Far-infrared vital signs monitoring method and device based on deep learning

Abstract

The present invention relates to the field of vital sign monitoring technology, and in particular to a far-infrared vital sign monitoring method and device based on deep learning. The method comprises the following steps: acquiring thermal imaging data of the monitoring area and generating a thermal map to obtain a thermal map of the monitoring area; integrating vital sign features of the thermal map of the monitoring area to obtain vital sign data of users in the monitoring area; performing center of gravity imbalance detection on the vital sign data of users in the monitoring area to obtain user center of gravity imbalance behavior data; acquiring a millimeter wave signal set in the monitoring area and performing posture change behavior analysis to obtain user rapid posture change behavior data; performing risk assessment based on user center of gravity imbalance behavior data and user rapid posture change behavior data to obtain user fall risk data in the monitoring area; performing fall risk warning strategy analysis based on user fall risk data in the monitoring area, thereby obtaining a user fall risk warning strategy. The present invention can improve the accuracy and efficiency of vital sign monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of vital sign monitoring, and in particular to a far-infrared vital sign monitoring method and device based on deep learning. Background Art

[0002] With the rapid development of medical technology, vital sign monitoring has become an important part of clinical medicine, emergency services and health management. Vital sign monitoring usually includes indicators such as heart rate, respiratory rate, body temperature and blood oxygen saturation, which are key to assessing the patient's health status and disease development. However, traditional vital sign monitoring methods mainly rely on contact sensors and instruments, which have many limitations. Traditional vital sign monitoring methods include electrocardiogram (ECG), pulse oximeter (SpO2) and thermometer. These methods usually require the sensor to be directly attached to the skin surface to obtain accurate physiological data. Although these technologies provide effective monitoring methods to some extent, their defects are also becoming increasingly apparent. Many traditional vital sign monitoring methods require patients to remain still during monitoring to reduce the impact of motion artifacts, which limits the patient's freedom of movement and affects daily life and rehabilitation. Traditional monitoring equipment usually requires direct contact with the human body, causing physiological discomfort to patients under long-term monitoring. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a far-infrared vital signs monitoring method and device based on deep learning to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a far-infrared vital signs monitoring method based on deep learning includes the following steps:

[0005] Step S1: acquiring thermal imaging data of the monitoring area through a far-infrared thermal imaging instrument, and generating a thermal map of the monitoring area based on the thermal imaging data of the monitoring area, thereby obtaining a thermal map of the monitoring area;

[0006] Step S2: integrating the heat distribution of the thermal map of the monitoring area to obtain a heat distribution image set of the monitoring area, and integrating the vital sign characteristics of users in the monitoring area according to the heat distribution image set of the monitoring area to obtain vital sign data of users in the monitoring area;

[0007] Step S3: performing user behavior pattern recognition based on the vital sign data of the user in the monitoring area, thereby obtaining user behavior pattern data, and performing user center of gravity imbalance detection on the temperature distribution image set in the monitoring area according to the user behavior pattern data, thereby obtaining user center of gravity imbalance behavior data;

[0008] Step S4: obtaining a millimeter wave signal set in the monitoring area through the millimeter wave module, and performing a millimeter wave fluctuation detection on the millimeter wave signal set in the monitoring area, thereby obtaining millimeter wave fluctuation data in the monitoring area; performing a user's rapid posture change behavior analysis based on the millimeter wave fluctuation data in the monitoring area, thereby obtaining the user's rapid posture change behavior data;

[0009] Step S5: Perform a fall risk assessment on the user based on the user's center of gravity imbalance behavior data and the user's rapid posture change behavior data, thereby obtaining the fall risk data of the users in the monitoring area; perform a fall risk warning strategy analysis based on the fall risk data of the users in the monitoring area, thereby obtaining the user fall risk warning strategy, and upload it to the monitoring area management platform to execute the risk warning task.

[0010] In the present invention, the far-infrared thermal imaging instrument can obtain thermal imaging data of the monitoring area in a non-contact manner, avoiding direct contact of traditional sensors with the skin and reducing the discomfort of the patient. Thermal imaging technology can capture changes in heat distribution in real time, so that the monitoring process is not affected by the patient's movement, improves the real-time and accuracy of the data, and does not require the patient to remain still during the monitoring process. By integrating the heat distribution image set, more accurate vital signs data, including heart rate, respiratory rate, etc., can be obtained to provide support for clinical diagnosis. By monitoring heat changes, abnormal signs can be discovered in time. Through behavioral pattern recognition, the patient's daily behavior can be automatically learned and analyzed to provide data support for personalized health management. The detection of center of gravity imbalance behavior can timely identify the risk of falling that may occur in patients, help medical staff take preventive measures, and reduce falling accidents. Millimeter wave technology can capture subtle posture changes, provide in-depth analysis of patients' rapid response ability, and is suitable for monitoring in various environments. Monitoring of rapid posture changes can provide real-time feedback for patients, help them improve their body posture, and reduce the risk of accidental injury. By comprehensively analyzing the data of imbalance and rapid posture change behavior, the patient's fall risk can be effectively assessed and preventive measures can be taken in advance. The generated fall risk warning strategy can be promptly transmitted to medical staff through the management platform to ensure rapid response and effectively protect patient safety. Systematic data analysis can provide a basis for medical management and help hospitals and medical staff to formulate more scientific monitoring and care plans. In summary, by utilizing the technical means of far-infrared thermal imaging and millimeter wave signal processing, a new way of vital signs monitoring is provided, which overcomes the limitations of traditional methods. The implementation of each step not only improves the accuracy and efficiency of monitoring, but also provides strong support for the safety and health management of patients. Through systematic data integration and analysis, more personalized and intelligent medical services can be achieved, improving the quality of life and sense of security of patients.

[0011] Optionally, step S1 specifically includes:

[0012] Step S11: Acquire thermal imaging data of the monitoring area through a far-infrared thermal imaging instrument;

[0013] Step S12: performing thermal imaging data median denoising on the thermal imaging data of the monitoring area, thereby obtaining thermal imaging denoised data of the monitoring area;

[0014] Step S13: Acquiring monitoring area environment sensing data through monitoring area environment sensors;

[0015] Step S14: performing environmental impact correction on the thermal imaging denoising data of the monitoring area according to the environmental sensing data of the monitoring area, thereby obtaining thermal imaging correction data of the monitoring area;

[0016] Step S15: mapping the temperature distribution of the monitoring area according to the thermal imaging correction data of the monitoring area, thereby obtaining a thermal map of the monitoring area.

[0017] In the present invention, the far-infrared thermal imager can provide real-time temperature distribution information, which is convenient for rapid response to environmental changes. It can capture subtle temperature changes in the monitoring area and improve the accuracy and reliability of monitoring, which avoids direct interference with the monitored object and is suitable for various sensitive places (sensitive places such as toilets). The median denoising algorithm effectively removes random noise, making the thermal imaging data clearer. Compared with other denoising methods, the median denoising performs better in retaining image details and ensuring accurate identification of temperature changes. Combining environmental sensor data (such as humidity, wind speed, temperature, etc.) can provide a more comprehensive monitoring background and enhance the depth of analysis. Environmental data provides important context for thermal imaging, which can help identify the cause of temperature changes and make decisions more effectively. Through environmental impact correction, the systematic errors caused by environmental factors are eliminated, and the accuracy of thermal imaging data is improved. The corrected data is more in line with the actual situation, can better reflect the real temperature distribution of the monitoring area, and enhance the application value of the data. The generated heat map can intuitively display the temperature distribution of the monitoring area, which is convenient for users to analyze and make decisions. The abnormal temperature area in the heat map is clear at a glance, which helps to quickly identify potential problems (such as equipment failure, heat loss, etc.) and improve response efficiency.

[0018] Optionally, step S14 is specifically:

[0019] Step S141: extracting the environmental temperature characteristics and the environmental humidity characteristics of the monitoring area environmental sensor data, thereby obtaining the monitoring area environmental temperature data and the monitoring area environmental humidity data;

[0020] Step S142: performing time series correlation on the monitoring area ambient temperature data and the monitoring area thermal imaging denoised data to obtain temperature-thermal radiation correlation data, and estimating the temperature influence factor based on the temperature-thermal radiation correlation data to obtain the temperature influence factor;

[0021] Step S143: performing time series correlation on the environmental humidity data of the monitoring area and the thermal imaging denoised data of the monitoring area to obtain humidity-thermal radiation correlation data, and estimating the humidity influence factor according to the humidity-thermal radiation correlation data to obtain the humidity influence factor;

[0022] Step S144: constructing a linear regression thermal radiation correction model according to the temperature influencing factor and the humidity influencing factor;

[0023] Step S145: performing environmental impact correction on the thermal imaging denoising data of the monitoring area through a linear regression thermal radiation correction model, thereby obtaining thermal imaging correction data of the monitoring area.

[0024] The present invention can obtain more accurate environmental parameters by extracting temperature and humidity data, and these data provide a basis for subsequent analysis. The acquired environmental temperature and humidity data provide rich information for analyzing environmental influencing factors, which is convenient for subsequent in-depth research. Through time series association, the relationship between temperature change and thermal radiation can be captured, providing an in-depth understanding of environmental thermal dynamics. The acquisition of temperature-thermal radiation correlation data helps to more accurately estimate the influencing factors of temperature on thermal radiation, and provides a basis for environmental management and regulation. The correlation analysis of humidity and thermal radiation can reveal the impact of humidity changes on thermal radiation and promote the comprehensive evaluation of environmental characteristics. The estimation of humidity influencing factors provides data support for the subsequent establishment of a more comprehensive thermal radiation correction model. The linear regression model can provide a scientific correction framework for thermal radiation data and effectively reduce the interference of external factors. By comprehensively considering the influencing factors of temperature and humidity, the model can adapt to the correction requirements under different environmental conditions and improve the reliability of monitoring data. The corrected thermal imaging data can more realistically reflect the environmental conditions and provide high-quality data support for subsequent environmental assessment and decision-making. The corrected data helps to formulate more scientific environmental management measures and improve the effectiveness of the monitoring system in practical applications.

[0025] Optionally, step S2 specifically includes:

[0026] Step S21: dividing the monitoring area heat map into grid areas, thereby obtaining a monitoring area grid heat map;

[0027] Step S22: performing pixel value statistics on the grid heat map of the monitoring area to obtain grid pixel value data, and performing grid heat calculation based on the grid pixel value data to obtain grid heat data;

[0028] Step S23: performing grid heat division according to the grid heat data, thereby obtaining high heat grid data and low heat grid data;

[0029] Step S24: integrating high heat distribution based on high heat grid data, thereby obtaining a high heat distribution image set of the monitoring area; integrating low heat distribution based on low heat grid data, thereby obtaining a low heat distribution image set of the monitoring area;

[0030] Step S25: spatially merging the high heat distribution image set of the monitoring area and the low heat distribution image set of the monitoring area to obtain a heat distribution image set of the monitoring area;

[0031] Step S26: integrating the vital sign features of users in the monitoring area according to the heat distribution image set in the monitoring area, thereby obtaining vital sign data of users in the monitoring area.

[0032] The present invention divides the monitoring area into grid areas, which can effectively improve the spatial resolution of the data and make the analysis more refined. After gridding, data management and analysis become more systematic, which is convenient for subsequent heat map statistics and heat calculation. Through unified grid division, it is convenient to compare and analyze the heat between different areas. By counting the pixel values ​​of the grid heat map, the information in the heat map can be converted into numerical data for further analysis. By calculating the heat through the grid pixel value, the heat distribution of each grid can be quantified, providing a basis for subsequent decision-making, and at the same time, it can timely reflect the heat changes in the monitoring area, which is convenient for rapid response. The division of high-heat grids and low-heat grids can help identify potential risk areas, such as high-heat areas may represent abnormal activities or potential safety hazards. By dividing the area into different heat levels, resources can be allocated more effectively and targeted intervention measures can be taken, which helps decision makers take appropriate management and control measures according to the heat distribution. The integrated high-heat and low-heat distribution image set can intuitively display the heat distribution characteristics of the monitoring area and help analysts quickly understand the data. By comparing high and low heat, the heat change trend is analyzed to reveal the changes in activities in the area and their potential causes. Integrating different data sets provides multi-dimensional information support for subsequent analysis and helps improve the reliability of data. Through spatial merging, a comprehensive set of heat distribution images of the monitoring area can be formed to provide users with an overall perspective. Different types of data can be compared in the same image, which enhances the explanatory power and applicability of the data. By integrating users' vital signs data, personalized health monitoring can be achieved, providing corresponding services for the needs of different users. The correlation analysis of vital sign characteristics and heat distribution can reveal the relationship between environment and health and help assess potential health risks.

[0033] Optionally, step S26 is specifically:

[0034] Step S261: extracting features of high heat distribution areas from the heat distribution image set of the monitoring area, thereby obtaining a high heat distribution image set of the monitoring area;

[0035] Step S262: performing morphological recognition of the user's thermal contour according to the high-calorie distribution image set in the monitoring area, thereby obtaining a user's thermal distribution image set, and estimating the user's body temperature according to the user's thermal distribution image set, thereby obtaining the user's body temperature data;

[0036] Step S263: dividing the user heat distribution image set into user part heat image sets, thereby obtaining a user neck heat image set and a user chest heat image set;

[0037] Step S264: performing heat change frequency statistics based on the user's neck heat image set to obtain the user's neck heat change frequency data, and estimating the user's heart rate based on the user's neck heat change frequency data to obtain the user's heart rate data;

[0038] Step S265: integrating periodic heat fluctuations according to the user's chest heat image set, thereby obtaining periodic heat fluctuation data of the user's chest, and estimating the user's respiratory frequency according to the periodic heat fluctuation data of the user's chest, thereby obtaining the user's respiratory frequency data;

[0039] Step S266: Integrate the user's body temperature data, the user's heartbeat data, and the user's respiratory rate data into user vital signs, thereby obtaining the user's vital signs data in the monitoring area.

[0040] The present invention can accurately identify the temperature abnormality of the user's body parts by extracting the high heat area in the heat distribution image, and detect potential health problems at an early stage. The generated high heat distribution image set provides a reliable data basis for subsequent analysis, ensuring the efficiency of subsequent processing. The morphological recognition method can clearly distinguish different heat distribution characteristics and help establish the user's heat distribution model. Accurate body temperature data can timely feedback the user's physical condition, provide important health information, and assist in diagnosis. The user's heat image set is divided into neck and chest image sets, so that the heat feature analysis of each part is more detailed, the monitoring accuracy is improved, and a more targeted health assessment can be performed according to the characteristics of different body parts. By counting the frequency of neck heat changes, the user's heart activity can be monitored in real time, and heart abnormalities can be detected in time. Heartbeat data can reflect the user's physiological state, provide early warnings, and help users take health measures in time. The integration of periodic fluctuations provides data support for the accurate estimation of respiratory rate, which helps to evaluate the user's respiratory health status. By monitoring the respiratory rate, the user's physiological response in different states can be evaluated, helping to optimize the health management plan. Integrating the user's body temperature, heartbeat and respiratory rate data can comprehensively evaluate the user's vital signs and provide a more comprehensive status report.

[0041] Optionally, step S3 specifically includes:

[0042] Step S31: extracting respiratory frequency features and heart rate features according to the vital sign data of the user in the monitoring area, thereby obtaining the respiratory frequency data and the heart rate data of the user;

[0043] Step S32: dividing the user's respiratory frequency data into user respiratory frequency time series, thereby obtaining the user's high respiratory frequency time series data and the user's low respiratory frequency time series data;

[0044] Step S33: dividing the user's heart rate data into user heart rate time series, thereby obtaining the user's high heart rate time series data and the user's low heart rate time series data;

[0045] Step S34: performing time series intersection on the user's high respiratory rate time series data and the user's high heart rate time series data, thereby obtaining the user's action time series data; performing time series intersection on the user's low respiratory rate time series data and the user's low heart rate time series data, thereby obtaining the user's static time series data;

[0046] Step S35: integrating the user behavior pattern according to the user action time series data and the user static time series data, thereby obtaining the user behavior pattern data;

[0047] Step S36: Performing a user gravity center imbalance detection on the monitoring area temperature distribution image set according to the user behavior pattern data, thereby obtaining the user gravity center imbalance behavior data.

[0048] The present invention obtains the vital signs data of the user through a sensor or a wearable device, and extracts the breathing rate and heart rate characteristics of the user by using an algorithm. These data generally include the number of breaths and heartbeats per minute. Accurately extracting the breathing and heartbeat data of the user can provide a reliable basis for subsequent analysis. These characteristics are important indicators for evaluating the health status of the user, and can detect abnormal conditions in a timely manner, providing data support for health management and disease prevention. Through time series division, the breathing pattern of the user in different situations can be clarified. For example, a high breathing rate may be associated with exercise, tension or anxiety, while a low breathing rate may be associated with a rest or relaxation state. This helps to understand the physiological and psychological state of the user. Time series division of the heart rate helps to analyze the changes in the heart rate of the user in different activity states, such as an increase in heart rate during exercise and a decrease in heart rate in a resting state. This information can be used for personalized health advice and exercise guidance. Clarifying the user's activity state (such as exercise or stillness) can provide accurate data support for behavior analysis. By identifying the user's activity and stillness states, a more detailed analysis can be provided for the health monitoring system to help evaluate the user's daily activity level. By integrating time series data in different states, the user's daily behavior pattern can be more comprehensively understood. This helps provide a basis for personalized health management, such as adjusting the user's exercise plan or improving their lifestyle. Imbalance detection can provide important safety warnings in real-time monitoring, especially for elderly or frail users. By promptly identifying imbalanced behaviors, the risk of falling can be reduced, a safe living environment can be promoted, and users can have a higher quality of life.

[0049] Optionally, step S36 is specifically:

[0050] Step S361: integrating the time-series changes of pixel temperatures of the temperature distribution image set of the monitoring area, thereby obtaining the temperature change data of the monitoring area;

[0051] Step S362: performing user behavior pattern recognition on the temperature change data of the monitoring area according to the user behavior pattern data, thereby obtaining user action behavior data and user stationary behavior data;

[0052] Step S363: performing a behavior conversion time point intersection operation on the user action behavior data and the user static behavior data, thereby obtaining user behavior conversion time point data;

[0053] Step S364: extracting the characteristics of the physical sign changes at the behavior conversion time point according to the user behavior conversion time point data, thereby obtaining the respiratory frequency change data at the behavior conversion time point and the heart beat rate change data at the behavior conversion time point;

[0054] Step S365: performing statistics of instantaneous high vital sign changes on the respiratory frequency change data at the behavior conversion time point and the heart beat rate change data at the behavior conversion time point, respectively, so as to obtain instantaneous high respiratory frequency change time point data and instantaneous high heart rate change time point data;

[0055] Step S366: Perform time point intersection operation on the instantaneous high respiratory frequency change time point data and the instantaneous high heart rate change time point data to obtain the user's center of gravity imbalance time point data, and integrate the vital sign characteristics of the user's center of gravity imbalance time point data to obtain the user's center of gravity imbalance behavior data.

[0056] The present invention can obtain detailed temperature change data by performing pixel-level time series integration on the temperature distribution image set of the monitoring area. This data can reflect the changes in ambient temperature in different time periods, provide important environmental background information for subsequent analysis of user behavior, and enhance the correlation between temperature and user behavior. Combining user behavior pattern data with temperature change data, the user's action and static behavior can be identified. This identification can help the system understand the user's activity patterns under specific environmental conditions, thereby providing data support for intelligent health monitoring, such as evaluating the user's exercise volume and static time, and identifying abnormal behavior patterns. By performing intersection operations on user action and static behavior data, the user's behavior conversion time point can be clarified. This data is particularly important for understanding the user's transition between active and static states, and helps to timely identify possible health risks (such as sudden activity cessation) and changes in activity patterns. Extracting respiratory rate and heart rate change data at the time point of behavior conversion can reveal changes in the user's physiological state during the behavior conversion process. Statistics on instantaneous high changes in respiratory rate and heart rate can help identify the user's physiological stress response at a specific time point. This is crucial for disease early warning and health management. It can issue an alarm in time when the user faces health risks and promote timely medical intervention. By analyzing the intersection of instantaneous high respiratory rate and heart rate change data, the time point when the user's center of gravity is unbalanced can be identified. The integration of this data can not only be used to monitor the user's balance state, but also provide a basis for sports rehabilitation, safety monitoring of the elderly, etc., and prevent accidents such as falls in time.

[0057] Optionally, step S4 is specifically:

[0058] Step S41: obtaining a millimeter wave signal set in the monitoring area through a millimeter wave module, and performing millimeter wave denoising on the millimeter wave signal set in the monitoring area, thereby obtaining a millimeter wave denoised signal set in the monitoring area;

[0059] Step S42: performing Fourier transform on the millimeter wave denoised signal set in the monitoring area to obtain the millimeter wave spectrum in the monitoring area, and performing millimeter wave instantaneous fluctuation detection based on the millimeter wave spectrum in the monitoring area to obtain millimeter wave fluctuation data in the monitoring area;

[0060] Step S43: performing feature selection according to the millimeter wave fluctuation data of the monitoring area, so as to obtain the millimeter wave fluctuation rate data and the millimeter wave fluctuation duration data of the monitoring area;

[0061] Step S44: performing high-value fluctuation rate statistics on the millimeter wave fluctuation rate data in the monitoring area, thereby obtaining high-value fluctuation rate time point data; performing short-duration statistics on the millimeter wave fluctuation duration data in the monitoring area, thereby obtaining short-duration fluctuation time point data;

[0062] Step S45: performing a time point intersection operation on the high-volume fluctuation rate time point data and the short-duration fluctuation time point data, thereby obtaining the user's rapid posture change time point data;

[0063] Step S46: Integrate the millimeter wave behavior features at the time point according to the user's rapid posture change time point data, so as to obtain the user's rapid posture change behavior data.

[0064] The present invention collects the signal set of the monitoring area through the millimeter wave module, and can obtain high-resolution and high-frequency signal data, which is helpful for detailed analysis of the dynamic changes of the monitoring area. Millimeter wave denoising technology effectively eliminates environmental noise and other interference signals, and improves the signal-to-noise ratio of the signal. This makes subsequent analysis more reliable and reduces the possibility of misjudgment and missed detection. The denoised signal set retains the essential characteristics of the signal, making subsequent spectrum analysis and fluctuation detection more accurate. Fourier transform converts time domain signals into frequency domain signals, which helps to identify and analyze the frequency components and their distribution in the signal, and reveal important features hidden in the signal. Spectrum-based analysis can quickly identify instantaneous fluctuations, provide real-time feedback, and support timely response to dynamic changes. Through rapid spectrum calculation and analysis, the fluctuation of the monitoring area can be monitored in real time, which is suitable for application scenarios that are sensitive to changes. Feature selection technology helps to extract the most representative information from the fluctuation data, reduce the data dimension, and reduce the subsequent calculation burden. Extracting fluctuation rate and duration data can deeply understand the dynamic behavior in the monitoring area, which helps to identify and classify behavior patterns. The results of feature selection can be optimized for specific application scenarios, such as behavior monitoring, anomaly detection, etc., to improve the applicability of the overall system. Statistics of high fluctuation rate and short duration fluctuation time can help quickly identify abnormalities or key events, providing important clues for subsequent analysis. Through statistical analysis, complex data can be converted into actionable information to help users understand the dynamic changes in the monitoring area. Timely triggering of response mechanisms based on high fluctuations and short duration fluctuations helps to achieve automated monitoring and intelligent alarms. Through time point intersection operations, data of different dimensions can be integrated to provide a more comprehensive view of dynamic changes. It can effectively capture the moment when the user's posture changes rapidly, providing a reliable basis for behavior analysis and prediction. Integrating the user's rapid posture change behavior data can provide a comprehensive analysis of user behavior, which helps to understand user needs and habits. By analyzing the user's rapid posture changes, it helps to identify potential user falls.

[0065] Optionally, step S5 specifically includes:

[0066] Step S51: classifying the user's center of gravity imbalance behavior data by the level of physical sign change, thereby obtaining imbalance behavior level classification data;

[0067] Step S52: classifying the user's rapid posture change behavior data by millimeter wave fluctuation amplitude, thereby obtaining rapid posture change behavior classification data;

[0068] Step S53: constructing a fall risk assessment model according to the imbalance behavior classification data and the rapid posture change behavior classification data;

[0069] Step S54: performing a fall risk assessment on the user's center of gravity imbalance behavior data and the user's rapid posture change behavior data through a fall risk assessment model, thereby obtaining the user's fall risk data in the monitoring area;

[0070] Step S55: Perform fall risk warning strategy analysis based on the fall risk data of users in the monitoring area, so as to obtain the user fall risk warning strategy, and upload it to the monitoring area management platform to execute the risk warning task.

[0071] The present invention can effectively identify imbalance behaviors of different degrees by classifying the amplitude of physical sign changes in the data of imbalance behavior of the center of gravity. The classification of different levels makes the subsequent analysis more targeted. Classifying the imbalance behavior data helps to establish a unified data standard and provide a consistent basis for subsequent model construction and analysis. By classifying the millimeter wave fluctuation amplitude of rapid posture changes, the user's body posture changes in dynamic situations can be captured, which helps to understand the behavioral characteristics of users in specific situations. Different levels of fluctuation amplitude can help identify potential risk behaviors. For example, rapid posture changes may be a precursor to falling. Through detailed analysis, risks can be assessed more accurately. The application of millimeter wave technology enhances the real-time monitoring capability of user dynamic behavior, which helps to provide instant feedback on the user's status, thereby improving the effectiveness of monitoring. Combining the data of imbalance behavior and rapid posture changes, a comprehensive fall risk assessment model is constructed, which makes the risk assessment more comprehensive and reduces omissions and errors. Using the data analysis results to build a model can provide managers with data-driven decision support and improve the scientificity and effectiveness of management. The construction of the model takes into account the behavioral characteristics of different users, has good adaptability, and can be widely used in different user groups. It can evaluate the user's imbalance and rapid posture changes in real time, provide instant feedback to the user, and help the user identify potential risks. By timely detecting the risk of falling, it is helpful to take preventive measures to reduce the incidence of actual falling events, thereby improving the safety and quality of life of users. By analyzing the fall risk data in the monitoring area, it is possible to formulate targeted early warning strategies to promptly remind users and relevant managers to take measures to prevent the occurrence of falls. The formulation of risk early warning strategies helps to rationally allocate resources in the monitoring area, such as increasing the number of guardians, improving environmental facilities, etc., to improve the overall safety management level. Uploading the early warning strategy to the management platform can form a good data feedback mechanism, which helps to continuously improve the early warning system and improve the flexibility and response speed of the system.

[0072] Optionally, the present specification further provides a far-infrared vital sign monitoring device, comprising a far-infrared vital sign monitoring device main body, a power supply unit and an electrical control unit, wherein the power supply unit is installed inside the far-infrared vital sign monitoring device main body, the electrical control unit is electrically connected to the power supply unit, and the electrical control unit is used to charge the far-infrared vital sign monitoring device main body and control the far-infrared vital sign monitoring device main body, and the electrical control unit comprises:

[0073] The monitoring area heat map generation module is used to obtain the monitoring area heat imaging data through the far-infrared thermal imaging instrument, and generate the monitoring area heat map based on the monitoring area heat imaging data, so as to obtain the monitoring area heat map;

[0074] A user vital signs feature integration module is used to integrate the heat distribution of the monitoring area heat map, so as to obtain a monitoring area heat distribution image set, and integrate the vital signs features of users in the monitoring area according to the monitoring area heat distribution image set, so as to obtain the vital signs data of users in the monitoring area;

[0075] A center of gravity imbalance detection module is used to perform user behavior pattern recognition based on the vital sign data of the user in the monitoring area, thereby obtaining user behavior pattern data, and perform user center of gravity imbalance detection on the temperature distribution image set of the monitoring area according to the user behavior pattern data, thereby obtaining user center of gravity imbalance behavior data;

[0076] The millimeter wave fluctuation analysis module is used to obtain the millimeter wave signal set of the monitoring area through the millimeter wave module, and perform millimeter wave fluctuation detection on the millimeter wave signal set of the monitoring area, so as to obtain the millimeter wave fluctuation data of the monitoring area; perform user rapid posture change behavior analysis based on the millimeter wave fluctuation data of the monitoring area, so as to obtain the user rapid posture change behavior data;

[0077] The fall risk assessment module is used to perform a fall risk assessment on the user based on the user's center of gravity imbalance behavior data and the user's rapid posture change behavior data, so as to obtain the fall risk data of the users in the monitoring area; perform a fall risk warning strategy analysis based on the fall risk data of the users in the monitoring area, so as to obtain the user fall risk warning strategy, and upload it to the monitoring area management platform to execute the risk warning task. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0079] Figure 1 It is a schematic diagram of the steps of the far-infrared vital signs monitoring method based on deep learning of the present invention;

[0080] Figure 2Detailed step flow diagram of step S1 in the present invention;

[0081] Figure 3 Detailed step flow diagram of step S2 in the present invention;

[0082] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0083] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0084] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0085] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0086] To achieve this, please refer to Figures 1 to 3 The present invention provides a far-infrared vital sign monitoring method based on deep learning, the method comprising the following steps:

[0087] Step S1: acquiring thermal imaging data of the monitoring area through a far-infrared thermal imaging instrument, and generating a thermal map of the monitoring area based on the thermal imaging data of the monitoring area, thereby obtaining a thermal map of the monitoring area;

[0088] In this embodiment, a high-precision far-infrared thermal imager (such as a FLIR series camera) is used to perform thermal imaging monitoring of the environment within the monitoring area. The instrument is set to capture images at a rate of 30 frames per second, which is suitable for indoor and outdoor environments. The captured data is processed by an image processing algorithm (such as image smoothing and edge detection in OpenCV) to generate a heat map, which displays different temperature areas in pseudo-color. Red in the heat map indicates high temperature areas, and blue indicates low temperature areas. Through this process, hot spots and cold spots in the monitoring area can be effectively identified, providing basic data for subsequent analysis.

[0089] Step S2: integrating the heat distribution of the thermal map of the monitoring area to obtain a heat distribution image set of the monitoring area, and integrating the vital sign characteristics of users in the monitoring area according to the heat distribution image set of the monitoring area to obtain vital sign data of users in the monitoring area;

[0090] In this embodiment, the generated heat map is integrated for heat distribution, and a clustering algorithm (such as K-means) is used to segment the heat map to identify different temperature areas. The heat distribution of each area (such as mean, standard deviation, etc.) is recorded to form a heat distribution image set of the monitoring area. Subsequently, a machine learning algorithm (such as a support vector machine SVM) is used to analyze the heat distribution image set to extract the user's vital signs features, such as heart rate and respiratory rate. These vital signs data will be used for subsequent behavioral pattern recognition to form a user vital signs data set.

[0091] Step S3: performing user behavior pattern recognition based on the vital sign data of the user in the monitoring area, thereby obtaining user behavior pattern data, and performing user center of gravity imbalance detection on the temperature distribution image set in the monitoring area according to the user behavior pattern data, thereby obtaining user center of gravity imbalance behavior data;

[0092] In this embodiment, based on the user's vital signs data, a deep learning model (such as LSTM) is used to perform pattern recognition on the user's behavior, analyze the user's static and dynamic behavior, and generate user behavior pattern data. At the same time, a gravity imbalance detection algorithm is applied to identify whether the user has a gravity imbalance by comparing the user's behavior pattern data with the preset normal behavior model. If a gravity imbalance is detected, the system will generate the user's gravity imbalance behavior data and display the trajectory of the user's behavior and the change of the center of gravity through a visualization tool to assist in subsequent risk assessment.

[0093] Step S4: obtaining a millimeter wave signal set in the monitoring area through the millimeter wave module, and performing a millimeter wave fluctuation detection on the millimeter wave signal set in the monitoring area, thereby obtaining millimeter wave fluctuation data in the monitoring area; performing a user's rapid posture change behavior analysis based on the millimeter wave fluctuation data in the monitoring area, thereby obtaining the user's rapid posture change behavior data;

[0094] In this embodiment, a millimeter wave radar module (such as an FMCW radar) is deployed in the monitoring area to obtain the millimeter wave signal set in the monitoring area in real time. The signal is subjected to spectrum analysis through FFT transformation to detect millimeter wave fluctuations. By setting a threshold, signal fluctuations are identified, and then the user's rapid posture changes (such as changing from sitting to standing or turning around quickly) are analyzed. A convolutional neural network (CNN) is used to extract features from millimeter wave fluctuation data to generate user rapid posture change behavior data, providing data support for subsequent fall risk assessment.

[0095] Step S5: Perform a fall risk assessment on the user based on the user's center of gravity imbalance behavior data and the user's rapid posture change behavior data, thereby obtaining the fall risk data of the users in the monitoring area; perform a fall risk warning strategy analysis based on the fall risk data of the users in the monitoring area, thereby obtaining the user fall risk warning strategy, and upload it to the monitoring area management platform to execute the risk warning task.

[0096] In this embodiment, a risk assessment model (such as a model based on logistic regression) is used to assess the user's fall risk by combining the user's center of gravity imbalance behavior data and rapid posture change behavior data. According to the assessment results, the user's fall risk level is classified (high, medium, and low), and a personalized fall risk warning strategy (such as push reminder notification) is generated for high-risk users. Finally, all risk assessment results and warning strategies are uploaded to the monitoring area management platform, and real-time monitoring and management are carried out through a visual dashboard to achieve an effective fall risk warning task.

[0097] Optionally, step S1 specifically includes:

[0098] Step S11: Acquire thermal imaging data of the monitoring area through a far-infrared thermal imaging instrument;

[0099] In this embodiment, a highly sensitive far-infrared thermal imager (such as the FLIR T1K series) is used to collect data from a specific monitoring area. The instrument should be set at a fixed position 5 meters away from the monitoring area to ensure that its field of view covers the entire area. The acquisition process should be carried out under stable environmental conditions, avoiding wind speeds exceeding 1 meter per second to reduce external interference. The instrument should capture a thermal imaging image every 1 second, and the data should be stored in RAW format to ensure the accuracy of subsequent processing.

[0100] Step S12: performing thermal imaging data median denoising on the thermal imaging data of the monitoring area, thereby obtaining thermal imaging denoised data of the monitoring area;

[0101] In this embodiment, after obtaining the thermal imaging data, a median filter algorithm is used for denoising. A 3x3 filter window is selected to process each thermal imaging data point. For each pixel value, the median of all pixel values ​​in its neighborhood is calculated, and the median is used as a new pixel value to replace the original pixel value. This can be done using the OpenCV library in Python, and the specific code is cv2.medianBlur(image, 3), thereby reducing random noise in the image and improving image quality.

[0102] Step S13: Acquiring monitoring area environment sensing data through monitoring area environment sensors;

[0103] In this embodiment, a variety of environmental sensors are installed in the monitoring area, including temperature sensors (such as DS18B20), humidity sensors (such as DHT22) and air pressure sensors (such as BMP280). These sensors should collect data every minute and use Internet of Things technology to transmit the data to the central database in real time. In order to ensure the accuracy of the data, the sensors should be calibrated regularly, the temperature sensor should work within the range of -40°C to 125°C, and the humidity sensor should have a measurement range of 0% to 100%.

[0104] Step S14: performing environmental impact correction on the thermal imaging denoising data of the monitoring area according to the environmental sensing data of the monitoring area, thereby obtaining thermal imaging correction data of the monitoring area;

[0105] In this embodiment, the collected environmental sensor data is used to correct the de-noised thermal imaging data. First, the real-time data of ambient temperature, humidity and air pressure are input into a linear regression model to determine the degree of their influence on the thermal imaging data.

[0106] Step S15: mapping the temperature distribution of the monitoring area according to the thermal imaging correction data of the monitoring area, thereby obtaining a thermal map of the monitoring area.

[0107] In this embodiment, a temperature distribution map is generated by processing the thermal imaging correction data. The corrected thermal imaging data is converted into a temperature heat map using the imshow function in Python's Matplotlib library. By setting the color scale and color mapping (such as using the 'hot' color map), different temperature values ​​are displayed as different colors to intuitively display the temperature distribution of the monitoring area. The heat map can be exported to PNG format and displayed on the visual interface of the monitoring system for real-time monitoring and analysis.

[0108] Optionally, step S14 is specifically:

[0109] Step S141: extracting the environmental temperature characteristics and the environmental humidity characteristics of the monitoring area environmental sensor data, thereby obtaining the monitoring area environmental temperature data and the monitoring area environmental humidity data;

[0110] In this embodiment, data is collected from multiple environmental sensors (such as temperature sensors and humidity sensors) arranged in the monitoring area. The sensor needs to have high precision and high response speed. For example, a digital temperature and humidity sensor DHT22 is used, and its temperature measurement range is -40°C to 80°C and the humidity range is 0% to 100%. The sensor records the ambient temperature and humidity at a frequency of collecting data once per minute. The extracted temperature data is analyzed by calculating statistical features such as the mean value and standard deviation to obtain the ambient temperature characteristics representing the area. At the same time, the humidity data is also processed similarly, and the noise is reduced by using data smoothing techniques (such as the moving average method) to ensure that the obtained humidity characteristics are more representative.

[0111] Step S142: performing time series correlation on the monitoring area ambient temperature data and the monitoring area thermal imaging denoised data to obtain temperature-thermal radiation correlation data, and estimating the temperature influence factor based on the temperature-thermal radiation correlation data to obtain the temperature influence factor;

[0112] In this embodiment, the ambient temperature data of the monitoring area is temporally associated with the thermal imaging data obtained by a thermal imaging device (such as a FLIR thermal imager). The thermal imaging device needs to have a resolution of at least 320x240 pixels and can accurately measure thermal radiation in the range of -20°C to 150°C. By synchronizing the timestamp, each temperature data point is matched with the thermal radiation value at the corresponding time to generate a temperature-thermal radiation association data set. Using a regression analysis method (such as multivariate linear regression), the temperature influence factor is estimated based on the association data. These factors can be used to evaluate the impact of temperature changes on thermal radiation and further extract the temperature influence factor. For example, the ambient temperature data and the thermal imaging data are aligned according to the timestamp. If the ambient temperature data is T(t1), T(t2), ..., T(tn), and the thermal imaging data is R(t1), R(t2), ..., R(tn), a temperature-thermal radiation association data set is formed. The data set is represented as (T(t1), R(t1)), (T(t2), R(t2)), ..., (T(tn), R(tn)). Use correlation analysis (such as Pearson's correlation coefficient) to preliminarily assess the relationship between temperature and thermal radiation. The calculation formula is ,in is the correlation coefficient, is the temperature at a certain moment, is the average temperature, is the thermal radiation value at a certain moment, is the average value of thermal radiation. The linear regression method is used to establish the model, assuming that the relationship between thermal radiation value R and temperature value T is R = aT + b. Use regression analysis software (such as scikit-learn in Python) to fit the data and obtain model parameters a and b. At this time, the temperature impact factor can be defined as the slope a of the model, which represents the change in thermal radiation value for each unit change in temperature.

[0113] Step S143: performing time series correlation on the environmental humidity data of the monitoring area and the thermal imaging denoised data of the monitoring area to obtain humidity-thermal radiation correlation data, and estimating the humidity influence factor according to the humidity-thermal radiation correlation data to obtain the humidity influence factor;

[0114] In this embodiment, the same method is used to perform time-series correlation between the ambient humidity data and the thermal imaging data of the monitoring area. A humidity sensor (such as Honeywell HIH-4030) is used here for data acquisition, and its accuracy must reach ±3%RH. Humidity-thermal radiation correlation data is formed by matching the humidity data with the timestamp of the thermal imaging denoised data. Regression analysis methods (such as multivariate linear regression) are used to estimate the humidity influencing factor based on this correlation data to quantify the impact of humidity changes on thermal radiation. For example, the humidity data and thermal imaging data are aligned according to the timestamp to form a humidity-thermal radiation correlation data set: (H(t1), R(t1)), (H(t2), R(t2)), ..., (H(tn), R(tn)). The relationship between humidity and thermal radiation is evaluated using a correlation analysis method. The Pearson correlation coefficient is calculated to obtain the degree of correlation between humidity and thermal radiation. The calculation formula is: ,in is the correlation coefficient, is the humidity at a certain moment, is the average humidity value, is the thermal radiation value at a certain moment, is the average value of thermal radiation. The linear regression method is used to establish the model, assuming that the relationship between humidity value H and thermal radiation value R is: R=cH+d. Use regression analysis software (such as scikit-learn in Python) to fit the data and obtain model parameters c and d. At this time, the humidity impact factor can be defined as the slope c of the model, which represents the change in thermal radiation value for each unit change in humidity. Use the holdout method (such as K-fold cross validation) to verify the accuracy of the model to ensure that the model can perform well on different data sets. The model performance is evaluated by calculating the mean square error (MSE).

[0115] Step S144: constructing a linear regression thermal radiation correction model according to the temperature influencing factor and the humidity influencing factor;

[0116] In this embodiment, after obtaining the temperature influence factor and the humidity influence factor, these two factors are used as independent variables to construct a linear regression thermal radiation correction model. The Scikit-learn library in Python can be used to build the model. The output of the model will be a corrected thermal radiation value, taking into account the effects of temperature and humidity. The model parameters need to be adjusted through the training data set to ensure that the model's fit reaches an R² value higher than 0.9, indicating that the model can well explain the changes in thermal radiation.

[0117] Step S145: performing environmental impact correction on the thermal imaging denoising data of the monitoring area through a linear regression thermal radiation correction model, thereby obtaining thermal imaging correction data of the monitoring area.

[0118] In this embodiment, the thermal imaging denoising data of the monitoring area is corrected for environmental impacts by constructing a linear regression thermal radiation correction model. The correction model is applied to the original thermal imaging data to generate a corrected thermal imaging data set. This step needs to ensure verification under multiple environmental conditions to evaluate the accuracy and stability of the corrected data. For example, the data under different time periods and different environmental conditions can be corrected and verified to ensure the universality of the correction model.

[0119] Optionally, step S2 specifically includes:

[0120] Step S21: dividing the monitoring area heat map into grid areas, thereby obtaining a monitoring area grid heat map;

[0121] In this embodiment, when dividing the monitoring area heat map into grid areas, it is first necessary to determine the boundaries of the monitoring area, which can be assisted by GIS (Geographic Information System) software. The monitoring area is evenly divided into several small grids, and the recommended grid size is 10 meters × 10 meters. The center point of each grid is recorded as a coordinate and assigned a unique identifier. Using the thermal information (such as temperature or radiation intensity) in the heat map, the pixel values ​​in each grid are averaged to generate a grid heat map for subsequent data processing.

[0122] Step S22: performing pixel value statistics on the grid heat map of the monitoring area to obtain grid pixel value data, and performing grid heat calculation based on the grid pixel value data to obtain grid heat data;

[0123] In this embodiment, when counting the pixel values ​​in each grid, statistical methods such as minimum value, maximum value and mean value are used. When calculating the heat of the grid, the pixel value is first multiplied by the grid area to obtain the heat value. If the pixel value unit is degrees Celsius and the grid area is 100 square meters, the heat calculation formula is: grid heat = mean pixel value × grid area. Finally, the heat data of all grids are organized into a table for further analysis.

[0124] Step S23: performing grid heat division according to the grid heat data, thereby obtaining high heat grid data and low heat grid data;

[0125] In this embodiment, a heat threshold is set based on the grid heat data to divide the grid into high heat and low heat grids. For example, the heat threshold is set to 5000 units, and grids with heat greater than 5000 are marked as "high heat", and those less than 5000 are marked as "low heat". In this way, areas with uneven heat distribution can be quickly identified, providing basic data for subsequent heat distribution integration.

[0126] Step S24: integrating high heat distribution based on high heat grid data, thereby obtaining a high heat distribution image set of the monitoring area; integrating low heat distribution based on low heat grid data, thereby obtaining a low heat distribution image set of the monitoring area;

[0127] In this embodiment, for high-calorie grid data, a clustering algorithm (such as K-Means) can be used to integrate high-calorie areas to form a high-calorie distribution image set. The calorie value, location and distribution range of each high-calorie area should be recorded. In the processing of low-calorie grid data, the clustering method is also used to obtain the overall distribution of low-calorie areas. The final generated image set must contain the calorie information of each area to ensure information integrity.

[0128] Step S25: spatially merging the high heat distribution image set of the monitoring area and the low heat distribution image set of the monitoring area to obtain a heat distribution image set of the monitoring area;

[0129] In this embodiment, when performing spatial merging, the overlay analysis function of the GIS software can be used to merge the high heat distribution image set with the low heat distribution image set. Select a suitable merging algorithm (such as intersection, union) to obtain a monitoring area heat distribution image set containing all grid heat information. The merged image should retain the visualization effect of different heat areas so that users can quickly identify the heat distribution trend.

[0130] Step S26: integrating the vital sign features of users in the monitoring area according to the heat distribution image set in the monitoring area, thereby obtaining vital sign data of users in the monitoring area.

[0131] In this embodiment, feature integration is performed based on the heat distribution image set of the monitoring area and combined with the user's vital signs data (such as heart rate, body temperature, etc.). Through data mining technology, such as association rule mining, the relationship between heat distribution and user's vital signs is found. This can be achieved by building a model to analyze heat data and vital signs data to generate a comprehensive assessment report on the user's health status. Finally, the integration results are visualized to facilitate decision-making by relevant personnel.

[0132] Optionally, step S26 is specifically:

[0133] Step S261: extracting features of high heat distribution areas from the heat distribution image set of the monitoring area, thereby obtaining a high heat distribution image set of the monitoring area;

[0134] In this embodiment, the features of the high heat distribution area in the monitoring area heat distribution image set are extracted to obtain the monitoring area high heat distribution image set including features such as area, perimeter, and thermal radiation value.

[0135] Step S262: performing morphological recognition of the user's thermal contour according to the high-calorie distribution image set in the monitoring area, thereby obtaining a user's thermal distribution image set, and estimating the user's body temperature according to the user's thermal distribution image set, thereby obtaining the user's body temperature data;

[0136] In this embodiment, a high heat distribution image set is used to perform morphological recognition of the user's heat profile, and morphological operations such as expansion and erosion are used to enhance the profile features and remove noise. Multiple morphological processing can be performed using structural elements, and a clear user profile can be extracted in combination with an edge detection algorithm (such as the Canny algorithm). The user's body temperature is estimated based on the average heat value of the profile area, and the user's body temperature data is finally generated. These data can be processed by a temperature conversion formula (for example, conversion between Celsius and Fahrenheit).

[0137] Step S263: dividing the user heat distribution image set into user part heat image sets, thereby obtaining a user neck heat image set and a user chest heat image set;

[0138] In this embodiment, the user's heat distribution image set will be divided into multiple sub-areas, specifically including the user's neck and chest. By defining ROI (region of interest) in the thermal image, the neck and chest areas are extracted from the entire image using image segmentation technology (such as K-means clustering). After extraction, data cleaning and preprocessing are required to ensure that the acquired neck thermal image set and chest thermal image set have consistent format and resolution, which is convenient for subsequent frequency statistics and fluctuation analysis. The user's heat distribution image set is divided to extract the user's neck and chest thermal image set. A coordinate-based partitioning algorithm can also be used to first determine the standard coordinate system of the user's body shape (for example, based on the positioning of the shoulder and chest center points). Then, through image segmentation technology, the image set is divided into the neck area (the area extending upward from the chest center point, such as 10-15 cm, and the specific distance can be adjusted according to the user's body shape standard) and the chest area (a certain area surrounding the chest center point, such as 8-12 cm). This can be automatically labeled using a deep learning model. The output results will include two independent sets of thermal image sets for subsequent analysis.

[0139] Step S264: performing heat change frequency statistics based on the user's neck heat image set to obtain the user's neck heat change frequency data, and estimating the user's heart rate based on the user's neck heat change frequency data to obtain the user's heart rate data;

[0140] In this embodiment, a time series analysis is performed on the user's neck thermal image set, which is specifically implemented by calculating the amplitude and frequency of the neck thermal changes over time. Fast Fourier Transform (FFT) can be used to identify the periodic changes in neck thermal and analyze its frequency components. Based on the statistical analysis results, the thermal change frequency data of the neck is calculated, and this data will be used to estimate the heart rate. The heart rate can be obtained by calculating the number of thermal change peaks per minute, and the commonly used physiological parameter conversion formula can be used to convert thermal changes into heart rate data.

[0141] Step S265: integrating periodic heat fluctuations according to the user's chest heat image set, thereby obtaining periodic heat fluctuation data of the user's chest, and estimating the user's respiratory frequency according to the periodic heat fluctuation data of the user's chest, thereby obtaining the user's respiratory frequency data;

[0142] In this embodiment, the periodic fluctuation of the user's chest thermal image set is analyzed. The chest thermal data is smoothed using the moving average method to eliminate the interference of short-term fluctuations. Next, the periodic characteristics of the chest thermal fluctuations are calculated using the autocorrelation function. By extracting the fluctuation period and amplitude, the user's chest thermal periodic fluctuation data is obtained. Subsequently, the respiratory frequency is estimated based on this data. Common methods include calculating the respiratory frequency by counting the number of fluctuation cycles per minute and adjusting it according to the required accuracy.

[0143] Step S266: Integrate the user's body temperature data, the user's heartbeat data, and the user's respiratory rate data into user vital signs, thereby obtaining the user's vital signs data in the monitoring area.

[0144] In this embodiment, the user's body temperature data, heart rate data, and respiratory rate data are integrated to form the user's vital signs data. The weighted average method or standardization method can be used to convert each physiological parameter into a unified quantitative index to facilitate a comprehensive assessment of the user's health status. The integrated vital signs data is presented through data visualization technology (such as a line chart or radar chart) to facilitate real-time monitoring and analysis, ensuring the effectiveness of the application in medical or health monitoring scenarios.

[0145] Optionally, step S3 specifically includes:

[0146] Step S31: extracting respiratory frequency features and heart rate features according to the vital sign data of the user in the monitoring area, thereby obtaining the respiratory frequency data and the heart rate data of the user;

[0147] In this embodiment, the acquired time series signal is analyzed in the frequency domain using Fast Fourier Transform (FFT) to extract the respiratory frequency (e.g., 12-20 times per minute is normal) and heart rate (normal heart rate range is 60-100 times per minute) data. These data are then stored in the database of the user device for subsequent analysis.

[0148] Step S32: dividing the user's respiratory frequency data into user respiratory frequency time series, thereby obtaining the user's high respiratory frequency time series data and the user's low respiratory frequency time series data;

[0149] In this embodiment, the sliding window method is used to divide the data into time series according to the user's respiratory rate data. By setting a time window (for example, 30 seconds), the average respiratory rate in the window is calculated and compared with the set threshold (for example, 20 times / minute), so as to divide the respiratory rate data into high-frequency and low-frequency time series. For example, if the respiratory rate in a certain period of time exceeds 20 times / minute, the data is marked as "high respiratory rate time series", otherwise it is "low respiratory rate time series". This process can be implemented with the help of Python programming, and the Pandas library is used for time series data processing.

[0150] Step S33: dividing the user's heart rate data into user heart rate time series, thereby obtaining the user's high heart rate time series data and the user's low heart rate time series data;

[0151] In this embodiment, in the heart rate data division stage, the heart rate is divided by a sliding window method similar to the processing method of the respiratory rate. Set a time window (for example, 1 minute), calculate the average heart rate in each window, and compare it with the normal range (60-100 times per minute). If the heart rate exceeds 100 beats / minute, it is marked as "high heart rate timing"; if it is less than 60 beats / minute, it is marked as "low heart rate timing". The Matplotlib visualization tool can be used to display the results of the heart rate timing division to help users understand the trend of heart rate changes.

[0152] Step S34: performing time series intersection on the user's high respiratory rate time series data and the user's high heart rate time series data, thereby obtaining the user's action time series data; performing time series intersection on the user's low respiratory rate time series data and the user's low heart rate time series data, thereby obtaining the user's static time series data;

[0153] In this embodiment, the user's action and resting state are identified through intersection analysis. The high breathing rate time series data and the high heart rate time series data are intersected to form "user action time series data". Relatively speaking, the low breathing rate and low heart rate time series data are intersected to form "user resting time series data". This process can be achieved by setting a threshold value, and using logical operators (such as AND) to match different states. For example, if the user's breathing rate and heart rate are both in a high state within the same time window, the time period is marked as an active state.

[0154] Step S35: integrating the user behavior pattern according to the user action time series data and the user static time series data, thereby obtaining the user behavior pattern data;

[0155] In this embodiment, the user's action time series data and static time series data are combined to analyze the user's behavior pattern through machine learning algorithms (such as K-means clustering). This involves extracting features from time series data, including time continuity, frequency, and intensity, and then building a user behavior pattern model. For example, by analyzing activities and static states in different time periods, the user's daily activity patterns (such as walking, sitting, or sleeping) can be identified. This pattern model can provide data support for health monitoring or personalized health recommendations.

[0156] Step S36: Performing a user gravity center imbalance detection on the monitoring area temperature distribution image set according to the user behavior pattern data, thereby obtaining the user gravity center imbalance behavior data.

[0157] In this embodiment, the temperature distribution image set of the monitoring area is analyzed based on the user behavior pattern data to detect the user's center of gravity imbalance. Through image processing technology, computer vision algorithms (such as edge detection and feature point matching) are applied to dynamically analyze the user's actions and static states to identify whether the user is in an unbalanced state. In specific implementation, the support vector machine (SVM) algorithm in machine learning can be used to classify the center of gravity imbalance behavior, and an alarm system can be implemented through real-time monitoring and data feedback mechanism. For example, when the user's center of gravity is detected to be unbalanced, the system can send a notification through a mobile phone APP to remind the user to take necessary corrective measures.

[0158] Optionally, step S36 is specifically:

[0159] Step S361: integrating the time-series changes of pixel temperatures of the temperature distribution image set of the monitoring area, thereby obtaining the temperature change data of the monitoring area;

[0160] In this embodiment, the temperature data of each time point in the temperature distribution image set of the monitoring area are integrated through time series analysis to form a temperature change data set. The sliding window technology can be used to calculate the average temperature in each time period, generate a temperature change curve, and draw a temperature change graph to obtain the temperature change trend of the monitoring area.

[0161] Step S362: performing user behavior pattern recognition on the temperature change data of the monitoring area according to the user behavior pattern data, thereby obtaining user action behavior data and user stationary behavior data;

[0162] In this embodiment, the user's dynamic behavior (such as walking, running) and static behavior (such as standing still, sitting down) are identified by analyzing the user's behavior pattern data. The changes in breathing frequency and heart rate in the user's behavior can be classified using a machine learning algorithm (such as a support vector machine or a decision tree), and the temperature changes caused by these feature changes can be obtained by combining the time series data. The temperature changes caused by these physical changes are associated with the temperature change data of the monitoring area, thereby identifying the user's action behavior data (such as the start and end time of each movement) and the user's static behavior data (such as the static duration).

[0163] Step S363: performing a behavior conversion time point intersection operation on the user action behavior data and the user static behavior data, thereby obtaining user behavior conversion time point data;

[0164] In this embodiment, the obtained user action behavior data and static behavior data are combined to first identify the time point of the user behavior transition. For example, when the user changes from an action state to a static state, the time point is recorded. Then, an intersection operation (for example, an intersection operation in set theory) is used to integrate all the behavior transition time points to obtain unified time point data, which will be used for subsequent analysis.

[0165] Step S364: extracting the characteristics of the physical sign changes at the behavior conversion time point according to the user behavior conversion time point data, thereby obtaining the respiratory frequency change data at the behavior conversion time point and the heart beat rate change data at the behavior conversion time point;

[0166] In this embodiment, by analyzing the obtained user behavior conversion time point data, the change characteristics of the respiratory frequency and heart rate are extracted. According to the physical signs before and after each behavior conversion time point, the change range of the respiratory frequency and heart rate is calculated to form a set of data.

[0167] Step S365: performing statistics of instantaneous high vital sign changes on the respiratory frequency change data at the behavior conversion time point and the heart beat rate change data at the behavior conversion time point, respectively, so as to obtain instantaneous high respiratory frequency change time point data and instantaneous high heart rate change time point data;

[0168] In this embodiment, the respiratory rate change data and heart rate change data extracted in step S364 are statistically analyzed for instantaneous high changes. First, a threshold is set, for example, an increase in respiratory rate of more than 20% or an increase in heart rate of more than 30 times / minute is considered to be "instantaneous high change". The data is analyzed point by point, and the time point data that meets the conditions is identified and recorded as the instantaneous high respiratory rate change time point and the instantaneous high heart rate change time point.

[0169] Step S366: Perform time point intersection operation on the instantaneous high respiratory frequency change time point data and the instantaneous high heart rate change time point data to obtain the user's center of gravity imbalance time point data, and integrate the vital sign characteristics of the user's center of gravity imbalance time point data to obtain the user's center of gravity imbalance behavior data.

[0170] In this embodiment, the intersection operation is performed on the instantaneous high respiratory frequency change time point data and the instantaneous high heart rate change time point data to determine the time point when the user may have a center of gravity imbalance. The time points in the two data sets are analyzed to find records that appear at the same time. These records are the time points of the center of gravity imbalance. Subsequently, the center of gravity imbalance time point is integrated with the user's vital sign data at that time point to generate the user's center of gravity imbalance behavior data, which is convenient for analyzing changes in their physiological state and possible risk warnings.

[0171] Optionally, step S4 is specifically:

[0172] Step S41: obtaining a millimeter wave signal set in the monitoring area through a millimeter wave module, and performing millimeter wave denoising on the millimeter wave signal set in the monitoring area, thereby obtaining a millimeter wave denoised signal set in the monitoring area;

[0173] In this embodiment, a millimeter wave signal set of the monitoring area is obtained through a millimeter wave module. An integrated millimeter wave sensor array can be used, which can operate in the 30 GHz to 300 GHz frequency band and is suitable for indoor or outdoor environments. First, the sensor is configured to collect signals at specific time intervals (for example, 1000 times per second), so that a complete signal set in the monitoring area can be obtained. After acquiring the signal, the collected signal is processed by an advanced denoising algorithm (such as wavelet transform denoising or Kalman filtering) to reduce environmental noise and multipath effects, thereby obtaining a clearer millimeter wave denoised signal set to ensure the accuracy of subsequent data processing.

[0174] Step S42: performing Fourier transform on the millimeter wave denoised signal set in the monitoring area to obtain the millimeter wave spectrum in the monitoring area, and performing millimeter wave instantaneous fluctuation detection based on the millimeter wave spectrum in the monitoring area to obtain millimeter wave fluctuation data in the monitoring area;

[0175] In this embodiment, the millimeter wave denoised signal set of the monitoring area is Fourier transformed to obtain the millimeter wave spectrum of the monitoring area. The fast Fourier transform (FFT) algorithm can be used to process the time domain signal to obtain its frequency domain representation. By setting an appropriate frequency range (such as 1 GHz to 100 GHz), the spectral characteristics of the signal can be analyzed. Subsequently, by analyzing the amplitude and phase of the spectrum, the instantaneous fluctuation characteristics, such as the peak value and frequency offset of the signal, are identified, and the instantaneous fluctuation of the millimeter wave is detected to form millimeter wave fluctuation data.

[0176] Step S43: performing feature selection according to the millimeter wave fluctuation data of the monitoring area, so as to obtain the millimeter wave fluctuation rate data and the millimeter wave fluctuation duration data of the monitoring area;

[0177] In this embodiment, feature selection is performed based on the millimeter wave fluctuation data of the monitoring area. Feature selection algorithms (such as principal component analysis (PCA) or random forest feature selection) can be applied to extract key features from the fluctuation data, such as fluctuation rate (the rate of change of fluctuation amplitude) and fluctuation duration (the length of time from the start to the end of the fluctuation). These features can help identify patterns of user behavior, such as fast movement or stationary behavior, thereby obtaining millimeter wave fluctuation rate data and duration data in the monitoring area.

[0178] Step S44: performing high-value fluctuation rate statistics on the millimeter wave fluctuation rate data in the monitoring area, thereby obtaining high-value fluctuation rate time point data; performing short-duration statistics on the millimeter wave fluctuation duration data in the monitoring area, thereby obtaining short-duration fluctuation time point data;

[0179] In this embodiment, high-value fluctuation rate statistics are performed on the millimeter wave fluctuation rate data to identify the time points of high-value fluctuations. In specific implementation, a threshold value can be set (such as when the fluctuation rate exceeds a preset value), and the rate data can be statistically analyzed to record all time points that exceed the threshold value. At the same time, short-duration statistics are performed on the fluctuation duration data, for example, a threshold value is set (such as when the duration is less than a certain number of seconds), and all time points of short-duration fluctuations are recorded. Through these statistics, high-value fluctuation rate time point data and short-duration fluctuation time point data can be formed to provide a basis for subsequent data analysis.

[0180] Step S45: performing a time point intersection operation on the high-volume fluctuation rate time point data and the short-duration fluctuation time point data, thereby obtaining the user's rapid posture change time point data;

[0181] In this embodiment, the time point intersection operation is performed on the high-volume fluctuation rate time point data and the short-duration fluctuation time point data to obtain the time points of the user's rapid posture change. Set operations can be used to find the time points that meet both high fluctuation and short duration, forming a new time point set. These time points correspond to the user's possible rapid posture change behaviors, such as turning around quickly, sitting down or standing up, etc. Through such analysis, the user behavior pattern can be further refined.

[0182] Step S46: Integrate the millimeter wave behavior features at the time point according to the user's rapid posture change time point data, so as to obtain the user's rapid posture change behavior data.

[0183] In this embodiment, the millimeter wave behavior characteristics of the time point are integrated according to the time point data of the user's rapid posture change to obtain the behavior data of the user's rapid posture change. For each time point, the fluctuation data before and after it can be analyzed to extract the millimeter wave behavior characteristics of the user at a specific moment, such as measuring the change in the amplitude of the millimeter wave signal to capture the strength and amplitude of the user's action; using the time information in the signal to calculate the speed and acceleration of the user's action; analyzing the spectrum of the signal to extract frequency domain features to identify different action modes. The extracted fluctuation features are then integrated into a feature vector, which includes the statistical features of the amplitude change (mean, standard deviation, maximum value, minimum value, etc.), the average value and change rate of the speed and acceleration, and the frequency features (such as the main frequency, frequency band energy, etc.). The feature vector can include a machine learning algorithm (such as a support vector machine or a neural network) to classify and identify the integrated behavior features, and finally generate the user's rapid posture change behavior data, which can be used in application scenarios such as smart home control, security monitoring or motion analysis.

[0184] Optionally, step S5 specifically includes:

[0185] Step S51: classifying the user's center of gravity imbalance behavior data by the level of physical sign change, thereby obtaining imbalance behavior level classification data;

[0186] In this embodiment, a threshold standard is set to divide the magnitude of the physical sign changes into several levels (for example, slight imbalance, moderate imbalance, and severe imbalance). Specifically, if the magnitude of the physical sign change is between 10% and 20%, it is classified as slight imbalance; if it is between 20% and 30%, it is classified as moderate imbalance; if it exceeds 30%, it is classified as severe imbalance. This classification method can help provide clear data basis for subsequent analysis.

[0187] Step S52: classifying the user's rapid posture change behavior data by millimeter wave fluctuation amplitude, thereby obtaining rapid posture change behavior classification data;

[0188] In this embodiment, by setting a threshold standard, the amplitude of millimeter wave fluctuations is divided into several levels (for example: slight posture change, moderate posture change, and severe posture change). Specifically, if the amplitude of millimeter wave fluctuations is between 3dB-5dB, it is classified as slight imbalance; between 5dB-10dB, it is classified as moderate imbalance; and more than 10dB is classified as severe imbalance. This classification method can help provide clear data basis for subsequent analysis.

[0189] Step S53: constructing a fall risk assessment model according to the imbalance behavior classification data and the rapid posture change behavior classification data;

[0190] In this embodiment, a fall risk assessment model is constructed by combining the obtained imbalance behavior grade classification data and the rapid posture change behavior grade classification data using a machine learning algorithm (such as a random forest or a support vector machine). The imbalance behavior grade and the posture change behavior grade are selected as model input features, and the model is trained in combination with the user's historical fall records or other users' fall records. The model performance is evaluated through cross-validation to ensure its generalization ability on unknown data. The model outputs a fall risk score for each user to form a personalized assessment result to help identify high-risk users. Identify how different levels of imbalance and posture changes affect the risk of falling, thereby obtaining personalized assessment results for each user.

[0191] Step S54: performing a fall risk assessment on the user's center of gravity imbalance behavior data and the user's rapid posture change behavior data through a fall risk assessment model, thereby obtaining the user's fall risk data in the monitoring area;

[0192] In this embodiment, the constructed fall risk assessment model is used to evaluate the real-time monitored user center of gravity imbalance behavior data and rapid posture change behavior data. The evaluation results will be presented in the form of scores ranging from 0 to 100, and the higher the score, the greater the risk of falling. A threshold can be set: if the score exceeds 70, it is judged as a high risk and corresponding early warning measures need to be taken.

[0193] Step S55: Perform fall risk warning strategy analysis based on the fall risk data of users in the monitoring area, so as to obtain the user fall risk warning strategy, and upload it to the monitoring area management platform to execute the risk warning task.

[0194] In this embodiment, after obtaining the fall risk data of users in the monitoring area, a risk warning strategy analysis is performed. Based on the user's fall risk score, a decision tree algorithm is used to design a corresponding warning strategy. The warning strategy may include sending an alarm notification (through a mobile phone application or wearable device vibration) to the guardian of a user with a high fall risk, and sending a warning message to the manager of the monitoring area. The strategy can also be optimized through machine learning so that it is continuously updated and improved based on historical fall event data, thereby improving the accuracy and timeliness of the warning. Ultimately, all strategies and warning information will be uploaded to the monitoring area management platform in real time for staff to review and take further action.

[0195] Optionally, the present specification further provides a far-infrared vital sign monitoring device, comprising a far-infrared vital sign monitoring device main body, a power supply unit and an electrical control unit, wherein the power supply unit is installed inside the far-infrared vital sign monitoring device main body, the electrical control unit is electrically connected to the power supply unit, and the electrical control unit is used to charge the far-infrared vital sign monitoring device main body and control the far-infrared vital sign monitoring device main body, and the electrical control unit comprises:

[0196] The monitoring area heat map generation module is used to obtain the monitoring area heat imaging data through the far-infrared thermal imaging instrument, and generate the monitoring area heat map based on the monitoring area heat imaging data, so as to obtain the monitoring area heat map;

[0197] A user vital signs feature integration module is used to integrate the heat distribution of the monitoring area heat map, so as to obtain a monitoring area heat distribution image set, and integrate the vital signs features of users in the monitoring area according to the monitoring area heat distribution image set, so as to obtain the vital signs data of users in the monitoring area;

[0198] A center of gravity imbalance detection module is used to perform user behavior pattern recognition based on the vital sign data of the user in the monitoring area, thereby obtaining user behavior pattern data, and perform user center of gravity imbalance detection on the temperature distribution image set of the monitoring area according to the user behavior pattern data, thereby obtaining user center of gravity imbalance behavior data;

[0199] The millimeter wave fluctuation analysis module is used to obtain the millimeter wave signal set of the monitoring area through the millimeter wave module, and perform millimeter wave fluctuation detection on the millimeter wave signal set of the monitoring area, so as to obtain the millimeter wave fluctuation data of the monitoring area; perform user rapid posture change behavior analysis based on the millimeter wave fluctuation data of the monitoring area, so as to obtain the user rapid posture change behavior data;

[0200] The fall risk assessment module is used to perform a fall risk assessment on the user based on the user's center of gravity imbalance behavior data and the user's rapid posture change behavior data, so as to obtain the fall risk data of the users in the monitoring area; perform a fall risk warning strategy analysis based on the fall risk data of the users in the monitoring area, so as to obtain the user fall risk warning strategy, and upload it to the monitoring area management platform to execute the risk warning task.

[0201] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0202] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A far-infrared vital signs monitoring method, characterized in that: The following steps are involved: Step S1: acquiring thermal imaging data of the monitoring area through a far-infrared thermal imaging instrument, and generating a thermal map of the monitoring area based on the thermal imaging data of the monitoring area, thereby obtaining a thermal map of the monitoring area; the thermal map of the monitoring area is a set of temperature distribution images of the monitoring area; Step S2: integrating the heat distribution of the thermal map of the monitoring area to obtain a heat distribution image set of the monitoring area, and integrating the vital sign characteristics of users in the monitoring area according to the heat distribution image set of the monitoring area to obtain vital sign data of users in the monitoring area; Step S3: Performing user behavior pattern recognition based on the user vital sign data in the monitoring area to obtain user behavior pattern data, and performing user center of gravity imbalance detection on the temperature distribution image set in the monitoring area according to the user behavior pattern data to obtain user center of gravity imbalance behavior data; the user center of gravity imbalance detection is specifically as follows: Integrate the pixel temperature time series changes of the temperature distribution image set of the monitoring area to obtain the temperature change data of the monitoring area; Obtain user action behavior data and user stationary behavior data based on user behavior pattern data and monitoring area temperature change data; Performing a behavior conversion time point intersection operation on the user's action behavior data and the user's static behavior data, thereby obtaining the user's behavior conversion time point data; Extract the features of the physical sign changes at the behavior conversion time point based on the user behavior conversion time point data, so as to obtain the respiratory frequency change data at the behavior conversion time point and the heart beat rate change data at the behavior conversion time point; The respiratory frequency change data at the behavior conversion time point and the heart beat rate change data at the behavior conversion time point are respectively statistically analyzed to obtain the instantaneous high respiratory frequency change time point data and the instantaneous high heart rate change time point data; Performing a time point intersection operation on the instantaneous high respiratory frequency change time point data and the instantaneous high heart rate change time point data to obtain the user's center of gravity imbalance time point data, and integrating the vital sign characteristics of the user's center of gravity imbalance time point data to obtain the user's center of gravity imbalance behavior data; Step S4: obtaining a millimeter wave signal set in the monitoring area through the millimeter wave module, and performing a millimeter wave fluctuation detection on the millimeter wave signal set in the monitoring area, thereby obtaining millimeter wave fluctuation data in the monitoring area; performing a user's rapid posture change behavior analysis based on the millimeter wave fluctuation data in the monitoring area, thereby obtaining the user's rapid posture change behavior data; Step S5: performing a fall risk assessment on the user based on the user's center of gravity imbalance behavior data and the user's rapid posture change behavior data, thereby obtaining the user's fall risk data in the monitoring area; The fall risk warning strategy analysis is performed based on the fall risk data of users in the monitoring area to obtain the user fall risk warning strategy and upload it to the monitoring area management platform to execute the risk warning task.

2. The far-infrared vital signs monitoring method according to claim 1, characterized in that: Step S1 is specifically as follows: Step S11: Acquire thermal imaging data of the monitoring area through a far-infrared thermal imaging instrument; Step S12: performing thermal imaging data median denoising on the thermal imaging data of the monitoring area, thereby obtaining thermal imaging denoised data of the monitoring area; Step S13: Acquiring monitoring area environment sensing data through monitoring area environment sensors; Step S14: performing environmental impact correction on the thermal imaging denoising data of the monitoring area according to the environmental sensing data of the monitoring area, thereby obtaining thermal imaging correction data of the monitoring area; Step S15: mapping the temperature distribution of the monitoring area according to the thermal imaging correction data of the monitoring area, thereby obtaining a thermal map of the monitoring area.

3. The far-infrared vital signs monitoring method according to claim 2, characterized in that: Step S14 is specifically as follows: Step S141: extracting the environmental temperature characteristics and the environmental humidity characteristics of the monitoring area environmental sensor data, thereby obtaining the monitoring area environmental temperature data and the monitoring area environmental humidity data; Step S142: performing time series correlation on the monitoring area ambient temperature data and the monitoring area thermal imaging denoised data to obtain temperature-thermal radiation correlation data, and estimating the temperature influence factor based on the temperature-thermal radiation correlation data to obtain the temperature influence factor; Step S143: performing time series correlation on the environmental humidity data of the monitoring area and the thermal imaging denoised data of the monitoring area to obtain humidity-thermal radiation correlation data, and estimating the humidity influence factor according to the humidity-thermal radiation correlation data to obtain the humidity influence factor; Step S144: constructing a linear regression thermal radiation correction model according to the temperature influencing factor and the humidity influencing factor; Step S145: performing environmental impact correction on the thermal imaging denoising data of the monitoring area through a linear regression thermal radiation correction model, thereby obtaining thermal imaging correction data of the monitoring area.

4. The far-infrared vital signs monitoring method according to claim 1, characterized in that: Step S2 is specifically as follows: Step S21: dividing the monitoring area heat map into grid areas, thereby obtaining a monitoring area grid heat map; Step S22: performing pixel value statistics on the grid heat map of the monitoring area to obtain grid pixel value data, and performing grid heat calculation based on the grid pixel value data to obtain grid heat data; Step S23: performing grid heat division according to the grid heat data, thereby obtaining high heat grid data and low heat grid data; Step S24: integrating high heat distribution based on high heat grid data, thereby obtaining a high heat distribution image set of the monitoring area; integrating low heat distribution based on low heat grid data, thereby obtaining a low heat distribution image set of the monitoring area; Step S25: spatially merging the high heat distribution image set of the monitoring area and the low heat distribution image set of the monitoring area to obtain a heat distribution image set of the monitoring area; Step S26: integrating the vital sign features of users in the monitoring area according to the heat distribution image set in the monitoring area, thereby obtaining vital sign data of users in the monitoring area.

5. The far-infrared vital signs monitoring method according to claim 4, characterized in that: Step S26 is specifically as follows: Step S261: extracting features of high heat distribution areas from the heat distribution image set of the monitoring area, thereby obtaining a high heat distribution image set of the monitoring area; Step S262: performing morphological recognition of the user's thermal contour according to the high-calorie distribution image set in the monitoring area, thereby obtaining a user's thermal distribution image set, and estimating the user's body temperature according to the user's thermal distribution image set, thereby obtaining the user's body temperature data; Step S263: dividing the user heat distribution image set into user part heat image sets, thereby obtaining a user neck heat image set and a user chest heat image set; Step S264: performing heat change frequency statistics based on the user's neck heat image set to obtain the user's neck heat change frequency data, and estimating the user's heart rate based on the user's neck heat change frequency data to obtain the user's heart rate data; Step S265: integrating periodic heat fluctuations according to the user's chest heat image set, thereby obtaining periodic heat fluctuation data of the user's chest, and estimating the user's respiratory frequency according to the periodic heat fluctuation data of the user's chest, thereby obtaining the user's respiratory frequency data; Step S266: Integrate the user's body temperature data, the user's heartbeat data, and the user's respiratory rate data into user vital signs, thereby obtaining the user's vital signs data in the monitoring area.

6. The far-infrared vital signs monitoring method according to claim 1, characterized in that: Step S3 is specifically as follows: Step S31: extracting respiratory frequency features and heart rate features according to the vital sign data of the user in the monitoring area, thereby obtaining the respiratory frequency data and the heart rate data of the user; Step S32: dividing the user's respiratory frequency data into user respiratory frequency time series, thereby obtaining the user's high respiratory frequency time series data and the user's low respiratory frequency time series data; Step S33: dividing the user's heart rate data into user heart rate time series, thereby obtaining the user's high heart rate time series data and the user's low heart rate time series data; Step S34: performing time series intersection on the user's high breathing rate time series data and the user's high heart rate time series data, thereby obtaining the user's action time series data; Performing time series intersection on the user's low breathing rate time series data and the user's low heart rate time series data to obtain the user's static time series data; Step S35: integrating the user behavior pattern according to the user action time series data and the user static time series data, thereby obtaining the user behavior pattern data; Step S36: Performing a user gravity center imbalance detection on the monitoring area temperature distribution image set according to the user behavior pattern data, thereby obtaining the user gravity center imbalance behavior data.

7. The far-infrared vital signs monitoring method according to claim 1, characterized in that: Step S4 is specifically as follows: Step S41: obtaining a millimeter wave signal set in the monitoring area through a millimeter wave module, and performing millimeter wave denoising on the millimeter wave signal set in the monitoring area, thereby obtaining a millimeter wave denoised signal set in the monitoring area; Step S42: performing Fourier transform on the millimeter wave denoised signal set in the monitoring area to obtain the millimeter wave spectrum in the monitoring area, and performing millimeter wave instantaneous fluctuation detection based on the millimeter wave spectrum in the monitoring area to obtain millimeter wave fluctuation data in the monitoring area; Step S43: performing feature selection according to the millimeter wave fluctuation data of the monitoring area, so as to obtain the millimeter wave fluctuation rate data and the millimeter wave fluctuation duration data of the monitoring area; Step S44: performing high-value fluctuation rate statistics on the millimeter wave fluctuation rate data in the monitoring area, thereby obtaining high-value fluctuation rate time point data; Perform short-duration statistics on the millimeter wave fluctuation duration data in the monitoring area to obtain short-duration fluctuation time point data; Step S45: performing a time point intersection operation on the high-volume fluctuation rate time point data and the short-duration fluctuation time point data, thereby obtaining the user's rapid posture change time point data; Step S46: Integrate the millimeter wave behavior features at the time point according to the user's rapid posture change time point data, so as to obtain the user's rapid posture change behavior data.

8. The far-infrared vital signs monitoring method according to claim 1, characterized in that: Step S5 is specifically as follows: Step S51: classifying the user's center of gravity imbalance behavior data by the level of physical sign change, thereby obtaining imbalance behavior level classification data; Step S52: classifying the user's rapid posture change behavior data by millimeter wave fluctuation amplitude, thereby obtaining rapid posture change behavior classification data; Step S53: constructing a fall risk assessment model according to the imbalance behavior classification data and the rapid posture change behavior classification data; Step S54: performing a fall risk assessment on the user's center of gravity imbalance behavior data and the user's rapid posture change behavior data through a fall risk assessment model, thereby obtaining the user's fall risk data in the monitoring area; Step S55: Perform fall risk warning strategy analysis based on the fall risk data of users in the monitoring area, so as to obtain the user fall risk warning strategy, and upload it to the monitoring area management platform to execute the risk warning task.

9. A far-infrared vital signs monitoring device, characterized in that: The device comprises a main body of a far-infrared vital signs monitoring device, a power supply unit and an electrical control unit. The power supply unit is installed inside the main body of the far-infrared vital signs monitoring device. The electrical control unit is electrically connected to the power supply unit. The electrical control unit is used to charge the main body of the far-infrared vital signs monitoring device and control the main body of the far-infrared vital signs monitoring device. The electrical control unit comprises: A monitoring area thermal map generation module is used to obtain the monitoring area thermal imaging data through a far-infrared thermal imaging instrument, and generate a monitoring area thermal map based on the monitoring area thermal imaging data, thereby obtaining a monitoring area thermal map; the monitoring area thermal map is a monitoring area temperature distribution image set; A user vital signs feature integration module is used to integrate the heat distribution of the monitoring area heat map, so as to obtain a monitoring area heat distribution image set, and integrate the vital signs features of users in the monitoring area according to the monitoring area heat distribution image set, so as to obtain the vital signs data of users in the monitoring area; The center of gravity imbalance detection module is used to perform user behavior pattern recognition based on the vital sign data of the user in the monitoring area, so as to obtain the user behavior pattern data, and perform user center of gravity imbalance detection on the temperature distribution image set in the monitoring area according to the user behavior pattern data, so as to obtain the user center of gravity imbalance behavior data; the user center of gravity imbalance detection is specifically: Integrate the pixel temperature time series changes of the temperature distribution image set of the monitoring area to obtain the temperature change data of the monitoring area; Obtain user action behavior data and user stationary behavior data based on user behavior pattern data and monitoring area temperature change data; Performing a behavior conversion time point intersection operation on the user's action behavior data and the user's static behavior data, thereby obtaining the user's behavior conversion time point data; Extract the features of the physical sign changes at the behavior conversion time point based on the user behavior conversion time point data, so as to obtain the respiratory frequency change data at the behavior conversion time point and the heart beat rate change data at the behavior conversion time point; The respiratory frequency change data at the behavior conversion time point and the heart beat rate change data at the behavior conversion time point are respectively statistically analyzed to obtain the instantaneous high respiratory frequency change time point data and the instantaneous high heart rate change time point data; Performing a time point intersection operation on the instantaneous high respiratory frequency change time point data and the instantaneous high heart rate change time point data to obtain the user's center of gravity imbalance time point data, and integrating the vital sign characteristics of the user's center of gravity imbalance time point data to obtain the user's center of gravity imbalance behavior data; The millimeter wave fluctuation analysis module is used to obtain the millimeter wave signal set of the monitoring area through the millimeter wave module, and perform millimeter wave fluctuation detection on the millimeter wave signal set of the monitoring area, so as to obtain the millimeter wave fluctuation data of the monitoring area; perform user rapid posture change behavior analysis based on the millimeter wave fluctuation data of the monitoring area, so as to obtain the user rapid posture change behavior data; The fall risk assessment module is used to perform a fall risk assessment on the user based on the user's center of gravity imbalance behavior data and the user's rapid posture change behavior data, so as to obtain the fall risk data of the users in the monitoring area; perform a fall risk warning strategy analysis based on the fall risk data of the users in the monitoring area, so as to obtain the user fall risk warning strategy, and upload it to the monitoring area management platform to execute the risk warning task.

Images