Medical vital sign monitoring system under 6G network based on artificial intelligence

Abstract

The invention discloses a medical vital sign monitoring system under a 6G network based on artificial intelligence, belongs to the technical field of medical monitoring, and solves the problems that an existing system is rigid in sampling, low in decision accuracy and free of closed loop execution. Comprising a terminal sensing module, a 6G edge AI collaborative decision module and an execution control module. The terminal sensing module collects physical sign data and extracts three types of features, and terminal and network parameters are integrated to form a state vector; the 6G edge AI collaborative decision-making module constructs an end-network integrated training sample, trains an exclusive AI decision-making model for various physical sign data, and is deployed at a 6G edge node to output personalized decision-making parameters; the execution control module completes refined management and control according to parameters, monitors and locally corrects operation abnormity in real time, and feeds back execution data to the decision module to realize model iteration; according to the system and the method, physical sign data dynamic acquisition, end-network collaborative intelligent decision and whole-process closed-loop optimization are realized, the acquisition quality and the terminal energy consumption are considered, and the monitoring real-time performance and the system stability are improved.

Description

Technical Field

[0001] This invention belongs to the field of medical monitoring technology, specifically relating to a medical vital signs monitoring system based on artificial intelligence under a 6G network. Background Technology

[0002] The field of medical vital sign monitoring is undergoing intelligent upgrades by incorporating 6G and AI. However, existing monitoring systems still have technical shortcomings in practical applications, with prominent issues such as poor data acquisition adaptability, low decision-making accuracy, and insufficient execution closure. Therefore, this invention urgently needs to solve the following technical problems: The fixed sampling strategy cannot dynamically adjust the density according to changes in vital sign data, and it does not integrate terminal and network parameters, resulting in a lack of data dimensions for decision-making. The lack of dedicated decision-making models for different vital sign data makes it difficult to balance multiple objectives such as data collection, power consumption, and transmission, resulting in high model deployment latency and insufficient real-time performance. The execution process is poorly managed, lacks real-time parameter monitoring and local adaptive correction, and has no execution data feedback mechanism, making it impossible to iteratively optimize the model. To address this, we propose an AI-based medical vital sign monitoring system under a 6G network. Summary of the Invention

[0003] The purpose of this invention is to provide a medical vital signs monitoring system based on artificial intelligence under a 6G network, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a medical vital signs monitoring system based on artificial intelligence under a 6G network, comprising: Terminal sensing module: Collects various vital sign data, records timestamps and forms time series, and extracts three types of features: data fluctuation, continuity and trend; synchronously collects terminal operation and 6G network basic parameters, and integrates features and parameters to form terminal state vectors corresponding to various vital sign data; 6G Edge AI Collaborative Decision Module: Acquires 6G operating parameters from the network side and forms a network state vector. Constructs training samples by splicing historical samples of terminal and network state vectors for various vital sign data. Sets three types of boundary constraints: sampling, power consumption, and transmission. Constructs an objective function and defines an output dimension formula. After training and deployment, a dedicated AI decision model is obtained. Substitutes the real-time terminal and network state vectors into the output sampling, sleep, and transmission decision parameters. The execution control module calculates the sampling cycle and collects data based on the sampling, hibernation, and transmission decision parameters of various vital sign data. It executes hibernation and transmission strategies according to rules, monitors the actual operating parameters throughout the process, initiates local adaptive correction when the parameters do not meet the constraints, and finally integrates the relevant execution data to generate feedback information for feedback, providing data support for model iterative optimization.

[0005] Preferably, the specific process of collecting various vital sign data, recording timestamps, and forming a time series is as follows: The data acquisition terminal collects various types of medical vital sign data; a baseline sampling period is set for each type of vital sign data, and the timestamp corresponding to each data point is recorded synchronously. The sampling points are numbered according to the collection order, and the collection window is a preset single data collection period containing a fixed number of sampling points. The data points collected for each type of vital signs are arranged in chronological order according to their corresponding timestamps, forming a time-series data sequence for each type of vital signs. Each data node in the time-series data sequence is the medical vital signs data collected from the corresponding sampling point.

[0006] The preferred process for extracting the three types of features—fluctuation, continuity, and trend—from data is as follows: The data fluctuation characteristic is obtained by calculating the mean value of the difference in vital sign data between adjacent sampling points within a single acquisition window. This characteristic value represents the magnitude of data change per unit time. The data continuity feature is obtained by calculating the ratio of the number of valid sampling points within a single acquisition window to the total number of sampling points. This feature value characterizes the completeness of data acquisition and ranges from 0 to 1. The data trend feature is obtained by calculating the slope of the change of vital sign data with the corresponding sampling timestamp within a single acquisition window. This feature value represents the overall direction of data change.

[0007] Preferably, the specific process for forming terminal state vectors corresponding to various vital sign data is as follows: Based on the baseline sampling period of various vital sign data, the operating parameters of the medical vital sign data acquisition terminal are collected synchronously. The operating parameters include: remaining power, device operating temperature, and signal reception strength. The collected terminal operating parameters are then normalized and dimensionless. Simultaneously, the network basic parameters of the 6G serving cell currently accessed by the terminal are obtained, including uplink signal strength and initial transmission delay, and the network basic parameters are normalized and dimensionless. The fluctuation, continuity, and trend characteristics of each type of vital sign data are integrated to form a feature sub-vector for the corresponding category. The feature sub-vector of each type of vital sign data is then integrated with the normalized terminal operating parameters and 6G network basic parameters to form a terminal state vector corresponding to each type of vital sign data.

[0008] Preferably, the specific process of acquiring 6G operating parameters from the network side and forming a network state vector, and constructing training samples by splicing historical samples of various vital sign data from the terminal and network state vectors, is as follows: The network side 6G network operation parameters of the current access serving cell of the terminal are obtained through the 6G base station standardized network management interface. The operation parameters include: cell load rate, uplink available bandwidth, link packet loss rate, and real-time transmission latency. All acquired network-side 6G network operation parameters are normalized and dimensionless, and then organized to form a network state vector. For each type of vital sign data, historical operational data of that type of data in medical monitoring scenarios is obtained. The terminal state vector corresponding to each sample in the historical operational data is concatenated with the network state vector to obtain the historical training sample vector corresponding to each type of vital sign data.

[0009] Preferably, the specific process of setting three types of boundary constraints—sampling, power consumption, and transmission—to construct the objective function and define the output dimension formula is as follows: Sampling constraints: The adjusted sampling period shall not be lower than the minimum sampling period for vital signs data, and shall not exceed the baseline sampling period for vital signs data; Power consumption constraint: The total power consumption of the terminal shall not exceed the preset dynamic power consumption threshold; Transmission constraints: Actual transmission delay and link packet loss rate shall not exceed the system's preset upper limit; For each historical training sample corresponding to vital sign data, an objective function is constructed with the goals of optimal data continuity, lowest terminal power consumption, and highest network transmission efficiency. The sampling adjustment coefficient, terminal global sleep duration, and transmission mode identifier are used as the core output dimensions. The corresponding feature values, parameters and thresholds are combined to complete the quantitative calculation and judgment of each dimension. The transmission mode identifier is 0 for instant transmission and 1 for buffered batch transmission.

[0010] The preferred approach involves obtaining a dedicated AI decision-making model, inputting it into the real-time terminal and network state vectors, and outputting the specific process of sampling, sleeping, and transmitting decision parameters as follows: Using a multi-objective optimization model as the baseline model, the training sample vectors are divided into training and validation sets. Model adaptation training is carried out by combining three types of global constraints, objective function, and output dimension. The training set is input into the model, and the parameters are iteratively adjusted with global constraints as the boundary and the objective function maximization as the direction. After each iteration, the model is validated with a validation set until the decision error is lower than the preset threshold to obtain a suitable AI decision model. The AI ​​decision model is then deployed on 6G edge nodes, and the dedicated model training and deployment for various vital sign data are completed in the same way. The real-time terminal and network state vectors of each type of data are concatenated into an input vector, which is then substituted into the corresponding model to output the sampling adjustment coefficients, the terminal global sleep duration, and the terminal global transmission mode.

[0011] Preferably, based on the sampling, dormancy, and transmission decision parameters of various vital sign data, the specific process of calculating the sampling cycle, collecting data, and executing the dormancy and transmission strategy according to the rules is as follows: The temporary execution sampling period is calculated by combining the sampling adjustment coefficient with the baseline sampling period of the corresponding vital sign data. This period is then limited to the range of the preset sampling period to obtain the final actual execution sampling period, which is used as the sampling period for the next acquisition window to carry out corresponding data acquisition. When the terminal has no corresponding data acquisition task and no related cached data to be transmitted, it enters a low-power sleep state according to the global sleep duration. It will automatically wake up when the sleep duration reaches the target or when there is a new data acquisition or transmission task. The execution strategy is based on the transmission mode identifier. When the identifier is 0, the current acquisition window data is sent in real time using instant transmission. When the identifier is 1, the data of multiple acquisition windows is cached and then sent as a unified frame.

[0012] Preferably, the specific process of monitoring actual operating parameters throughout the process and initiating local adaptive correction when parameters do not meet constraints is as follows: Throughout the sampling, hibernation, and transmission process, the terminal monitors the execution status of vital sign data in real time, collects and calculates actual operating parameters, including actual continuous characteristics of data collection, actual transmission latency, actual link packet loss rate, and actual power consumption per unit time of the terminal. The actual operating parameters monitored are compared with the preset constraints. If they are not met, local adaptive correction is initiated: if the continuous features collected do not meet the standards, the sampling adjustment coefficient is adjusted by a preset step size; if the packet loss rate or latency exceeds the upper limit, the system is temporarily switched to buffered batch transmission; if the actual power consumption of the terminal per unit time exceeds the threshold, the global sleep time is adjusted by a preset step size.

[0013] Preferably, relevant data is integrated and feedback information is generated and sent back to provide data support for model iterative optimization. The specific process is as follows: The terminal integrates the final actual execution sampling cycle, actual sleep duration, actual transmission mode, real-time monitoring parameters, and adaptive correction results corresponding to each type of vital sign data to generate execution status feedback information. This feedback information is then transmitted back to the 6G edge AI collaborative decision-making module via the 6G uplink control link. This module uses the feedback information to update the historical training sample library for the corresponding category of vital sign data, providing real and effective operational data support for the iterative optimization of AI decision-making models corresponding to various types of vital sign data.

[0014] Compared with the prior art, the beneficial effects of the present invention are: (1) The AI-based 6G network medical vital signs monitoring system extracts three types of features of vital signs data—fluctuation, continuity, and trend—through the terminal perception module and dynamically adjusts the sampling density according to the feature values. At the same time, it collects the terminal operation and basic parameters of the 6G network and normalizes and integrates them into the terminal state vector, realizing the dynamic intelligent collection of vital signs data. It takes into account both the collection quality and the terminal energy consumption, and also provides complete and standardized basic data for subsequent decision-making.

[0015] (2) The AI-based 6G network medical vital signs monitoring system constructs a dedicated AI decision model for various vital signs data through the 6G edge AI collaborative decision module. It combines the terminal network state vector to construct training samples and set three types of global constraints. It constructs an objective function with multi-objective optimization as the direction and deploys the model on the 6G edge node to realize intelligent decision-making of terminal network integration. It takes into account the multi-objective optimization needs of data collection, terminal power consumption and network transmission. At the same time, it meets the real-time requirements of medical vital signs monitoring by relying on the low latency characteristics of 6G.

[0016] (3) The AI-based 6G network medical vital signs monitoring system achieves refined management of sampling, hibernation, and transmission through the execution control module based on decision parameters. It monitors the full-process operation parameters in real time and performs targeted local adaptive corrections. It also integrates and feeds back the execution and correction data to the collaborative decision module to optimize the model. It realizes intelligent management of the entire process of acquisition-transmission-hibernation and closed-loop optimization of system decision execution, which greatly improves the stability of system operation and continuously optimizes the scenario adaptability of the decision model. Attached Figure Description

[0017] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1; Please see Figure 1 This invention provides a medical vital signs monitoring system based on artificial intelligence under a 6G network, comprising: Terminal sensing module: Collects various vital sign data through multi-channel sensors, records timestamps and forms time-series sequences, and extracts three types of features: data fluctuation, continuity, and trend. Simultaneously, it collects terminal operation and 6G network basic parameters, integrates features and parameters to form terminal state vectors corresponding to various vital sign data. The specific process is as follows: The medical vital signs data acquisition terminal has a built-in multi-channel acquisition sensor to collect medical vital signs data; For each type of vital sign data, a baseline sampling period is set. For each data point collected, the corresponding timestamp is recorded synchronously. ,in, k is the sampling point number, and N is the sampling window length; The collected vital signs data are arranged in chronological order according to timestamps to form a time-series data sequence for this type of data. ,in, The medical vital signs data (such as heart rate and blood oxygen corresponding data values) of the kth sampling point are collected to complete the raw data collection; Extracting the data fluctuation characteristics, data continuity characteristics, and data trend characteristics of time series data sequence X, specifically: The data fluctuation characteristic (characterizing the magnitude of data change per unit time) is calculated as follows: , in: : Data fluctuation characteristic value; the larger the value, the more drastic the data change. N: Number of sampling points within a single acquisition window; Medical vital signs data at the kth sampling point; Medical vital signs data at the (k-1)th sampling point; like A large value indicates drastic data changes, requiring an increase in sampling density to capture details; like A smaller value indicates that the data changes gradually, which can reduce the sampling density to save power. This feature is directly used for AI sampling strategy decision-making. The data continuity feature (characterizing the completeness of data collection) is calculated as follows: , in: Continuous characteristic values ​​of data The closer the value is to 1, the more complete the data collection. : Number of missing sampling points within a single acquisition window (positive integer, counted in real time by the terminal, including missing points due to acquisition failure or signal interruption); like If the data is below the set threshold, it indicates that the data collection is incomplete and the terminal's working mode needs to be adjusted. like A higher value indicates a good data collection status, allowing the current strategy to be maintained or the sampling density to be reduced. This feature is used to constrain data collection continuity in AI decision-making. Data trend characteristics (representing the overall direction of data change) are calculated as follows: , in: Data trend characteristic value (slope) >0 indicates that the data is on an upward trend. <0 indicates a downward trend. =0 represents a trend towards stability; : The timestamp corresponding to the kth sampling point; If | A larger value indicates a significant trend change, requiring a high sampling density to track the changes. If | A smaller value indicates stable data, allowing for a reduction in sampling density; Baseline sampling period based on vital signs data Simultaneously, the operating parameters of the medical vital signs data acquisition terminal are collected, including but not limited to: remaining power, device operating temperature, signal reception strength, etc., and the collected parameter data are normalized and dimensionless.

[0020] The medical vital signs data acquisition terminal interacts directly with the base station through the 6G communication module to obtain the basic network parameters of the 6G service cell it is currently connected to (i.e., the coverage area of ​​the base station to which the terminal currently belongs), including but not limited to: uplink signal strength and initial transmission delay, and performs normalization and dimensionless processing on the basic 6G network parameters. The fluctuation characteristics, continuity characteristics, and trend characteristics of this type of vital sign data are integrated to form corresponding feature sub-vectors. The feature vectors of vital signs data are integrated with the normalized terminal operating parameters and 6G network basic parameters to form the terminal state vector corresponding to the current vital signs data. .

[0021] It should be noted that by using multi-channel sensors to achieve standardized collection of various vital sign data, and combining the baseline sampling period and timestamp to complete the time-series processing of the data, we can not only ensure the comprehensive collection of various types of vital sign data, but also ensure the time sequence and integrity of the original data, thus laying a high-quality data foundation for subsequent feature extraction and analysis. By extracting three types of features—fluctuation, continuity, and trend—and quantifying the feature values, we can accurately characterize the dynamic changes, completeness of data collection, and overall trend of vital signs data. Simultaneously, we dynamically adjust the sampling density based on the feature values, increasing the sampling density to ensure data detail when data changes drastically and reducing it to save terminal power consumption when data is stable. This achieves an optimal balance between data acquisition quality and terminal energy consumption. Furthermore, the extracted features can directly provide accurate feature basis for AI sampling strategy decisions. Based on the benchmark sampling period, terminal operating parameters and 6G network basic parameters are synchronously collected, and the two types of parameters are normalized and dimensionless, which effectively eliminates the dimensional differences between different types of parameters and ensures the comparability of parameter data. Then, the feature sub-vectors are integrated with the normalized parameters to form a terminal state vector, which can comprehensively and objectively reflect the terminal's hardware acquisition status and real-time network environment. This provides a complete and standardized basic data for the subsequent model training and real-time decision-making of the 6G edge AI collaborative decision-making module, effectively improving the accuracy and scenario adaptability of subsequent decision parameter output.

[0022] The 6G edge AI collaborative decision-making module acquires 6G operating parameters from the network side and forms a network state vector. It constructs training samples by splicing historical samples of various vital sign data from the terminal and network state vectors. It sets three types of boundary constraints: sampling, power consumption, and transmission. It constructs an objective function and defines an output dimension formula. After training and deployment, it obtains a dedicated AI decision-making model. Substituting the real-time terminal and network state vectors, it outputs sampling, sleep, and transmission decision parameters. The specific process is as follows: By using the standardized network management interface of 6G base stations, the network-side 6G network operation parameters of the currently accessing serving cell of the terminal can be obtained, including but not limited to: cell load rate, uplink available bandwidth, link packet loss rate, real-time transmission latency, etc. After normalizing and dedimensionalizing all acquired network operating parameters, they are organized to form the network state vector. ; For each type of vital sign data, a dedicated AI decision-making model is built. The specific process is as follows: Obtain historical operational data for this type of medical monitoring scenario, and extract the terminal state vector corresponding to each sample in the historical data. With network state vector By concatenating the vectors, we can obtain the historical training sample vectors for this type of data. ; , in, This is a vector concatenation symbol; Three types of global constraints are defined as boundary conditions for model training and real-time decision-making, including: Sampling constraints: The adjusted sampling period shall not be lower than the minimum sampling period for this type of data, nor shall it exceed the baseline sampling period for this type of data, to ensure the integrity of data collection; Power consumption constraint: The total power consumption of the terminal shall not exceed the preset dynamic power consumption threshold (the preset dynamic power consumption threshold is adaptively adjusted according to the remaining power of the terminal, and the lower the power, the lower the threshold) to avoid the power being depleted too quickly; Transmission constraints: Actual transmission delay and link packet loss rate shall not exceed the system's preset upper limit to ensure data transmission reliability; For each historical training sample corresponding to vital sign data, with the core optimization directions of optimal data continuity, lowest terminal power consumption, and highest network transmission efficiency, the objective function of the AI ​​decision-making model is constructed, and its expression is: , in: The objective function value corresponding to the sample. The larger the value, the better the decision-making effect; For the preset weighting coefficients, satisfy ; These are the continuous feature values ​​of vital sign data in the current sample (taken from the terminal state vector corresponding to the sample). The terminal's normalized power consumption per unit time (taken from the terminal operating parameters in the sample terminal state vector). This is the link packet loss rate (taken from the network operating parameters in the sample terminal state vector). The sampling adjustment coefficient of vital sign data, terminal global sleep duration, and terminal global transmission mode identifier are used as the core output dimensions of the AI ​​decision-making model. The quantitative calculation formulas for each output dimension for each historical training sample are as follows: The formula for calculating the sampling adjustment factor is: , in: This is the data sampling adjustment factor corresponding to the current sample; The minimum sampling adjustment factor preset for this type of data; The fluctuation characteristic value of vital sign data in the current sample (taken from the terminal state vector corresponding to the current sample); The trend feature value of this type of vital sign data in the current sample (taken from the terminal state vector corresponding to the current sample); a1 and a2 are preset weighting coefficients; The formula for calculating the global sleep duration of the terminal is: , in, This represents the global sleep duration for the terminal corresponding to the current sample. This is the maximum allowed sleep duration for the system; The remaining battery power of the terminal corresponding to the current sample (taken from the terminal state vector corresponding to the current sample). The cell load rate corresponding to the current sample (taken from the network state vector corresponding to the current sample). The formula for calculating the terminal's global transmission mode is: , in: M=0 indicates instant transmission, M=1 indicates buffered batch transmission; Set a preset packet loss rate threshold; This is a preset uplink bandwidth threshold; P represents the packet loss rate of the link corresponding to the current sample; B represents the normalized uplink available bandwidth corresponding to the current sample; A multi-objective optimization model (a well-known model) is used as the baseline model. The previously constructed training sample vectors are divided into a training set and a validation set. Model adaptation training is carried out based on the training set, three types of global constraints, the constructed objective function, and the defined output dimension. The training set is input into the baseline model, with global constraints as boundary conditions and maximizing the decision objective function as the optimization direction. The model parameters are adjusted through an iterative optimization algorithm. After each iteration, the model output results are verified using a validation set (to determine whether the sampling adjustment coefficient, terminal global sleep duration, and terminal global transmission mode meet the constraints and whether the decision objective function value is optimal). The training continues iteratively until the decision error on the validation set is below a preset threshold (or the number of iterations reaches a preset upper limit), at which point the training stops and an AI decision model adapted to the current type of vital sign data is obtained. The trained model was deployed on a 6G edge node for real-time decision-making based on vital sign data. Following the same model building, training, and deployment methods, the AI ​​decision-making models for the remaining categories of vital sign data were trained and deployed. For each type of vital sign data, its real-time terminal state vector within the current acquisition window is calculated. With real-time network state vector The data are concatenated into a real-time input vector and substituted into the corresponding AI decision-making model to output the sampling adjustment coefficients of vital signs data, the global sleep duration of the terminal, and the global transmission mode of the terminal.

[0023] It should be noted that by obtaining network-side operating parameters through the standardized interface of 6G base stations and normalizing them to form network state vectors, and combining them with terminal state vectors on the terminal side, standardized fusion of terminal and network data is achieved. This provides integrated terminal and network data support that fits the actual medical monitoring scenario for model training and decision-making, ensuring the uniformity and adaptability of the data. We build dedicated AI decision-making models for various vital sign data. By integrating terminal and network status information of historical samples through vector concatenation, the model training data can comprehensively reflect the terminal status and network environment under different collection scenarios. At the same time, we set three types of global constraints: sampling, power consumption, and transmission. This not only defines reasonable boundaries for model training and real-time decision-making, but also ensures the core requirements of system operation from three dimensions: collection integrity, terminal low power consumption, and transmission reliability, taking into account both data collection quality and equipment and network operating status. The objective function is constructed with the core optimization directions of optimal data continuity, lowest terminal power consumption, and highest network transmission efficiency. Targeted quantitative calculation formulas are designed for sampling adjustment coefficients, sleep duration, and transmission mode identification, so that the decision parameters output by the model can accurately match the actual needs of medical monitoring and achieve a synergistic balance of multi-objective optimization. The model is adapted and trained by separating the training set and the validation set. The model parameters are continuously adjusted through iterative optimization and validation, which effectively improves the model's decision accuracy and generalization ability. The trained, dedicated model is deployed on 6G edge nodes, leveraging the low latency of the 6G network to achieve real-time output of decision parameters. This significantly reduces data transmission and decision latency, meeting the high real-time requirements of medical vital sign monitoring. Simultaneously, a dedicated model is matched for each type of vital sign data, combining real-time terminal and network state vectors to output personalized decision parameters. This allows sampling, sleep, and transmission strategies to accurately adapt to the collection characteristics of different vital sign data and the real-time terminal network environment, improving the accuracy of system decision-making and scenario adaptability.

[0024] The execution control module: Based on the sampling, dormancy, and transmission decision parameters of various vital sign data, it calculates the sampling cycle and collects data, executes dormancy and transmission strategies according to rules, monitors actual operating parameters throughout the process, initiates local adaptive correction when parameters do not meet constraints, and finally integrates relevant execution data to generate feedback information for data support for model iterative optimization. The specific process is as follows: For each type of vital sign data, based on the corresponding sampling adjustment coefficient, terminal global sleep duration, and terminal global transmission mode, sampling control, sleep control, and transmission control are executed respectively, as follows: The temporary sampling period is calculated based on the sampling adjustment factor and the baseline sampling period. The calculation process is as follows: , in: To temporarily execute the sampling period; The baseline sampling period for vital sign data; The sampling adjustment coefficients corresponding to the vital signs data; The temporary execution sampling period is limited to a preset sampling period range to obtain the final actual execution sampling period. The constraint method is as follows: , Where T is the actual sampling period; The minimum sampling period corresponding to vital sign data; The terminal uses the final actual execution sampling period as the sampling period for the next collection window of the current vital sign data to complete the collection of the corresponding category of vital sign data; When there is no current vital sign data collection task to be executed and no cached data corresponding to the current vital sign data to be transmitted, the terminal uses the global sleep duration obtained by parsing as the actual sleep duration to enter a low-power sleep state. When the sleep time reaches the global sleep time or a new acquisition / transmission task is triggered, the terminal will automatically wake up and resume normal operation; Execute the corresponding transmission strategy based on the transmission mode identifier of the current vital signs data: If M=0, instant transmission is used, and the data of the current acquisition window is encapsulated and sent in real time; If M=1, use buffered batch transmission, buffer the data from multiple acquisition windows and then send them as a unified frame; Throughout the sampling, hibernation, and transmission processes, the terminal monitors the execution status of the current vital signs data in real time, and synchronously collects and calculates actual operating parameters, including actual continuous characteristics collected, actual transmission latency, actual link packet loss rate, and actual power consumption per unit time of the terminal. The specific calculation process is as follows: The actual calculation process for continuous features is as follows: , in: For actual continuous features, N is the length of the acquisition window. This represents the actual number of missing sampling points. The actual transmission delay is calculated as follows: , in: This represents the actual transmission delay. The time when vital sign data is sent. The moment of confirmation of receipt of vital sign data; The calculation process for the actual link packet loss rate is as follows: , in: This represents the actual packet loss rate of the link. The number of lost data packets. The total number of data packets sent; The calculation process for the actual power consumption of the terminal per unit time is as follows: , in: This represents the actual power consumption of the terminal per unit time. The remaining power at the start of the cycle. The remaining battery power at the end of the cycle. For statistical time intervals; For each type of vital sign data, the actual operating parameters obtained from real-time monitoring are compared with preset constraints. If the actual parameters do not meet the constraints, local adaptive correction is immediately initiated, specifically as follows: If continuous features are actually collected ( If a preset continuous feature threshold is used, the sampling adjustment coefficient is adjusted according to the preset step size to increase the sampling density and ensure the integrity of data acquisition. Such as actual link packet loss rate Or actual transmission delay ( To preset the packet loss rate threshold, If the preset latency limit is set, the transmission mode will be temporarily switched to buffered batch transmission to reduce transmission pressure. If the actual power consumption of the terminal per unit time ( If the preset power consumption threshold is used, the global sleep time will be adjusted by a preset step size to extend the sleep time, thereby reducing terminal power consumption and ensuring stable terminal operation. The terminal integrates the final actual sampling period, actual sleep duration, actual transmission mode, real-time monitoring parameters, and adaptive correction results corresponding to each type of vital sign data to generate execution status feedback information. This execution status feedback information is then transmitted back to the 6G edge AI collaborative decision-making module via the 6G uplink control link. This is used to update the historical training sample library of the corresponding category of vital sign data, providing real and effective operational data support for the iterative optimization of the corresponding AI decision-making model.

[0025] It should be noted that personalized sampling, hibernation, and transmission control are performed for each type of vital sign data based on exclusive decision parameters. By calculating and constraining the sampling cycle, the system ensures that the collection rhythm is adapted to the data characteristics. It intelligently switches between hibernation and wake-up modes based on the terminal's working status. At the same time, it selects instant or batch transmission strategies based on the transmission identifier, thus achieving refined management and control of the entire process of collection, hibernation, and transmission, and making the system operation strategy highly matched with actual decision-making needs. The system performs real-time monitoring and quantitative calculation of core operating parameters throughout the entire process of sampling, hibernation, and transmission. It comprehensively captures the integrity of data acquisition, network transmission status, and terminal power consumption. By comparing actual parameters with preset constraints in real time, it can quickly identify abnormal problems in system operation. For different types of parameter non-compliance, it performs targeted local adaptive correction, adjusts sampling coefficients, transmission modes, and hibernation duration in a timely manner, and can quickly resolve problems without remote interaction, effectively ensuring the stability of system operation, the integrity of data acquisition, and the reliability of network transmission. The system comprehensively integrates execution data and correction results across the entire process, generating complete execution status feedback information. This feedback is then transmitted back to the 6G edge AI collaborative decision-making module via a 6G uplink control link, providing real and effective operational data for the historical training sample library. This feedback data accurately reflects the application effect of model decision parameters in real-world scenarios, allowing the AI ​​decision-making model to continuously iterate and optimize based on actual operating conditions. This continuously improves the accuracy of decision parameters and scenario adaptability, achieving closed-loop optimization of system decision-making and execution, and driving the continuous upgrading of the overall monitoring system.

[0026] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A medical vital signs monitoring system based on artificial intelligence under a 6G network, characterized in that, include: Terminal sensing module: Collects various vital sign data, records timestamps and forms time series, and extracts three types of features: data fluctuation, continuity and trend; Synchronously collect terminal operation and 6G network basic parameters, and integrate features and parameters to form terminal status vectors corresponding to various vital sign data; 6G Edge AI Collaborative Decision Module: Acquires 6G operating parameters from the network side and forms a network state vector. Constructs training samples by splicing historical samples of terminal and network state vectors for various vital sign data. Sets three types of boundary constraints: sampling, power consumption, and transmission. Constructs an objective function and defines an output dimension formula. After training and deployment, a dedicated AI decision model is obtained. Substitutes the real-time terminal and network state vectors into the output sampling, sleep, and transmission decision parameters. Execution control module: Based on the sampling, hibernation, and transmission decision parameters of various vital sign data, it completes the calculation of the sampling cycle and carries out data collection, and executes the hibernation and transmission strategies according to the rules; The system monitors the actual operating parameters throughout the process, initiates local adaptive correction when the parameters do not meet the constraints, and finally integrates the relevant execution data to generate feedback information for data support for model iterative optimization.

2. The medical vital signs monitoring system based on artificial intelligence under a 6G network according to claim 1, characterized in that: The specific process of collecting various vital sign data, recording timestamps, and forming a time series is as follows: The data acquisition terminal collects various types of medical vital sign data; a baseline sampling period is set for each type of vital sign data, and the timestamp corresponding to each data point is recorded synchronously. The sampling points are numbered according to the collection order, and the collection window is a preset single data collection period containing a fixed number of sampling points. The data points collected for each type of vital signs are arranged in chronological order according to their corresponding timestamps, forming a time-series data sequence for each type of vital signs. Each data node in the time-series data sequence is the medical vital signs data collected from the corresponding sampling point.

3. The medical vital signs monitoring system based on artificial intelligence under a 6G network according to claim 1, characterized in that: The specific process for extracting the three types of features—fluctuation, continuity, and trend—from data is as follows: The data fluctuation characteristic is obtained by calculating the mean value of the difference in vital sign data between adjacent sampling points within a single acquisition window. This characteristic value represents the magnitude of data change per unit time. The data continuity feature is obtained by calculating the ratio of the number of valid sampling points within a single acquisition window to the total number of sampling points. This feature value characterizes the completeness of data acquisition and ranges from 0 to 1. The data trend feature is obtained by calculating the slope of the change in vital signs data with the corresponding sampling timestamp within a single acquisition window. This feature value characterizes the overall direction of data change.

4. The medical vital signs monitoring system based on artificial intelligence under a 6G network according to claim 1, characterized in that: The specific process for generating terminal state vectors corresponding to various vital sign data is as follows: Based on the baseline sampling period of various vital sign data, the operating parameters of the medical vital sign data acquisition terminal are collected synchronously. The operating parameters include: remaining power, device operating temperature, and signal reception strength. The collected terminal operating parameters are then normalized and dimensionless. Simultaneously, the network basic parameters of the 6G serving cell currently accessed by the terminal are obtained, including uplink signal strength and initial transmission delay, and the network basic parameters are normalized and dimensionless. The fluctuation, continuity, and trend characteristics of each type of vital sign data are integrated to form a feature sub-vector for the corresponding category. The feature sub-vector of each type of vital sign data is then integrated with the normalized terminal operating parameters and 6G network basic parameters to form a terminal state vector corresponding to each type of vital sign data.

5. The medical vital signs monitoring system based on artificial intelligence under a 6G network according to claim 1, characterized in that: The specific process of acquiring 6G operating parameters from the network side and forming a network state vector, and then constructing training samples by splicing historical samples of various vital sign data from the terminal and network state vectors, is as follows: The network side 6G network operation parameters of the current access serving cell of the terminal are obtained through the 6G base station standardized network management interface. The operation parameters include: cell load rate, uplink available bandwidth, link packet loss rate, and real-time transmission latency. All acquired network-side 6G network operation parameters are normalized and dimensionless, and then organized to form a network state vector. For each type of vital sign data, historical operational data of that type of data in medical monitoring scenarios is obtained. The terminal state vector corresponding to each sample in the historical operational data is concatenated with the network state vector to obtain the historical training sample vector corresponding to each type of vital sign data.

6. The medical vital signs monitoring system based on artificial intelligence under a 6G network according to claim 1, characterized in that: The specific process of setting three types of boundary constraints—sampling, power consumption, and transmission—to construct the objective function and define the output dimension formula is as follows: Sampling constraints: The adjusted sampling period shall not be lower than the minimum sampling period for vital signs data, and shall not exceed the baseline sampling period for vital signs data; Power consumption constraint: The total power consumption of the terminal shall not exceed the preset dynamic power consumption threshold; Transmission constraints: Actual transmission delay and link packet loss rate shall not exceed the system's preset upper limit; For each historical training sample corresponding to vital sign data, an objective function is constructed with the goals of optimal data continuity, lowest terminal power consumption, and highest network transmission efficiency. The sampling adjustment coefficient, terminal global sleep duration, and transmission mode identifier are used as the core output dimensions. The corresponding feature values, parameters and thresholds are combined to complete the quantitative calculation and judgment of each dimension. The transmission mode identifier is 0 for instant transmission and 1 for buffered batch transmission.

7. The medical vital signs monitoring system based on artificial intelligence under a 6G network according to claim 1, characterized in that: The specific process of obtaining a dedicated AI decision-making model, substituting it with the real-time terminal and network state vectors, and outputting sampling, sleep, and transmission decision parameters is as follows: Using a multi-objective optimization model as the baseline model, the training sample vectors are divided into training and validation sets. Model adaptation training is carried out by combining three types of global constraints, objective function, and output dimension. The training set is input into the model, and the parameters are iteratively adjusted with global constraints as the boundary and the objective function maximization as the direction. After each iteration, the model is validated with a validation set until the decision error is lower than the preset threshold to obtain a suitable AI decision model. The AI ​​decision model is then deployed on 6G edge nodes, and the dedicated model training and deployment for various vital sign data are completed in the same way. The real-time terminal and network state vectors of each type of data are concatenated into an input vector, which is then substituted into the corresponding model to output the sampling adjustment coefficients, the terminal global sleep duration, and the terminal global transmission mode.

8. The medical vital signs monitoring system based on artificial intelligence under a 6G network according to claim 1, characterized in that: Based on the sampling, dormancy, and transmission decision parameters of various vital sign data, the specific process of calculating the sampling cycle, collecting data, and executing the dormancy and transmission strategies according to the rules is as follows: The temporary execution sampling period is calculated by combining the sampling adjustment coefficient with the baseline sampling period of the corresponding vital sign data. This period is then limited to the range of the preset sampling period to obtain the final actual execution sampling period, which is used as the sampling period for the next acquisition window to carry out corresponding data acquisition. When the terminal has no corresponding data acquisition task and no related cached data to be transmitted, it enters a low-power sleep state according to the global sleep duration. It will automatically wake up when the sleep duration reaches the target or when there is a new data acquisition or transmission task. The execution strategy is based on the transmission mode identifier. When the identifier is 0, the current acquisition window data is sent in real time using instant transmission. When the identifier is 1, the data of multiple acquisition windows is cached and then sent as a unified frame.

9. The medical vital signs monitoring system based on artificial intelligence under a 6G network according to claim 1, characterized in that: The specific process of monitoring actual operating parameters throughout the process and initiating local adaptive correction when parameters do not meet constraints is as follows: Throughout the sampling, hibernation, and transmission process, the terminal monitors the execution status of vital sign data in real time, collects and calculates actual operating parameters, including actual continuous characteristics of data collection, actual transmission latency, actual link packet loss rate, and actual power consumption per unit time of the terminal. The actual operating parameters monitored are compared with the preset constraints. If they are not met, local adaptive correction is initiated: if the continuous features collected do not meet the standards, the sampling adjustment coefficient is adjusted by a preset step size; if the packet loss rate or latency exceeds the upper limit, the system is temporarily switched to buffered batch transmission; if the actual power consumption of the terminal per unit time exceeds the threshold, the global sleep time is adjusted by a preset step size.

10. The medical vital signs monitoring system based on artificial intelligence under a 6G network according to claim 1, characterized in that: The process involves integrating relevant execution data to generate feedback information, providing data support for model iteration and optimization. The specific steps are as follows: The terminal integrates the final actual execution sampling cycle, actual sleep duration, actual transmission mode, real-time monitoring parameters, and adaptive correction results corresponding to each type of vital sign data to generate execution status feedback information. This feedback information is then transmitted back to the 6G edge AI collaborative decision-making module via the 6G uplink control link. This module uses the feedback information to update the historical training sample library for the corresponding category of vital sign data, providing real and effective operational data support for the iterative optimization of AI decision-making models corresponding to various types of vital sign data.

Images