A vital signs monitoring wristband, monitoring system and method

Abstract

The application discloses a kind of vital signs monitoring wristband, monitoring system and method, specifically related to vital signs monitoring technical field, realize by the following steps: by wristband built-in sensor real-time monitoring multiple vital signs parameters and upload to cloud, generate data set;Based on these data, respectively to each parameter is short-term fluctuation and long-term trend evaluation, obtain evaluation result;Evaluation result is input into trained machine learning model, output fluctuation risk type and generate initial classification data group;According to initial data group and the cross-correlation feedback of each parameter, fluctuation state level division is carried out, and according to preset rule, the sampling frequency of each parameter is adjusted, to realize efficient intelligent health monitoring;The application is dynamically self-adapting sampling mechanism, according to the fluctuation situation and health risk type of vital signs parameter, intelligently adjusts sampling frequency, while guaranteeing the real-time of data, effectively reduces the energy consumption of data acquisition, prolongs the use time of equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vital sign monitoring, more particularly, the present application relates to a vital sign monitoring wristband, a monitoring system and a method. BACKGROUND

[0002] With the increasing demand for health management in modern society, more and more people begin to pay attention to their own health status, especially the monitoring of vital signs. Vital signs (such as heart rate, blood pressure, blood oxygen saturation, body temperature, etc.) are important indicators for assessing the health status of the human body, which can reflect the basic physiological state of the body. Traditional vital sign monitoring relies on hospital equipment or occasional handheld measurement tools. These devices usually can only provide instant single measurement, and it is difficult to conduct long-term and continuous health monitoring.

[0003] In recent years, with the development of smart wearable device technology, vital sign monitoring based on wearable devices has become possible. These smart devices are equipped with various sensors that can collect multiple vital sign data in real time and upload them to the cloud platform for analysis by users and medical professionals. Despite this, existing smart monitoring devices still face the following challenges in practical application:

[0004] Conflict between real-time and energy consumption: Real-time data collection and uploading require high frequency, but this will cause the battery of the device to be quickly consumed. How to ensure the real-time and accuracy of data while reducing energy consumption and prolonging the battery life of the device is still an important technical problem.

[0005] Insufficient data analysis: Existing health monitoring devices often lack efficient data analysis mechanisms and are difficult to identify health risks in real time, especially in the case of multiple vital sign parameters interacting with each other. Lack of accurate analysis methods and systems leads to an inability to comprehensively assess the health risks of individuals. Therefore, a vital sign monitoring wristband, a monitoring system and a method are proposed to solve the above problems. SUMMARY

[0006] To achieve the above purpose, the present application provides the following technical solutions:

[0007] A vital sign monitoring method, comprising the following steps:

[0008] Real-time monitoring of multiple preset vital sign parameters by the built-in sensors of the monitoring wristband and uploading to the cloud to obtain a vital sign parameter data set;

[0009] Based on the vital sign parameter data set, short-term fluctuation evaluation and long-term trend evaluation are performed on each preset vital sign parameter respectively to obtain short-term fluctuation evaluation results and long-term trend evaluation results;

[0010] The short-term fluctuation assessment result and the long-term trend assessment result are input into a pre-trained machine learning model to obtain a fluctuation risk type of each preset vital sign parameter, and the fluctuation risk types are summarized to obtain an initial classification data set;

[0011] Based on the initial classification data set and cross-correlation feedback results of each preset vital sign parameter, the fluctuation state of each preset vital sign parameter is hierarchically divided, and the sampling frequency of each preset vital sign parameter is adjusted according to a preset sampling optimization rule.

[0012] In a preferred embodiment, the short-term fluctuation assessment refers to:

[0013] Based on the change of the preset vital sign parameter within the preset time window, a dynamic fluctuation index is calculated to measure the instantaneous fluctuation.

[0014] In a preferred embodiment, the long-term trend assessment refers to:

[0015] Based on the historical data of the preset vital sign parameter, a health trend index is calculated to assess the long-term health condition.

[0016] In a preferred embodiment, the machine learning model is a convolutional neural network, the dynamic fluctuation index and the health trend index are input into the convolutional neural network model, the convolutional neural network outputs a result of 0 or 1, 0 represents that the fluctuation risk type of the vital sign parameter is a low risk type, 1 represents that the fluctuation risk type of the vital sign parameter is a high risk type, and the vital sign parameters are sorted according to the preset corresponding number, and all convolutional neural network output results are summarized to obtain the initial classification data set.

[0017] In a preferred embodiment, the logic for obtaining the dynamic fluctuation index is:

[0018] The data of the preset vital sign parameter X within the preset time window is obtained from the sensor in real time, assuming that the data collected within the preset time window is a time series, the standard deviation is used as the fluctuation measure, and the weighted standard deviation is introduced to improve the sensitivity to recent changes:

[0019] wi represents the weight coefficient corresponding to the i-th numbered data in the time series, N represents the total number of data in the time series, and σ(X) represents the fluctuation measure value of the time series;

[0020] The fluctuation amplitude change is calculated as a supplementary indicator of dynamic fluctuation:

[0021] The difference between the maximum value and the minimum value of the data in the current time window is calculated to obtain a fluctuation amplitude value A(X); the change amount of the fluctuation amplitude of the current time window and the fluctuation amplitude of the previous time window is compared to obtain an amplitude change value, and the calculation formula is:

[0022] ΔA(X)=A(X)-A(Xprevious); A(Xprevious) represents the fluctuation amplitude value corresponding to the previous time window, and ΔA(X) represents the amplitude change value;

[0023] The calculation formula of the dynamic fluctuation index is:

[0024] ∈ represents a preset non-zero constant, and DVI represents the dynamic fluctuation index.

[0025] In a preferred embodiment, the weight coefficient corresponding to the i-th numbered data in the time series is obtained by the following logic:

[0026] Based on the exponential decay factor, the fluctuation rate of the current vital sign parameter is dynamically adjusted, and the formula is:

[0027] The change rate of the current vital sign parameter sampling data at time ti, β is a preset non-zero influence coefficient of controlling the fluctuation rate on the decay, t0 represents the time starting point of the time series X, ti represents the time point of the current vital sign parameter sampling data in the time series X, and λ is a preset decay rate constant.

[0028] In a preferred embodiment, the health trend index is obtained by the following logic:

[0029] The historical data of the preset vital sign parameter X in the preset historical window is obtained in real time from the sensor, and the long-term change trend is extracted by regression analysis:

[0030] X(j)=a+b·Tj+ε; Tj represents the sampling data of the vital sign parameter X corresponding to the time point j in the historical window, a and b are both regression coefficients, representing the initial value and change rate of the vital sign parameter X in the historical window, and ε is the regression residual error, X(Tj) represents the regression predicted value of the vital sign parameter X corresponding to the time point j in the historical window, the trend intensity coefficient TSI is defined to measure the change trend of the health state, TSI=|b|, and the root mean square error between the sampling value and the regression predicted value of the vital sign parameter X in the historical window is introduced to obtain the long-term stability coefficient LTSI reflecting the long-term stability of the health parameter;

[0031] The calculation formula of the health trend index is:

[0032] LTSImax is a preset long-term stability coefficient standard value, used for normalizing the long-term stability coefficient LTSI, the normalized long-term stability coefficient LTSI is located in the range of [0, 1], f1 and f2 are both preset non-zero adjustment coefficients, and the sum of f1 and f2 is one, and HTI represents a health trend index.

[0033] In a preferred embodiment, the hierarchical division of the fluctuation state of each preset vital sign parameter based on the initial classification data set and the cross-correlation feedback result of each preset vital sign parameter refers to:

[0034] Using fuzzy reasoning, the result of the preset vital sign parameter X in the initial classification data set, the cross-correlation value of the preset vital sign parameter X and other vital sign parameters calculated by the Pearson correlation coefficient calculation formula, and the preset correlation threshold value are used as input variables of fuzzy logic, the hierarchical level corresponding to the fluctuation state of the vital sign parameter X is used as the output variable of fuzzy logic, the input variables are fuzzified, the values of the input variables are converted into fuzzy sets, the output variables are fuzzified, the output variables are converted into fuzzy sets, fuzzy rules are formulated to describe the adaptation degree of each level under different data category combinations, and the fuzzified input variables are inferred by the fuzzy rules to obtain the hierarchical level corresponding to the fluctuation state of the vital sign parameter X.

[0035] In a preferred embodiment, a vital sign monitoring wristband comprises:

[0036] a built-in sensor module for real-time monitoring of a plurality of preset vital sign parameters;

[0037] a data processing unit for receiving and processing vital sign parameter data collected by the built-in sensor module, and uploading the data to a cloud data processing platform to generate a vital sign parameter data set;

[0038] a short-term fluctuation evaluation module for evaluating the short-term fluctuation of each preset vital sign parameter based on the vital sign parameter data set, and obtaining the short-term fluctuation evaluation result of each parameter;

[0039] a long-term trend evaluation module for evaluating the long-term trend of each preset vital sign parameter based on the vital sign parameter data set, and obtaining the long-term trend evaluation result of each parameter;

[0040] a machine learning decision module for inputting the short-term fluctuation evaluation result and the long-term trend evaluation result into a pre-trained machine learning model, outputting the fluctuation risk type of each preset vital sign parameter, and generating an initial classification data set;

[0041] a cross-correlation feedback module for analyzing the cross-correlation between each of the preset vital sign parameters;

[0042] a sampling frequency optimization module for performing hierarchical division on the fluctuation state of each of the preset vital sign parameters according to the preliminary classification data set and the cross-correlation feedback information, and adjusting the sampling frequency of each of the vital sign parameters according to a preset sampling optimization rule.

[0043] In a preferred embodiment, a vital sign monitoring system comprises:

[0044] a vital sign monitoring terminal comprising a sensor module for monitoring a plurality of preset vital sign parameters in real time, and a data processing unit for receiving and processing vital sign parameter data collected by the built-in sensor module, and uploading the data to a cloud data processing platform to generate a vital sign parameter data set;

[0045] a cloud data processing platform for receiving the vital sign data set uploaded from the vital sign monitoring wristband, and performing the following steps:

[0046] performing short-term fluctuation assessment on each of the vital sign parameters to obtain a short-term fluctuation assessment result;

[0047] performing long-term trend assessment on each of the vital sign parameters to obtain a long-term trend assessment result;

[0048] inputting the assessment results into a pre-trained machine learning model to output the fluctuation risk type of each of the vital sign parameters, and performing hierarchical division on the fluctuation state of each of the parameters according to the cross-correlation between different parameters;

[0049] optimizing the sampling frequency of each of the preset vital sign parameters according to the hierarchical division result.

[0050] Technical effects and advantages of the present application:

[0051] The present application intelligently adjusts the sampling frequency according to the fluctuation of the vital sign parameters and the health risk type through a dynamic adaptive sampling mechanism. While ensuring real-time data, it can effectively reduce the energy consumption of data collection and transmission, thereby significantly prolonging the use time of the device and solving the contradiction between high-frequency data collection and device endurance. The present application combines short-term fluctuation assessment and long-term trend assessment to comprehensively analyze the fluctuation of vital sign parameters by calculating dynamic fluctuation index and health trend index. Short-term fluctuation assessment can timely capture sudden health changes, while long-term trend assessment can help discover gradual health change trends and provide early warning of potential health risks. Through the application of a machine learning model, the present application can accurately determine the fluctuation risk type of vital signs, helping to achieve efficient health monitoring and early warning.

[0052] The present application can conduct personalized health management according to the historical health data and real-time monitoring results of individuals. By introducing fuzzy reasoning and cross-correlation analysis, the system can dynamically adjust health assessment and intervention strategies. For example, the system will adjust the monitoring frequency of each vital sign parameter according to the health characteristics of individuals, ensuring timely monitoring of high-risk parameters while saving unnecessary resource consumption. The present application further optimizes health risk assessment by calculating the Pearson correlation coefficients between multiple vital sign parameters to analyze their cross-correlation. The correlation between different parameters can provide deeper information for health warning, avoid ignoring the potential impact between certain parameters, and improve the accuracy and comprehensiveness of health monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to facilitate the understanding of those skilled in the art, the present application will be further described below in conjunction with the drawings;

[0054] Figure 1 A schematic diagram of a vital sign monitoring method according to the present application.

[0055] Figure 2 A schematic diagram of a vital sign monitoring wristband according to the present application.

[0056] Figure 3 A schematic diagram of a vital sign monitoring system according to the present application. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0058] Reference Figure 1 - Figure 3 The following embodiments are obtained:

[0059] Embodiment 1:

[0060] With the continuous development of technology and the improvement of people's health awareness, vital sign monitoring technology has become an important part in the fields of daily health management, disease prevention, and elderly care. Vital signs such as heart rate, blood pressure, blood oxygen saturation, and body temperature reflect the basic health status of the human body and are important indicators for judging disease risk and health status. Traditional vital sign monitoring relies on hospital equipment or relatively simple handheld measuring instruments, which usually have the disadvantages of low use frequency, long measurement period, and poor immediacy.

[0061] In recent years, with the popularity of smart wearable devices, vital sign monitoring based on smart bands, watches, and other devices has gradually emerged. These devices usually have built-in sensors that can collect vital sign data in real time and upload it to the cloud, making it convenient for users to manage their long-term health. However, current monitoring technologies still have some problems in the following aspects:

[0062] Conflict between real-time and energy consumption: High-frequency data collection and transmission consume a lot of energy, leading to shorter device battery life. Reducing the sampling frequency may affect the real-time nature of the data, reducing the accuracy and response speed of the monitoring.

[0063] Insufficient data analysis and processing: Although modern wearable devices can collect vital sign data in real time, how to efficiently analyze these data and adjust the sampling frequency and identify abnormal fluctuations in a timely manner is still a challenge. Current monitoring methods mostly rely on single parameter evaluation, lacking cross-correlation analysis of multiple vital sign parameters and dynamic adaptive adjustment mechanism.

[0064] Lack of personalized fluctuation risk assessment: Most monitoring devices cannot dynamically adjust the sampling strategy according to individual differences of users, but only sample according to fixed rules. This makes it difficult for the device to accurately judge the fluctuation risk according to the user's health status in some specific scenarios.

[0065] Data fusion and optimization problems: Traditional vital sign monitoring systems often ignore the internal correlation between vital signs and fail to effectively utilize the cross-information between various vital sign parameters, resulting in low monitoring efficiency and inability to achieve intelligent sampling frequency adjustment and health assessment.

[0066] The present application provides a vital sign monitoring method, aiming to solve the contradiction between real-time and energy consumption in the prior art, insufficient data analysis and processing, lack of personalized fluctuation risk assessment, and data fusion and optimization. Through the innovative multi-level adaptive sampling mechanism, combined with short-term fluctuation assessment and long-term trend assessment, the machine learning model is introduced to classify and evaluate the fluctuation risk, and the cross-correlation feedback is used to divide the fluctuation state level, the present application dynamically adjusts the sampling frequency, while ensuring the real-time of data, reduces the energy consumption of the device, and improves the endurance of the device. According to the individual vital sign fluctuation and long-term trend, combined with the user's historical health data, the personalized fluctuation risk assessment is carried out, and the monitoring result is more targeted and accurate. By analyzing the cross-correlation between multiple vital sign parameters, the fluctuation state of each parameter is comprehensively evaluated, which provides more basis for dynamically adjusting the sampling frequency and optimizing the monitoring effect. Based on the machine learning model, the vital sign fluctuation risk type is classified, combined with the cross-correlation feedback mechanism, the different fluctuation states are hierarchically divided, and the sampling frequency is intelligently adjusted according to the hierarchical change, so as to realize the dynamic optimization of the adaptive sampling strategy.

[0067] The present application provides a vital sign monitoring method, comprising the following steps:

[0068] The present application provides a vital sign monitoring method, comprising the following steps:

[0069] Based on the vital sign parameter data set, short-term fluctuation assessment and long-term trend assessment are respectively carried out on each preset vital sign parameter to obtain short-term fluctuation assessment results and long-term trend assessment results; short-term fluctuation assessment: aiming to assess the changes of vital signs in a short time, such as rapid changes of heart rate, short-term fluctuations of blood pressure, etc. These changes may reflect the abnormality of instantaneous health status or the response to external stimuli. Long-term trend assessment: by analyzing the vital sign data in a period of time, the long-term trend is evaluated, such as whether the heart rate is continuously increasing, whether the blood sugar has an upward trend, etc. This helps to judge the health change trend of the individual and predict potential health risks. The combination of the two assessment results can more comprehensively describe the individual's health status, which can reveal both instantaneous fluctuation and provide a long-term health change perspective.

[0070] The short-term fluctuation assessment results and long-term trend assessment results are input into a pre-trained machine learning model to obtain the fluctuation risk type of each preset vital sign parameter, and the fluctuation risk types are summarized to obtain an initial classification data set. This step uses a machine learning model to comprehensively analyze the results of short-term fluctuation assessment and long-term trend assessment. The machine learning model (such as a convolutional neural network) can understand the relationship between different fluctuations and health risks through pre-trained data. Through the output of the model, each vital sign parameter is classified as high risk or low risk, providing a reference for subsequent health intervention and decision-making. This step can automatically identify and classify the health risk type of each parameter, reducing the need for manual intervention and improving the accuracy of judgment.

[0071] Based on the initial classification data set and the cross-correlation feedback results of each preset vital sign parameter, the fluctuation state of each preset vital sign parameter is hierarchically divided, and the sampling frequency of each preset vital sign parameter is adjusted according to the preset sampling optimization rules. In this step, the system not only relies on single vital sign data, but also considers the cross-correlation between multiple vital sign parameters. For example, there may be a correlation between heart rate and blood oxygen, and the fluctuations of blood pressure and body temperature may also affect each other. Cross-correlation feedback assesses the overall health status by calculating the relationship between vital sign parameters, making risk assessment more comprehensive and accurate. Through this hierarchical division, the system can refine the classification of health status and further optimize risk management strategies. This makes monitoring more flexible, allowing different levels of intervention for parameter fluctuations according to specific fluctuation states. According to the previous evaluation results and hierarchical division, the system can dynamically adjust the sampling frequency. For example, if a vital sign parameter is assessed as a high-risk fluctuation state, the system will increase the sampling frequency of that parameter to more finely monitor its changes; for parameters with smaller fluctuations, the system can reduce the sampling frequency to reduce energy consumption. This adaptive sampling mechanism can ensure real-time data while optimizing device energy consumption, prolonging battery life, and ensuring the efficiency of health monitoring. This step is the core of the intelligent monitoring system, which can intelligently adjust strategies based on data feedback, avoiding the resource waste caused by traditional fixed sampling frequencies.

[0072] The short-term fluctuation evaluation refers to calculating a dynamic fluctuation index based on the changes of the preset vital sign parameters within a preset time window to measure the instantaneous fluctuation. The long-term trend evaluation refers to calculating a health trend index based on the historical data of the preset vital sign parameters to evaluate the long-term health condition. The short-term fluctuation evaluation aims to capture the instantaneous fluctuation of the vital sign parameters within a short time window. This process quantifies the change amplitude of parameters such as heart rate, blood pressure, blood oxygen, etc. by calculating the dynamic fluctuation index. This evaluation helps identify immediate health abnormalities or sudden fluctuations, such as sudden heart rate increase, sudden blood pressure rise, etc., which may be caused by factors such as exercise, emotional fluctuations, external environment, etc., or even potential warning signals of underlying health problems. Through this real-time evaluation method, the instantaneous changes in health condition can be effectively tracked, providing a basis for subsequent intervention and treatment.

[0073] The long-term trend evaluation is based on the historical data of the preset vital sign parameters, and by calculating the health trend index, it can help evaluate the individual's health change trend and reveal the potential long-term changes in health status. The goal of this process is to identify long-term persistent health problems or improvement trends, such as early signs of chronic diseases, tracking of health recovery processes, etc. The health trend index reflects the long-term stability of vital signs, helping to understand the individual's health performance over a longer period of time.

[0074] The machine learning model is a convolutional neural network, which inputs the dynamic fluctuation index and the health trend index into the convolutional neural network model. The convolutional neural network outputs a result of 0 or 1, where 0 represents a low-risk type of fluctuation risk corresponding to the vital sign parameter, and 1 represents a high-risk type of fluctuation risk corresponding to the vital sign parameter. The results are sorted according to the preset vital sign parameter corresponding number, and all the convolutional neural network output results are summarized to obtain the initial classification data set.

[0075] In the present application, the convolutional neural network model is based on a hierarchical structure, in which the input data is processed step by step through multiple convolutional layers. The main function of the convolutional layer is to extract local features from the input data, which represent certain regularities in space or time. In vital sign monitoring, these convolutional layers can capture the subtle fluctuations of vital sign parameters over time, such as the instantaneous change trend of parameters such as heart rate, blood pressure, etc.

[0076] The convolution operation calculates the weighted sum of a local region by sliding a small window over the input data, generating a new feature map. This process effectively preserves the local correlation in the data and reduces the complexity of the calculation. As the number of layers increases, the convolutional neural network gradually extracts simple local features, to more complex pattern recognition, and finally forms a deep understanding of the entire data set. In this invention, these feature maps help the model identify the regularity between vital sign parameters, thereby evaluating the fluctuation degree of each parameter.

[0077] After the convolution layer, the pooling layer performs dimensionality reduction on the extracted features, further compressing the size of the data while preserving the most important feature information. Finally, the features extracted by the convolution layer and the pooling layer are integrated through the fully connected layer to perform the final classification or regression task. Through this hierarchical learning method, the convolutional neural network can efficiently analyze the short-term fluctuations and long-term trends of vital sign parameters, and finally output the risk level of each parameter, helping to achieve more accurate health assessment and monitoring.

[0078] The convolutional neural network takes the dynamic fluctuation index and health trend index as input, analyzes the complex patterns and potential health risks in the vital sign data. Through the trained machine learning model, it can efficiently and automatically classify the risk of each vital sign parameter, outputting a result of 0 or 1, indicating the type of fluctuation risk of the parameter. 0 represents low risk, and 1 represents high risk. This binary classification can quickly determine the health status of vital signs and provide timely feedback. By aggregating the output results of each parameter, an initial classification data set is obtained, which can integrate the risk types of each vital sign parameter and provide effective basis for subsequent health intervention, data analysis and decision-making.

[0079] The logic for obtaining the dynamic fluctuation index is as follows:

[0080] Real-time data of the preset vital sign parameter X in the preset time window is obtained from the sensor, assuming that the data collected in the preset time window is a time series, and the standard deviation is used as the measure of fluctuation, while considering the introduction of weighted standard deviation to improve the sensitivity to recent changes:

[0081] wi represents the weight coefficient corresponding to the i-th number of data in the time series, N represents the total number of data in the time series, and σ(X) represents the fluctuation measure value of the time series;

[0082] Calculate the fluctuation amplitude change as a complementary indicator of dynamic fluctuation:

[0083] The difference between the maximum value and the minimum value of the data in the current time window is calculated to obtain a fluctuation amplitude value A(X); the change amount of the fluctuation amplitude of the current time window and the fluctuation amplitude of the previous time window is compared to obtain an amplitude change value, and the calculation formula is:

[0084] ΔA(X)=A(X)-A(Xprevious); A(Xprevious) represents the fluctuation amplitude value corresponding to the previous time window, and ΔA(X) represents the amplitude change value;

[0085] The calculation formula of the dynamic fluctuation index is:

[0086] ε represents a preset non-zero constant, and DVI represents the dynamic fluctuation index. By introducing the weight, the model can better focus on the recent data changes, especially in the case of rapid fluctuation, the changes can be immediately reflected on the impact on the health status. By comparing the change of the current and past fluctuation amplitude, it can better judge whether there is abnormal fluctuation or change trend, so as to provide more accurate health warning. On the basis of comprehensively considering the fluctuation measure and amplitude change, the dynamic fluctuation index can more comprehensively evaluate the health fluctuation and provide more accurate data support for the sampling frequency adjustment. The larger the dynamic fluctuation index, the stronger the volatility of the vital sign data. That is, the vital sign parameter changes greatly in a short time, which may imply a greater health risk or unstable health condition. For example, the sharp fluctuation of heart rate, the rapid change of blood pressure, etc. may indicate potential health problems, which may need further monitoring.

[0087] The weight coefficient corresponding to the i-th numbered data in the time series is obtained as follows:

[0088] Based on the exponential decay factor, the fluctuation rate of the current vital sign parameter is dynamically adjusted, and the formula is:

[0089] represents the change rate of the current vital sign parameter sampling data at time t i, β is a preset non-zero influence coefficient of controlling the fluctuation rate on the decay, t 0 represents the time starting point of the time series X, t i represents the time point of the current vital sign parameter sampling data in the time series X, and λ is a preset decay rate constant. The exponential decay factor makes the influence of the earlier data gradually decrease over time. The decay rate constant λ controls the rate of this process. Part shows that as the amplitude of the vital sign parameter changes (such as heart rate, blood pressure, etc.) increases, the weight of the data at that moment is increased, enhancing the sensitivity to parameter changes. t0 represents the starting time point of the time series, which is used to calculate the time difference. The time-based decay factor makes the model pay more attention to recent data, thereby responding quickly to current vital sign fluctuations, especially when large fluctuations occur. According to the rate of change of the vital sign parameter, dynamic adjustment is made, so that in the case of rapid or severe fluctuations, more weight can be given to more accurately monitor and assess the current health status. The formula, by considering the feedback mechanism of fluctuation rate and change amplitude, helps to capture abnormal fluctuations in vital signs in a timely manner, providing more accurate basis for subsequent sampling frequency adjustment and health intervention.

[0090] The acquisition logic of the health trend index is as follows:

[0091] The historical data of the preset vital sign parameter X under the preset historical window is obtained in real time from the sensor, and the long-term trend is extracted through regression analysis:

[0092] X(j) = a + b · Tj + ε; Tj represents the sampling data of the vital sign parameter X corresponding to the time point j under the historical window, a and b are regression coefficients, representing the initial value and change rate of the vital sign parameter X under the historical window, and ε is the regression residual. X(Tj) represents the regression predicted value of the vital sign parameter X corresponding to the time point j under the historical window. The trend intensity coefficient TSI is defined to measure the change trend of the health state, TSI = |b|, and the root mean square error between the sampling value and the regression predicted value of the vital sign parameter X under the historical window is introduced to obtain the long-term stability coefficient LTSI reflecting the long-term stability of the health parameter.

[0093] The calculation formula of the health trend index is as follows:

[0094] LTSImax is a preset long-term stability coefficient standard value, which is used to normalize the long-term stability coefficient LTSI. The normalized long-term stability coefficient LTSI is located in the range of [0, 1], f1 and f2 are both preset non-zero adjustment coefficients, and the sum of f1 and f2 is one. HTI represents the health trend index.

[0095] The health trend index predicts the future trend of vital signs through regression analysis. Regression analysis helps to find the change rule of data, ensuring that the trend of health change can be effectively extracted from historical data. For example, the initial value of vital signs, the rate of change, etc. are estimated by regression, providing a mathematical basis for future health prediction, which can identify the long-term change rule of health parameters, not just instantaneous fluctuations. Through the evaluation of long-term trends, it can help to judge the gradual changes of individual health status, and timely discover potential chronic diseases or long-term unstable health conditions.

[0096] The health trend index introduces the fluctuation intensity coefficient (TSI), which reflects the frequency and intensity of vital sign changes. When the vital signs fluctuate greatly, the health risk may also be high, so it is necessary to evaluate the health risk by weighting the fluctuation intensity. This helps the system to more sensitively capture the fluctuations of health parameters and respond quickly to health changes. The significance of this design is to ensure that the monitoring system can respond to changes in real time when facing situations with greater health fluctuations, and timely adjust the health assessment and sampling strategy.

[0097] By measuring the long-term stability of health parameters, the stability and trend of health conditions can be effectively reflected. The introduction of the long-term stability coefficient helps to judge whether the health condition is in a long-term unstable state, providing prediction of chronic diseases or potential health problems. It can evaluate the health trend from a long-term perspective and identify chronic health risks or gradually deteriorating health conditions.

[0098] The health trend index uses a preset weighting coefficient to comprehensively evaluate the health status by combining the fluctuation intensity and long-term stability coefficient. In this way, the system can flexibly adjust the weight of different factors while ensuring the accuracy of health assessment, giving more or less attention to different parameters according to the current health status. Weighted integration makes health assessment not only a single indicator evaluation, but a dynamic and comprehensive evaluation process that can adaptively adjust according to different health conditions.

[0099] Regression analysis is typically fitted by least squares, which is widely used in linear regression analysis, especially in such problems. Least squares fits the data by minimizing the differences between actual data points and the predicted values of the regression model, resulting in optimal regression coefficients. Specifically, by adjusting the parameters of the model (such as initial values and rates of change) to minimize the sum of squared prediction errors, the trend of vital sign data is predicted. The larger the health trend index, the more significant the change in vital signs in the long-term trend, indicating a sharp change or deterioration in health status. For patients with chronic diseases, a sharp change in long-term health trend may indicate an exacerbation or unstable state of the disease, and a large health trend index value may indicate poor long-term stability of health status, with large fluctuations, which usually indicates that the individual's health status is unstable and requires more monitoring and intervention.

[0100] Based on the initial classification data set and the cross-correlation feedback results of each preset vital sign parameter, the fluctuation state of each preset vital sign parameter is hierarchically divided, which means:

[0101] Using fuzzy reasoning, the results of the preset vital sign parameter X in the initial classification data set, the cross-correlation values of the preset vital sign parameter X and other vital sign parameters calculated by the Pearson correlation coefficient formula, and the preset correlation threshold are used as input variables of fuzzy logic, the hierarchical level corresponding to the fluctuation state of the vital sign parameter X is used as the output variable of fuzzy logic, the input variables are fuzzified, the values of the input variables are converted to fuzzy sets, the output variables are fuzzified, the output variables are converted to fuzzy sets, the fuzzy rules are formulated to describe the adaptation degree of each level under different data category combinations, and the fuzzified input variables are inferred by fuzzy rules to obtain the hierarchical level corresponding to the fluctuation state of the vital sign parameter X.

[0102] In health monitoring, cross-correlation refers to the relationship between different vital sign parameters. For example, there may be a certain positive correlation between heart rate and blood pressure, and a negative correlation between body temperature and blood oxygen saturation. By calculating the Pearson correlation coefficient, the correlation between these parameters can be quantitatively described: the Pearson correlation coefficient is a value between -1 and 1, indicating the degree of linear relationship between two variables. When the correlation coefficient is close to 1, it indicates a strong positive correlation between the two variables; close to -1, it indicates a negative correlation; close to 0, it indicates almost no correlation.

[0103] For example, in health monitoring: there might be a strong negative correlation between heart rate and blood oxygen, meaning that an increase in heart rate usually accompanies a decrease in blood oxygen. In this case, when the heart rate fluctuates greatly, the change in blood oxygen should also be given more attention. There might be a certain positive correlation between blood pressure and body temperature, meaning that when the body temperature rises, the blood pressure might also tend to rise. In the fuzzy reasoning system, the role of cross-correlation feedback is reflected in the mutual influence between multiple vital sign parameters, and the system can intelligently adjust health assessment and risk prediction according to the correlation of these parameters.

[0104] After the input variables are calculated, they enter the fuzzy logic system. In the fuzzy reasoning system, the input variables are fuzzified, converting specific numerical values into fuzzy sets. This is because vital sign parameters and their fluctuation states often do not have clear boundaries, but rather exhibit gradual properties.

[0105] For example: the fluctuation intensity can be fuzzified into several levels such as "low fluctuation", "moderate fluctuation", and "high fluctuation". Cross-correlation can also be fuzzified into levels such as "weak correlation", "moderate correlation", and "strong correlation". This fuzzification approach allows the system to more flexibly handle uncertainty or incomplete data, avoiding the limitations of overly precise classification. Once the input variables are fuzzified, they will serve as inputs for fuzzy logic reasoning. Next, according to the set fuzzy rules, different combinations of input variables are reasoned to obtain the output results. These fuzzy rules are formulated based on historical data and empirical rules, describing the fluctuation state levels under different combinations of vital sign parameters.

[0106] For example, suppose there are the following fuzzy rules: if the fluctuation intensity is "high fluctuation" and the cross-correlation is "strong correlation", then the fluctuation state level should be "high risk". If the fluctuation intensity is "low fluctuation" and the cross-correlation is "weak correlation", then the fluctuation state level is "low risk". These rules determine the level of fluctuation state based on the actual fluctuations of vital sign parameters and their mutual relationships.

[0107] After reasoning through the fuzzy rules, the system will obtain a fuzzy output result, which is usually the fluctuation state level of the vital sign parameter. The fluctuation state level will indicate the current health risk level of the parameter.

[0108] For example, after reasoning, if the system concludes that the fluctuation state is "high risk", it means that there may be significant abnormal fluctuations in the current health state, which requires further attention and intervention. The fuzzification of the output variable will convert the reasoning level (such as "low risk" or "high risk") into specific numerical values or levels, and finally feedback to the user or health monitoring system.

[0109] In the fuzzy inference process, the Max-Min Method is often used for inference. This method is a common inference method in fuzzy logic systems, which infers the output result based on the fuzzy sets of input variables and pre-set fuzzy rules.

[0110] Max method: For the input of fuzzy rules, in each rule, the smallest input membership value is selected. Specifically, under multiple input conditions, the minimum value represents the minimum value of the current rule activation degree. For example, if the rule is "if heart rate is high and blood oxygen is low, then fluctuation state is high risk", then for this rule, the fluctuation degree of heart rate is "high" and blood oxygen is "low", the system takes the minimum value of the two as the indicator of the activation degree.

[0111] Min method: For the output of the inference result, the minimum method is used to determine the membership of the output variable. That is, among the output results of multiple rules, the minimum value is selected as the final inference output. This can ensure that under the influence of multiple rules, the system output result will not be too intense, but will be constrained by the minimum of all rules.

[0112] For example, in fuzzy rule 1, the membership of heart rate is 0.8, the membership of blood pressure is 0.7, and the activation value of rule 1 is 0.7 (taking the minimum value). Similarly, in fuzzy rule 2, the membership of heart rate is 0.8, the membership of blood oxygen is 0.6, and the activation value of rule 2 is 0.6. Through the Max-Min method, the final output of health risk is the minimum value of the two, i.e. 0.6.

[0113] Suppose the system is monitoring a patient's four vital sign parameters: heart rate, blood pressure, blood oxygen saturation, and body temperature. These data have been evaluated for short-term fluctuations and long-term trends, and the system classifies them into low-risk and high-risk categories. The risk type of each parameter can be represented by 0 for low risk and 1 for high risk. The initial classification data set: heart rate: 1 (high risk); blood pressure: 0 (low risk); blood oxygen saturation: 1 (high risk); body temperature: 0 (low risk); this classification data set indicates that heart rate and blood oxygen saturation fluctuate greatly, belonging to a higher health risk situation, while blood pressure and body temperature are relatively stable, with lower risk.

[0114] To further understand the cross-correlation between these parameters, the system calculates the Pearson correlation coefficient between each pair of parameters, which is a measure of their linear relationship. The Pearson correlation coefficient between heart rate and blood pressure is assumed to be 0.2, indicating a weak positive correlation. The Pearson correlation coefficient between heart rate and blood oxygen saturation is assumed to be -0.7, indicating a strong negative correlation, meaning that an increase in heart rate is usually accompanied by a decrease in blood oxygen saturation. The Pearson correlation coefficient between blood pressure and body temperature is assumed to be 0.3, indicating a weak positive correlation. The Pearson correlation coefficient between blood oxygen saturation and body temperature is assumed to be -0.5, indicating a negative correlation.

[0115] Based on the initial classification data set and the Pearson correlation coefficients, the system categorizes these parameters into different volatility levels and assigns different sampling frequencies to each level. The system's level categorization is based on the following criteria: High volatility state (high-risk level): If a parameter is assessed as high-risk and has a strong correlation with other parameters, it is categorized as a high volatility state. Medium volatility state (medium-risk level): If a parameter is assessed as medium-risk and has a weak correlation with other parameters, it is categorized as a medium volatility state. Low volatility state (low-risk level): If a parameter is assessed as low-risk and has a low correlation with other parameters, it is categorized as a low volatility state. Specific level categorization: Heart rate: Since heart rate is assessed as high-risk and has a strong negative correlation with blood oxygen saturation (-0.7), it is categorized as a high-risk level. Blood pressure: Blood pressure is assessed as low-risk and has weak correlations with other parameters (0.2 and 0.3), so it is categorized as a low-risk level. Blood oxygen saturation: Blood oxygen saturation has a strong negative correlation with heart rate (-0.7) and is assessed as high-risk, so it is categorized as a high-risk level. Body temperature: Body temperature is assessed as low-risk, has a weak positive correlation with blood pressure (0.3), and a negative correlation with blood oxygen saturation (-0.5), so it is categorized as a low-risk level.

[0116] According to the hierarchical division, the system presets different sampling frequencies for each level. The following is the corresponding sampling frequency of each level: high-risk level (high volatility state): for example, heart rate and blood oxygen saturation are classified as high-risk level, which has larger fluctuations and requires more frequent sampling to monitor changes in real time. Therefore, high-frequency sampling (multiple times per minute) is set for these parameters. Medium-risk level (medium volatility state): for example, when the fluctuations of certain parameters are relatively stable, but there is still some volatility, the system will allocate them to medium-frequency sampling (once per minute). Low-risk level (low volatility state): blood pressure and body temperature are evaluated as low-risk, which have smaller fluctuations, so a lower sampling frequency can be selected to ensure that the system saves energy without affecting the effectiveness of the data. For these parameters, the system can set low-frequency sampling (once per hour).

[0117] Embodiment 2:

[0118] A vital sign monitoring wristband, comprising:

[0119] a built-in sensor module for real-time monitoring of multiple preset vital sign parameters;

[0120] a data processing unit for receiving and processing vital sign parameter data collected by the built-in sensor module, and uploading the data to a cloud data processing platform to generate a vital sign parameter data set;

[0121] a short-term fluctuation evaluation module that evaluates the short-term fluctuation of each preset vital sign parameter based on the vital sign parameter data set, and obtains the short-term fluctuation evaluation result of each parameter;

[0122] a long-term trend evaluation module that evaluates the long-term trend of each preset vital sign parameter based on the vital sign parameter data set, and obtains the long-term trend evaluation result of each parameter;

[0123] a machine learning decision module that inputs the short-term fluctuation evaluation result and the long-term trend evaluation result into a pre-trained machine learning model, outputs the fluctuation risk type of each preset vital sign parameter, and generates a preliminary classification data set;

[0124] a cross-correlation feedback module that analyzes the cross-correlation between each preset vital sign parameter;

[0125] a sampling frequency optimization module that classifies the fluctuation state of each preset vital sign parameter according to the preliminary classification data set and cross-correlation feedback information, and adjusts the sampling frequency of each vital sign parameter according to the preset sampling optimization rules.

[0126] Embodiment 3:

[0127] A vital sign monitoring system, comprising:

[0128] A vital sign monitoring terminal, which is internally provided with a sensor module for monitoring a plurality of preset vital sign parameters in real time, and a data processing unit for receiving and processing vital sign parameter data collected by the internal sensor module, and uploading the data to a cloud data processing platform to generate a vital sign parameter data set;

[0129] A cloud data processing platform, which is configured to receive the vital sign data set uploaded from the vital sign monitoring wristband, and perform the following steps:

[0130] Performing short-term fluctuation assessment on each vital sign parameter to obtain a short-term fluctuation assessment result;

[0131] Performing long-term trend assessment on each vital sign parameter to obtain a long-term trend assessment result;

[0132] Inputting the assessment results into a pre-trained machine learning model to output the fluctuation risk type of each vital sign parameter, and performing hierarchical division on the fluctuation state of each parameter according to the cross-correlation between different parameters;

[0133] According to the result of hierarchical division, optimizing the sampling frequency of each preset vital sign parameter.

[0134] The above formulas are all dimensionless numerical calculations, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the latest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0135] It should be understood that the sequence of the above processes in various embodiments of the present application does not mean the order of execution, and the execution order of the processes should be determined according to their functions and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0136] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0137] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0138] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A vital signs monitoring method, characterized by, The method comprises the following steps: Real-time monitoring of a plurality of preset vital sign parameters through monitoring of the built-in sensors in the wristband and uploading to the cloud to obtain a vital sign parameter data set; Based on the vital sign parameter data set, short-term fluctuation assessment and long-term trend assessment are respectively performed on each preset vital sign parameter to obtain short-term fluctuation assessment results and long-term trend assessment results; The short-term fluctuation assessment results and the long-term trend assessment results are input into a pre-trained machine learning model to obtain the fluctuation risk type of each preset vital sign parameter, and the fluctuation risk types are summarized to obtain an initial classification data set; Based on the initial classification data set and the cross-correlation feedback results of each preset vital sign parameter, the fluctuation state of each preset vital sign parameter is hierarchically divided, and the sampling frequency of each preset vital sign parameter is adjusted according to a preset sampling optimization rule; The short-term fluctuation assessment refers to calculating a dynamic fluctuation index based on the changes of the preset vital sign parameter within a preset time window to measure the instantaneous fluctuation; The long-term trend assessment refers to calculating a health trend index based on the historical data of the preset vital sign parameter to assess the long-term health condition; The machine learning model is a convolutional neural network, and the dynamic fluctuation index and the health trend index are input into the convolutional neural network model, and the convolutional neural network outputs a result of 0 or 1, where 0 indicates that the fluctuation risk type of the vital sign parameter is a low-risk type, and 1 indicates that the fluctuation risk type of the vital sign parameter is a high-risk type, and the results are sorted according to the preset vital sign parameter corresponding number, and all the convolutional neural network output results are summarized to obtain the initial classification data set; Based on the initial classification data set and the cross-correlation feedback results of each preset vital sign parameter, the fluctuation state of each preset vital sign parameter is hierarchically divided, which means: Using fuzzy reasoning, obtaining the result of the preset vital sign parameter X in the initial classification data set, calculating the cross-correlation value of the preset vital sign parameter X and other vital sign parameters by the Pearson correlation coefficient calculation formula, and the preset correlation threshold value are used as the input variables of the fuzzy logic, the fluctuation state corresponding to the level of the vital sign parameter X is used as the output variable of the fuzzy logic, the input variables are fuzzed, the values of the input variables are converted into fuzzy sets, the output variables are fuzzed, the output variables are converted into fuzzy sets, the fuzzy rules are formulated to describe the adaptation degree of each level under different data category combinations, and the fuzzed input variables are inferred by the fuzzy rules to obtain the fluctuation state corresponding to the level of the vital sign parameter X.

2. The vital sign monitoring method of claim 1, wherein, The logic for obtaining the dynamic fluctuation index is: Real-time acquisition of the data of the preset vital sign parameter X within the preset time window from the sensor, the data collected within the preset time window is a time series, and the standard deviation is used as the measure of fluctuation, and the weighted standard deviation is introduced to improve the sensitivity to recent changes: ; denotes the weight coefficient corresponding to the i-th numbered data in the time series, denotes the total number of data in the time series, denotes the fluctuation metric value of the time series; The fluctuation amplitude change is calculated as a supplementary index of dynamic fluctuation: Calculate the difference between the maximum value and the minimum value of the data in the current time window to obtain a fluctuation amplitude value ; compare the fluctuation amplitude of the current time window with the change amount of the fluctuation amplitude of the previous time window to obtain an amplitude change value, and the calculation formula is: ; represents a fluctuation amplitude value corresponding to the previous time window, represents an amplitude change value; The calculation formula of the dynamic fluctuation index is: ; represents a preset non-zero constant, represents a dynamic volatility index.

3. The vital signs monitoring method of claim 2, wherein, The logic for obtaining the weight coefficient corresponding to the i-th numbered data in the time series is: Based on the exponential decay factor, dynamically adjust according to the fluctuation rate of the current vital sign parameter, the formula is: ; represents a rate of change of the current vital sign parameter sampling data over time , is a non-zero influence coefficient of the preset control fluctuation rate on the decay, represents a time starting point of the time series, represents a time point of the current vital sign parameter sampling data in the time series, is a preset decay rate constant.

4. The vital signs monitoring method of claim 3, wherein, The acquisition logic of the health trend index is: Real-time acquisition of historical data of the preset vital sign parameter X in the preset historical window from the sensor, and extraction of the long-term changing trend through regression analysis: ; denotes the time point of the vital sign parameter X under the history window corresponding sampling data, and are regression coefficients, representing the starting value and the change rate of the vital sign parameter X under the history window, is the regression residual, denotes the time point of the vital sign parameter X under the history window corresponding regression prediction value, defining the trend intensity coefficient carries out the change trend of the metric health state, = At the same time, the root mean square error between the sampling value and the regression prediction value of the vital sign parameter X under the history window is introduced to obtain the long-term stability coefficient reflecting the long-term stability of the health parameter ; The calculation formula of the health trend index is: ; is a preset long-term stability coefficient standard value, used for normalizing the long-term stability coefficient , the normalized long-term stability coefficient is located in the range of [0, 1], and are preset non-zero adjustment coefficients, and and the sum of which is 1, represents a health trend index.

5. A vital sign monitoring wristband for implementing a vital sign monitoring method as claimed in claims 1-4, characterized in that, Including: Built-in sensor module for real-time monitoring of multiple preset vital sign parameters; Data processing unit for receiving and processing vital sign parameter data collected by the built-in sensor module, and uploading data to a cloud data processing platform to generate a vital sign parameter data set; Short-term fluctuation assessment module, based on the vital sign parameter data set, respectively assesses the short-term fluctuation of each preset vital sign parameter to obtain the short-term fluctuation assessment result of each parameter; Long-term trend assessment module, based on the vital sign parameter data set, respectively assesses the long-term trend of each preset vital sign parameter to obtain the long-term trend assessment result of each parameter; Machine learning decision module, input the short-term fluctuation assessment result and long-term trend assessment result into the pre-trained machine learning model, output the fluctuation risk type of each preset vital sign parameter, and generate a preliminary classification data set; Cross-correlation feedback module, analyze the cross-correlation between each preset vital sign parameter; Sampling frequency optimization module, according to the preliminary classification data set and cross-correlation feedback information, divide the fluctuation state of each preset vital sign parameter into levels, and adjust the sampling frequency of each vital sign parameter according to the preset sampling optimization rule.

6. A vital signs monitoring system for implementing a vital signs monitoring method according to any one of claims 1 to 4, characterized by Including: Vital sign monitoring terminal, built-in sensor module for real-time monitoring of multiple preset vital sign parameters, and data processing unit for receiving and processing vital sign parameter data collected by the built-in sensor module, and uploading data to a cloud data processing platform to generate a vital sign parameter data set; Cloud data processing platform, for receiving vital sign data set uploaded from vital sign monitoring wristband, and performing the following steps: Assess the short-term fluctuation of each vital sign parameter to obtain the short-term fluctuation assessment result; Assess the long-term trend of each vital sign parameter to obtain the long-term trend assessment result; Input the assessment result into the pre-trained machine learning model, output the fluctuation risk type of each vital sign parameter, and divide the fluctuation state of each parameter into levels according to the cross-correlation between different parameters; According to the results of hierarchical division, optimize the sampling frequency of each preset vital sign parameter.

Images