, but only marks and stores the data. If three consecutive comparison results exceed the personalized safety threshold range, the patient's vital signs are determined to be in an abnormal state, and the edge computing processing module immediately generates an abnormal status indicator. If the comparison results remain within the personalized safety threshold range, a normal status indicator is generated. After the local judgment result is generated, the edge computing processing module integrates the respiratory rate, heart rate, body movement and bed abduction monitoring data, and the corresponding status indicators, and transmits them in real time to the central monitoring and display module via the hospital intranet.

[0029] The central monitoring display module is the core unit for system display and early warning. It can be flexibly deployed at hospital nurse stations or cloud servers. If deployed at a nurse station, it adopts a hardware architecture combining a central monitoring screen with a local data server. The central monitoring screen is a high-definition touch screen that supports multi-touch operation, enabling visual display of data and manual interaction. The local data server is used to store monitoring data for the ward or floor. If deployed on a cloud server, it adopts a distributed architecture of cloud storage and cloud computing, enabling unified management and processing of monitoring data across wards, floors, and hospital areas. It also supports remote access and operation by medical staff through various mobile terminals. The central monitoring and display module establishes a stable encrypted communication connection with the edge computing processing modules of all wards through the hospital's intranet. It can receive real-time vital sign monitoring data, body movement and bed alighting monitoring data, and local judgment results from patients in each ward and bed. It also stores the received data in real time, displays it intelligently and visually, and provides tiered abnormality warnings. It also has functions such as historical data query, intelligent statistical analysis, and automatic report generation, providing comprehensive and objective data analysis support for medical staff to judge the condition and carry out nursing work. Its core functions have been innovatively optimized, including intelligent panoramic display, tiered intelligent warning, alarm information tracing, and intelligent data management, which effectively improves the efficiency of clinical monitoring.

[0030] The central monitoring display module features an intelligent panoramic view design, simultaneously displaying monitoring information for all intensive care units and patients in all beds within the hospital. It supports intelligent clustering by ward, floor, and disease risk level, allowing medical staff to switch and filter viewing angles via touch controls. Each bed has an independent display area, showing the patient's name, bed number, current respiratory rate, current heart rate, body movement status, and corresponding status indicator in real time. The status indicator uses color differentiation: green for normal status and red for abnormal status. A yellow warning indicator is displayed if a patient's vital signs are approaching personalized safety thresholds but not yet reaching abnormal standards. All indicators are dynamically updated based on real-time data transmitted from the edge computing processing module, enabling medical staff to quickly and intuitively grasp the vital signs of all patients. After receiving the local judgment results from the edge computing processing module, the central monitoring display module generates and updates the status indicator in the corresponding bed's display area in real time, while simultaneously storing the corresponding vital sign data to a local or cloud server.

[0031] When the local judgment result received by the central monitoring display module is an abnormal status indicator, it will immediately trigger the corresponding alarm signal and accurately push the alarm signal to the designated medical staff's mobile terminal. This stage adopts an innovative early warning logic of intelligent priority scheduling, nighttime silent adaptation, and targeted alarm information push + source tracing, which not only ensures the timely transmission of alarm information, but also avoids disturbing the patient's rest, while providing medical staff with comprehensive abnormal information support. The alarm signal triggering process first involves intelligent priority scheduling. The system presets different alarm priorities based on the type of abnormal signs. For example, severe abnormalities such as apnea and sudden changes in heart rate have higher priority than general abnormalities such as slightly slower respiratory rate and slightly faster heart rate. Higher priority abnormalities will trigger stronger alarm prompts and will be prominently displayed on the central monitoring screen. Subsequently, the system identifies whether the current system time is within the preset nighttime quiet period. This nighttime quiet period can also be adaptively adjusted according to the actual rest status of the patients in the ward. If most patients in the ward are asleep, the system will automatically extend the nighttime quiet period. If the current system time is within the nighttime quiet period, the central monitoring display module will only trigger the flashing of the red alarm light and disable the sound alarm function of the audible and visual alarm to avoid disturbing the patients' rest in the ward. If the current system time is not within the nighttime quiet period, the central monitoring display module will trigger both the flashing of the red alarm light and a continuous sound alarm to remind the medical staff at the nursing station to pay attention in a timely manner. The alarm signal is pushed out via a targeted push method. The system can bind the mobile terminals of medical staff to specific beds. Each ward or bed corresponds to a designated on-duty nurse's mobile terminal. Mobile terminals include nurses' PDAs, smartphones, smart bracelets, vibration call devices, etc. When an abnormal status is generated in a bed, the central monitoring and display module will accurately push the abnormal alarm information of that bed to the bound mobile terminal. The alarm information not only includes core information such as the patient's name, bed number, type of abnormal sign, and value of abnormal sign, but also includes a waveform diagram of the signs during the abnormal period and information on the patient's movement and alighting from the bed. This enables the traceability of alarm information and allows medical staff to make a preliminary judgment on the patient's abnormal condition while en route to the ward. If the mobile terminal is a smart bracelet or vibration call device, it will trigger multiple vibration alerts simultaneously to ensure that medical staff can receive the alarm information as soon as possible, quickly locate the abnormal bed, and go to handle it.

[0032] The central monitoring and display module also boasts comprehensive intelligent data management capabilities. It can securely and long-term store all patients' vital sign monitoring data, status indicators, alarm records, and body movement / absence status data. The storage period can be flexibly set according to hospital needs. It supports precise querying and filtering of historical data based on various criteria such as patient name, bed number, monitoring time, vital sign type, and abnormality type. Simultaneously, it can perform intelligent statistical analysis on historical data, automatically generating trend curves of patients' vital signs, intuitively displaying the changes in patients' respiratory rate and heart rate at different times, and automatically marking abnormal points in the curves and corresponding alarm information. This provides objective data analysis support for medical staff in diagnosing patients' conditions and adjusting treatment plans. Furthermore, the system can automatically extract monitoring data from nighttime periods, generating standardized nighttime vital sign monitoring reports. These reports include core information such as the patient's average nighttime respiratory rate, average heart rate, number of body movements, number of ambulations, and records of abnormalities. The reports can be directly printed or exported as electronic documents, significantly reducing the paperwork burden on medical staff.

[0033] Example 2 like Figure 2 As shown, the non-contact intelligent monitoring method for vital signs in this embodiment is based on the non-contact intelligent monitoring system for vital signs in Embodiment 1. This method relies on the collaborative work of the system's non-contact sensors, edge computing processing module, and central monitoring and display module. Through eight continuous and interconnected steps—signal acquisition, signal preprocessing, signal separation, vital sign value calculation, threshold comparison and judgment, result transmission, status display, and abnormal alarm push—it achieves non-contact, continuous, and highly accurate real-time monitoring of the patient's vital signs such as respiration and heart rate. It also provides graded, targeted, and intelligent early warning prompts when the patient's vital signs become abnormal. The entire methodology is designed around the core logic of non-intrusive data collection, localized intelligent processing, and centralized display and early warning. All steps are executed automatically and continuously by the system without human intervention. Each step adds innovative sub-steps and optimized processing methods on top of the basic execution logic, effectively improving the method's anti-interference ability, monitoring accuracy, and timely early warning. It completely solves the problems of intrusiveness, intermittency, and subjectivity of traditional manual rounds, and overcomes the shortcomings of contact monitors, such as discomfort and inability to be widely promoted in general wards. While ensuring the quality of patients' rest, it significantly improves the safety and efficiency of vital sign monitoring in wards, freeing medical staff from mechanical, timed rounds and allowing them to focus on observing patients' conditions and handling abnormalities.

[0034] The first step involves acquiring raw micro-motion signals caused by the patient's breathing and heartbeat using non-contact sensors. Based on the ward's layout and specific clinical monitoring needs, each bed is independently equipped with a corresponding non-contact sensor. The sensors selected are either millimeter-wave radar sensors or high-sensitivity piezoelectric thin-film sensors. Both types of sensors are independently deployed in the monitoring area of ​​their respective beds, and patients do not need to wear any devices, achieving completely contactless signal acquisition. This step adds two innovative sub-steps to the basic signal acquisition logic: initial sensor calibration and precise detection area positioning, ensuring the accuracy and stability of the raw micro-motion signal acquisition. If a millimeter-wave radar sensor is selected, the initial sensor calibration sub-step is performed first to accurately calibrate the sensor's transmit power and receive sensitivity, ensuring the sensor is in optimal working condition. Then, the sensor is installed on the ceiling above the bed or on the headboard wall. After adjusting the installation height, the sensor's built-in angle adjustment structure is used to perform a precise detection area positioning sub-step, ensuring the sensor's detection beam is precisely aligned with the patient's torso area. The detection beam employs a narrow beamforming design to avoid signal interference from the surrounding environment. After calibration and positioning, the sensor automatically emits electromagnetic waves with wavelengths from 4mm to 12mm. Upon contact with the human body surface, these electromagnetic waves generate an echo signal, which is received by the sensor in real time. The sensor transmits the echo signal to the edge computing processing module via encrypted wireless communication. The edge computing processing module performs precise phase demodulation processing on the echo signal, extracting the micro-displacement information of the chest cavity caused by the patient's breathing and heartbeat. This displacement information is the raw micro-displacement signal required for subsequent system processing. During demodulation, stray phase information is automatically filtered to improve signal purity. If a high-sensitivity piezoelectric thin-film sensor is selected, the initial sensor calibration sub-step is performed first. This involves zero-point calibration of each piezoelectric sensing unit to eliminate signal errors caused by initial bias. Next, the sensor is placed under the mattress, ensuring a seamless fit between the sensor's sensing surface and the mattress's underside. A precise detection area positioning sub-step is then performed, accurately positioning the sensor to correspond to the area below the patient's chest and abdomen when lying flat. After calibration and positioning, the minute fluctuations in the chest and abdomen caused by the patient's breathing and heartbeat will generate periodic, minute pressure changes on the mattress. These pressure changes trigger the piezoelectric effect of the sensor, causing it to generate a charge signal that matches the rhythm of breathing and heartbeat. The sensor transmits this charge signal to the edge computing processing module via a shielded wired link. The edge computing processing module performs low-noise charge amplification and precise voltage conversion on the charge signal, converting it into a recognizable voltage signal. This voltage signal is the raw micro-motion signal required for subsequent system processing. In this step, the signal acquisition by the non-contact sensor is performed continuously 24 / 7, with the acquisition frequency matching the processing frequency of the subsequent edge computing processing module to ensure real-time capture of changes in the patient's vital signs.

[0035] The second step involves receiving the raw micro-motion signals through the edge computing processing module and preprocessing them to obtain purified micro-motion signals. As the core data processing unit in the ward, the edge computing processing module receives raw micro-motion signals transmitted from all non-contact sensors within the ward in real time via an encrypted communication link. Since the raw micro-motion signals are easily contaminated with environmental clutter and random noise during acquisition and transmission, and suffer from baseline drift, directly affecting the accuracy of subsequent processing results, this step performs multi-dimensional fusion preprocessing on the raw micro-motion signals. The preprocessing process sequentially includes three innovative sub-steps: baseline drift correction, adaptive filtering, and multi-scale wavelet denoising. All three sub-steps are automatically executed by the edge computing processing module, synchronously and in real-time with the signal acquisition process, ensuring timely signal processing. First, the baseline drift correction sub-step is executed. The edge computing processing module uses an adaptive polynomial fitting algorithm to perform baseline fitting on the original micro-motion signal, accurately extracting the baseline drift trend and performing reverse compensation to completely eliminate the baseline drift problem in the original micro-motion signal, making the signal waveform return to the standard baseline. Then, the adaptive filtering sub-step is executed, using an adaptive filtering algorithm based on minimum mean square error to filter the corrected signal. This algorithm can dynamically adjust the filtering parameters according to the frequency characteristics and amplitude changes of the original micro-motion signal, accurately identifying and removing electromagnetic interference, equipment vibration, and other noise signals unrelated to breathing and heartbeat from the signal, retaining only the effective signal components. Finally, the multi-scale wavelet denoising sub-step is executed, using a multi-scale wavelet decomposition algorithm to denoise the filtered signal, decomposing the signal into wavelet spaces of different scales, thresholding the noise coefficients in each scale space, and then performing wavelet reconstruction, effectively suppressing random noise in the signal and making the signal waveform smoother. Through the coordinated processing of the above three innovative sub-steps, a clean micro-motion signal with complete removal of baseline drift, clutter signals, and random noise is finally obtained, providing a high-quality signal foundation for subsequent signal separation steps.

[0036] The third step involves using an edge computing processing module to perform signal separation on the pure micro-motion signal to obtain respiratory and heart rate waveform data. The pure micro-motion signal is a mixture of respiratory and heart rate micro-motion signals; direct calculation of vital signs would result in severe mutual interference. Therefore, this step requires high-precision signal separation to achieve completely independent extraction of respiratory and heart rate waveform data. This step employs a core algorithm combining empirical mode decomposition with sample entropy analysis, and adds two innovative sub-steps: endpoint effect suppression and component correlation verification. This significantly improves the accuracy of signal separation and effectively avoids distortion and misjudgment problems during the signal separation process. First, an endpoint effect suppression sub-step is executed. The edge computing processing module uses the mirror extension method to extend the data at both ends of the pure micro-motion signal, effectively suppressing the endpoint effect that is prone to occur in the subsequent empirical mode decomposition. Then, empirical mode decomposition is performed on the extended pure micro-motion signal, decomposing the continuous signal into multiple intrinsic mode function components with different time scales and frequency characteristics. Each component corresponds to an independent frequency component in the mixed signal; the respiratory micro-motion signal corresponds to the low-frequency component, and the heart rate micro-motion signal corresponds to the high-frequency component. Next, a sample entropy calculation sub-step is executed, calculating the sample entropy value of each intrinsic mode function component. The sample entropy value reflects the complexity and irregularity of the signal. The sample entropy value of the component corresponding to the respiratory waveform is calculated. The respiratory waveform is relatively small, while the entropy value of the component corresponding to the heart rate waveform is relatively large. The system presets two different sample entropy thresholds, with the first threshold being less than the second threshold, which serve as the criteria for judging the respiratory and heart rate waveform components, respectively. Components with sample entropy values ​​less than the first threshold are initially reconstructed to obtain preliminary respiratory waveform data, and components with sample entropy values ​​greater than the second threshold are initially reconstructed to obtain preliminary heart rate waveform data. Finally, a component correlation verification sub-step is executed to calculate the correlation coefficient between the preliminary respiratory waveform data, the preliminary heart rate waveform data, and the original pure micro-motion signal. If the correlation coefficient does not reach the preset standard, the sample entropy threshold is readjusted and reconstruction is performed again until the correlation coefficient reaches the standard, ultimately obtaining completely independent and highly accurate respiratory and heart rate waveform data. While completing the separation of respiratory and heart rate waveform data, the edge computing processing module also performs auxiliary analysis on the amplitude changes and signal continuity of the pure micro-motion signal. The amplitude changes are used to determine the intensity of the patient's body movement, and the absence of a valid signal for a long period of time is used to determine the patient's status at bedside. The analysis results of body movement and status at bedside are transmitted along with subsequent vital sign data.

[0037] The fourth step involves calculating the respiratory rate based on the respiratory waveform data and the heart rate based on the heart rate waveform data using an edge computing processing module. This step employs adapted algorithms to calculate the respiratory rate and heart rate values ​​separately, taking into account the different characteristics of the respiratory and heart rate waveform data. Furthermore, an innovative sub-step for removing motion interference is added to the calculation of both vital signs. The core calculation process is also adaptively optimized to ensure the accuracy and real-time performance of the results. The calculation of respiratory rate and heart rate values ​​is performed synchronously and independently, with the computation cycle controlled within a short time to ensure timely capture of subtle changes in the patient's vital signs. For the calculation of respiratory rate, the first step is to perform a body movement interference removal sub-step. The edge computing processing module monitors the patient's body movement amplitude in real time. When the body movement amplitude exceeds a preset body movement threshold, the calculation of respiratory rate is temporarily paused. The calculation continues only after the body movement amplitude recovers below the threshold and the signal stabilizes, thus avoiding calculation errors caused by body movement. Subsequently, wavelet transform processing is performed on the respiratory waveform data without body movement interference to convert the one-dimensional time domain signal into a two-dimensional respiratory signal time-frequency graph, clearly reflecting the change law of respiratory signal frequency over time. Next, an adaptive dominant frequency band identification algorithm is used to extract the curve of the dominant frequency band over time from the time-frequency graph, automatically adapting to slight changes in the patient's respiratory rhythm. Then, an adaptive peak detection algorithm is used to detect peaks in this curve, identifying all characteristic peaks and removing false peaks. Finally, the time interval between two adjacent characteristic peaks is calculated, which is one respiratory cycle. The reciprocal of this time interval is calculated to obtain the real-time respiratory rate value. For heart rate calculation, the same body motion interference removal sub-step is first performed to avoid calculation errors caused by body motion. Then, autocorrelation analysis is performed on the heart rate waveform data without body motion interference. A noise threshold filtering algorithm is used to eliminate random interference and highlight the periodic characteristics of the heart rate signal to obtain the autocorrelation function of the heart rate waveform data. Next, an extreme point precise location algorithm is used to extract the first maximum point other than the zero point from the autocorrelation function. The time delay corresponding to this maximum point is one heartbeat cycle. Finally, the reciprocal of this time delay is calculated and the unit conversion is completed to obtain the real-time heart rate value.

[0038] The fifth step involves using an edge computing processing module to compare respiratory rate and heart rate values ​​with preset safety threshold ranges, generating local judgment results that include normal or abnormal status indicators. To avoid misjudgments and missed judgments caused by uniform thresholds, this step uses personalized safety threshold ranges based on the patient's historical vital signs data for comparison and judgment. It also incorporates three innovative sub-steps: sliding window mean statistics, dynamic threshold updates, and continuous comparison verification. These improvements make the judgment of abnormal states more closely aligned with the patient's individual physiological characteristics, resulting in a more scientific and accurate assessment. First, a sliding window mean statistics sub-step is executed. The local data storage chip of the edge computing processing module continuously records the respiratory rate and heart rate monitoring data of each patient. A sliding window statistical method is used to process the historical data, obtaining the historical mean respiratory rate and historical mean heart rate of the current patient under normal physiological conditions, reflecting the patient's baseline vital signs in real time. Next, a threshold dynamic update sub-step is executed. Based on the historical mean respiratory rate and historical mean heart rate, and combined with clinical monitoring safety standards, the upper and lower limits of the preset general safety threshold range are dynamically adjusted to generate a personalized safety threshold range adapted to the individual physiological characteristics of the patient. This threshold range is updated every [percentage missing]. After each statistical cycle, the mean is recalculated and adjusted based on new historical data to ensure it always matches the patient's actual physiological condition. Next, a continuous comparison and verification sub-step is executed, comparing the real-time calculated respiratory rate and heart rate values ​​with the personalized safety threshold range one by one in real time. Each comparison result is marked and stored. If a single comparison result exceeds the threshold range, it is not immediately considered abnormal, but only marked. If three consecutive comparison results exceed the personalized safety threshold range, the patient's vital signs are immediately determined to be in an abnormal state, and an abnormal state identifier is generated. If the comparison results remain within the threshold range, the patient's vital signs are determined to be in a normal state, and a normal state identifier is generated.

[0039] The sixth step involves sending the local assessment results to the central monitoring and display module via the edge computing processing module. After generating the local assessment results, the edge computing processing module first integrates the real-time calculated respiratory rate, heart rate, body movement and bed agitation monitoring data, and corresponding status identifiers to form a standardized monitoring data message. This message is then encrypted using a hospital-specific encryption algorithm to ensure data transmission security and prevent leakage of patient vital signs data. After encryption, the encrypted monitoring data message is transmitted in real-time via the hospital intranet to the central monitoring and display module deployed at the nurses' station or cloud server using wired or wireless connections. A data verification sub-step is added during transmission; if the central monitoring and display module fails to receive the data, the edge computing processing module automatically retransmits it to ensure data integrity and stability. Simultaneously, the edge computing processing module performs real-time local backups of the integrated monitoring data, storing it on a local data storage chip to prevent data loss due to network failures. Once the network is restored, the local backup data is synchronized to the central monitoring and display module.

[0040] Step 7: The central monitoring and display module receives local judgment results and generates corresponding status identifiers on the display interface. The central monitoring and display module receives encrypted monitoring data packets transmitted in real-time from the edge computing processing modules of each ward via the hospital intranet. First, it performs data decryption and verification sub-steps, using the corresponding decryption algorithm to decrypt the packets and verifying the integrity of the decrypted data to ensure it is not damaged during transmission. After successful data verification, the monitoring data is analyzed in real-time to extract core information such as patient name, bed number, respiratory rate, heart rate, body movement / alighting status, and local judgment results. Subsequently, it displays this information intelligently and visually in the intelligent panoramic view of the display interface. The display interface has an independent display area for each bed, displaying the analyzed core information in real-time within the corresponding area. Simultaneously, status identifiers are generated and updated in the corresponding area based on the local judgment results. Status identifiers are color-coded: green for normal status, red for abnormal status, and yellow for signs approaching threshold boundaries. All identifiers are dynamically updated based on the real-time analyzed data. The central monitoring display module also supports intelligent clustering of beds by ward, floor, and disease risk level. Medical staff can switch and filter the display perspective through touch operation to quickly and intuitively grasp the vital signs of patients in all monitored beds. If the central monitoring display module is deployed on a cloud server, medical staff can also remotely access the display interface through mobile terminals to achieve mobile monitoring.

[0041] Step 8: When the local assessment result indicates an abnormal status, the central monitoring and display module triggers the corresponding alarm signal and pushes the alarm signal to the designated mobile terminal. After parsing the local assessment result as an abnormal status, the central monitoring and display module immediately initiates an intelligent alarm process. This process includes three innovative sub-steps: intelligent priority scheduling, nighttime silent adaptive response, and targeted alarm information push + source tracing. These three sub-steps are executed simultaneously, ensuring timely transmission of alarm information while avoiding interference with the patient's rest, and providing comprehensive abnormal information support for medical staff. First, the intelligent priority scheduling sub-step is executed. The system presets different alarm priorities based on the type of abnormal signs. Severe abnormalities such as apnea and sudden changes in heart rate have higher priority than general abnormalities such as slightly slower respiratory rate and slightly faster heart rate. High-priority abnormalities are prominently displayed on the central monitoring screen and trigger stronger alarm prompts. Next, the nighttime silent adaptive sub-step is executed. The system identifies whether the current system time is within the preset nighttime silent period, which can be adaptively adjusted according to the actual rest status of the ward. If the current time is within the nighttime silent period, the central monitoring display module only triggers the flashing of the red alarm light, and the audible alarm function of the sound and light alarm is turned off. If the current time is not within the nighttime silent period, both the flashing of the red alarm light and continuous audible alarm are triggered. The system first triggers an audible alarm. Finally, it executes a targeted alarm information push and source tracing sub-step. Based on the preset binding relationship between beds and medical staff's mobile terminals, the system accurately pushes abnormal alarm information to the designated on-duty medical staff's mobile terminals. These mobile terminals include PDAs, smartphones, smart bracelets, and vibration call devices. The alarm information not only includes core information such as the patient's name, bed number, type of abnormal sign, and abnormal sign value, but also includes a waveform diagram of the abnormal sign during the abnormal period and information on the patient's movement and alighting from the bed. This allows for source tracing of the alarm information, facilitating medical staff's initial assessment of the patient's abnormal condition. For portable devices such as smart bracelets and vibration call devices, multiple vibration alerts will be triggered simultaneously, ensuring that medical staff receive the alarm information immediately, quickly locate the abnormal bed, and proceed to the scene for treatment.

[0042] The non-contact intelligent vital sign monitoring method of this embodiment, while performing the above eight core steps, also includes auxiliary steps of intelligent data management. The central monitoring and display module will store all received vital sign monitoring data, status identifiers, alarm records, and body movement and bed leave status data in a long-term and secure manner. It supports accurate querying and filtering of historical data by various conditions such as patient name, bed number, monitoring time, and vital sign type. At the same time, it can perform intelligent statistical analysis on historical data, automatically generate the patient's vital sign trend change curve, and mark the abnormal points in the curve and the corresponding alarm information. It can also automatically extract monitoring data during the nighttime period and generate a standardized nighttime vital sign monitoring report, which can be directly printed or exported as an electronic document. All steps of the entire monitoring method are executed automatically and continuously by the system without human intervention, achieving 24 / 7 uninterrupted vital sign monitoring. This effectively fills the gap in monitoring during traditional manual rounds, ensuring patients' rest quality and protecting their privacy while significantly improving the safety, accuracy, and efficiency of vital sign monitoring in the ward. It effectively frees up the manpower of medical staff, allowing them to devote more energy to core nursing work such as patient observation and abnormal management. It is suitable for large-scale application in various wards such as general wards, geriatric wards, and psychiatric wards in hospitals.

[0043] As is known from common technical knowledge, this invention can be implemented through other embodiments that do not depart from its spirit or essential characteristics. Therefore, the disclosed embodiments described above are merely illustrative in all respects and are not the only ones. All modifications within the scope of this invention or its equivalents are included in this invention.

Claims

1. A non-contact intelligent vital sign monitoring system, characterized in that, include: At least one non-contact sensor, said at least one non-contact sensor being used to acquire raw micro-motion signals caused by the patient's breathing and heartbeat; An edge computing processing module connected to the at least one non-contact sensor, the edge computing processing module being deployed locally in the ward, the edge computing processing module receiving the original micro-motion signal and preprocessing the original micro-motion signal to obtain a purified micro-motion signal, the edge computing processing module performing signal separation on the purified micro-motion signal to obtain respiratory waveform data and heart rate waveform data, the edge computing processing module calculating a respiratory rate value based on the respiratory waveform data and a heart rate value based on the heart rate waveform data, the edge computing processing module comparing the respiratory rate value and the heart rate value with a preset safety threshold range to generate a local judgment result; The central monitoring and display module is communicatively connected to the edge computing processing module. The central monitoring and display module is deployed at the nurse station or cloud server. The central monitoring and display module receives the local judgment result and generates a corresponding status identifier on the display interface. When the local judgment result is an abnormal status identifier, the central monitoring and display module triggers the corresponding alarm signal and pushes the alarm signal to the designated mobile terminal.

2. The non-contact intelligent vital sign monitoring system according to claim 1, characterized in that, The at least one non-contact sensor is a millimeter-wave radar sensor, which is installed on the ceiling or headboard above the patient's bed with the detection area of ​​the millimeter-wave radar sensor facing the patient's torso area. The millimeter-wave radar sensor emits electromagnetic waves with a wavelength of 4mm to 12mm and receives echo signals reflected from the human body surface. The edge computing processing module performs phase demodulation on the echo signals to extract the chest cavity micro-movement displacement information caused by breathing and heartbeat as the original micro-movement signal. Alternatively, the at least one non-contact sensor is a high-sensitivity piezoelectric thin-film sensor, which is laid under the mattress with its sensing surface in contact with the underside of the mattress. The high-sensitivity piezoelectric thin-film sensor senses minute pressure changes caused by the patient's breathing and heartbeat and generates an electrical charge signal. The edge computing processing module amplifies and converts the electrical charge signal to obtain the original micro-motion signal.

3. The non-contact intelligent vital sign monitoring system according to claim 1, characterized in that, When the edge computing processing module performs signal separation on the pure micro-motion signal, it first performs empirical mode decomposition on the pure micro-motion signal to obtain multiple intrinsic mode function components. The edge computing processing module calculates the sample entropy value of each intrinsic mode function component. The edge computing processing module reconstructs the respiratory waveform data into the intrinsic mode function components whose sample entropy value is less than a first threshold, and reconstructs the heart rate waveform data into the intrinsic mode function components whose sample entropy value is greater than a second threshold, where the first threshold is less than the second threshold.

4. The non-contact intelligent vital sign monitoring system according to claim 1, characterized in that, When the edge computing processing module calculates the respiratory frequency value based on the respiratory waveform data, the edge computing processing module first performs wavelet transform on the respiratory waveform data to obtain a respiratory signal time-frequency graph. The edge computing processing module extracts the curve of the main frequency band changing with time from the respiratory signal time-frequency graph. The edge computing processing module performs peak detection on the curve and calculates the time interval between adjacent peaks. The edge computing processing module generates the respiratory frequency value based on the reciprocal of the time interval. When the edge computing processing module calculates the heart rate value based on the heart rate waveform data, the edge computing processing module first performs autocorrelation analysis on the heart rate waveform data to obtain an autocorrelation function. The edge computing processing module extracts the time delay corresponding to the first maximum point other than zero from the autocorrelation function. The edge computing processing module generates the heart rate value based on the reciprocal of the time delay.

5. The non-contact intelligent vital sign monitoring system according to claim 1, characterized in that, In the step of comparing the respiratory rate value and the heart rate value with the preset safety threshold range by the edge computing processing module, the edge computing processing module first obtains the historical average respiratory rate and historical average heart rate of the current patient. The edge computing processing module dynamically adjusts the upper and lower limits of the safety threshold range based on the historical average respiratory rate to generate a personalized safety threshold range. The edge computing processing module compares the respiratory rate value and heart rate value calculated in real time with the personalized safety threshold range one by one. When the results of three consecutive comparisons all exceed the personalized safety threshold range, the edge computing processing module generates the abnormal status identifier.

6. A non-contact intelligent monitoring method for vital signs, based on the non-contact intelligent monitoring system for vital signs as described in any one of claims 1-5, characterized in that, Includes the following steps: Raw micro-motion signals caused by the patient's breathing and heartbeat are collected using non-contact sensors; The edge computing processing module receives the original micro-motion signal and preprocesses it to obtain a clean micro-motion signal. The edge computing processing module performs signal separation on the pure micro-motion signal to obtain respiratory waveform data and heart rate waveform data. The edge computing processing module calculates the respiratory rate value based on the respiratory waveform data and the heart rate value based on the heart rate waveform data. The edge computing processing module compares the respiratory rate value and the heart rate value with a preset safety threshold range to generate a local judgment result containing a normal state identifier or an abnormal state identifier. The edge computing processing module sends the local judgment result to the central monitoring and display module. The central monitoring and display module receives the local judgment result and generates the corresponding status identifier on the display interface. When the local judgment result is an abnormal status identifier, the corresponding alarm signal is triggered through the central monitoring and display module and the alarm signal is pushed to the designated mobile terminal.

7. The non-contact intelligent monitoring method for vital signs according to claim 6, characterized in that, In the step of acquiring the original micro-motion signal caused by the patient's breathing and heartbeat using a non-contact sensor, when a millimeter-wave radar sensor is used as the non-contact sensor, it includes emitting electromagnetic waves with a wavelength of 4mm to 12mm through the millimeter-wave radar sensor and receiving the echo signal reflected from the human body surface. The edge computing processing module performs phase demodulation on the echo signal to extract the chest cavity micro-motion displacement information caused by breathing and heartbeat as the original micro-motion signal. When a high-sensitivity piezoelectric thin-film sensor is used as the non-contact sensor, it includes sensing the minute pressure changes caused by the patient's breathing and heartbeat through the high-sensitivity piezoelectric thin-film sensor and generating a charge signal. The edge computing processing module amplifies and converts the charge signal to obtain the original micro-motion signal.

8. The non-contact intelligent monitoring method for vital signs according to claim 6, characterized in that, The step of performing signal separation on the pure micro-motion signal through the edge computing processing module to obtain respiratory waveform data and heart rate waveform data includes performing empirical mode decomposition on the pure micro-motion signal through the edge computing processing module to obtain multiple intrinsic mode function components, calculating the sample entropy value of each intrinsic mode function component through the edge computing processing module, reconstructing the respiratory waveform data by the intrinsic mode function components whose sample entropy value is less than a first threshold through the edge computing processing module, and reconstructing the heart rate waveform data by the intrinsic mode function components whose sample entropy value is greater than a second threshold through the edge computing processing module, wherein the first threshold is less than the second threshold.

9. A non-contact intelligent monitoring method for vital signs according to claim 6, characterized in that, The steps of calculating the respiratory frequency value based on the respiratory waveform data and the heart rate value based on the heart rate waveform data by the edge computing processing module include: performing wavelet transform on the respiratory waveform data to obtain a respiratory signal time-frequency graph by the edge computing processing module; extracting the main frequency band change curve over time from the respiratory signal time-frequency graph by the edge computing processing module; performing peak detection on the change curve and calculating the time interval between adjacent peaks by the edge computing processing module; generating the respiratory frequency value based on the reciprocal of the time interval by the edge computing processing module; performing autocorrelation analysis on the heart rate waveform data to obtain an autocorrelation function by the edge computing processing module; extracting the time delay corresponding to the first maximum point other than zero from the autocorrelation function by the edge computing processing module; and generating the heart rate value based on the reciprocal of the time delay by the edge computing processing module.

10. A non-contact intelligent monitoring method for vital signs according to claim 6, characterized in that, The step of comparing the respiratory rate value and the heart rate value with a preset safety threshold range through the edge computing processing module to generate a local judgment result containing a normal state identifier or an abnormal state identifier includes obtaining the current patient's historical average respiratory rate and historical average heart rate through the edge computing processing module; dynamically adjusting the upper and lower limits of the safety threshold range based on the historical average respiratory rate through the edge computing processing module to generate a personalized safety threshold range; comparing the real-time calculated respiratory rate value and heart rate value with the personalized safety threshold range through the edge computing processing module; and generating the abnormal state identifier through the edge computing processing module when the results of three consecutive comparisons all exceed the personalized safety threshold range. The step of triggering the corresponding alarm signal through the central monitoring display module and pushing the alarm signal to the designated mobile terminal includes, when the abnormal status identifier is generated, identifying whether the current system time is in a preset nighttime silent period through the central monitoring display module. If the current system time is in the nighttime silent period, the central monitoring display module only triggers the flashing of the alarm light without triggering the sound alarm of the audible and visual alarm, and pushes the alarm information corresponding to the abnormal status identifier to the vibrating call device or smart bracelet carried by the nurse.

Images