A method and system for temporal fusion of multidimensional physiological data and prediction of health status

Abstract

This invention provides a method and system for time-series fusion of multidimensional physiological data and prediction of health status, relating to the field of electronic digital data processing technology. The invention employs high-precision sensing devices and parameter calibration models to calibrate the data, and uses a weighted fusion algorithm to achieve time synchronization and error compensation, improving the accuracy of physiological parameters. Simultaneously, a closed-loop feedback mechanism is used to correct fatigue and stress indices in real time, and a comprehensive status index is generated based on the evaluation results for further analysis and prediction. Machine learning technology is used to predict the changing trend of the comprehensive status index, and combined with the user's physiological data, targeted dynamic rehabilitation plans are provided, including training intensity optimization, dietary adjustments, and rest cycle planning. A timed feedback mechanism is established to allow users to obtain real-time changes in their health status and adjustment suggestions, thereby improving the scientific rigor and timeliness of rehabilitation plans, significantly optimizing rehabilitation effects and efficiency, and simultaneously achieving intelligent health management.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing technology, specifically to a method and system for temporal fusion of multidimensional physiological data and prediction of health status. Background Technology

[0002] Methods for estimating rehabilitation status originated in sports medicine and aim to assess patients' rehabilitation progress. With the development of exercise physiology and biomechanics, scientific assessment tools, such as functional assessment standards and biomarkers, have gradually been formed, promoting the personalization and precision of rehabilitation treatment.

[0003] The existing technology, with publication number CN106570346A, entitled "Method and System for Determining Physiological Condition Assessment Factors," includes the following steps: acquiring multiple first risk factors for a specific physiological condition based on knowledge; acquiring multiple second risk molecules related to the same physiological condition based on clinical data; performing logistic regression model analysis on the second risk molecules; calculating the correlation coefficient between the first risk molecules and the second risk molecules after logistic regression model analysis to determine the correlation between the first and second risk molecules; and selecting assessment factors for a specific physiological condition based on the correlation between the first and second risk molecules to determine the assessment factors valuable for that physiological condition. This method can effectively improve the quality of risk factors, reduce the modeling dimensions of risk prediction models, provide effective risk factor assessment, and offer a valid basis for risk prediction of physiological conditions.

[0004] The above technical solutions, when applied to postoperative rehabilitation physiological status assessment methods, often have the following technical drawbacks:

[0005] The lack of real-time dynamic acquisition and fusion of multidimensional physiological data under multiple exercise states, and the focus on the screening and assessment of static risk factors, makes it impossible to achieve automatic calibration, time synchronization and error compensation of key physiological indicators such as fatigue and stress under different exercise states. This results in insufficient capture of dynamic changes in the user's health status and makes it difficult to accurately reflect the actual physiological condition.

[0006] Meanwhile, the above-mentioned technical solutions lack the functions of closed-loop feedback correction based on time-series data and comprehensive assessment and prediction of health status. Since they have failed to achieve real-time closed-loop adjustment of fatigue index and stress index, and have not built an analysis model of comprehensive status index and its changing trend, they cannot provide users with dynamic health status prediction and personalized rehabilitation plan guidance, thus limiting the timeliness and personalization of health management.

[0007] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0008] The purpose of this invention is to provide a method and system for temporal fusion of multidimensional physiological data and prediction of health status, so as to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] A method for temporal fusion of multidimensional physiological data and prediction of health status, comprising the following steps:

[0011] Step 1: Collect multidimensional physiological data of the user in different exercise states during the current exercise and rehabilitation stage in real time, and analyze the multidimensional physiological data of different exercise states to build a parameter calibration model; and the multidimensional physiological data includes fatigue-related indicators and stress-related indicators.

[0012] The parameter calibration model is used to automatically calibrate multidimensional physiological data under different motion states;

[0013] Step 2: Acquire the automatically calibrated multidimensional physiological data, and use a weighted fusion algorithm to synchronize the time and compensate for errors in the multidimensional physiological data of different motion states to generate physiological parameter sequences for different motion states.

[0014] Step 3: Based on the physiological parameter sequences of different exercise states, calculate the fatigue index and stress index under different exercise states and perform closed-loop correction; preset thresholds and compare and evaluate the corrected fatigue index and stress index.

[0015] Step 4: Based on the assessment content of the fatigue index and stress index, further generate a comprehensive state index under different exercise states;

[0016] Simultaneously, the comprehensive state index under adjacent motion states is analyzed to generate a difference index;

[0017] Step 5: Based on the difference index, predict the trend of changes in the comprehensive status index over a preset time period; and set a feedback cycle to periodically provide users with advance feedback on the trend of changes in the comprehensive status index through wearable devices, while also issuing dynamic rehabilitation plan adjustment suggestions.

[0018] Furthermore, step one specifically includes:

[0019] The system is equipped with wearable devices to collect multidimensional physiological data in real time, including heart rate (HR), respiratory rate (BR), blood oxygen saturation (Xyb), heart rate variability (HRV), skin conductance response (GSR), and cortisol level (Cor). The data is then transmitted in real time to the main control unit of the wearable device via BLE communication for unified processing. At the same time, by adding timestamps, the multidimensional physiological data is made time-correlated.

[0020] Based on the motion analysis results of the triaxial accelerometer in the wearable device, the different motion states in the current exercise and rehabilitation stage are sequentially labeled to form a label sequence {1,2,...,i,...,n}; i represents the sequential label of the motion state at the current moment, and n represents the total number of motion states finally identified in the entire exercise and rehabilitation stage.

[0021] Based on the labeled sequence, the collected multidimensional physiological data are associated with the corresponding motion states to form a multidimensional dataset.

[0022] Furthermore, step one specifically includes:

[0023] During the user's movement, the main control MCU reads the raw multidimensional physiological data currently labeled i in real time and outputs calibration values ​​through the trained vector regression model;

[0024] If a change in the labeled sequence is detected, i.e. the motion state switches from i to i+1, the online fine-tuning mechanism is triggered to perform incremental learning using real-time physiological data and labeled sequences to update the model parameters.

[0025] The calibrated data is transmitted to the application in real time via wireless communication.

[0026] Furthermore, step two specifically includes:

[0027] A weighted fusion algorithm is used to perform error compensation and time synchronization processing on the automatically calibrated multidimensional physiological data; linear interpolation is used to synchronize data with inconsistent timestamps, including the calculation of interpolated heart rate data.

[0028] Interpolation methods are used to synchronize data from different timestamps to compensate for signal loss or delay during the acquisition process; linear interpolation is selected to achieve a smooth transition of time data and ensure the temporal continuity of the fused data.

[0029] Error compensation is performed on the collected data. Error calculation is based on the statistical differences of historical data, and dynamic error correction is performed using a Kalman filter. The Kalman filter dynamically adjusts the error estimation according to the difference between the real-time observed value and the predicted value.

[0030] The generated physiological parameter sequence is processed through time synchronization and error compensation to form a continuous physiological parameter sequence {P1, P2, ..., P...}. i , ..., P n Each value represents a fusion physiological parameter value under a certain movement state, where P i P represents the physiological parameter value of the i-th motion state. n The physiological parameter value representing the nth motion state;

[0031] Furthermore, step three specifically includes:

[0032] Based on the heart rate, respiratory rate, and blood oxygen saturation from the physiological parameter sequence, after standardization, let {HR} i ,BR i ,Xyb i} represent heart rate, respiratory rate, and blood oxygen saturation in the i-th exercise state, respectively. The fatigue index is calculated using the following formula:

[0033]

[0034] in, Let be the fatigue index under the i-th exercise state, and k1, k2 and k3 are the weighting coefficients of heart rate, respiratory rate and blood oxygen saturation respectively, and k1+k2+k3=1;

[0035] Based on the heart rate variability, skin conductance response, and cortisol levels in the physiological parameter sequences, standardization was performed, and {HRV} was defined as follows: i GSR i Cor i} represent the heart rate variability, skin conductance response, and cortisol level in the i-th exercise state, respectively. The stress index is calculated using the following formula:

[0036]

[0037] in, Let k be the stress index under the i-th exercise state, and k4, k5 and k6 are the weighting coefficients of the corresponding heart rate variability, skin conductance response and cortisol level, respectively, and k4+k5+k6=1.

[0038] Furthermore, step three specifically includes:

[0039] Real-time closed-loop feedback correction is performed on the fatigue index and stress index, and two consecutive timestamps t are extracted from the data stream. The fatigue index corresponding to 1 and t, i.e., Apl(t) 1) with Apl(t), and the pressure exponent, i.e., Ayl(t) 1) The fatigue deviation ΔApl and pressure deviation ΔAyl are obtained by calculating the difference between Ayl(t);

[0040] The specific content of the closed-loop feedback adjustment includes a comprehensive evaluation and visualization output of the corrected fatigue index and stress index; the evaluation logic is as follows:

[0041] Set thresholds for the fatigue index and the pressure index, namely the fatigue threshold Tpl and the pressure threshold Tyl;

[0042] Normal range: If both the fatigue index and the stress index are below their respective thresholds, it is considered normal, and the user can continue the current activity.

[0043] Mild Abnormality: If either the fatigue index or the stress index exceeds the corresponding threshold, a warning of mild abnormality will be issued, and suggestions for adjusting the activity will be made.

[0044] Severe Abnormality: If both the fatigue index and stress index exceed the corresponding threshold, it is marked as a severe abnormality. In this case, it is recommended that the user adjust the exercise intensity in time or request a necessary health check.

[0045] Furthermore, step four specifically includes:

[0046] The fatigue index and stress index are combined to generate a comprehensive condition index. The specific calculation formula is as follows:

[0047]

[0048] in, The index represents the comprehensive state index under the current motion state i. k7 and k8 are the weight parameters of the combined fatigue index and stress index, respectively, and k7+k8=1.

[0049] After generating the comprehensive state index, the difference index between adjacent motion states is analyzed. By comparing the difference in the comprehensive state index between adjacent motion states i and i+1, a difference index is generated. The specific calculation formula is as follows:

[0050]

[0051] Difference Index This reflects the magnitude of changes in stress and fatigue from the current state to the next state. Furthermore, step five specifically includes:

[0052] A comprehensive state index under the above motion state is established using machine learning methods. and difference index As input, a state prediction model is used to predict the trend of changes in the comprehensive state index over a future period of time.

[0053] Set the timing for data synchronization and feedback cycles according to the user's specific needs and activity types;

[0054] At the end of each cycle, the predicted results of the comprehensive state index are transmitted to the data transmission module on the user's wearable device;

[0055] Among them, machine learning methods include linear regression and time series analysis for prediction.

[0056] Furthermore, step five specifically includes:

[0057] Based on the predicted trend of the comprehensive health index, personalized rehabilitation or training adjustment suggestions are automatically generated. By analyzing each user's historical health data, current health status and predicted future health trends, targeted suggestions are generated, including adjusting training intensity, optimizing diet plans and adjusting rest cycles.

[0058] In addition, a user feedback mechanism should be established to allow users to provide feedback on the suggestions they receive, which can then be used to optimize the predictive model and adjust future health management strategies.

[0059] A system for temporal fusion of multidimensional physiological data and prediction of health status includes:

[0060] The physiological data acquisition module collects multidimensional physiological data of users under different exercise states in real time during the current exercise and rehabilitation stage, and analyzes the multidimensional physiological data of different exercise states to build a parameter calibration model; and the multidimensional physiological data includes fatigue-related indicators and stress-related indicators.

[0061] The parameter calibration model is used to automatically calibrate multidimensional physiological data under different motion states;

[0062] The data fusion and calibration module acquires multidimensional physiological data after automatic calibration, and uses a weighted fusion algorithm to perform time synchronization and error compensation on the multidimensional physiological data of different motion states, generating physiological parameter sequences for different motion states.

[0063] The index calculation and correction module calculates the fatigue index and stress index under different exercise states based on the physiological parameter sequences of different exercise states, and performs closed-loop correction; preset thresholds are used to compare and evaluate the corrected fatigue index and stress index.

[0064] The state index analysis module, based on the assessment of fatigue index and stress index, further generates a comprehensive state index under different exercise states;

[0065] Simultaneously, the comprehensive state index under adjacent motion states is analyzed to generate a difference index;

[0066] The trend prediction and feedback module, based on the difference index, predicts the changing trend of the comprehensive status index over a preset period of time; and sets a feedback cycle to periodically provide users with advance feedback on the changing trend of the comprehensive status index through wearable devices, while also issuing dynamic rehabilitation plan adjustment suggestions.

[0067] Compared with the prior art, the beneficial effects of the present invention are:

[0068] This invention achieves dynamic acquisition and correlation of multidimensional physiological data under multiple motion states by configuring wearable devices to collect multidimensional physiological data in real time and forming a label sequence based on triaxial accelerometers for sequential labeling of different motion states. It uses a trained vector regression model to automatically calibrate the original multidimensional physiological data, dynamically updates the model parameters with an online fine-tuning mechanism, and uses a weighted fusion algorithm, linear interpolation method and Kalman filter to perform time synchronization and error compensation of the data to generate a continuous and accurate physiological parameter sequence, effectively solving the problems of insufficient dynamic change capture and imperfect data fusion in the prior art.

[0069] This invention calculates fatigue and stress indices, adjusts them in real time through a closed-loop feedback correction mechanism, and performs a comprehensive evaluation based on preset thresholds to generate a comprehensive state index with fusion weights k7 and k8. It further calculates the difference index between adjacent movement states and establishes a state prediction model based on machine learning methods, using the comprehensive state index and difference index as inputs. This model predicts future trends in the comprehensive state index and provides feedback to the user via wearable devices, automatically generating personalized rehabilitation and training adjustment suggestions. This achieves dynamic closed-loop management and personalized guidance of health status, significantly improving the timeliness and personalization of health management. Attached Figure Description

[0070] Figure 1 This is a schematic diagram of the overall method flow of the present invention;

[0071] Figure 2 This is a schematic diagram of the overall system framework of the present invention. Detailed Implementation

[0072] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0073] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship will also change accordingly.

[0074] Example 1:

[0075] Please see Figure 1 This invention provides a technical solution: a method for temporal fusion of multidimensional physiological data and prediction of health status, the specific steps of which include:

[0076] Step 1: Collect multidimensional physiological data of the user in different exercise states during the current exercise and rehabilitation stage in real time, and analyze the multidimensional physiological data of different exercise states to build a parameter calibration model; and the multidimensional physiological data includes fatigue-related indicators and stress-related indicators.

[0077] The parameter calibration model is used to automatically calibrate multidimensional physiological data under different motion states;

[0078] Step 2: Acquire the automatically calibrated multidimensional physiological data, and use a weighted fusion algorithm to synchronize the time and compensate for errors in the multidimensional physiological data of different motion states to generate physiological parameter sequences for different motion states.

[0079] Step 3: Based on the physiological parameter sequences of different exercise states, calculate the fatigue index and stress index under different exercise states and perform closed-loop correction; preset thresholds and compare and evaluate the corrected fatigue index and stress index.

[0080] Step 4: Based on the assessment content of the fatigue index and stress index, further generate a comprehensive state index under different exercise states;

[0081] Simultaneously, the comprehensive state index under adjacent motion states is analyzed to generate a difference index;

[0082] Step 5: Based on the difference index, predict the trend of changes in the comprehensive status index over a preset time period; and set a feedback cycle to periodically provide users with advance feedback on the trend of changes in the comprehensive status index through wearable devices, while also issuing dynamic rehabilitation plan adjustment suggestions.

[0083] Step one specifically includes:

[0084] A wearable device is configured to collect multidimensional physiological data in real time, including heart rate (HR), respiratory rate (BR), blood oxygen saturation (Xyb), heart rate variability (HRV), skin conductance response (GSR), and cortisol level (Cor). This data is transmitted in real time to the main control unit of the wearable device via BLE communication for unified processing. Simultaneously, timestamps are added to ensure the multidimensional physiological data has temporal correlation. In this embodiment, "BLE" refers to Bluetooth Low Energy technology.

[0085] Based on the motion analysis results of the triaxial accelerometer in the wearable device, the different motion states in the current exercise and rehabilitation stage are sequentially labeled to form a label sequence {1,2,...,i,...,n}; i represents the sequential label of the motion state at the current moment, and n represents the total number of motion states finally identified in the entire exercise and rehabilitation stage.

[0086] Furthermore, the specific procedures for collecting multidimensional physiological data include:

[0087] Heart rate and blood oxygen data are collected using a photoplethysmography (PPG) sensor; motion status information, such as rest, walking, and running, is measured using a triaxial accelerometer; respiratory rate is monitored using a chest sensor; muscle activity is measured using electrode sensors; and a continuous sequence of markers is associated with the collected data to indicate the current user's motion stage.

[0088] Furthermore, assuming that during the user's exercise and rehabilitation phase, the motion state is classified and sequentially labeled using a motion state recognition algorithm:

[0089] The user starts from "still", with the state marked as 1;

[0090] The user then enters "walking" mode, marked as 2;

[0091] Then enter "Running" mode and mark it as 3;

[0092] Finally, it enters the "rest" state and is marked as 4;

[0093] In this case, i represents the current marker of the motion state, that is, at a certain moment the user is in a motion state marked as 2, which is "walking";

[0094] n represents the total number of all different motion states identified. In this example, n=4, which includes motion states such as stationary, walking, running, and resting.

[0095] The structure of the labeled sequence helps the model accurately track changes in the user's motion state, and in subsequent analysis and calibration, it links the physiological data of each state with its motion stage, thus forming a complete data recording and analysis process.

[0096] Based on the labeled sequence, the collected multidimensional physiological data are associated with the corresponding motion states to form a multidimensional dataset.

[0097] Preliminary analysis of movement status and user physiological performance using physiological data parameters:

[0098] The dynamic threshold algorithm is used to assess changes in heart rate (HR) and blood oxygen saturation (Xyb) under specific exercise conditions to determine whether the patient is in a state of fatigue or stress; muscle activity levels are analyzed based on electromyography (EMG) data, and exercise intensity is determined by combining triaxial accelerometer (ACC) data.

[0099] By using state recognition algorithms, including decision trees, motion states are classified and the label sequence is continuously updated, thereby evaluating the user's continuous motion state in real time.

[0100] By combining the labeled sequence {1,2,…,i,…,n} with the corresponding physiological data set, a calibration model of motion state and physiological parameters is established through regression analysis or machine learning algorithms:

[0101] The calibration of physiological indicators was performed using multiple linear regression:

[0102]

[0103] In the formula, Y represents a quantitative indicator of the exercise state, namely fatigue level or exercise performance score;

[0104] HR stands for Heart Rate Variable, which is the heart rate value measured under specific exercise conditions;

[0105] Xyb represents the blood oxygen saturation variable, that is, the blood oxygen saturation measured under a specific exercise state;

[0106] α represents the intercept in the regression model, which is the expected value of the target variable when all independent variables are zero;

[0107] β1 and β2 represent regression coefficients, which describe the strength of the influence of heart rate and blood oxygen saturation on quantitative indicators of exercise status, respectively.

[0108] The error term represents the deviation between the observed value and the regression line, reflecting the influence of unobserved factors other than heart rate and blood oxygen saturation.

[0109] The operating logic of this formula is as follows: Under different exercise states, corresponding heart rate and blood oxygen saturation data, along with corresponding exercise state scores or labels, are collected; using statistical analysis software or programming environments, such as R or Python, the collected data is applied to the above regression model equation, and the regression parameters of the optimal solution are calculated using the least squares method; model validation: cross-validation of the model is performed to check its predictive accuracy and generalization ability; this step helps to confirm whether the model is effective and how reliable it is; after the model is established, it can be used to predict the exercise state of an exerciser at a given heart rate and blood oxygen saturation, or to provide personalized exercise suggestions to the exerciser.

[0110] Step one also includes:

[0111] During the user's movement, the microcontroller-unit (MCU) reads the raw multidimensional physiological data currently labeled i in real time and outputs calibration values ​​through the trained vector regression model;

[0112] If a change in the labeled sequence is detected, i.e. the motion state switches from i to i+1, the online fine-tuning mechanism is triggered to perform incremental learning using real-time physiological data and labeled sequences to update the model parameters.

[0113] The calibrated data is transmitted to the application in real time via wireless communication.

[0114] Step two specifically includes:

[0115] A weighted fusion algorithm is used to perform error compensation and time synchronization processing on the automatically calibrated multidimensional physiological data; linear interpolation is used to synchronize data with inconsistent timestamps, including the calculation of interpolated heart rate data.

[0116] Interpolation methods are used to synchronize data from different timestamps to compensate for signal loss or delay during the acquisition process; linear interpolation is selected to achieve a smooth transition of time data and ensure the temporal continuity of the fused data.

[0117] Error compensation is performed on the collected data. Error calculation is based on the statistical differences of historical data, and dynamic error correction is performed using a Kalman filter. The Kalman filter dynamically adjusts the error estimation according to the difference between the real-time observed value and the predicted value.

[0118] The generated physiological parameter sequence is processed through time synchronization and error compensation to form a continuous physiological parameter sequence {P1, P2, ..., P...}. i , ..., P n Each value represents a fusion physiological parameter value under a certain movement state, where P i P represents the physiological parameter value of the i-th motion state. n The physiological parameter value representing the nth motion state;

[0119] Application of weighted fusion algorithm: To achieve the fusion of different data sources, a weighted fusion model is defined, which involves adjusting the weight coefficients of multiple data sources; the weight settings are optimized based on the data source acquisition quality and its correlation with the target motion state.

[0120] Based on the formula:

[0121]

[0122] Where: P iis the physiological parameter value of the i-th exercise state; M3 is the number of data source types, namely heart rate, blood oxygen saturation, and respiratory rate; w k X is the weight coefficient corresponding to the k-th data source; k,i It is the raw multidimensional physiological data collected from the k-th data source in the i-th motion state;

[0123] In this formula, the weighting coefficient w k and data source X k,i The specific quantity and type depend on the system settings and the actual data collected, and no longer describe individual physiological indicators one by one, thereby simplifying the form to adapt to the scalability and flexibility of multi-source data;

[0124] In this embodiment, the physiological parameter sequence undergoes time synchronization and error compensation processing as follows:

[0125] In processing the physiological parameter sequence, linear interpolation is first used to correct for inconsistencies in timestamps from different data sources, addressing signal loss or delays that may occur during sensor acquisition and ensuring data temporal continuity. The interpolation formula generates a smoothly transitioning interpolated data set by calculating the difference between timestamps. Next, a Kalman filter is used for dynamic error compensation of the interpolated data, adjusting the error between observed and predicted values ​​in real time to dynamically correct data accuracy. The Kalman filter updates the Kalman gain and combines prediction and observation errors to generate accurate corrected physiological parameter values. Finally, a weighted fusion algorithm is used to adjust weighting coefficients based on data source acquisition quality and correlation with motion state. The fused multi-source data forms a continuous physiological parameter sequence that is temporally consistent, with fully compensated errors, ensuring high data accuracy and real-time performance, providing reliable support for subsequent rehabilitation assessments and program adjustments.

[0126] Step three specifically includes:

[0127] Based on the heart rate, respiratory rate, and blood oxygen saturation from the physiological parameter sequence, after standardization, let {HR} i ,BR i ,Xyb i} represent heart rate, respiratory rate, and blood oxygen saturation in the i-th exercise state, respectively. The fatigue index is calculated using the following formula:

[0128]

[0129] in, Let be the fatigue index under the i-th exercise state, and k1, k2 and k3 are the weighting coefficients of heart rate, respiratory rate and blood oxygen saturation respectively, and k1+k2+k3=1;

[0130] Furthermore, the normalized range of heart rate, respiratory rate and blood oxygen saturation in the i-th exercise state is (0,1).

[0131] Based on the heart rate variability, skin conductance response, and cortisol levels in the physiological parameter sequences, standardization was performed, and {HRV} was defined as follows: i GSR i Cor i} represent the heart rate variability, skin conductance response, and cortisol level in the i-th exercise state, respectively. The stress index is calculated using the following formula:

[0132]

[0133] in, Let k be the stress index under the i-th exercise state, and k4, k5 and k6 are the weighting coefficients of the corresponding heart rate variability, skin conductance response and cortisol level, respectively, and k4+k5+k6=1.

[0134] Step three also includes:

[0135] Real-time closed-loop feedback correction is performed on the fatigue index and stress index, and two consecutive timestamps t are extracted from the data stream. The fatigue index corresponding to 1 and t, i.e., Apl(t) 1) with Apl(t), and the pressure exponent, i.e., Ayl(t) 1) The fatigue deviation ΔApl and pressure deviation ΔAyl are obtained by calculating the difference between Ayl(t);

[0136] The specific formula for calculating the deviation is as follows:

[0137]

[0138]

[0139] Real-time closed-loop feedback correction is performed based on fatigue deviation ΔApl and pressure deviation ΔAyl; based on the analytic hierarchy process (AHP) adopted by the expert group, early warning thresholds are set for the fatigue index and pressure index, which are defined as fatigue threshold Tpl and pressure threshold Tyl, respectively.

[0140] The specific content of the closed-loop feedback adjustment includes a comprehensive evaluation and visualization output of the corrected fatigue index and stress index; the evaluation logic is as follows:

[0141] Normal range: If both the fatigue index and the stress index are lower than their respective thresholds, namely the fatigue threshold Tpl and the stress threshold Tyl, then it is judged as normal, and the user continues the current activity unchanged;

[0142] Mild Abnormality: If either the fatigue index or the stress index exceeds the corresponding threshold, a warning of mild abnormality will be issued, and suggestions for adjusting the activity will be made.

[0143] Severe Abnormality: If both the fatigue index and stress index exceed the corresponding threshold, it is marked as a severe abnormality. In this case, it is recommended that the user adjust the exercise intensity in time or request a necessary health check.

[0144] Step four specifically includes:

[0145] The fatigue index and stress index are combined to generate a comprehensive condition index. The specific calculation formula is as follows:

[0146]

[0147] in, The index represents the comprehensive state index under the current motion state i. k7 and k8 are the weight parameters of the combined fatigue index and stress index, respectively, and k7+k8=1.

[0148] After generating the comprehensive state index, the difference index between adjacent motion states is analyzed. By comparing the difference in the comprehensive state index between adjacent motion states i and i+1, a difference index is generated. The specific calculation formula is as follows:

[0149]

[0150] Difference Index This reflects the magnitude of changes in stress and fatigue from the current state to the next. In sports training and health management, monitoring the fatigue and stress of athletes or users under different exercise conditions is crucial. This method provides a simple and effective way to integrate these indicators and, by analyzing changes between adjacent states, helps coaches and athletes better understand the dynamic changes in their physical and emotional state, thereby optimizing training and recovery strategies.

[0151] This approach is superior to other complex models such as neural networks because it is not only computationally simple but also highly transparent, making it easy for athletes and coaches to understand and use.

[0152] Step five specifically includes:

[0153] A comprehensive state index under the above motion state is established using machine learning methods. and difference index As input, a state prediction model is used to predict the trend of changes in the comprehensive state index over a future period of time.

[0154] Set the timing for data synchronization and feedback cycles according to the user's specific needs and activity types;

[0155] At the end of each cycle, the predicted results of the comprehensive state index are transmitted to the data transmission module on the user's wearable device;

[0156] Among them, machine learning methods include linear regression and time series analysis for prediction.

[0157] Step five also includes:

[0158] Based on the predicted trend of the comprehensive health index, personalized rehabilitation or training adjustment suggestions are automatically generated. By analyzing each user's historical health data, current health status and predicted future health trends, targeted suggestions are generated, including adjusting training intensity, optimizing diet plans and adjusting rest cycles.

[0159] In addition, a user feedback mechanism should be established to allow users to provide feedback on the suggestions they receive, which can then be used to optimize the predictive model and adjust future health management strategies.

[0160] In this embodiment, to meet the needs of real-time prediction of patients' health status and generation of personalized rehabilitation suggestions in postoperative rehabilitation management, this technical solution establishes a status prediction model using machine learning methods. Using a comprehensive status index and a difference index as input variables, it predicts the trend of changes in the comprehensive status index of patients over a future period based on linear regression and time series analysis. The calculation formula of the prediction model optimizes the model parameters using a training dataset, thereby achieving accurate prediction of the trend of patients' health status.

[0161] Based on the status prediction results, a data synchronization and feedback cycle is set according to the patient's specific activity type and recovery status. At the end of the cycle, the predicted comprehensive status index result is fed back to the patient's worn terminal device via the data transmission module, ensuring real-time data transmission and facilitating patient viewing and tracking of their health status. Subsequently, based on the changing trend of the comprehensive status index, combined with the patient's historical health data and current health status, personalized rehabilitation suggestions will be generated for the patient. These suggestions include adjusting training intensity, optimizing diet plans and rest cycles, providing comprehensive and targeted support for the patient's postoperative rehabilitation.

[0162] To improve the accuracy of the predictive model and the user experience, this solution also includes a user feedback mechanism. After receiving health advice, patients can provide feedback on their devices. This feedback will be used to dynamically optimize the machine learning model, thereby improving future health management strategies for patients. This feedback loop further enhances the personalization of rehabilitation recommendations and the reliability of model predictions, forming a health management system that is adjusted in real time and continuously optimized.

[0163] The overall technical solution effectively meets the needs for accurate prediction and dynamic suggestions in postoperative rehabilitation management through status prediction and real-time adjustment mechanisms, while ensuring the feasibility and innovation of the model. Its results can significantly improve the patient's rehabilitation experience and enhance the quality of rehabilitation.

[0164] Example 2:

[0165] Please see Figure 2 A system for temporal fusion of multidimensional physiological data and prediction of health status, comprising:

[0166] The physiological data acquisition module collects multidimensional physiological data of users under different exercise states in real time during the current exercise and rehabilitation stage, and analyzes the multidimensional physiological data of different exercise states to build a parameter calibration model; and the multidimensional physiological data includes fatigue-related indicators and stress-related indicators.

[0167] The parameter calibration model is used to automatically calibrate multidimensional physiological data under different motion states;

[0168] The data fusion and calibration module acquires multidimensional physiological data after automatic calibration, and uses a weighted fusion algorithm to perform time synchronization and error compensation on the multidimensional physiological data of different motion states, generating physiological parameter sequences for different motion states.

[0169] The index calculation and correction module calculates the fatigue index and stress index under different exercise states based on the physiological parameter sequences of different exercise states, and performs closed-loop correction; preset thresholds are used to compare and evaluate the corrected fatigue index and stress index.

[0170] The state index analysis module, based on the assessment of fatigue index and stress index, further generates a comprehensive state index under different exercise states;

[0171] Simultaneously, the comprehensive state index under adjacent motion states is analyzed to generate a difference index;

[0172] The trend prediction and feedback module, based on the difference index, predicts the changing trend of the comprehensive status index over a preset period of time; and sets a feedback cycle to periodically provide users with advance feedback on the changing trend of the comprehensive status index through wearable devices, while also issuing dynamic rehabilitation plan adjustment suggestions.

[0173] Based on the wearable device-based exercise and rehabilitation status estimation system in this embodiment, the following is a simulated operation example of the wearable device-based exercise and rehabilitation status estimation system, demonstrating the device's role in the exercise management of postoperative rehabilitation patients:

[0174] Table 1 shows the physiological data of users under different exercise states:

[0175]

[0176] A weighted fusion algorithm was used to analyze the automatically calibrated physiological data; time synchronization was performed using linear interpolation.

[0177] The physiological parameter value P of the i-th motion state is calculated using a weighted fusion algorithm. i :

[0178] P i =0.3 0.6 + 0.2 0.5 + 0.5 0.8 = 0.67;

[0179] Calculate the fatigue index under the i-th motion state. The specific calculation formula is as follows:

[0180]

[0181] The pressure index for the i-th motion state is calculated using the following formula:

[0182]

[0183] The formula for calculating the comprehensive state index under motion state i is:

[0184]

[0185] Difference Index The calculation formula is: =0.415 0.37;

[0186] Trend Prediction and Feedback: Use machine learning methods to predict the trend of changes in the comprehensive state index and provide user feedback;

[0187] Data feedback: At the end of the cycle, the synthesized data will be fed back to the user, along with adjustment suggestions, including increasing rest time and adjusting diet.

[0188] It should be noted that all calculation formulas in this application employ regression analysis, including but not limited to machine learning algorithms, to deeply analyze the collected parameters and identify their natural trends and interrelationships. Specialized software, such as Python's Scikit-learn library or the R language, is used to automatically generate mathematical models that match the data. Then, cross-validation and other methods are used to objectively evaluate the model performance, and continuous feedback and optimization are combined to ensure that the created formulas truly reflect the inherent laws of the data, thereby guaranteeing their effectiveness and accuracy. In all calculation formulas in this application, the parameters in each formula undergo dimensionless processing within a consistent range to ensure that different physical quantities are compared on the same scale; dimensionless processing techniques include, but are not limited to, min-max normalization and Z-score standardization.

[0189] The technical solution of this invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of this invention.

[0190] The logic and / or steps represented in the flowchart or otherwise described herein, i.e., a ordered list of executable instructions for implementing logical functions, may be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from or in conjunction with such an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0191] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0192] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for temporal fusion of multidimensional physiological data and prediction of health status, characterized in that, The specific steps include: Step 1: Real-time collection of multidimensional physiological data of users under different exercise states during the current exercise and rehabilitation phase, including heart rate (HR), respiratory rate (BR), blood oxygen saturation (Xyb), heart rate variability (HRV), skin conductance response (GSR), and cortisol level (Cor); and analysis of multidimensional physiological data under different exercise states to construct a parameter calibration model; the multidimensional physiological data includes fatigue-related indicators and stress-related indicators. The parameter calibration model is used to automatically calibrate multidimensional physiological data under different motion states; Step 2: Acquire the automatically calibrated multidimensional physiological data, and use a weighted fusion algorithm to synchronize the time and compensate for errors in the multidimensional physiological data of different motion states to generate physiological parameter sequences for different motion states. Step 3: Based on the physiological parameter sequences of different exercise states, calculate the fatigue index and stress index under different exercise states. Specifically: Based on the heart rate, respiratory rate, and blood oxygen saturation from the physiological parameter sequence, after standardization, let {HR} i ,BR i ,Xyb i } represent heart rate, respiratory rate, and blood oxygen saturation in the i-th exercise state, respectively. The fatigue index is calculated using the following formula: in, Let be the fatigue index under the i-th exercise state, and k1, k2 and k3 are the weighting coefficients of heart rate, respiratory rate and blood oxygen saturation respectively, and k1+k2+k3=1; Based on the heart rate variability, skin conductance response, and cortisol levels in the physiological parameter sequences, standardization was performed, and {HRV} was defined as follows: i GSR i Cor i } represent the heart rate variability, skin conductance response, and cortisol level in the i-th exercise state, respectively. The stress index is calculated using the following formula: in, Let k be the stress index under the i-th exercise state, and k4, k5 and k6 are the weighting coefficients of the corresponding heart rate variability, skin conductance response and cortisol level, respectively, and k4+k5+k6=1; The fatigue index and stress index are corrected in a closed loop; a preset threshold is set, and the corrected fatigue index and stress index are compared and evaluated. Step 4: Based on the assessment content of the fatigue index and stress index, further generate a comprehensive state index under different motion states; simultaneously, analyze the comprehensive state index under adjacent motion states to generate a difference index; specifically including: The fatigue index and stress index are combined to generate a comprehensive condition index. The specific calculation formula is as follows: in, The index represents the comprehensive state index under the current motion state i. k7 and k8 are the weight parameters of the combined fatigue index and stress index, respectively, and k7+k8=1. After generating the comprehensive state index, the difference index between adjacent motion states is analyzed. By comparing the difference in the comprehensive state index between adjacent motion states i and i+1, a difference index is generated. The specific calculation formula is as follows: Difference Index It reflects the magnitude of changes in stress and fatigue from the current state to the next state; Step 5: Based on the difference index, predict the trend of changes in the comprehensive status index over a preset time period; and set a feedback cycle to periodically provide users with advance feedback on the trend of changes in the comprehensive status index through wearable devices, while also issuing dynamic rehabilitation plan adjustment suggestions.

2. The method for temporal fusion of multidimensional physiological data and prediction of health status according to claim 1, characterized in that: Step one specifically includes: Configure wearable devices to collect multidimensional physiological data in real time and transmit it to the main control unit of the wearable device in real time via BLE communication for unified processing; at the same time, by adding timestamps, the multidimensional physiological data has time correlation. Based on the motion analysis results of the triaxial accelerometer in the wearable device, the different motion states in the current exercise and rehabilitation stage are sequentially labeled to form a label sequence {1,2,...,i,...,n}; i represents the sequential label of the motion state at the current moment, and n represents the total number of motion states finally identified in the entire exercise and rehabilitation stage. Based on the labeled sequence, the collected multidimensional physiological data are associated with the corresponding motion states to form a multidimensional dataset.

3. The method for temporal fusion of multidimensional physiological data and prediction of health status according to claim 2, characterized in that: Step one also includes: During the user's movement, the main control MCU reads the raw multidimensional physiological data currently labeled i in real time and outputs calibration values ​​through the trained vector regression model; If a change in the label sequence is detected, i.e. the sequence label of the motion state switches from i to i+1, the online fine-tuning mechanism is triggered to perform incremental learning using real-time physiological data and label sequence to update the model parameters. The calibrated multidimensional physiological data is transmitted to the application in real time via wireless communication.

4. The method for temporal fusion of multidimensional physiological data and prediction of health status according to claim 3, characterized in that: Step two specifically includes: A weighted fusion algorithm is used to perform error compensation and time synchronization processing on the automatically calibrated multidimensional physiological data; a continuous physiological parameter sequence {P1,P2,...,Pj} is formed, where each value represents the fused physiological parameter value under a certain motion state, where P represents the fused physiological parameter value and j represents the position index in the labeled sequence; Synchronize data with inconsistent timestamps by using linear interpolation, including calculating interpolated heart rate data; Error compensation is performed on the automatically calibrated multidimensional physiological data. Error calculation is based on the statistical differences of historical data, and dynamic error correction is performed using a Kalman filter. The Kalman filter dynamically adjusts the error estimation according to the difference between the real-time observed value and the predicted value.

5. The method for temporal fusion of multidimensional physiological data and prediction of health status according to claim 4, characterized in that: Step three specifically includes: Real-time closed-loop feedback correction is performed on the fatigue index and stress index, and two consecutive timestamps t are extracted from the data stream. The fatigue index and pressure index corresponding to 1 and t are obtained by calculating the difference between them to obtain the fatigue deviation ΔApl and the pressure deviation ΔAyl, respectively. The specific content of the closed-loop feedback adjustment includes a comprehensive evaluation and visualization output of the corrected fatigue index and stress index; the evaluation logic is as follows: Normal range: If both the fatigue index and stress index are below their respective thresholds, it is considered normal, and the user can continue their current activity. Mild Abnormality: If either the fatigue index or the stress index exceeds the corresponding threshold, a warning of mild abnormality will be issued, and suggestions for adjusting the activity will be made. Severe Abnormality: If both the fatigue index and stress index exceed the threshold, it is marked as a severe abnormality. In this case, it is recommended that the user adjust the exercise intensity in time.

6. The method for temporal fusion of multidimensional physiological data and prediction of health status according to claim 5, characterized in that: Step five specifically includes: A comprehensive state index under the above motion state is established using machine learning methods. and difference index As input, a state prediction model is used to predict the trend of changes in the comprehensive state index over a future period of time. Set the timing for data synchronization and feedback cycles according to the user's specific needs and activity types; At the end of each cycle, the predicted results of the comprehensive state index are transmitted to the data transmission module on the user's wearable device; Among them, machine learning methods for prediction include linear regression and time series analysis.

7. The method for temporal fusion of multidimensional physiological data and prediction of health status according to claim 6, characterized in that: Step five also includes: Based on the predicted trend of the comprehensive health index, personalized rehabilitation or training adjustment suggestions are automatically generated. By analyzing each user's historical health data, current health status and predicted future health trends, targeted suggestions are generated, including adjusting training intensity, optimizing diet plans and adjusting rest cycles. In addition, a user feedback mechanism should be established to allow users to provide feedback on the suggestions they receive, which can then be used to optimize the predictive model and adjust future health management strategies.

8. A system for temporal fusion of multidimensional physiological data and prediction of health status, characterized in that: The system is used to execute the method for temporal fusion and health status prediction of multidimensional physiological data as described in any one of claims 1-7, including: The physiological data acquisition module collects multidimensional physiological data of users in different exercise states during the current exercise and rehabilitation phase in real time, including heart rate (HR), respiratory rate (BR), blood oxygen saturation (Xyb), heart rate variability (HRV), skin conductance response (GSR), and cortisol level (Cor); and analyzes the multidimensional physiological data of different exercise states to construct a parameter calibration model; and the multidimensional physiological data includes fatigue-related indicators and stress-related indicators. The parameter calibration model is used to automatically calibrate multidimensional physiological data under different motion states; The data fusion and calibration module acquires multidimensional physiological data after automatic calibration, and uses a weighted fusion algorithm to perform time synchronization and error compensation on multidimensional physiological data of different motion states, generating physiological parameter sequences of different motion states. The index calculation and correction module calculates the fatigue index and stress index under different exercise states based on physiological parameter sequences. Specifically: Based on the heart rate, respiratory rate, and blood oxygen saturation from the physiological parameter sequence, after standardization, let {HR} i ,BR i ,Xyb i } represent heart rate, respiratory rate, and blood oxygen saturation in the i-th exercise state, respectively. The fatigue index is calculated using the following formula: in, Let be the fatigue index under the i-th exercise state, and k1, k2 and k3 are the weighting coefficients of heart rate, respiratory rate and blood oxygen saturation respectively, and k1+k2+k3=1; Based on the heart rate variability, skin conductance response, and cortisol levels in the physiological parameter sequences, standardization was performed, and {HRV} was defined as follows: i GSR i Cor i } represent the heart rate variability, skin conductance response, and cortisol level in the i-th exercise state, respectively. The stress index is calculated using the following formula: in, Let k be the stress index under the i-th exercise state, and k4, k5 and k6 are the weighting coefficients of the corresponding heart rate variability, skin conductance response and cortisol level, respectively, and k4+k5+k6=1; The fatigue index and stress index are corrected in a closed loop; a preset threshold is set, and the corrected fatigue index and stress index are compared and evaluated. Step 4: Based on the assessment content of the fatigue index and stress index, further generate a comprehensive state index under different motion states; simultaneously, analyze the comprehensive state index under adjacent motion states to generate a difference index; specifically including: The fatigue index and stress index are combined to generate a comprehensive condition index. The specific calculation formula is as follows: in, The index represents the comprehensive state index under the current motion state i. k7 and k8 are the weight parameters of the combined fatigue index and stress index, respectively, and k7+k8=1. After generating the comprehensive state index, the difference index between adjacent motion states is analyzed. By comparing the difference in the comprehensive state index between adjacent motion states i and i+1, a difference index is generated. The specific calculation formula is as follows: Difference Index It reflects the magnitude of changes in stress and fatigue from the current state to the next state; Step 5: Based on the difference index, predict the trend of changes in the comprehensive status index over a preset time period; and set a feedback cycle to periodically provide users with advance feedback on the trend of changes in the comprehensive status index through wearable devices, while also issuing dynamic rehabilitation plan adjustment suggestions.

Images