Health monitoring method and system based on data fusion of medical wearable equipment

Abstract

The invention discloses a health monitoring method and system based on medical wearable equipment data fusion. The monitoring method comprises the following steps: collecting physiological parameters from various medical wearable devices; preprocessing the collected physiological parameter data by using a signal processing technology, calculating a standardized score of each physiological parameter, and calculating a health index according to the standardized scores of the physiological parameters; on the basis of the change trend of the comprehensive health index values, a health state prediction model is constructed by adopting a machine learning algorithm, and the prediction efficiency is evaluated by calculating the accuracy of the model; applying the prediction model to the real-time data stream, triggering an alarm mechanism when the health index value exceeds a set threshold, and calculating the proportion of the health index value exceeding the threshold; and according to the size of the proportion value and feedback, a feedback result is used for optimizing the prediction model. The health state prediction model is constructed by adopting a machine learning algorithm, so that health monitoring is more accurate and convenient.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a health monitoring method and system based on medical wearable device data fusion. Background Art

[0002] With the continuous advancement of science and technology, the field of medical health has undergone a significant transformation, moving from the traditional medical service model to a more intelligent, personalized and preventive direction. In the past, medical health services mainly focused on the diagnosis and treatment of diseases, and medical help was usually sought only after symptoms appeared.

[0003] With the development of medical wearable devices and remote monitoring technology, especially with the development of the Internet of Things, big data, cloud computing and artificial intelligence technology, health monitoring has become more frequent and convenient, and medical wearable devices have gradually become an important tool for health management. It is also a general trend to apply machine learning methods to health monitoring of medical wearable devices. In view of this, the present invention proposes a health monitoring method and system based on medical wearable device data fusion. Summary of the invention

[0004] The present invention provides a health monitoring method based on medical wearable device data fusion, comprising:

[0005] S10, collect physiological parameters from multiple medical wearable devices and ensure data synchronization through timestamp alignment;

[0006] S20, preprocessing the collected physiological parameter data using signal processing technology, calculating a standardized score for each physiological parameter, and calculating a health index based on the standardized score of the physiological parameter;

[0007] S30. Based on the changing trend of the comprehensive health index values, a health status prediction model is constructed using a machine learning algorithm, and its prediction effectiveness is evaluated by calculating the accuracy of the model;

[0008] S40, applying the prediction model to the real-time data stream, triggering an alarm mechanism when the health indicator value exceeds a set threshold, and calculating the proportion exceeding the threshold;

[0009] S50: Based on the size of the ratio value and the feedback, the feedback result is used to optimize the prediction model.

[0010] As described above, a health monitoring method based on medical wearable device data fusion, in which data synchronization is ensured by timestamp alignment, requires enabling the timestamp recording function in the settings of the medical wearable device to ensure that the timestamp is automatically recorded each time data is collected. The collected data needs to be corrected to ensure that all data points correspond to the same time point.

[0011] In the health monitoring method based on medical wearable device data fusion as described above, the collected physiological parameter data are preprocessed using signal processing technology, including noise filtering and outlier detection.

[0012] As described above, a health monitoring method based on medical wearable device data fusion, in which a health status prediction model is constructed using a machine learning algorithm, requires first selecting the most representative features from the standardized physiological parameter samples, and constructing feature vectors and training data sets.

[0013] As described above, a health monitoring method based on medical wearable device data fusion is described, in which the training of the health status prediction model is specifically divided into the following sub-steps: setting up an objective function, defining constraints, obtaining a complete minimized objective function, solving weight vectors and bias terms, and obtaining a predicted health status model.

[0014] As described above, a health monitoring method based on medical wearable device data fusion is described, in which the complete minimization objective function needs to be solved by converting it into a dual problem using the Lagrange multiplier method, and the dual problem is solved to obtain the Lagrange multiplier, thereby calculating the weight vector and the bias term.

[0015] In the health monitoring method based on medical wearable device data fusion as described above, the prediction performance of the model is evaluated by calculating the accuracy of the model through cross-validation, which is evaluated by comparing the difference between the predicted label and the true label.

[0016] The present invention also provides a health monitoring system based on medical wearable device data fusion, including: a collection module, a processing and calculation module, a training model module, an alarm module, and a feedback optimization module.

[0017] Collection module: used to collect physiological parameters from various medical wearable devices and ensure data synchronization through timestamp alignment.

[0018] Processing and calculation module: used to pre-process the collected physiological parameter data using signal processing technology, calculate the standardized score of each physiological parameter, and calculate the health index based on the standardized score of the physiological parameter.

[0019] Training model module: Based on the changing trend of comprehensive health indicator values, a health status prediction model is constructed using machine learning algorithms, and its prediction effectiveness is evaluated by calculating the accuracy of the model.

[0020] Alarm module: used to apply the prediction model to real-time data streams. When the health indicator value exceeds the set threshold continuously, the alarm mechanism is triggered and the proportion exceeding the threshold is calculated.

[0021] Feedback optimization module: Provides personalized health advice based on the size of the ratio value, and uses the feedback results to optimize the prediction model.

[0022] The present invention also provides a computer storage medium, comprising at least one memory and at least one processor;

[0023] A memory for storing one or more program instructions;

[0024] A processor is used to run one or more program instructions to execute a health monitoring method based on medical wearable device data fusion.

[0025] The beneficial effects achieved by the present invention are as follows: the present invention adopts a machine learning algorithm to construct a health status prediction model, which can monitor health status changes in real time, making health monitoring more accurate and convenient. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0027] Figure 1 This is a flow chart of a health monitoring method based on medical wearable device data fusion provided in Example 1 of the present application.

[0028] Figure 2 This is a schematic diagram of a health monitoring system based on medical wearable device data fusion provided in Example 2 of the present application. DETAILED DESCRIPTION

[0029] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0030] Embodiment 1

[0031] like Figure 1 As shown, the first embodiment of the present application provides a health monitoring method based on medical wearable device data fusion, including:

[0032] S10. Collect physiological parameters from multiple medical wearable devices and ensure data synchronization through timestamp alignment.

[0033] S11. Collect physiological parameters from various medical wearable devices.

[0034] Medical wearable devices are divided into vital signs monitoring, sports rehabilitation, and disease diagnosis and management. Common forms include smart bracelets, smart watches, smart patches, etc. with high-precision sensors.

[0035] To obtain physiological parameters and other features from a variety of medical wearable devices, it is necessary to first set the data collection frequency of the medical wearable device according to the data needs to ensure that the data points are dense enough to capture changes in physiological parameters. The specific physiological parameters collected include heart rate, blood pressure, blood oxygen saturation, body temperature, number of steps, sleep quality, acceleration, skin temperature, skin conductance, respiratory rate, body composition, electrocardiogram, etc. These parameters are stored in the form of time series. Other features include height, age, weight, gender, etc.

[0036] S12. Ensure data synchronization through timestamp alignment.

[0037] Enable the timestamp recording function in the settings of the medical wearable device to ensure that the timestamp is automatically recorded each time data is collected, design the data storage format, and transfer the collected physiological parameter data to the storage device via Bluetooth, Wi-Fi or other wireless communication methods.

[0038] Extract timestamps for each physiological parameter record to ensure that each data point has an accurate time stamp. Parse the timestamp field of the physiological parameter sample, synchronize the data collected by different devices to the same time axis, and perform corrections to ensure that all data points correspond to the same time point. corrected =timestamp iso +Δt for time correction, timestamp corrected Indicates the corrected timestamp, timestamp iso Indicates the standard time format, and Δt indicates the time deviation.

[0039] S20. Preprocess the collected physiological parameter data using signal processing technology, calculate the standardized score of each physiological parameter, and calculate the health index based on the standardized score of the physiological parameter.

[0040] S21. Preprocess the collected physiological parameter data using signal processing technology.

[0041] The collected physiological parameter data is subjected to noise filtering, and a low-pass filter is used to remove high-frequency noise and retain useful physiological signals. The specific formula is y n =αx n +(1-α)y n-1 ,y nrepresents the filter output value at the current time n, x n represents the input value of n at the current moment, α represents the smoothing factor, and its value range is 0<α<1. When α is close to 0, the filter responds slowly to new data and the filtering effect is stronger. When α is close to 1, the filter responds faster to new data, but the filtering effect is weaker. n-1 Indicates the filter output value at the previous moment n-1, indicating the filter result that has been calculated before.

[0042] The collected physiological parameter data are tested for outliers, and data points with a value exceeding 3 times the standard deviation are detected and removed. The standard deviation σ and mean μ of each physiological parameter are calculated to detect and remove data points with a value exceeding 3 times the standard deviation. The standard deviation calculation formula is: Where N represents the number of data points, e i represents the i-th data point, and μ represents the average value of the data point. The average value calculation formula is Where N represents the number of data points, e i represents the i-th data point.

[0043] If a data point e i If the value exceeds 3 times the standard deviation of the mean value μ, the data point is considered an outlier. The upper threshold formula is Limit Upper =μ+3σ, the lower threshold formula is Limit Lower =μ-3σ, if e i >Limit Upper or e i <Limit Lower , then e i For outliers, e i Removed.

[0044] S22. Calculate the standardized score of each physiological parameter.

[0045] The standardized score reflects the relative position of a physiological parameter in its distribution. The specific formula is: where w i represents the weighting coefficient of the ith data point of a single physiological parameter, reflecting the importance and reliability of the data point, y i represents the i-th data point of a single physiological parameter, μ represents the average value of a single physiological parameter data point, σ represents the standard deviation of a single physiological parameter data point, and R represents the robustness adjustment coefficient, which can reduce the impact of outliers on the standardized score and make the result more stable. U represents the dynamic threshold, which can be dynamically adjusted according to the distribution of the data to better adapt to the changes in the data. i Close to 0, indicating that the parameter value is close to the average level, z i If it is greater than 0, it means that the parameter value is higher than the average level.i If it is less than 0, it means that the parameter value is below the average level.

[0046] S23. Calculate health indicators based on the standardized scores of each physiological parameter.

[0047] The standardized score of a single physiological parameter can reflect the health level in a certain aspect, and the health index can comprehensively reflect the health status. The health index Z is calculated based on the standardized score of each physiological parameter. The specific formula is: where g i represents the weighting coefficient of the ith physiological parameter, M represents the number of physiological parameters, z i represents the standardized score of the ith physiological parameter, δ i Represents the time decay factor, taking into account the newness of the data, and the newer data is more representative. Indicates the motion state factor, reflecting the influence of a specific state. The data response is different in different motion states. θ represents the weight of environmental quality, θ represents the impact of environmental quality on health indicators, and w η It represents the weight of mental health, and η represents the impact of mental health on health indicators, including the degree of psychological stress and emotional level.

[0048] S30. Based on the changing trend of the comprehensive health index values, a health status prediction model is constructed using a machine learning algorithm, and its prediction effectiveness is evaluated by calculating the accuracy of the model.

[0049] Select the most representative features from the standardized physiological parameter samples and construct the feature vector s i , according to the feature vector, the health status prediction model is trained using the machine learning algorithm. The training data set is s i is the feature vector of the i-th sample, I is the number of samples, s i =[z i , Z i , ΔZ i , f i ], where z i represents the standardized score of a physiological parameter in a sample, z i =[z i1 ,z i2 ,...,z id ], d means each sample contains the standardized scores of d physiological parameters. i Represents the health index value, ΔZ i Indicates the trend of health index value change, f i Indicates other characteristics such as age, gender, etc. i is the health status label of the i-th sample, a i∈{-1,1}, 1 means healthy and -1 means unhealthy.

[0050] Training the health status prediction model is divided into the following sub-steps:

[0051] S31. Establish an objective function.

[0052] The objective function defines the goal of model optimization, finding a hyperplane w in the feature space T s+b=0, so that the data points of different categories are separated to the maximum extent. By minimizing the objective function, the optimal hyperplane can be found, thereby achieving accurate prediction of health status. The specific formula for minimizing the objective function is: in represents the regularization term, represents the loss term. w, b are the unknown quantities to be solved, w represents the weight vector, b represents the bias term, and ξ represents the slack variable, which is used to process data that cannot be linearly separated and allows some samples to be wrong in classification. T represents the transpose of the weight vector w. The transpose operation converts a column vector into a row vector. T w represents the square norm of the weight vector w, which is used to control the complexity of the model. C represents the penalty parameter, which is used to balance the regularization term and the loss term in the objective function. I represents the number of samples, ξ i represents the degree of classification error or insufficient interval of the i-th sample. λ represents the regularization parameter, balancing the importance of the health indicator value, L(F(q i ) represents the loss function of the health index value, which is used to measure the deviation degree of the health index value.

[0053] S32. Define constraints to obtain a complete minimization objective function.

[0054] The constraints are Constraints are used to ensure that each sample meets the classification requirements.

[0055] The complete minimization objective function is Where w represents the weight vector, b represents the bias term, ξ represents the slack variable, and w T represents the transpose of the weight vector w, w T w represents the square norm of the weight vector w, C represents the penalty parameter, which is used to balance the regularization term and the loss term in the objective function, I represents the number of samples, ξ i represents the degree of classification error or insufficient interval of the i-th sample. λ represents the regularization parameter, balancing the importance of the health indicator value, L(F(q i ) represents the loss function of the health index value, which is used to measure the deviation of the health index value. i is the health status label of the i-th sample, si is the feature vector of the i-th sample, w T s i +b determines sample s i On which side of the hyperplane is w T s i +b>0, the sample belongs to the positive class, w T s i +b<0, the sample belongs to the negative class. If the sample is correctly classified and has enough margin, then a i (w T s i +b)≥1,ξ i =0.

[0056] S33. Solve the weight vector and bias term.

[0057] Solve the above complete minimization objective function and transform it into a dual problem using the Lagrange multiplier method. The Lagrange function is: where γ is the Lagrange multiplier, i is the Lagrange multiplier of the i-th sample, used to represent the classification constraint a i (w T s i +b)≥1-ξ i The importance of , β is another Lagrange multiplier used to express the non-negative constraint ξ i Importance ≥0.

[0058] The dual problem is Among them, γ is the Lagrange multiplier, γ i is the Lagrange multiplier of the i-th sample, γ j is the Lagrange multiplier of the jth sample, a i represents the health status label of the i-th feature vector, a j represents the health status label of the jth feature vector, The feature vector s represents the i-th sample i and the feature vector s of the jth sample j The inner product of It represents the interaction between samples. By subtracting this term, the influence of the similarity between samples and the combination of labels on the results is minimized. Indicates the abnormality of health index value, L(F(q i ))=|F(q i )-V|,F(q i ) represents the health indicator value output by the model, and V represents the health threshold.

[0059] Solving the dual problem, we can obtain the Lagrange multiplier γi , and then calculate the weight vector w and bias term b, Among them, i:γ i >0 indicates the set of support vectors, which are samples on the boundary of the interval, i.e., γ i >0 samples, s represents the feature vector, k h denotes the label of the hth support vector, s h Denotes the eigenvector of the h-th support vector.

[0060] Thus, the prediction model f(s person )=(sign(w T s person +b),F(q)).sign(w T s person +b) indicates the output health status label, including healthy label and unhealthy label, s person represents the input feature vector consisting of individual physiological parameters and other characteristics, w T represents the transpose of the weight vector, b represents the bias term, and F(q) represents the output health indicator value.

[0061] The accuracy of the model is calculated through cross-validation to evaluate the predictive effectiveness of the model. The specific formula is: I represents the number of test samples, τ represents the indicator function, and a i represents the predicted label, represents the true label, q i represents the predicted health index value, Indicates the true health indicator value.

[0062] S40, applying the prediction model to the real-time data stream, when the health indicator value exceeds the set threshold continuously, triggering the alarm mechanism, and calculating the proportion exceeding the threshold at the same time.

[0063] The trained model is applied to real-time data streams to monitor health status. When the health indicator value exceeds the set threshold V continuously, the alarm mechanism is triggered. The specific formula is: Alert represents the alert flag, F(q) represents the health indicator value, V represents the threshold, and k represents the number of consecutive points.

[0064] The formula P = |F(q)-V| / V is used to calculate the proportion P exceeding or falling below the threshold value, which serves as the key basis for evaluating the user's health risk level.

[0065] S50. Provide personalized health advice based on the size of the ratio value, and use the feedback results to optimize the prediction model.

[0066] According to the size of the ratio value P, the user's health risk level is evaluated and personalized health advice is provided to the user. If the ratio value is 0<|P|<10%, it means that the fluctuation is within the normal range. If the ratio value is 10%<|P|<20%, it means that the fluctuation is abnormal and attention should be paid and measures should be taken. |P|>20% means that the fluctuation is beyond the normal range and medical treatment should be sought in time. At the same time, the feedback results are used to optimize the prediction model. The model optimization formula is: w new represents the new weight vector, w old represents the old weight vector, η represents the learning rate, Represents the gradient.

[0067] Embodiment 2

[0068] like Figure 2 As shown, Embodiment 2 of the present application provides a health monitoring system based on medical wearable device data fusion, including:

[0069] Collection module: includes collection parameter submodule and timestamp submodule.

[0070] Parameter collection submodule: collects physiological parameters from various medical wearable devices.

[0071] Medical wearable devices are divided into vital signs monitoring, sports rehabilitation, and disease diagnosis and management. Common forms include smart bracelets, smart watches, smart patches, etc. with high-precision sensors.

[0072] To obtain physiological parameters and other features from a variety of medical wearable devices, it is necessary to first set the data collection frequency of the medical wearable device according to the data needs to ensure that the data points are dense enough to capture changes in physiological parameters. The specific physiological parameters collected include heart rate, blood pressure, blood oxygen saturation, body temperature, number of steps, sleep quality, acceleration, skin temperature, skin conductance, respiratory rate, body composition, electrocardiogram, etc. These parameters are stored in the form of time series. Other features include height, age, weight, gender, etc.

[0073] Timestamp submodule: ensures data synchronization through timestamp alignment.

[0074] Enable the timestamp recording function in the settings of the medical wearable device to ensure that the timestamp is automatically recorded each time data is collected, design the data storage format, and transfer the collected physiological parameter data to the storage device via Bluetooth, Wi-Fi or other wireless communication methods.

[0075] Extract timestamps for each physiological parameter record to ensure that each data point has an accurate time stamp. Parse the timestamp field of the physiological parameter record, synchronize the data collected by different devices to the same time axis, and perform corrections to ensure that all data points correspond to the same time point.corrected =timestamp iso +Δt for time correction, timestamp corrected Indicates the corrected timestamp, timestamp iso Indicates the standard time format, and Δt indicates the time deviation.

[0076] Processing and calculation module: including preprocessing submodule, outlier detection submodule, calculation standardization submodule, and calculation health indicator submodule.

[0077] Preprocessing submodule: Use signal processing technology to preprocess the collected physiological parameter data.

[0078] The collected physiological parameter data is subjected to noise filtering, and a low-pass filter is used to remove high-frequency noise and retain useful physiological signals. The specific formula is y n =αx n +(1-α)y n-1 ,y n represents the filter output value at the current time n, x n represents the input value of n at the current moment, α represents the smoothing factor, and its value range is 0<α<1. When α is close to 0, the filter responds slowly to new data and the filtering effect is stronger. When α is close to 1, the filter responds faster to new data, but the filtering effect is weaker. n-1 Indicates the filter output value at the previous moment n-1, indicating the filter result that has been calculated before.

[0079] Outlier detection submodule: performs outlier detection on the collected physiological parameter data, and detects and removes data points that exceed 3 times the standard deviation.

[0080] By calculating the standard deviation σ and mean μ of each physiological parameter, data points exceeding 3 times the standard deviation are detected and removed. The standard deviation calculation formula is: Where N represents the number of data points, e i represents the i-th data point, and μ represents the average value of the data point. The average value calculation formula is Where N represents the number of data points, e i represents the i-th data point.

[0081] If a data point e i If the value exceeds 3 times the standard deviation of the mean value μ, the data point is considered an outlier. The upper threshold formula is Limit Upper =μ+3σ, the lower threshold formula is Limit Lower =μ-3σ, if e i >Limit Upper or e i <Limit Lower, then e i For outliers, e i Removed.

[0082] Calculate standardized score submodule: Calculate the standardized score of each physiological parameter.

[0083] The standardized score reflects the relative position of a physiological parameter in its distribution. The specific formula is: where w i represents the weighting coefficient of the ith data point of a single physiological parameter, reflecting the importance and reliability of the data point, y i represents the i-th data point of a single physiological parameter, μ represents the average value of a single physiological parameter data point, σ represents the standard deviation of a single physiological parameter data point, and R represents the robustness adjustment coefficient, which can reduce the impact of outliers on the standardized score and make the result more stable. U represents the dynamic threshold, which can be dynamically adjusted according to the distribution of the data to better adapt to the changes in the data. i Close to 0, indicating that the parameter value is close to the average level, z i If it is greater than 0, it means that the parameter value is higher than the average level. i If it is less than 0, it means that the parameter value is below the average level.

[0084] Calculate health index submodule: Calculate health index based on the standardized scores of each physiological parameter.

[0085] The standardized score of a single physiological parameter can reflect the health level in a certain aspect, and the health index can comprehensively reflect the health status. The health index Z is calculated based on the standardized score of each physiological parameter. The specific formula is: where g i represents the weighting coefficient of the ith physiological parameter, M represents the number of physiological parameters, z i represents the standardized score of the ith physiological parameter, δ i Represents the time decay factor, taking into account the newness of the data, and the newer data is more representative. Indicates the motion state factor, reflecting the influence of a specific state. The data response is different in different motion states. θ represents the weight of environmental quality, θ represents the impact of environmental quality on health indicators, and w η It represents the weight of mental health, and η represents the impact of mental health on health indicators, including the degree of psychological stress and emotional level.

[0086] Training model module: including feature vector sub-module, training health status prediction model sub-module, feature vector sub-module: construct training data set according to feature vector.

[0087] Select the most representative features from the standardized physiological parameter samples and construct the feature vector s i , according to the feature vector, the health status prediction model is trained using the machine learning algorithm. The training data set is s i is the feature vector of the i-th sample, I is the number of samples, s i =[z i , Z i , ΔZ i , f i ], where z i represents the standardized score of a physiological parameter in a sample, z i =[z i1 ,z i2 ,...,z id ], d means each sample contains the standardized scores of d physiological parameters. i Represents the health index value, ΔZ i Indicates the trend of health index value change, f i Indicates other characteristics such as age, gender, etc. i is the health status label of the i-th sample, a i ∈{-1,1}, 1 means healthy and -1 means unhealthy.

[0088] Training health status prediction model submodule: including objective function submodule, constraint condition submodule, solution submodule and verification submodule.

[0089] Objective function submodule: set up the objective function.

[0090] The objective function defines the goal of model optimization, finding a hyperplane w in the feature space T s+b=0, so that the data points of different categories are separated to the maximum extent. By minimizing the objective function, the optimal hyperplane can be found, thereby achieving accurate prediction of health status. The specific formula for minimizing the objective function is: in represents the regularization term, represents the loss term. w, b are the unknown quantities to be solved, w represents the weight vector, b represents the bias term, and ξ represents the slack variable, which is used to process data that cannot be linearly separated and allows some samples to be wrong in classification. T represents the transpose of the weight vector w. The transpose operation converts a column vector into a row vector. T w represents the square norm of the weight vector w, which is used to control the complexity of the model. C represents the penalty parameter, which is used to balance the regularization term and the loss term in the objective function. I represents the number of samples, ξ irepresents the degree of classification error or insufficient interval of the i-th sample. λ represents the regularization parameter, balancing the importance of the health indicator value, L(F(q i ) represents the loss function of the health index value, which is used to measure the deviation degree of the health index value.

[0091] Constraint submodule: Define constraints to obtain the complete minimization objective function.

[0092] The constraints are Constraints are used to ensure that each sample meets the classification requirements.

[0093] The complete minimization objective function is Where w represents the weight vector, b represents the bias term, ξ represents the slack variable, and w T represents the transpose of the weight vector w, w T w represents the square norm of the weight vector w, C represents the penalty parameter, which is used to balance the regularization term and the loss term in the objective function, I represents the number of samples, ξ i represents the degree of classification error or insufficient interval of the i-th sample. λ represents the regularization parameter, balancing the importance of the health indicator value, L(F(q i ) represents the loss function of the health index value, which is used to measure the deviation of the health index value. i is the health status label of the i-th sample, s i is the feature vector of the i-th sample, w T s i +b determines sample s i On which side of the hyperplane is w T s i +b>0, the sample belongs to the positive class, w T s i +b<0, the sample belongs to the negative class. If the sample is correctly classified and has enough margin, then a i (w T s i +b)≥1,ξ i =0.

[0094] Solving submodule: solving weight vector and bias term.

[0095] Solve the above complete minimization objective function and transform it into a dual problem using the Lagrange multiplier method. The Lagrange function is: where γ is the Lagrange multiplier, i is the Lagrange multiplier of the i-th sample, used to represent the classification constraint a i (w T s i +b)≥1-ξ iThe importance of , β is another Lagrange multiplier used to express the non-negative constraint ξ i Importance ≥0.

[0096] The dual problem is Among them, γ is the Lagrange multiplier, γ i is the Lagrange multiplier of the i-th sample, γ j is the Lagrange multiplier of the jth sample, a i represents the health status label of the i-th feature vector, a j represents the health status label of the jth feature vector, The feature vector s represents the i-th sample i and the feature vector s of the jth sample j The inner product of It represents the interaction between samples. By subtracting this term, the influence of the similarity between samples and the combination of labels on the results is minimized. Indicates the abnormality of health index value, L(F(q i ))=|F(q i )-V|,F(q i ) represents the health indicator value output by the model, and V represents the health threshold.

[0097] Solving the dual problem, we can obtain the Lagrange multiplier γ i , and then calculate the weight vector w and bias term b, Among them, i:γ i >0 indicates the set of support vectors, which are samples on the boundary of the interval, i.e., γ i >0 samples, s represents the feature vector, k h denotes the label of the hth support vector, s h Denotes the eigenvector of the h-th support vector.

[0098] Thus, the prediction model f(s person )=(sign(w T s person +b),F(q)).sign(w T s person +b) indicates the output health status label, including healthy label and unhealthy label, s person represents the input feature vector consisting of individual physiological parameters and other characteristics, w T represents the transpose of the weight vector, b represents the bias term, and F(q) represents the output health indicator value.

[0099] Verification submodule: Calculate the accuracy of the model through cross-validation and evaluate the predictive effectiveness of the model.

[0100] The specific formula is I represents the number of test samples, τ represents the indicator function, and a i represents the predicted label, represents the true label, q i represents the predicted health index value, Indicates the true health indicator value.

[0101] Alarm module: includes alarm mechanism submodule and calculation ratio submodule.

[0102] Alarm mechanism submodule: Apply the trained model to the real-time data stream to monitor the health status. When the health indicator value exceeds the set threshold V continuously, the alarm mechanism is triggered.

[0103] The specific formula is Alert represents the alert flag, F(q) represents the health indicator value, V represents the threshold, and k represents the number of consecutive points.

[0104] Calculate ratio submodule: Calculate the ratio exceeding or falling below the threshold.

[0105] The formula P = |F(q)-V| / V is used to calculate the proportion P exceeding or falling below the threshold value, which serves as the key basis for evaluating the user's health risk level.

[0106] Feedback optimization module: Provides personalized health advice based on the size of the ratio value, and uses the feedback results to optimize the prediction model.

[0107] According to the size of the ratio value P, the user's health risk level is evaluated and personalized health advice is provided to the user. If the ratio value is 0<|P|<10%, it means that the fluctuation is within the normal range. If the ratio value is 10%<|P|<20%, it means that the fluctuation is abnormal and attention should be paid and measures should be taken. |P|>20% means that the fluctuation is beyond the normal range and medical treatment should be sought in time. At the same time, the feedback results are used to optimize the prediction model. The model optimization formula is: w new represents the new weight vector, w old represents the old weight vector, η represents the learning rate, Represents the gradient.

[0108] Corresponding to the above embodiment, an embodiment of the present invention provides a computer storage medium, including: at least one memory and at least one processor;

[0109] The memory is used to store one or more program instructions;

[0110] A processor is used to run one or more program instructions to execute a health monitoring method based on medical wearable device data fusion.

[0111] Corresponding to the above embodiments, an embodiment of the present invention provides a computer-readable storage medium, which contains one or more program instructions, and the one or more program instructions are used by a processor to execute a health monitoring method based on medical wearable device data fusion.

[0112] The embodiments disclosed in the present invention provide a computer-readable storage medium, in which computer program instructions are stored. When the computer program instructions are executed on a computer, the computer executes the above-mentioned health monitoring method based on medical wearable device data fusion.

[0113] In the embodiment of the present invention, the processor may be an integrated circuit chip having the ability to process signals. The processor may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0114] The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor for execution, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The processor reads the information in the storage medium and completes the steps of the above method in combination with its hardware.

[0115] The storage medium may be a memory, which may be, for example, a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memory.

[0116] Among them, the non-volatile memory can be a read-only memory (ROM), a programmable ROM (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory.

[0117] The volatile memory may be a random access memory (RAM) which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDRSDRAM), enhanced synchronous DRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus DRAM (DRRAM).

[0118] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0119] Those skilled in the art will appreciate that in one or more of the above examples, the functions described in the present invention can be implemented using a combination of hardware and software. When software is used, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any media that facilitates the transmission of computer programs from one place to another. Storage media can be any available media that can be accessed by general or special-purpose computers.

[0120] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solution of the present invention should be included in the scope of protection of the present invention.

Claims

1. A health monitoring method based on medical wearable device data fusion, characterized in that: include: S10, collect physiological parameters from multiple medical wearable devices and ensure data synchronization through timestamp alignment; S20, preprocessing the collected physiological parameter data using signal processing technology, calculating a standardized score for each physiological parameter, and calculating a health index based on the standardized score of the physiological parameter; S30. Based on the changing trend of the comprehensive health index values, a health status prediction model is constructed using a machine learning algorithm, and its prediction effectiveness is evaluated by calculating the accuracy of the model; S40, applying the prediction model to the real-time data stream, triggering an alarm mechanism when the health indicator value exceeds a set threshold, and calculating the proportion exceeding the threshold; S50: Based on the size of the ratio value and the feedback, the feedback result is used to optimize the prediction model.

2. A health monitoring method based on medical wearable device data fusion as claimed in claim 1, characterized in that: Ensuring data synchronization through timestamp alignment requires enabling the timestamp recording function in the settings of the medical wearable device to ensure that the timestamp is automatically recorded each time data is collected. The collected data needs to be corrected to ensure that all data points correspond to the same time point.

3. A health monitoring method based on medical wearable device data fusion as claimed in claim 1, characterized in that: The collected physiological parameter data are preprocessed using signal processing technology, including noise filtering and outlier detection.

4. A health monitoring method based on medical wearable device data fusion as claimed in claim 1, characterized in that: To build a health status prediction model using a machine learning algorithm, it is necessary to first select the most representative features from the standardized physiological parameter samples and construct feature vectors and training data sets.

5. A health monitoring method based on medical wearable device data fusion as claimed in claim 1, characterized in that: Training the health status prediction model is specifically divided into the following sub-steps: setting up the objective function, defining constraints, obtaining the complete minimized objective function, solving the weight vector and bias term, and obtaining the predicted health status model.

6. A health monitoring method based on medical wearable device data fusion as claimed in claim 5, characterized in that: To solve the complete minimization objective function, the Lagrange multiplier method is used to transform it into a dual problem. By solving the dual problem, the Lagrange multiplier is obtained, and the weight vector and bias term are calculated.

7. A health monitoring method based on medical wearable device data fusion as claimed in claim 1, characterized in that: The accuracy of the model is calculated by cross-validation to evaluate the predictive performance of the model by comparing the difference between the predicted label and the true label.

8. A health monitoring system based on medical wearable device data fusion, characterized in that: include: Collection module: used to collect physiological parameters from various medical wearable devices and ensure data synchronization through timestamp alignment. Processing and calculation module: used to pre-process the collected physiological parameter data using signal processing technology, calculate the standardized score of each physiological parameter, and calculate the health index based on the standardized score of the physiological parameter. Training model module: Based on the changing trend of comprehensive health indicator values, a health status prediction model is constructed using machine learning algorithms, and its prediction effectiveness is evaluated by calculating the accuracy of the model. Alarm module: used to apply the prediction model to real-time data streams. When the health indicator value exceeds the set threshold continuously, the alarm mechanism is triggered and the proportion exceeding the threshold is calculated. Feedback optimization module: Provides personalized health advice based on the size of the ratio value, and uses the feedback results to optimize the prediction model.

9. A computer storage medium, characterized in that include: at least one memory and at least one processor; A memory for storing one or more program instructions; The processor is used to run one or more program instructions to execute a method for implementing a health monitoring system based on medical wearable device data fusion.

Images