Real-time data driven oxygen therapy exercise training and rehabilitation method and apparatus

Abstract

The application discloses a real-time data driven oxygen therapy exercise training and rehabilitation method and device, belongs to the technical field of health management and oxygen-enriched rehabilitation training, and aims to solve the problems of lack of individual adaptability of traditional oxygen therapy exercise schemes, difficulty in dynamically matching the physical state of users, and lag of abnormal monitoring response. The device integrates a rehabilitation intelligent monitoring and evaluation platform, an exercise carrier, an oxygen inhalation machine, a physiological monitoring component, and a brain-computer interface component. Through cooperation of the components, basic physiological and health data of a user are collected, a standard behavior baseline is matched and corrected to an individualized behavior baseline. An individualized oxygen-enriched rehabilitation exercise scheme is generated in combination with parameters of the exercise carrier. During execution, data are collected in real time, AI abnormality detection is performed to determine abnormalities and generate a wake-up prompt, and the baseline is corrected according to the type of the abnormality. The application realizes dynamic adaptation of oxygen therapy exercise schemes to individuals, and improves training safety and rehabilitation accuracy.

Description

Technical Field

[0001] This invention relates to the field of health management and oxygen-enriched rehabilitation training technology, and more specifically to a real-time data-driven oxygen therapy exercise training and rehabilitation method and device. Background Technology

[0002] Oxygen therapy exercise training has been widely used in the field of health intervention, especially for elderly people recovering from Alzheimer's disease or post-tumor surgery. However, current technology still faces significant bottlenecks: Firstly, training programs are mostly based on general population data, failing to adequately consider individual differences such as age and baseline health status, resulting in insufficient adaptability. Secondly, they do not take into account the characteristics of the target exercise vehicle, easily leading to a mismatch between the program and the vehicle's capabilities, directly affecting the smooth execution of training. Furthermore, there is a lack of active exercise functions that do not require user self-adjustment, making it impossible to automatically set target gas concentration, exercise speed, intensity, etc., based on the health indicators of elderly people with limited cognitive or operational abilities. The current rehabilitation equipment relies on fixed thresholds for time-related exercises, making it difficult to use and convenient. Furthermore, its single-dimensional approach to monitoring abnormal training makes it difficult to accurately identify physiological abnormalities and deviations in exercise execution. It also lacks a dynamic adjustment mechanism; once training parameters are fixed, they cannot be corrected based on abnormal situations. Long-term use can easily lead to changes in the user's physical condition, reducing training safety and rehabilitation effectiveness. In addition, existing rehabilitation equipment has limited functionality and fails to integrate and link qi generation, exercise, and multi-parameter monitoring. It cannot optimize parameters based on individual physiological differences in different rehabilitation environments, nor does it have precise control over the neurofeedback dimension, resulting in low rehabilitation efficiency and high safety risks.

[0003] Chinese patent application CN117497131A discloses a smart oxygen therapy cloud platform and management method for COPD based on the Internet of Things (IoT). This system includes an output terminal, a smart application layer, and a sensing IoT layer. The smart application layer analyzes input data and existing platform data to formulate the next oxygen therapy plan, achieving standardized management and online-offline collaboration for COPD patients' home oxygen therapy. However, it still fails to solve the problems of single monitoring dimensions, static plans, lack of neurofeedback support, and weak remote collaboration. Even though research has pointed out the potential value of combining neurofeedback and oxygen therapy for rehabilitation, and that oxygen therapy exercise training for special populations such as COPD urgently requires multi-dimensional monitoring and dynamic adaptation, existing patent technologies have not yet formed a systematic solution. Therefore, to overcome these limitations, this invention proposes a real-time data-driven oxygen therapy exercise training and rehabilitation method and device. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a real-time data-driven method and device for oxygen therapy exercise training and rehabilitation. It solves the problems of how to accurately acquire rehabilitation exercise-related parameters suitable for individual users for oxygen therapy exercise training, generate rehabilitation exercise plans suitable for individual users and target exercise carriers, promptly identify execution anomalies during training, and dynamically correct relevant parameters based on anomaly types to ensure the suitability and safety of training.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Real-time data-driven oxygen therapy exercise training and rehabilitation methods include:

[0007] The system acquires the user's basic physiological information and health monitoring indicators, which are then used to perform stepwise intersection calculations and matching with a standard behavioral baseline database to obtain the user's standard behavioral baseline.

[0008] Based on the user's standard behavioral baseline obtained through matching, a graded testing and rehabilitation exercise program is constructed to modify the user's standard behavioral baseline by analyzing the differences in the user's response characteristics between non-oxygen and oxygen-inhaled environments, thereby obtaining the user's personalized behavioral baseline.

[0009] Based on a personalized behavioral baseline, combined with the movement mode parameters of the target exercise vehicle selected by the user, physiological state goals and exercise execution goals are set for each training stage, and a multi-factor weighted decision-making method is used to generate a personalized oxygen-enriched rehabilitation exercise program.

[0010] When a user triggers the execution of a personalized oxygen-enriched rehabilitation exercise program, real-time data of the user is collected through physiological monitoring components, brain-computer interface components, exercise carriers and oxygen inhalers. Through deviation standardization processing and abnormal judgment threshold correction, AI anomaly detection is performed to identify deviation data points, determine whether the user has execution abnormalities, quantify the degree of abnormality and classify the abnormality type, and then generate an AI wake-up prompt.

[0011] When a user is found to have an execution anomaly, the user's personalized behavior baseline is classified and corrected according to the anomaly type.

[0012] Specifically, the steps to obtain the user's standard behavioral baseline include:

[0013] The system acquires the user’s basic physiological information and health monitoring indicators. The basic physiological information includes age, resting heart rate, and type of underlying disease. The health monitoring indicators include susceptibility gene data and telomere test results.

[0014] The basic physiological information and health test indicators that have passed the validity verification are processed to unify the format and normalize the values ​​to form a structured data feature set;

[0015] Based on the age information in the structured data feature set, and combined with the dynamic age segmentation rules of the standard behavior baseline library, all standard behavior baselines corresponding to the age segment are extracted from the standard behavior baseline library to form the first baseline subset.

[0016] Based on the basic disease type information in the structured data feature set, a standard behavioral baseline consistent with the basic disease type is extracted from the first baseline subset; and the matching tolerance is adjusted according to the variance of the baseline sample distribution corresponding to the basic disease type to form a second baseline subset.

[0017] Based on the health monitoring indicator information in the structured data feature set, the interval mapping matching method is used to extract the standard behavioral baselines from the second baseline subset where the mapping overlap between the health monitoring indicator interval and the user's health monitoring indicator is greater than a preset overlap threshold, forming the third baseline subset.

[0018] Specifically, the steps for obtaining a user's standard behavioral baseline also include:

[0019] Multi-dimensional similarity fusion calculation is performed on each standard behavioral baseline of the third baseline subset, and mapping weights are assigned according to the mapping position of the user's health detection indicators in the corresponding behavioral baseline indicator interval.

[0020] Based on this mapping weight, the degree of fit between the user's health monitoring indicators and the corresponding intervals of each standard behavioral baseline is calculated; and the degree of match between the resting heart rate, the exercise physiological parameters associated with underlying diseases and the corresponding parameters of each standard behavioral baseline is calculated.

[0021] The fit of health monitoring indicators, the matching degree of resting heart rate, and the matching degree of exercise physiological parameters related to underlying diseases are integrated to obtain the comprehensive similarity of each standard behavioral baseline in the third baseline subset;

[0022] If there is a standard behavior baseline in the third baseline subset with a comprehensive similarity greater than the preset high fit standard, then it is determined as the standard behavior baseline that matches the user.

[0023] If none exist, then remove the abnormal standard behavior baselines in the third baseline subset whose comprehensive similarity is lower than the mean of the comprehensive similarity of the third baseline subset minus the standard deviation of the comprehensive similarity. Recalculate the comprehensive similarity of the remaining standard behavior baselines, and select the standard behavior baselines whose comprehensive similarity after recalculation is greater than the preset adaptation threshold to determine the standard behavior baselines that match the user.

[0024] Specifically, the standard behavioral baseline is a collection of multiple standard behavioral baselines formed by hierarchical clustering based on age, underlying disease type, and health test indicator range using a density clustering algorithm. Each set of standard behavioral baselines is a multi-dimensional parameter set reflecting the rehabilitation exercise ability of the corresponding population, including exercise tolerance range parameters, physiological response threshold range parameters, and exercise adaptation range parameters.

[0025] Specifically, the steps for constructing a graded testing and rehabilitation exercise program include:

[0026] Extract core parameters from the user's standard behavioral baseline, including exercise tolerance range parameters, physiological response threshold range parameters, and exercise adaptation range parameters;

[0027] Based on the exercise tolerance interval parameters of the standard behavioral baseline, multiple continuous gradient intervals are divided in ascending order. The gradient intervals are divided according to the distribution density of exercise tolerance in the standard behavioral baseline. Combining the exercise adaptation interval parameters of the standard behavioral baseline, the total duration of a single test exercise is determined, and the duration is allocated according to the importance of each gradient interval. Referring to the physiological response threshold interval parameters of the standard behavioral baseline, the movement frequency change pattern is set for each gradient interval. A gradient test rehabilitation exercise program is thus formed.

[0028] Specifically, the steps to obtain a user's personalized behavioral baseline include:

[0029] In both non-oxygenated and oxygenated environments, the graded rehabilitation exercise was divided into four phases—warm-up, steady-state exercise, intensity escalation, and recovery—based on a synchronous acquisition mechanism of exercise phases. The exercise physiological data of the user was collected synchronously at the same time interval for each exercise phase.

[0030] Outlier removal was performed on the exercise physiological data collected in both non-oxygen and oxygen-inhaled environments. The exercise physiological data of each exercise phase in both environments were aligned with the exercise initiation time as the time zero point and the same time step to construct time-series datasets for both non-oxygen and oxygen-inhaled environments.

[0031] Segmented temporal clustering and differential quantification analysis were used to extract response difference features from the temporal datasets of non-oxygen-inhaling and oxygen-inhaling environments. The response difference features include static difference features, dynamic change features, and motion adaptation association features.

[0032] Based on the response difference characteristics, the user's standard behavioral baseline is modified in a targeted manner, including: adjusting the physiological response threshold range parameters of the standard behavioral baseline using a dynamic migration algorithm based on static difference characteristics; modifying the exercise tolerance range using a tolerance limit mapping algorithm with dynamic change characteristics and exercise adaptation association characteristics as joint inputs; and modifying the exercise adaptation range using an adaptation dynamic weight allocation algorithm based on the action standardization data and intensity tolerance duration data in the exercise adaptation association characteristics, combined with the fluctuation coefficient data in the dynamic change characteristics.

[0033] The revised physiological response threshold range parameters, exercise tolerance range parameters, and exercise adaptation range parameters are integrated to form a personalized behavioral baseline for the user.

[0034] Specifically, a multi-factor weighted decision-making method is used to generate personalized oxygen-enriched rehabilitation exercise programs, including:

[0035] Obtain motion mode parameters of the user's target motion carrier, including motion carrier action execution method parameters, motion intensity adjustment range parameters, and motion rhythm controllable range parameters;

[0036] Set the weight coefficients for each parameter of the personalized behavior baseline. The parameters of the personalized behavior baseline include exercise tolerance interval parameters, physiological response threshold interval parameters, and exercise adaptation interval parameters.

[0037] By combining the user's personalized behavioral baseline parameters with the movement pattern parameters of the target motion vehicle, physiological state goals and movement execution goals are set for each training stage;

[0038] Based on the physiological state goals and exercise execution goals of each training phase, the parameters of the dynamic adjustment curve of exercise intensity, exercise cycle division, oxygen supply mode, and rest interval setting are obtained for each training phase through the phase adaptability parameter calculation method.

[0039] The parameters for the dynamic adjustment curve of exercise intensity are calculated using the parameter intersection mapping method, based on the exercise tolerance interval parameters and the exercise intensity adjustment range parameters of the target exercise vehicle. The exercise cycle division parameters are determined using the functional stage matching method, based on the exercise tolerance interval parameters, the exercise mode parameters of the target exercise vehicle, and the physiological state goals and exercise execution goals of the training stage, setting the exercise cycle duration. The oxygen enrichment mode parameters are determined using the threshold trigger adaptation method, based on the physiological response threshold interval parameters and combined with the physiological state goals of the training stage, setting differentiated oxygen enrichment rules. The rest interval setting parameters are determined using the intensity recovery correlation method, based on the exercise adaptation interval parameters and the post-exercise recovery law, dividing the rest nodes according to the intensity levels of the dynamic adjustment curve of exercise intensity.

[0040] Following the sequential logic of warm-up, main training, and recovery phases, parameters such as dynamic adjustment curve of exercise intensity, exercise cycle division, oxygen supply mode, and rest interval setting are embedded one by one to form a personalized oxygen-enriched rehabilitation exercise plan.

[0041] Specifically, the steps for deviation standardization and anomaly detection threshold correction include:

[0042] When a user triggers the execution of a personalized oxygen-enriched rehabilitation exercise program, real-time data of the user is collected simultaneously. The real-time data includes: dynamic exercise physiological data, brain-computer interface data, exercise execution data, and real-time oxygen supply parameters.

[0043] Add a unified timestamp to the real-time data, divide the data into segments according to the training stage, and construct a real-time data sequence corresponding to each training stage;

[0044] Based on the physiological state goals and exercise execution goals of each training stage, and combined with the current real-time oxygen supply parameters, a dynamic threshold system for stage goals is constructed to set standard threshold ranges for exercise physiological dynamic data and exercise execution data for each training stage.

[0045] The real-time data sequences of exercise physiological dynamic data and exercise execution data are standardized for deviation. The absolute difference between the exercise physiological dynamic data and exercise execution data and the standard threshold interval of the corresponding training stage is calculated. The deviation value is converted into a corrected deviation value according to the proportion of the absolute difference to the total span of the standard threshold interval, thus forming the deviation sequence of exercise physiological dynamic data and exercise execution data.

[0046] Based on the real-time data sequence of brain-computer interface data, the dynamic changing trends of the proportion of alpha wave energy and the proportion of beta wave energy are extracted, and the user's real-time fatigue and concentration values ​​are quantified.

[0047] Based on fatigue and focus values, a two-dimensional error tolerance adjustment is performed to dynamically correct the user's preset error tolerance and generate the user's real-time error tolerance; and the adaptation deviation between the real-time oxygen supply parameters and the current stage preset oxygen supply standard is calculated to generate an oxygen supply adaptation value.

[0048] Set anomaly detection threshold for each training stage, and adjust the anomaly detection threshold for each training stage based on the oxygen supply fit value and the user's real-time fault tolerance.

[0049] Specifically, the steps for performing AI anomaly detection, identifying data points that deviate from their target range, quantifying the degree of anomaly, classifying anomalies, and then generating AI wake-up prompts include:

[0050] The deviation sequence of dynamic exercise physiological data and exercise execution data is compared one by one with the anomaly judgment threshold after correction in the corresponding training stage. Deviation data points that exceed the anomaly judgment threshold are marked. The proportion of deviation data points within the preset detection period is calculated. If it is greater than the preset proportion threshold, it is determined that the current user has an execution abnormality.

[0051] The degree of anomaly is classified based on the degree of deviation of each deviation data point from the anomaly judgment threshold and the proportion of deviation data points within the detection period.

[0052] An AI algorithm combining decision tree and gradient boosting tree was used to extract anomaly features from real-time data that deviated from the time window of the data point and to classify the anomaly types. The anomaly features included: the deviation between the dynamic data of exercise physiology and the data of exercise execution, the real-time fault tolerance, and the oxygen supply adaptation value.

[0053] Establish a dynamic mapping library of anomaly types, anomaly severity, user status, and wake-up strategies. This library is used to match wake-up strategies based on the user's current anomaly type, anomaly severity, fatigue level, and focus level, and to generate AI wake-up prompts through multimodal interaction components.

[0054] Specifically, the steps for classifying and correcting a user's personalized behavioral baseline based on the type of anomaly include:

[0055] When an abnormality is detected in the user's execution, the user's personalized behavior baseline correction operation is triggered to construct an abnormal dataset of the current personalized oxygen-enriched rehabilitation exercise program execution.

[0056] Based on the preset association mapping rules between anomaly types and baseline parameters, the target correction dimension parameters that do not fit in the personalized behavior baseline are located.

[0057] Feature extraction is performed on various time series data in the abnormal dataset. Combined with the functional attributes of the target correction dimension parameters, feature indicators of the target correction dimension parameters are selected to construct a feature set for correction.

[0058] Based on the characteristic indicators of the target correction dimension parameters, the degree of adaptation deviation of the target correction dimension parameters is quantified by multi-dimensional feature weighted deviation calculation. This is used to determine the adjustment direction and adjustment magnitude of the target correction dimension parameters and to make targeted adjustments to the target correction dimension parameters.

[0059] The corrected target dimension parameters are subjected to safety boundary verification and correction. The corrected target dimension parameters are then integrated with the remaining parameters of the personalized behavior baseline to form the corrected personalized behavior baseline.

[0060] The real-time data-driven oxygen therapy exercise training and rehabilitation device consists of a rehabilitation intelligent monitoring and assessment platform, an exercise carrier, an oxygen concentrator, physiological monitoring components, and a brain-computer interface component.

[0061] The intelligent rehabilitation monitoring and assessment platform is used to build a personalized behavioral baseline for users, generate personalized oxygen-enriched rehabilitation exercise plans, and diagnose execution abnormalities to provide feedback and correct the personalized behavioral baseline; the exercise vehicle is used to provide an exercise execution environment adapted to the personalized oxygen-enriched rehabilitation exercise plan; the oxygen concentrator is used to dynamically output adapted oxygen supply parameters based on the personalized oxygen-enriched rehabilitation exercise plan and real-time monitoring data; the physiological monitoring component is used to collect dynamic physiological data of users during exercise; and the brain-computer interface component is used to collect EEG characteristic data related to fatigue and concentration.

[0062] The beneficial effects of this invention are:

[0063] This application uses multi-dimensional basic data and a standard parameter set formed by hierarchical clustering to gradually match and correct through dual-environment testing. This accurately obtains rehabilitation exercise-related parameters suitable for individual users. Combined with the characteristics of the target exercise vehicle, a multi-factor weighted decision is used to generate personalized training plans, ensuring that the plans not only conform to the individual's physiological response patterns and exercise tolerance, but also adapt to the operating characteristics of the exercise vehicle. Through real-time data acquisition, standardized processing, and dynamic threshold correction, AI anomaly detection is achieved, which can promptly and accurately identify execution anomalies, quantify their degree, classify them, and generate targeted intervention prompts. At the same time, based on the anomaly type, relevant parameters are dynamically corrected to construct a closed-loop optimization mechanism for the entire process. This effectively improves the accuracy, smoothness of execution, and targeted intervention of oxygen therapy exercise training, ensuring the safety of the training process and long-term rehabilitation effects. Attached Figure Description

[0064] Figure 1 This is a flowchart of the real-time data-driven oxygen therapy exercise training and rehabilitation method of the present invention.

[0065] Figure 2 This is a flowchart illustrating how the present invention obtains a standard behavioral baseline of the user;

[0066] Figure 3 A flowchart for obtaining a user's personalized behavioral baseline according to the present invention;

[0067] Figure 4 This is a flowchart of the deviation standardization process and anomaly detection threshold correction of the present invention;

[0068] Figure 5 This is a flowchart illustrating the AI ​​anomaly detection process used in this invention. Detailed Implementation

[0069] Please see Figure 1 This embodiment introduces a real-time data-driven oxygen therapy exercise training and rehabilitation method, including:

[0070] Step S1: Obtain the user's basic physiological information and health monitoring indicators, and perform multi-dimensional preprocessing operations. Specifically, this includes: firstly, verifying data validity through multi-source data cross-validation technology to eliminate abnormal data caused by detection equipment errors and data transmission interference; secondly, mapping exercise physiological parameters of different units to the [0,1] interval to eliminate dimensional differences and construct a structured data feature set. A standard behavioral baseline library is pre-established. This library is based on rehabilitation exercise data from a large-scale population of different ages and basic health states. Density clustering algorithms are used to perform hierarchical clustering by age segment, underlying disease type, and health monitoring indicator interval to form a set of multiple standard behavioral baselines. Each standard behavioral baseline is a multi-dimensional parameter set reflecting the rehabilitation exercise ability of the corresponding population, including core parameters such as exercise tolerance interval parameters, physiological response threshold interval parameters, and exercise adaptation interval parameters. The standard behavioral baseline is obtained by performing standard behavioral baseline matching through a rehabilitation intelligent monitoring and evaluation platform using dynamic threshold filtering and a stepwise intersection operation matching method based on multi-dimensional similarity fusion.

[0071] In this example, the standardized integration and precise matching of user basic data were achieved, eliminating the dimensional differences and redundant information of heterogeneous data. The constructed structured data feature set provided reliable data support for baseline matching. Furthermore, the standard behavioral baseline selected by the stepwise intersection matching method is highly consistent with the user's age, underlying disease type, and health test indicators, which can accurately reflect the basic characteristics of the rehabilitation exercise ability of the user's population. This lays a scientific and suitable reference foundation for the subsequent generation of personalized behavioral baselines.

[0072] Please see Figure 2 Preferably, the specific steps for obtaining the user's standard behavioral baseline include:

[0073] The system acquires users' basic physiological information and health monitoring indicators. Basic physiological information includes age, resting heart rate, and underlying disease types, all related to exercise physiological parameters. Health monitoring indicators include susceptibility gene data and telomere detection results. For example, for elderly rehabilitation users, the system focuses on collecting data on resting heart rate, blood pressure, arterial oxygen saturation, 6-minute walk test, continuous glucose monitoring, and renal function. For pediatric users, feeding status, respiratory rhythm, and transcutaneous oxygen saturation are additionally recorded. For users requiring nighttime monitoring, nighttime oxygen saturation monitoring data is supplemented. Simultaneously, the system uniformly records users' pulmonary signs, level of consciousness, pupillary response, and other clinical assessment information to provide comprehensive data support for subsequent baseline matching.

[0074] Multi-dimensional preprocessing is performed on basic physiological information and health test indicators. The basic physiological information and health test indicators that have passed the validity verification are subjected to format unification and numerical normalization processing. That is, the format is unified according to the pre-set data dictionary field format, and the min-max standardization formula is used to map the numerical data to the [0,1] interval to form a structured data feature set.

[0075] Based on age information from the structured data feature set and combined with the dynamic age segmentation rules of the standard behavioral baseline library, all standard behavioral baselines for the corresponding age segment are extracted from the standard behavioral baseline library to form the first baseline subset. Simultaneously, a population data distribution density verification method is used. If the number of baseline samples for that age segment in the standard behavioral baseline library is lower than a preset population data threshold, the baseline data of adjacent age segments are dynamically merged to ensure the representativeness of the samples in the first baseline subset. The preset population data threshold refers to the minimum number of baseline samples pre-set to ensure that the standard behavioral baseline samples for a certain age segment have statistical representativeness and matching reliability, based on the population data distribution characteristics, density clustering effectiveness requirements, and baseline matching accuracy targets of the standard behavioral baseline library. This is configured based on the population data distribution patterns, clustering effectiveness verification results, and target matching accuracy requirements of the standard behavioral baseline library.

[0076] Based on the basic disease type information in the structured data feature set, and referring to the ICD-11 disease classification mapping table, a standard behavioral baseline consistent with the basic disease type is extracted from the first baseline subset. A dynamic matching threshold adjustment mechanism is introduced. Based on the variance of the baseline sample distribution corresponding to the basic disease type, the matching tolerance is dynamically adjusted according to the degree of deviation of the variance from the preset variance benchmark, thereby balancing the sample coverage and matching accuracy of the second baseline subset. This ensures that the second baseline subset contains a sufficient number of representative baseline samples and accurately matches the rehabilitation exercise ability characteristics of the population corresponding to the basic disease type, thus forming the second baseline subset. The baseline set is the intersection of the first baseline subset and the corresponding baseline subsets for the underlying disease type in the standard behavioral baseline library. The preset variance benchmark is determined based on statistical analysis of rehabilitation exercise data of large-scale populations with different underlying disease types, reflecting the conventional difference level of rehabilitation exercise ability under that underlying disease type. When the variance of the baseline sample distribution is greater than the preset variance benchmark, it indicates that there is greater individual difference in rehabilitation exercise ability under that disease type, and the matching tolerance is increased accordingly to expand the sample coverage and avoid missing potential suitable baselines. When the variance of the baseline sample distribution is less than the preset variance benchmark, it indicates that the rehabilitation exercise ability characteristics of the population under that disease type are more concentrated, and the matching tolerance is decreased accordingly to narrow the screening range and improve the accuracy of baseline matching.

[0077] Based on health monitoring indicator information in the structured data feature set, an interval mapping matching method is used to extract standard behavioral baselines from the second baseline subset where the mapping overlap between the health monitoring indicator intervals and the user's health monitoring indicators is greater than a preset overlap threshold, forming a third baseline subset. This third baseline subset is the weighted intersection of the second baseline subset and the corresponding health monitoring indicator interval baseline subsets in the standard behavioral baseline library. The mapping overlap is calculated by mapping the normalized user health monitoring indicator values ​​in the structured data feature set to the interval boundaries of the corresponding health monitoring indicator intervals of each standard behavioral baseline in the second baseline subset. The overlap ratio for a single indicator is 1 when the indicator value is completely within the interval, and partially overlapping indicates a different ratio. The overlap is calculated as a proportion of the overlap length to the total span of the indicator interval, with a value of 0 when there is no overlap. This is then combined with the dynamic weight proportion of each health monitoring indicator in the structured data feature set, and the comprehensive overlap value is obtained through weighted summation. The preset overlap threshold is the minimum mapping overlap standard determined based on the importance of health monitoring indicators to rehabilitation exercise ability and the distribution characteristics of corresponding indicator intervals in the standard behavioral baseline library, used to screen effective matching baselines. The preset overlap threshold is dynamically set by combining the dynamic weight proportion of health monitoring indicators in the structured data feature set, the correlation strength between each health monitoring indicator interval in the standard behavioral baseline library and the rehabilitation exercise ability of the population, and through statistical analysis of the adaptation effectiveness data of each indicator interval.

[0078] Multi-dimensional similarity fusion calculations were performed on each standard behavioral baseline of the third baseline subset. First, an interval mapping weighted algorithm was used to assign mapping weights based on the mapping position of the user's health test indicators within the corresponding behavioral baseline indicator intervals. The closer the mapping position is to the center of the interval, the higher the weight, and the closer the mapping position is to the boundary of the interval, the lower the weight. Based on this mapping weight, the fit between the user's health test indicators and the corresponding intervals of each standard behavioral baseline was calculated. Second, the matching degree between the static heart rate, the underlying disease-related exercise physiological parameters, and the corresponding parameters of each standard behavioral baseline was calculated separately. The underlying disease-related exercise physiological parameters refer to exercise physiological parameters that are directly related to the pathogenesis, disease progression, and rehabilitation prognosis of the user's underlying disease based on the conclusions of clinical rehabilitation medicine research, and can objectively reflect the specific impact of the underlying disease on the individual's rehabilitation exercise tolerance and physiological response patterns. The fit of health test indicators, the matching degree of static heart rate, and the matching degree of underlying disease-related exercise physiological parameters were fused to obtain the comprehensive similarity of each standard behavioral baseline. Among them, the fit of health test indicators had the highest weight, while the weight of the matching degree of static heart rate and the matching degree of underlying disease-related exercise physiological parameters were evenly distributed.

[0079] If a standard behavioral baseline with a comprehensive similarity greater than the preset high fit standard exists in the third baseline subset, it is directly identified as the standard behavioral baseline matched with the user. This preset high fit standard is determined based on the core needs of rehabilitation exercise ability matching and combined with the effectiveness analysis of large-scale matching data. It belongs to a high fit standard that can directly meet the individual's basic fit needs. If it does not exist, abnormal standard behavioral baselines with a comprehensive similarity lower than the mean of the comprehensive similarity of the third baseline subset minus the standard deviation of the comprehensive similarity are removed from the third baseline subset. The comprehensive similarity of the remaining standard behavioral baselines is recalculated. The standard behavioral baseline with a comprehensive similarity greater than the preset fit threshold after recalculation is selected as the standard behavioral baseline matched with the user. This preset fit threshold is the minimum similarity standard to ensure that the baseline has basic fit. It is determined based on clinical rehabilitation data and population fit effectiveness verification.

[0080] Step S2: Based on the matched user's standard behavioral baseline, a graded testing rehabilitation exercise program is constructed using the rehabilitation intelligent monitoring and assessment platform. A dual-environment testing scenario is built using an exercise vehicle and an oxygen concentrator. By analyzing the differences in the user's response in non-oxygenated and oxygenated environments, the user's standard behavioral baseline is corrected, resulting in a personalized behavioral baseline. Specifically, the graded testing rehabilitation exercise program includes parameters for exercise intensity gradient, exercise duration allocation, and exercise rhythm variation. The exercise intensity gradient parameter divides the exercise into multiple continuous gradient intervals in ascending order. The exercise duration allocation parameter specifies the total duration of a single test exercise and the duration percentage of each gradient interval. The exercise rhythm variation parameter specifies the frequency variation pattern of movements within each gradient interval, ensuring that the testing exercise comprehensively covers all parameters of the standard behavioral baseline.

[0081] In this embodiment, step S2 ensures a comprehensive assessment of the user's rehabilitation exercise capabilities by constructing a graded testing and rehabilitation exercise program that covers the range of parameters of the standard behavioral baseline. Real-time heart rate, blood oxygen saturation, and blood pressure data are simultaneously collected in both oxygen- and non-oxygen-supply environments, providing complete physiological response data for both conditions. In-depth analysis of the numerical differences and trend similarities of the same parameters in oxygen- and non-oxygen-supply environments accurately extracts the physiological response differences and exercise capacity boundary characteristics of the user under different oxygen supply conditions, providing sufficient and individualized basis for revising the standard behavioral baseline. Targeted dynamic revision of the standard behavioral baseline based on these characteristics, eliminating incompatible parameter ranges and optimizing parameter weight allocation, results in a personalized behavioral baseline that accurately reflects the user's individual rehabilitation exercise capabilities and physiological response patterns, providing a highly suitable core reference for the subsequent generation of personalized oxygen-enriched rehabilitation exercise programs.

[0082] Preferably, the specific steps for constructing a graded testing and rehabilitation exercise program based on the matched user standard behavioral baseline include:

[0083] Extract core parameters from the user's standard behavioral baseline, including exercise tolerance interval parameters, physiological response threshold interval parameters, and exercise adaptation interval parameters, clarify the numerical range and distribution characteristics of each core parameter, and use this as the design benchmark for the graded testing and rehabilitation exercise program to ensure that the program can fully cover the boundaries and intermediate intervals of each parameter of the standard behavioral baseline.

[0084] Based on the exercise tolerance interval parameters of the standard behavioral baseline, multiple continuous gradient intervals are divided in order from low to high. The division of gradient intervals is based on the distribution density of exercise tolerance in the standard behavioral baseline. Intervals with higher distribution density correspond to finer gradients, while intervals with lower distribution density correspond to coarser gradients, ensuring that each gradient can accurately detect the user's exercise response at different intensity levels.

[0085] Based on the motion adaptation interval parameters of the standard behavioral baseline, the total duration of a single test motion is determined. The total duration must ensure that stable physiological response data can be obtained in each gradient interval. The duration is allocated according to the importance of each gradient interval, with the intensity gradient covering the core parameter interval of the standard behavioral baseline having the highest proportion, followed by the intensity gradient covering the edge parameter interval, to ensure the integrity of the test data in the key intervals.

[0086] Referring to the physiological response threshold range parameters of the standard behavioral baseline, the movement frequency change pattern is set for each gradient range. The gradient range corresponding to the core parameters adopts a stable movement frequency, while the gradient range corresponding to the edge parameters adopts a gradually increasing or decreasing movement frequency to simulate the rhythm changes in actual rehabilitation exercises, while avoiding abnormal physiological responses caused by sudden rhythm changes.

[0087] The designed gradient-based rehabilitation exercise program is compared with the parameters of the standard behavioral baseline to verify whether the program fully covers the parameter range of the standard behavioral baseline. Clinical rehabilitation exercise safety guidelines are introduced to verify whether each parameter in the program meets the requirements for safe exercise. If any parameters exceed the safe range or do not cover the baseline parameters, the exercise intensity gradient, duration allocation, or rhythm changes are adjusted to finally form a complete gradient-based rehabilitation exercise program.

[0088] Please see Figure 3 Preferably, the specific steps for obtaining a user's personalized behavioral baseline include:

[0089] A standardized testing environment was established. The non-oxygen-absorbing environment maintained normal atmospheric oxygen supply conditions, while the oxygen-absorbing environment achieved stable control of oxygen supply parameters through an oxygen concentrator. The oxygen concentration and pressure were kept constant according to preset standards, ensuring that the only difference between the non-oxygen-absorbing and oxygen-absorbing environments was the oxygen supply conditions, and eliminating other environmental interference factors.

[0090] A multimodal monitoring device integrating physiological monitoring components is used to collect data according to a synchronous acquisition mechanism based on motion phases. The graded rehabilitation exercise test is divided into four motion phases: warm-up, steady-state exercise, intensity escalation, and recovery. For each motion phase, the user's exercise physiological data is collected synchronously at the same time interval. Specifically, the exercise physiological data includes real-time heart rate data, blood oxygen saturation data, blood pressure data, exercise amplitude data, and exercise energy consumption data. During the acquisition process, the integrity and validity of the collected data are ensured by monitoring device signal strength, judging the continuity of data sequence, and screening for reasonable ranges of exercise physiological parameters, avoiding data distortion caused by device detachment or motion interference.

[0091] Outlier removal was performed on the exercise physiological data collected in both oxygen-inhaled and oxygen-inhaled environments. First, the 3σ criterion was used to initially screen out abnormal data that exceeded the statistical distribution range. Then, based on the movement characteristics and physiological patterns of the corresponding movement phase, isolated outliers caused by non-standard movements or sudden body tremors were removed.

[0092] Using the start of exercise as the zero point, the exercise physiological data of each exercise phase in both non-oxygenated and oxygenated environments are aligned with the same time step to ensure that the physiological data of the same exercise stage and the same time node correspond one-to-one. The missing data segments are supplemented using a phase feature interpolation algorithm. Based on the changing trend of adjacent data in the same exercise phase and the general law of the exercise physiological parameters in that exercise phase, supplementary data that fits the reality is generated, and a complete time series dataset of non-oxygenated and oxygenated environments is constructed.

[0093] Segmented temporal clustering and differential quantification analysis were used to extract response difference features from the temporal datasets of non-oxygen-inhaling and oxygen-inhaling environments. Specifically, the response difference features include static difference features, dynamic change features, and motion adaptation association features.

[0094] Static difference characteristics reflect the direct impact of changes in oxygen supply conditions on the user's basic physiological state, such as the decrease in peak heart rate and the increase in steady-state blood oxygen saturation under oxygen inhalation conditions; the mean difference, peak difference, and steady-state difference of the same exercise physiological parameter between non-oxygen inhalation and oxygen inhalation conditions are calculated according to the exercise phase and used as static difference characteristics.

[0095] Dynamic change characteristics reflect the speed and stability of the user's physiological response to changes in exercise intensity under different oxygen supply conditions, such as the difference in the rate of increase of heart rate with increasing exercise intensity and the difference in the amplitude of blood pressure fluctuations; dynamic change characteristics are obtained by calculating the temporal change rate, fluctuation coefficient, and peak occurrence time difference of exercise physiological parameters in non-oxygen-suction environment and oxygen-suction environment.

[0096] Exercise adaptation correlation features reflect the impact of oxygen supply conditions on users' exercise performance and intensity adaptation limits, such as the proportion of improvement in movement standardization and the extent of increase in endurance duration under the same exercise intensity in an oxygen-supplied environment. Based on exercise movement amplitude data and exercise energy consumption data, the movement standardization, energy consumption efficiency, and intensity endurance duration of users in non-oxygen-supplied and oxygen-supplied environments at each exercise intensity level are calculated as exercise adaptation correlation features.

[0097] Based on the characteristics of response differences, the user's standard behavioral baseline is specifically modified to ensure that the modified personalized behavioral baseline accurately matches the user's actual physiological response patterns, exercise intensity tolerance limits, and exercise execution adaptation characteristics under oxygen inhalation conditions. Specifically, this includes:

[0098] Using static difference features as the core input, the physiological response threshold interval parameters of the standard behavioral baseline are adjusted by the interval boundary dynamic migration algorithm. In the static difference features, the peak heart rate difference is the difference between the peak heart rate in the oxygen-inhaled environment and the peak heart rate in the non-oxygen-inhaled environment, and the peak blood pressure difference is the difference between the peak blood pressure in the oxygen-inhaled environment and the peak blood pressure in the non-oxygen-inhaled environment. When both the peak heart rate difference and the peak blood pressure difference are less than zero and reach the preset difference threshold, the upper limit of the original physiological response threshold interval is shifted downward according to the proportion of the absolute difference to the corresponding peak in the non-oxygen-inhaled environment. The steady-state blood oxygen saturation difference is the difference between the steady-state blood oxygen saturation value in the oxygen-inhaled environment and the steady-state blood oxygen saturation value in the non-oxygen-inhaled environment. When the difference is greater than the preset steady-state value threshold, the lower limit of the original physiological response threshold interval is shifted upward according to the proportion of the absolute difference to the steady-state blood oxygen saturation value in the non-oxygen-inhaled environment. At the same time, the time-series change rate data in the dynamic change features are extracted, and the division ratio of each sub-interval within the original physiological response threshold interval is optimized according to the distribution ratio of each time-series change rate in the whole motion phase. The original parameter sub-intervals whose sub-interval boundaries do not match the time-series change rate distribution features are removed to obtain the corrected physiological response threshold interval parameters. The distribution proportion is obtained by statistically analyzing the temporal change rate of each exercise physiological parameter in the dynamic change characteristics, calculating its frequency or contribution value in the four exercise phases of warm-up, steady-state exercise, intensity escalation, and recovery, and then dividing the cumulative value of a single temporal change rate across all phases by the total cumulative value of all temporal change rates across all phases to obtain the proportion of a single temporal change rate in the entire exercise phase; the preset difference threshold is based on evidence-based data on exercise safety in clinical rehabilitation, the peak heart rate and blood pressure distribution characteristics of the corresponding population in the standard behavioral baseline database, and the regulatory patterns of the oxygen inhalation environment on cardiovascular physiological indicators, and is preset accordingly. The minimum difference quantification standard for determining whether the relevant peak difference value triggers the downward migration of the upper limit of the physiological response threshold interval is based on evidence-based data on blood oxygen safety during clinical rehabilitation exercise, the distribution characteristics of blood oxygen saturation steady-state values ​​of the corresponding population in the baseline database, and the law of blood oxygen saturation improvement by the oxygen inhalation environment. Both are configured based on the variance of the target parameter distribution of the corresponding population in the baseline database, combined with the safe control and effective improvement range of relevant indicators under the clinical oxygen inhalation environment.

[0099] Using dynamic change features and exercise adaptation correlation features as joint inputs, the exercise tolerance range is corrected through a tolerance limit mapping algorithm. In the exercise adaptation correlation features, the intensity tolerance duration difference is the difference between the intensity tolerance duration in the oxygen-inhaled environment and the intensity tolerance duration in the non-oxygen-inhaled environment; the energy consumption efficiency difference is the difference between the energy consumption efficiency in the oxygen-inhaled environment and the energy consumption efficiency in the non-oxygen-inhaled environment. When the intensity tolerance duration difference is greater than a preset duration threshold and the energy consumption efficiency difference is greater than a preset efficiency threshold, the corresponding intensity level is included in the core tolerance range, and the weight of this range is increased according to the proportion of the intensity tolerance duration difference to the intensity tolerance duration in the non-oxygen-inhaled environment. In the static difference features, the criterion for abnormally elevated exercise physiological parameters is that the peak value of the corresponding exercise physiological parameter in the oxygen-inhaled environment exceeds the sum of the peak value of the corresponding exercise physiological parameter in the non-oxygen-inhaled environment and the preset elevation threshold. Intensity levels that meet this criterion are removed from the exercise tolerance range, or their weight is reduced by a preset weight reduction ratio. Based on the above adjustments, the upper and lower limits of the corrected exercise tolerance range and the core tolerance range are determined, resulting in the corrected exercise tolerance range parameters. Among them, the preset increase threshold is a minimum superimposed quantitative standard for judging whether the peak value of the parameter is abnormally increased under oxygen inhalation environment, based on evidence-based data on the safety of clinical rehabilitation exercise, the peak distribution characteristics of exercise physiological parameters of the corresponding population in the baseline database, and the normal fluctuation range of physiological indicators in the oxygen inhalation environment. The preset weight reduction ratio is a fixed or intervalized ratio standard for reducing the weight of abnormal intensity levels, based on the priority of clinical rehabilitation exercise safety, the negative impact coefficient of abnormal intensity levels in the baseline database, and the exercise tolerance interval adaptation rules.

[0100] Based on motion standardization data and intensity tolerance duration data from the motion fit correlation features, combined with fluctuation coefficient data from the dynamic change features, a dynamic weight allocation algorithm for fit is used to correct the motion fit interval. The differences in motion standardization and intensity tolerance duration for each intensity level are calculated between non-oxygenated and oxygenated environments. When both the difference in motion standardization and intensity tolerance duration exceed a preset threshold, the weight of the corresponding intensity level in the motion fit interval is increased according to the combined weight of these two differences. The combined weight is calculated by weighted summation after normalizing the differences in motion standardization and intensity tolerance duration for a certain intensity level between non-oxygenated and oxygenated environments, and then combining their dynamic weights in the motion fit assessment. The proportion of each individual difference to the total weighted sum is used to quantify the contribution of each difference to the increase in the weight ratio of the intensity level. When the motion standardization of a certain intensity level is lower than the preset minimum standardization threshold or the energy consumption efficiency is lower than the preset minimum efficiency threshold, the parameter range of the intensity level is reduced by a preset range reduction ratio, or its weight is reduced by a preset weight reduction ratio. The fluctuation coefficient data in the dynamic change characteristics are extracted, and the intensity levels with fluctuation coefficients less than the preset fluctuation threshold are marked as the core adaptation interval. The parameter range and weight ratio of the core adaptation interval are defined to obtain the corrected motion adaptation interval parameters. Among them, the preset range reduction ratio refers to a fixed ratio standard that is pre-set to reduce the parameter range of intensity levels that do not meet the standards for movement standardization or energy consumption efficiency, based on the priority requirements for safety and effectiveness of clinical rehabilitation exercise, the distribution characteristics of parameters of each intensity level in the standard behavioral baseline library, and the influence weight of movement standardization and energy consumption efficiency on exercise adaptability; the preset weight reduction ratio refers to a fixed ratio standard that is pre-set to reduce the weight of intensity levels that do not meet the standards for movement standardization or energy consumption efficiency, based on the priority of safety of clinical rehabilitation exercise, the negative impact coefficient of abnormal intensity levels on rehabilitation effect in the standard behavioral baseline library, and the parameter adaptation rules of the exercise adaptation interval.

[0101] By integrating the revised physiological response threshold range parameters, exercise tolerance range parameters, and exercise adaptation range parameters, a personalized behavioral baseline that accurately reflects an individual's rehabilitation exercise ability, physiological response patterns, and exercise adaptation characteristics is obtained.

[0102] Step S3: Using a rehabilitation intelligent monitoring and assessment platform, based on a personalized behavioral baseline and combined with the exercise mode parameters of the target exercise vehicle selected by the user, a personalized oxygen-enriched rehabilitation exercise plan is generated using a multi-factor weighted decision-making method. The parameters of the personalized behavioral baseline include exercise tolerance range parameters, physiological response threshold range parameters, and exercise adaptation range parameters. The exercise mode parameters of the target exercise vehicle include the movement execution method parameters, exercise intensity adjustment range parameters, and controllable exercise rhythm range parameters. First, weight coefficients are set for each parameter of the personalized behavioral baseline, determined based on the safety and effectiveness priorities of rehabilitation training. Then, based on the numerical range of each parameter and its weight coefficients, the optimal combination of the following parameters is calculated: the dynamic adjustment curve parameter of exercise intensity, the exercise cycle division parameter, the oxygen-enriched supply mode parameter, and the rest interval setting parameter. The dynamic adjustment curve parameter of exercise intensity is calculated based on the exercise tolerance range parameter and the exercise intensity adjustment range parameter of the target exercise vehicle; the exercise cycle division parameter is determined by combining the exercise tolerance range parameter with the exercise duration adaptation rule; the oxygen-enriched supply mode parameter is determined based on the physiological response threshold range parameter; and the rest interval setting parameter is calculated based on the exercise adaptation range parameter and the post-exercise recovery rule.

[0103] In this embodiment, step S3 uses a personalized behavioral baseline that accurately reflects an individual's rehabilitation ability as the core basis, combined with the specific movement mode parameters of the target exercise vehicle selected by the user, to generate a personalized oxygen-enriched rehabilitation exercise plan using a multi-factor weighted decision-making method. This ensures from the source that the plan not only conforms to the individual's physiological response patterns and exercise tolerance limits, but also adapts to the movement execution characteristics and intensity adjustment range of the target exercise vehicle. By setting weight coefficients for each parameter based on the priority of rehabilitation training safety and effectiveness, priority is given to ensuring exercise tolerance and physiological response safety, while also taking into account exercise adaptability and exercise vehicle characteristics, effectively balancing safety and effectiveness goals in the rehabilitation process. The optimal combination of the dynamic adjustment curve of exercise intensity, exercise cycle division, oxygen enrichment mode, and rest interval settings calculated based on the weight coefficients and the numerical range of each parameter ensures that each execution parameter in the plan matches the individual's rehabilitation needs in an oxygen-inhaled environment, avoiding problems of low rehabilitation efficiency caused by inappropriate intensity, unreasonable cycle, or unsuitable oxygen supply. The final personalized oxygen-enriched rehabilitation exercise plan has high individual adaptability, execution safety, and rehabilitation effectiveness, providing a directly implementable basis for subsequent standardized and precise oxygen-enriched rehabilitation training.

[0104] Preferably, the specific steps for generating a personalized oxygen-enriched rehabilitation exercise program using a multi-factor weighted decision-making method include:

[0105] Personalized behavioral baseline parameters are obtained, including exercise tolerance range parameters, physiological response threshold range parameters, and exercise adaptation range parameters. The exercise tolerance range parameters include the upper limit, lower limit, and core tolerance intensity levels that an individual can tolerate. The physiological response threshold range parameters include the safe threshold ranges for heart rate, blood oxygen saturation, and blood pressure during exercise. The exercise adaptation range parameters include the exercise intensity range where the individual's movement standardization is met and the intensity range where energy consumption efficiency meets rehabilitation needs. The exercise mode parameters of the target exercise vehicle are also obtained, including movement execution method parameters, exercise intensity adjustment range parameters, and controllable exercise rhythm range parameters. The movement execution method parameters include the types of movements that the exercise vehicle can perform and the range of movement amplitude adjustment. The exercise intensity adjustment range parameters include the minimum and maximum exercise intensity values ​​that the exercise vehicle can output, and the intensity adjustment step size. The controllable exercise rhythm range parameters include the frequency range of movements supported by the exercise vehicle and the time interval between rhythm changes.

[0106] Based on the principle that safety is the priority over effectiveness in rehabilitation training, the weighting levels of personalized behavioral baseline parameters are divided. The first weighting level is the exercise tolerance range parameter, with the core objective of avoiding sports injuries and maintaining individual physiological stability. The second weighting level is the physiological response threshold range parameter and the exercise intensity adjustment range parameter of the target exercise vehicle, with the core objective of balancing rehabilitation effects and the intensity adaptability of the exercise vehicle. The third weighting level is the exercise adaptability range parameter, the action execution mode parameter of the target exercise vehicle, and the controllable range parameter of exercise rhythm, with the core objective of optimizing the smoothness of exercise execution and individual adaptability.

[0107] The weights of each personalized behavioral baseline parameter are determined using a hierarchical importance allocation method. The total weight of the first weight level is set according to the principle of ensuring safety, the total weight of the second weight level is set according to the principle of balancing effect and adaptation, and the total weight of the third weight level is set according to the principle of optimizing execution details. Within the same level, the weights of individual parameters are evenly allocated according to the degree of influence of the parameter on the corresponding goal, forming a quantifiable weight coefficient matrix to ensure that the weight allocation is directly related to the rehabilitation goal and has no logical contradiction.

[0108] By combining the user's personalized behavioral baseline parameters with the movement pattern parameters of the target exercise vehicle, physiological state goals and exercise execution goals are set for each training phase. The training phase includes a warm-up phase, a main training phase, and a recovery phase. Specifically, the goal of the warm-up phase is to activate the individual's basic physiological state, such as raising the heart rate to the lower limit of the physiological response threshold range and maintaining blood oxygen saturation within a safe range, while adapting to the low-intensity movement requirements of the exercise vehicle. The goal of the main training phase is to maintain the individual's physiological state within the core safe range, such as stabilizing the heart rate and blood pressure within the core segment of the physiological response threshold range, while covering the intensity range adapted to the exercise vehicle, thus achieving a rehabilitation effect. The goal of the recovery phase is to guide the individual's physiological state back to the pre-training safe range and adapt to the low-intensity recovery movements of the exercise vehicle.

[0109] Based on the physiological state goals and exercise execution goals of each training phase, the parameters for dynamic adjustment curves of exercise intensity, exercise cycle division, oxygen supply mode, and rest interval settings for each training phase are obtained through phase adaptability parameter calculation methods; specifically including:

[0110] The parameter intersection mapping method is adopted, taking the exercise tolerance interval parameters of the first weight level and the exercise intensity adjustment range parameters of the target exercise vehicle of the second weight level as inputs. Combining the physiological state goals and exercise execution goals of the warm-up stage, main training stage and recovery stage, the intersection of the intensity range of the corresponding stage of the exercise tolerance interval parameters and the intensity range of the corresponding stage of the exercise vehicle intensity adjustment range parameters is taken to form the basic intensity level of each training stage. Then, according to the preset weight ratio, the core tolerance intensity level of the exercise tolerance interval is mapped to the core segment of the basic intensity level of each stage, and the intermediate intensity level of the exercise vehicle intensity adjustment range is mapped to the transition segment of the basic intensity level of each stage. Finally, according to the goal requirements of each stage, the acceleration rate or deceleration rate of the intensity within the basic intensity level of each stage is set to ensure that the intensity change conforms to the individual physiological adaptation law and the operating characteristics of the exercise vehicle. Finally, the parameters of the continuous dynamic adjustment curve of exercise intensity covering the three training stages are integrated to form a parameter.

[0111] Using the functional phase matching method, combining the exercise tolerance range parameters of the first weight level, the movement mode parameters of the target exercise vehicle, and the physiological state and exercise execution goals of the training phase, an overall training cycle framework of warm-up, main training, and recovery phases is divided according to the principle of adapting to individual tolerance limits and vehicle performance capabilities. Then, for each phase, the duration of the exercise cycle is set according to its corresponding intensity range and target requirements. For example, the duration of the warm-up phase should meet the needs of gradual physiological activation and not exceed the adaptation duration of the low-intensity segment of exercise tolerance. The duration of the main training phase should ensure that the core intensity segment continuously accounts for no less than 60% of the total duration of the phase and does not exceed the smaller value between the longest stable duration of the core tolerance intensity and the longest running duration of the intermediate intensity segment of the vehicle. The duration of the recovery phase should meet the needs of smooth physiological decline and adapt to the recovery adaptation duration of the lowest intensity segment of exercise tolerance. Finally, a complete set of exercise cycle division parameters is formed.

[0112] A threshold-triggered adaptation method is adopted, using the physiological response threshold range parameters of the second weight level as the core basis, and combining the physiological state goals of the three training stages to set differentiated oxygen-enriched supply rules. Among them, the warm-up and recovery stages mainly use the basic oxygen supply mode, and low-concentration oxygen-enriched supply is triggered only when the real-time monitored blood oxygen saturation is lower than the preset proportion of the lower limit of the physiological response threshold range. The main training stage sets the trigger conditions as the real-time heart rate is higher than the upper limit of the core segment of the physiological response threshold range by a preset proportion or the real-time blood oxygen saturation is lower than the lower limit of the core segment of the physiological response threshold range by a preset proportion. The supply concentration is set according to the goal of maintaining the individual's blood oxygen saturation within the core segment of the physiological response threshold range, and the supply pressure is set according to the principle of adapting to the current output intensity of the target exercise vehicle. At the same time, the stopping conditions of oxygen-enriched supply are defined, and finally, the parameters of the oxygen-enriched supply mode covering the three training stages are integrated.

[0113] The intensity-recovery correlation method was adopted. Based on the exercise adaptation interval parameters of the third weight level, the post-exercise recovery pattern, and the exercise execution goals and intensity characteristics of the three training stages, rest nodes were divided according to the principle that the higher the intensity level of the exercise intensity dynamic adjustment curve, the greater the density of rest nodes and the longer the duration of a single rest. Among them, the warm-up stage, due to its low intensity and the goal of activating the physiological state, does not have a separate rest node, but only reserves a short transition interval after the stage to connect with the main training stage. The main training stage is divided into multiple rest nodes according to the intensity level, and the duration of a single rest is set according to the goal of allowing the individual's physiological state to fall back to the exercise adaptation interval, and basal oxygen supply is maintained during the rest period. The recovery stage only has one short rest node in the middle, and the duration is set according to the principle of not interrupting the physiological decline rhythm and assisting the physiological state to recover smoothly. Basal oxygen supply is also maintained during the rest period. Finally, the rest interval setting parameters adapted to the three training stages were formed.

[0114] Following the sequential logic of warm-up, main training, and recovery phases, the parameters of the dynamic adjustment curve of exercise intensity, the division parameters of exercise cycle, the parameters of oxygen supply mode, and the setting parameters of rest interval are embedded into the corresponding phases one by one to ensure that the parameters of each phase are connected and there are no temporal conflicts, thus forming a personalized oxygen-enriched rehabilitation exercise plan.

[0115] Step S4: When the user triggers the personalized oxygen-enriched rehabilitation exercise program, the rehabilitation intelligent monitoring and evaluation platform, in conjunction with the physiological monitoring component, brain-computer interface component, exercise vehicle, and oxygen concentrator, continuously and synchronously collects dynamic exercise physiological data, brain-computer interface data, exercise execution data, and real-time oxygen supply parameters during the training process. This data is used to identify abnormal deviations of the user through AI anomaly detection and generate AI wake-up prompts based on the type of abnormal deviation. The dynamic exercise physiological data includes real-time heart rate data, real-time blood oxygen saturation data, and real-time blood pressure data, collected by the physiological monitoring component. The brain-computer interface data consists of real-time EEG characteristic data related to fatigue and concentration, collected by the brain-computer interface component through a lightweight headband or ear-hook EEG acquisition device, using contact electrodes to capture EEG signals on the scalp surface in real time. Targeted acquisition frequency bands are set for alpha waves related to fatigue and beta waves related to concentration, and the target EEG signals are separated by the built-in frequency band filtering module of the device. The exercise execution data includes actual exercise intensity data, actual exercise rhythm data, and actual completion progress data, collected by the exercise vehicle. Real-time oxygen supply parameters, including oxygen concentration, oxygen pressure, and oxygen flow rate, are collected by the oxygen concentrator. The rehabilitation intelligent monitoring and evaluation platform uses AI anomaly detection, which combines statistical analysis of time-series data with feature matching, to perform real-time data analysis. AI anomaly detection is used to determine if there are any abnormalities such as physiological indicators exceeding safe ranges or excessive deviations in exercise execution. A tiered early warning response mechanism is established. When an anomaly is detected, an AI wake-up prompt is generated based on the type of deviation, pushing targeted prompts to the user, such as adjusting exercise intensity and correcting exercise rhythm. If the anomaly level reaches a preset warning level, a reminder containing the anomaly type, corresponding training stage, and data segment is simultaneously pushed to the medical staff.

[0116] In this embodiment, relying on the AI ​​anomaly detection based on time-series statistics and feature matching of the rehabilitation intelligent monitoring and assessment platform, it can accurately identify risk situations such as physiological indicators exceeding limits, abnormal fluctuations in EEG characteristics, and excessive deviations in motor execution in real time, avoiding the lag and missed judgment problems of traditional manual monitoring; combined with the hierarchical early warning response mechanism, it can guide users to adjust their exercise status in real time through AI wake-up prompts to ensure that the training fits the requirements of the plan, and simultaneously link with medical staff in case of serious abnormalities, thus building a dual safety protection system.

[0117] Preferably, the specific steps for identifying abnormal deviations of the user through AI anomaly detection and generating AI wake-up prompts based on the type of abnormal deviation include:

[0118] Please see Figure 4 When a user triggers the execution of a personalized oxygen-enriched rehabilitation exercise program, the rehabilitation intelligent monitoring and evaluation platform simultaneously receives real-time data collected by the physiological monitoring component, brain-computer interface component, exercise vehicle, and oxygen concentrator, including: dynamic exercise physiological data, brain-computer interface data, exercise execution data, and real-time oxygen supply parameters; it adds a unified timestamp to the real-time data, divides the data into segments according to the training stage, and constructs a real-time data sequence that corresponds one-to-one with each training stage to ensure the continuity of data time sequence and the relevance of the scenario.

[0119] Based on the physiological state goals and exercise execution goals of each training stage in the personalized oxygen-enriched rehabilitation exercise program, and combined with the current real-time oxygen supply parameters, a dynamic threshold system for stage goals is constructed, and standard threshold ranges for exercise physiological dynamic data and exercise execution data for each training stage are set. Specifically, for exercise physiological dynamic data, the standard threshold ranges for exercise physiological dynamic data in each training stage are dynamically adjusted according to the current oxygen supply concentration, based on the physiological response threshold range in the personalized behavioral baseline. For example, the upper limit of the safe heart rate in the main training stage is adaptively increased in an oxygen-enriched environment. For exercise execution data, the standard threshold ranges for exercise execution data are set based on the preset exercise intensity, rhythm, and progress parameters of each stage, combined with the operating characteristics of the target exercise vehicle.

[0120] The real-time data sequences of exercise physiological dynamic data and exercise execution data are standardized for deviation. The absolute deviation of each individual data point in the exercise physiological dynamic data and exercise execution data from the corresponding standard threshold interval of the training stage is calculated: if an individual data point is within the standard threshold interval, the absolute deviation is 0; if an individual data point is below the lower limit of the interval, the absolute deviation is the difference between the lower limit of the interval and the data point; if an individual data point is above the upper limit of the interval, the absolute deviation is the difference between the data point and the upper limit of the interval. Based on the proportion of this absolute difference to the total span of the standard threshold interval, a corrected deviation value in the [0,1] interval is generated, forming a deviation sequence of exercise physiological dynamic data and exercise execution data. The closer the corrected deviation value is to 1, the more severe the deviation from the target stage; the closer it is to 0, the closer it is to the target requirement. Specifically, when the real-time fault tolerance is increased, the corrected deviation value corresponding to the same original data decreases synchronously to avoid over-judging abnormalities in users under high load; when the real-time fault tolerance is decreased, the corrected deviation value corresponding to the same original data increases synchronously to ensure timely capture of subtle deviations in low-focus states.

[0121] For the real-time data sequence of brain-computer interface component data, the dynamic change trend of alpha wave energy ratio and beta wave energy ratio is extracted from the EEG signal through the temporal feature extraction algorithm, and the user's real-time fatigue value and concentration value are quantified. The value range is [0,1]. The closer the fatigue value is to 1, the higher the fatigue level, and the closer the concentration value is to 1, the better the concentration state.

[0122] Based on fatigue and focus values, a two-dimensional error tolerance adjustment is implemented to dynamically correct the user's preset error tolerance and generate the user's real-time error tolerance. For example, based on the statistical average of rehabilitation exercise error tolerance for people of corresponding age groups and underlying disease types in the standard behavioral baseline library, a preset error tolerance benchmark value is established. If the fatigue value is higher than the preset fatigue benchmark threshold, the error tolerance is increased proportionally to the fatigue value exceeding the benchmark: the greater the fatigue value exceeds the fatigue benchmark threshold, the greater the increase in real-time error tolerance, to avoid over-intervention for users under high load conditions. If the focus value is lower than the preset focus benchmark threshold, the error tolerance is decreased proportionally to the focus value being lower than the benchmark: the greater the focus value is lower than the focus benchmark threshold, the greater the decrease in real-time error tolerance, to ensure timely correction of execution deviations caused by inattention. If the fatigue value is higher than the fatigue benchmark threshold and the focus value is lower than the focus benchmark threshold, the lower value of the corresponding error tolerance is taken according to the principle of safety priority, balancing the timeliness of intervention and the adaptability to physical load. If both the fatigue value and the focus value are within the corresponding benchmark threshold range, the real-time error tolerance remains at the error tolerance benchmark value.

[0123] The deviation between the real-time oxygen supply parameters collected by the oxygen concentrator and the current stage preset oxygen supply standard is calculated to generate an oxygen supply fit value, which ranges from [0,1]. The closer the value is to 1, the better the oxygen supply parameters fit the stage target.

[0124] First, based on personalized behavioral baselines and clinical safety guidelines, abnormal judgment thresholds are set for each training stage according to the functional positioning of the main training, warm-up, and recovery stages. Then, the abnormal judgment thresholds for each training stage are corrected according to the oxygen supply fit value and the user's real-time fault tolerance. That is, the thresholds are dynamically adjusted according to the oxygen supply fit value. The lower the fit, the lower the threshold is adjusted to improve the sensitivity of abnormal judgment, and the higher the fit, the higher the threshold is adjusted to reduce the false judgment rate of normal fluctuations. Combined with real-time fault tolerance, the threshold is further calibrated. The threshold is raised to relax the standard in the high fatigue state and lowered to tighten the standard in the low focus state. At the same time, boundary constraints are used to ensure that the thresholds are within a reasonable range, so as to achieve accurate matching between the judgment standard and the oxygen-rich scenario and the user's state.

[0125] Please see Figure 5The system compares the deviation sequences of dynamic exercise physiological data and exercise execution data with the corresponding anomaly detection thresholds after correction for each training stage, marking data points that deviate from the thresholds. It also calculates the percentage of deviating data points within a preset detection period; if this percentage exceeds a preset threshold, the user is deemed to have an execution anomaly. Based on the degree of deviation of each data point from the anomaly detection threshold and the percentage of deviating data points within the detection period, the degree of anomaly is classified. For example, a slight anomaly is defined as a data point with a deviation percentage less than 1.2 times the preset threshold, and a single data point exceeding the threshold by a slightly elevated margin; a moderate anomaly is defined as a data point with a deviation percentage between 1.2 and 1.5 times the preset threshold, and most data points exceeding the threshold by a moderate margin; and a severe anomaly is defined as a data point with a deviation percentage greater than 1.5 times the preset threshold. The deviation is classified as severe if it is multiple times higher than normal and multiple deviation data points have a deviation value close to 1. The preset detection period refers to the fixed number of data collection periods set based on the rhythm characteristics and data collection frequency of each training stage to capture continuous abnormal trends. Its length is adapted to the change pattern of exercise intensity in the corresponding training stage, ensuring a comprehensive reflection of the deviation state during a continuous training process, avoiding misjudgment due to instantaneous fluctuations caused by too short a period or delay in abnormality due to too long a period. The preset percentage threshold refers to the minimum percentage standard of deviation data points set based on the reliability requirements of abnormality judgment in clinical oxygen-enriched rehabilitation training and the normal fluctuation range in personalized behavioral baselines, to distinguish between occasional deviations and persistent abnormalities. Its value balances the timeliness of abnormality capture with the risk of misjudgment, ensuring that only when the deviation state exhibits persistent characteristics within the preset detection period is it judged as abnormal, excluding interference from single or a few occasional deviation data points.

[0126] An AI algorithm fusion of decision tree and gradient boosting tree is employed to extract anomaly features from real-time data within the time window of the data point that deviates from its position, and to classify anomaly types. Anomaly features include: deviation between exercise physiological dynamic data and exercise execution data, real-time fault tolerance, and oxygen supply adaptation value. The time window refers to a fixed data segment range centered on the deviation data point and covering consecutive data collection cycles before and after it, adapted to the preset detection cycle of each training phase. Its length is sufficient to completely encompass the contextual data generated by the deviation state, ensuring the comprehensiveness and relevance of anomaly feature extraction and avoiding classification bias caused by a single data point or local segment. For example, deviations in exercise physiological dynamic data exceeding the corrected anomaly judgment threshold and deviations in exercise execution data... If the intensity is within the normal range, the oxygen supply adaptation value is low, and the real-time error tolerance remains at the baseline value, it is classified as an oxygen deficiency type physiological abnormality. If the deviation of exercise execution data is greater than the corrected abnormality judgment threshold, the deviation of exercise physiological dynamic data is within the normal range, the oxygen supply adaptation value is high, and the real-time error tolerance is lower than the baseline value, it is classified as an inattentive type execution abnormality. If the deviation of both exercise physiological dynamic data and exercise execution data is greater than the corrected abnormality judgment threshold, the oxygen supply adaptation value is within the normal range, and the real-time error tolerance is higher than the baseline value, it is classified as a fatigue accumulation type coordination abnormality. If both exercise physiological dynamic data and exercise execution data are within the normal range, the oxygen supply adaptation value is significantly lower than the stage target, and the real-time error tolerance remains at the baseline value, it is classified as an oxygen supply parameter deviation type abnormality.

[0127] A dynamic mapping library is established, linking anomaly type, severity, user status, and wake-up strategy. This library matches wake-up strategies based on the user's current anomaly type, severity, fatigue level, and focus level. AI wake-up prompts are generated through multimodal interaction components, synchronously linking with the oxygen concentrator for coordinated control, ensuring precise adaptation of intervention logic to the oxygen-enriched rehabilitation scenario and user status. For minor anomalies, lightweight guidance prompts are pushed, such as voice guidance to adjust breathing rhythm and screen status reminders, without interfering with core exercise and oxygen supply parameters, only assisting the user in adapting to training requirements. For moderate anomalies, specific intervention logic is matched based on the anomaly type. For example, for physiological anomalies due to insufficient oxygen supply, dynamic fine-tuning of oxygen concentrator parameters is triggered simultaneously, with voice notification of the adjustment results; for execution anomalies due to insufficient focus, vibration rhythm guidance from the exercise carrier and focus-enhancing prompts are used; for coordinating anomalies due to accumulated fatigue, suggestions to reduce exercise intensity and short rest prompts are pushed, achieving dual intervention of parameter optimization and behavioral guidance. For severe anomalies, an emergency training pause prompt and audible / visual alarm are immediately pushed, simultaneously sending attention reminders to medical staff including the anomaly type, corresponding training stage, data segment, user status, and oxygen supply adaptation status, maximizing training safety. Among them, the awakening strategy refers to a personalized intervention plan that integrates multimodal prompts and equipment collaborative control logic, with safety intervention and precise adaptation as the core, based on different abnormality types, abnormality levels and real-time status of users in oxygen-enriched rehabilitation training. It is based on clinical oxygen-enriched rehabilitation safety guidelines, training scenario characteristics and user personalized behavioral baselines, classifies abnormal scenarios, matches corresponding intervention intensities and methods, and integrates multimodal interactive resources and oxygen supply and exercise equipment control rules.

[0128] Step S5: When the rehabilitation intelligent monitoring and assessment platform determines that the user has an execution abnormality, it indicates that the physiological response threshold range, exercise tolerance range, or exercise adaptation range parameters in the current personalized behavioral baseline deviate from the actual adaptation needs of the user's current physical state and the oxygen-enriched training scenario. This leads to physiological indicators exceeding limits, exercise execution deviations, or oxygen supply imbalances during training execution. Targeted corrections are needed to achieve accurate matching between the baseline and the individual's dynamic state. Its core function is: based on the abnormality type, abnormality degree, and corresponding real-time data obtained in step S4, to construct an abnormal baseline deviation mapping relationship, classify and locate the baseline mismatch dimension according to the abnormality type, classify and correct the user's personalized behavioral baseline, and ensure that the subsequently generated personalized oxygen-enriched rehabilitation exercise plan continues to fit the user's actual state.

[0129] In this embodiment, the intelligent rehabilitation monitoring and assessment platform first establishes a precise correlation between abnormal data and personalized behavioral baseline parameters. Then, based on this correlation, it classifies and corrects the mismatched parameter dimensions in the baseline, realizing real-time adaptation of the personalized behavioral baseline to the user's dynamic physical state and oxygen-rich training scenario. This effectively eliminates the deviation between the baseline and actual needs, ensuring that subsequent plans are more in line with the individual's tolerance limits and execution capabilities. This not only ensures training safety but also improves the accuracy and effectiveness of rehabilitation, perfecting the closed-loop optimization mechanism of the entire process.

[0130] Preferably, the specific steps for classifying and correcting the user's personalized behavioral baseline include:

[0131] When an abnormality is detected in the user's performance, the user's personalized behavior baseline correction operation is triggered, and an abnormal dataset is constructed during the execution of the current personalized oxygen-enriched rehabilitation exercise program. This dataset includes the deviation sequence corresponding to the abnormality type, the duration of the abnormality, the fatigue and concentration fluctuation curves, and the time series data of the oxygen supply adaptation value.

[0132] Based on pre-defined association mapping rules between abnormality types and baseline parameters, the target correction dimension parameters that are not compatible in the personalized behavioral baseline are located, and the baseline correction direction corresponding to the abnormalities in the personalized behavioral baseline is determined. The association mapping rules refer to a fixed correspondence system between abnormality types and core parameters of the personalized behavioral baseline, established in advance based on clinical oxygen-enriched rehabilitation safety guidelines, the functional attributes of each parameter in the personalized behavioral baseline, and the physiological execution correlation mechanism of abnormality occurrence. This system clarifies the specific parameter dimensions in the physiological response threshold range, exercise tolerance range, and exercise adaptation range that need to be corrected for different abnormality types, providing precise directional guidance for baseline correction.

[0133] Feature extraction is performed on various time series data in the abnormal dataset to obtain the statistical characteristics of the deviation sequence, the trend characteristics of the fluctuation curve, the steady-state characteristics of the time series data, and the dynamic characteristics of the occurrence and development of anomalies. Combined with the functional attributes of the target correction dimension parameter, feature indicators directly related to the adaptability of the parameter are selected, a correction basis feature set is constructed, and the correlation logic between the feature indicators and the target correction dimension parameter is clarified to provide quantitative support for parameter correction.

[0134] Based on the characteristic indicators of the target correction dimension parameters, the degree of adaptation deviation of the target correction dimension parameters is quantified through multi-dimensional feature weighted deviation calculation. Specifically, based on the correlation importance between the characteristic indicators and the target correction dimension parameters, corresponding weights are assigned to each characteristic indicator. The local adaptation deviation reflected by each characteristic indicator is weighted and fused to obtain a comprehensive adaptation deviation quantification value for the target correction dimension parameters. This quantification value includes the deviation direction and magnitude; a positive deviation indicates that the parameter is higher than the actual adaptation requirement, and a negative deviation indicates that the parameter is lower than the actual adaptation requirement. This determines the adjustment direction and adjustment magnitude of the target correction dimension parameters, enabling targeted adjustment of the target correction dimension parameters to ensure... The adjusted parameters can compensate for the adaptation deviation between the original baseline and the user's actual state. For example, if the target correction dimension parameter is a blood oxygen saturation-related parameter in the physiological response threshold range, and the characteristic indicators are the deviation trend of the steady-state blood oxygen saturation value from the baseline preset value and the steady-state characteristics of the oxygen supply adaptation value, after assigning weights according to the importance of their correlation with the blood oxygen saturation parameter, a weighted calculation is performed to obtain the comprehensive adaptation deviation quantification value. If the quantification value is negative and exceeds the preset deviation range, it indicates that the blood oxygen saturation-related parameter is higher than the actual adaptation requirement. The adjustment direction is to lower the boundary of the parameter range. The adjustment magnitude is positively correlated with the absolute value of the quantification value, that is, the larger the absolute value, the larger the adjustment magnitude.

[0135] The corrected target dimension parameters are checked for safety boundaries and corrected accordingly. The parameters are verified to be within the physiological safety range and the range suitable for rehabilitation training. If the corrected parameters exceed the safety boundary, a fallback adjustment is made according to the safety boundary threshold. At the same time, the adjustment range logic of the correction algorithm is optimized to ensure that the parameter adjustment meets both the adaptation requirements and the safety requirements. The corrected target dimension parameters are integrated with the remaining parameters of the personalized behavior baseline to form a complete corrected personalized behavior baseline.

[0136] This embodiment introduces a real-time data-driven oxygen therapy exercise training and rehabilitation device, which consists of a rehabilitation intelligent monitoring and assessment platform, an exercise carrier, an oxygen inhaler, physiological monitoring components, and a brain-computer interface component.

[0137] The intelligent monitoring and evaluation platform for rehabilitation is the core control and decision-making unit of real-time data-driven oxygen therapy exercise training and rehabilitation devices. Its core functions are to coordinate the entire process of data processing, baseline matching, program generation, anomaly detection, and baseline correction. Specifically, it includes: receiving and integrating the user's basic physiological information and health test indicators, performing multi-dimensional preprocessing and stepwise intersection calculation matching of standard behavioral baselines; driving the construction of gradient-based testing rehabilitation exercise programs based on the matching results, and completing the generation of personalized behavioral baselines by combining the differences in response characteristics of dual environments; fusing personalized behavioral baselines with target exercise carrier parameters, and generating personalized oxygen-enriched rehabilitation exercise programs through a multi-factor weighted decision-making method; synchronously receiving real-time data collected by various components during training, performing deviation standardization processing, anomaly judgment threshold correction, and AI anomaly detection, quantifying the degree of anomalies, classifying anomaly types, and generating AI wake-up prompts; when execution anomalies are detected, constructing an anomaly dataset and locating the baseline mismatch dimension based on association mapping rules, completing the classification, correction, and integration of personalized behavioral baselines, and ensuring the accuracy, dynamic adaptability, and execution safety of the training program throughout the process.

[0138] The exercise vehicle is a physical execution platform for users to carry out personalized oxygen-enriched rehabilitation exercise programs. Its core functions are to provide an exercise execution environment, oxygen-enriched collaborative support and data collection capabilities that are adapted to personalized oxygen-enriched rehabilitation exercise programs, and to adapt to different rehabilitation needs and exercise scenarios. Specifically, it includes: adjustable and controllable movement execution methods, exercise intensity, and exercise rhythm; the ability to dynamically adjust the exercise intensity curve, exercise cycle division, and rhythm change parameters according to the program settings, accurately outputting corresponding intensity, frequency, and amplitude of exercise movements; integrated oxygen inhalation pipeline interface or oxygen supply adapter structure, supporting stable connection with the oxygen inhalation pipeline device of the oxygen concentrator, ensuring uninterrupted coordination between oxygen supply and exercise movements, and avoiding pipeline entanglement affecting exercise execution or ventilation; real-time collection of actual exercise intensity, exercise rhythm, completion progress, and other exercise execution data during its own operation, synchronously uploading them to the rehabilitation intelligent monitoring and evaluation platform, providing data support for program optimization and anomaly detection; support for integrated linkage with physiological monitoring components and brain-computer interface components, ensuring the time synchronization of exercise execution, oxygen supply, physiological monitoring, and neural feedback collection, adapting to the exercise needs of each stage of warm-up, main training, and recovery; its structural design fits the oxygen therapy rehabilitation scenario, taking into account both exercise stability and oxygen compatibility, providing reliable hardware support for program implementation and real-time optimization. Specifically, it covers a variety of equipment types adapted to different rehabilitation scenarios: smart treadmills offer a linear running platform mode, with handrail supports integrating oxygen supply pipe interfaces, and a telescopic oxygen supply pipe device to meet stable oxygen supply and data collection under dynamic running postures; smart power bikes and rehabilitation pedal bikes use seated pedaling as the core exercise form, adapting to lower limb rehabilitation needs, and the equipment body has a reserved structure for fixing oxygen supply pipes to ensure smooth and tangle-free oxygen supply during seated exercise; rehabilitation elliptical trainers adapt to mild to moderate rehabilitation needs through a low-impact, full-body coordinated exercise mode, with a range of adjustable exercise rhythm and intensity. Personalized behavioral baseline parameters support linkage with multimodal monitoring devices; the upper limb rehabilitation trainer focuses on the recovery of upper limb motor function, with exercise intensity and step length precisely adapted to the upper limb tolerance limit, and integrates a lightweight oxygen supply interface and data acquisition module; the exercise bike focuses on highly adaptable rhythm adjustment, supports dynamic linkage with oxygen-enriched supply mode parameters, and meets the oxygen supply adaptation needs under different intensity training conditions; the oxygen-enriched stepping fitness device achieves low-load rehabilitation training through stepping exercises, and the device integrates an oxygen supply adaptation structure and physiological data acquisition port, suitable for use in home and rehabilitation institution scenarios.

[0139] An oxygen concentrator is a core device for providing oxygen-enriched nutrition. Its core function is to dynamically output appropriate oxygen supply parameters based on personalized oxygen-enriched rehabilitation exercise plans and real-time monitoring data, ensuring synergistic adaptation between oxygen therapy and exercise training. Specifically, this includes: precisely controlling parameters such as oxygen concentration, pressure, and flow rate; implementing differentiated oxygen supply rules at different training stages according to the pre-set oxygen-enriched supply mode; collecting its own oxygen supply parameter data in real time and uploading it to a rehabilitation intelligent monitoring and evaluation platform as the basis for calculating oxygen supply suitability and detecting anomalies; and supporting the receipt of dynamic control commands from the system. When AI anomaly detection identifies oxygen supply-related anomalies, it synchronously adjusts oxygen supply parameters to adapt to the user's physiological state and exercise needs, ensuring the stability and safety of the oxygen inhalation environment and providing reliable oxygen therapy support for improving exercise tolerance and rehabilitation effects.

[0140] The oxygen inhaler in this device can also be replaced with a hydrogen inhaler. As the core equipment for providing hydrogen therapy, its core functions are the same as the oxygen inhaler, only adaptable to hydrogen-related therapeutic needs. The hydrogen inhaler is based on water electrolysis technology and can precisely control parameters such as hydrogen concentration, flow rate, and pressure. It executes differentiated hydrogen supply rules at different training stages according to the hydrogen supply mode set in a personalized rehabilitation exercise plan. It collects its own hydrogen supply parameter data in real time and uploads it to the rehabilitation intelligent monitoring and evaluation platform as the basis for calculating hydrogen supply suitability and detecting anomalies. It supports receiving dynamic adjustment commands from the system and, when AI anomaly detection identifies hydrogen supply-related anomalies, synchronously adjusts the hydrogen supply parameters to adapt to the user's physiological state and exercise needs. It also has safety mechanisms such as hydrogen concentration over-limit warning, water shortage protection, and overheat protection to ensure the stability and safety of the hydrogen supply environment. Through the selective antioxidant and anti-inflammatory effects of hydrogen, it helps reduce exercise-induced oxidative stress damage and accelerate fatigue recovery, providing reliable hydrogen therapy support for the exercise training of rehabilitation patients.

[0141] The physiological monitoring component is a key module for collecting dynamic physiological data of users during exercise. Its core function is to synchronously capture multi-dimensional physiological indicators throughout the entire process, providing accurate and effective data support for baseline generation, program optimization, and anomaly detection. Specifically, it includes: integrated monitoring units for multiple parameters such as heart rate, blood oxygen saturation, blood pressure, exercise range, and exercise energy consumption; a synchronous acquisition mechanism based on exercise phase; and stable data collection at fixed time intervals in both oxygen-inhaled and oxygen-inhaled environments, and at each training stage. It also has a data validity verification function, ensuring the integrity and accuracy of the collected data through signal strength monitoring, temporal continuity judgment, and screening and elimination of abnormal data within a reasonable range. The processed physiological data is uploaded in real time to the rehabilitation intelligent monitoring and evaluation platform, serving as the core basis for response difference feature extraction, physiological response threshold calculation, and anomaly judgment, accurately reflecting the changes in the user's physiological state and tolerance limits during exercise.

[0142] The brain-computer interface component is a dedicated module for capturing user neurofeedback data. Its core function is to collect real-time EEG characteristic data related to fatigue and concentration, providing supplementary support for dynamic error correction, abnormality classification, and awakening strategy matching. Specifically, it includes: adopting a lightweight headband or ear hook design, capturing EEG signals on the scalp surface in real time through contact electrodes, setting directional acquisition frequency bands for alpha waves related to fatigue and beta waves related to concentration, and separating the target signals through a built-in filtering module; quantifying the user's real-time fatigue and concentration values ​​through a time-series feature extraction algorithm, and simultaneously uploading them to a rehabilitation intelligent monitoring and evaluation platform; and supporting the system to dynamically correct the preset error tolerance based on a dual-dimensional error correction logic, providing key data for abnormality judgment threshold correction and abnormality classification, helping to improve the targeting and accuracy of intervention measures.

[0143] Working principle and its effects:

[0144] The working principle and effects of this invention revolve around constructing a full-process oxygen therapy exercise training management logic centered on individual data and aiming for dynamic adaptation. Through multi-stage data integration, environmental comparison testing, real-time monitoring, and closed-loop correction, it achieves precision and safety in training programs. Overall, it first matches an initial set of standard parameters based on the user's basic physiological and health data. Then, through gradient testing under dual oxygen supply environments, it obtains core parameters suitable for the individual. Furthermore, it generates personalized training programs by combining the characteristics of the exercise vehicle. Simultaneously, it collects multi-dimensional data in real time to detect training anomalies and dynamically optimizes core parameters based on anomaly types, forming a closed-loop management system of data input, program generation, monitoring feedback, and parameter optimization. This effectively solves the problems of insufficient individual adaptation, singular monitoring, and static programs in existing technologies.

[0145] In summary, this invention improves the accuracy of initial parameter matching through multi-dimensional data integration, enhances individual adaptability through dual-environment comparative testing, improves anomaly identification efficiency through multi-dimensional real-time monitoring, and achieves dynamic parameter optimization through closed-loop correction. In principle, it constructs a full-link personalized management system from individual data to solutions, and from monitoring to optimization. In terms of effect, it significantly improves the accuracy, safety, and effectiveness of oxygen therapy exercise training, ensuring that the solutions are tailored to individual and carrier characteristics, providing timely warnings and mitigating abnormal risks, and guaranteeing long-term rehabilitation effects through dynamic optimization. This provides a systematic and personalized management solution for oxygen therapy exercise training.

[0146] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A real-time data-driven oxygen therapy exercise training and rehabilitation method, characterized in that, include: The system acquires the user's basic physiological information and health monitoring indicators, which are then used to perform stepwise intersection calculations and matching with a standard behavioral baseline database to obtain the user's standard behavioral baseline. The standard behavioral baseline is a set of multi-dimensional parameters reflecting the rehabilitation exercise ability of the corresponding population, formed by hierarchical clustering according to age segment, underlying disease type, and health test index range using density clustering algorithm. It includes exercise tolerance range parameters, physiological response threshold range parameters, and exercise adaptation range parameters. The stepwise intersection operation is a hierarchical intersection operation method that, based on the user's basic physiological information and health test indicators, selects matching standard behavioral baselines from the standard behavioral baseline library layer by layer in the order of age segment, underlying disease type, and health test indicator range. Based on the user's standard behavioral baseline obtained through matching, a graded testing and rehabilitation exercise program is constructed to modify the user's standard behavioral baseline by analyzing the differences in the user's response in non-oxygen and oxygen-inhaled environments, thereby obtaining the user's personalized behavioral baseline. Based on a personalized behavioral baseline, and combined with the movement mode parameters of the target exercise vehicle selected by the user, physiological state goals and exercise execution goals are set for each training stage. A multi-factor weighted decision-making method is used to generate a personalized oxygen-enriched rehabilitation exercise program. When a user triggers the execution of a personalized oxygen-enriched rehabilitation exercise program, real-time data of the user is collected through physiological monitoring components, brain-computer interface components, exercise carriers and oxygen concentrators. Through deviation standardization processing and abnormal judgment threshold correction, AI anomaly detection is performed to identify deviation data points, determine whether the user has execution abnormalities, quantify the degree of abnormality and classify the abnormality type, and then generate AI wake-up prompts. When a user is found to have an execution anomaly, the user's personalized behavior baseline is classified and corrected according to the anomaly type.

2. The real-time data-driven oxygen therapy exercise training and rehabilitation method as described in claim 1, characterized in that, The specific steps for obtaining the user's standard behavioral baseline include: The system acquires the user's basic physiological information and health monitoring indicators, including age, resting heart rate, and type of underlying disease; the health monitoring indicators include susceptibility gene-related data and telomere detection results. The basic physiological information and health test indicators that have passed the validity verification are processed to unify the format and normalize the values ​​to form a structured data feature set; Based on the age information in the structured data feature set, and combined with the dynamic age segmentation rules of the standard behavior baseline library, all standard behavior baselines corresponding to the age segment are extracted from the standard behavior baseline library to form the first baseline subset. Based on the basic disease type information in the structured data feature set, a standard behavioral baseline consistent with the basic disease type is extracted from the first baseline subset; and the matching tolerance is adjusted according to the variance of the baseline sample distribution corresponding to the basic disease type to form a second baseline subset. Based on the health monitoring indicator information in the structured data feature set, the interval mapping matching method is used to extract the standard behavioral baselines from the second baseline subset where the mapping overlap between the health monitoring indicator interval and the user's health monitoring indicator is greater than a preset overlap threshold, forming the third baseline subset.

3. The real-time data-driven oxygen therapy exercise training and rehabilitation method as described in claim 2, characterized in that, The specific steps for obtaining the user's standard behavioral baseline also include: Multi-dimensional similarity fusion calculation is performed on each standard behavioral baseline of the third baseline subset, and mapping weights are assigned according to the mapping position of the user's health detection indicators in the corresponding behavioral baseline indicator interval. Based on this mapping weight, the degree of fit between the user's health monitoring indicators and the corresponding intervals of each standard behavioral baseline is calculated; and the degree of match between the resting heart rate, the exercise physiological parameters associated with underlying diseases and the corresponding parameters of each standard behavioral baseline is calculated. The fit of health monitoring indicators, the matching degree of resting heart rate, and the matching degree of exercise physiological parameters related to underlying diseases are integrated to obtain the comprehensive similarity of each standard behavioral baseline in the third baseline subset; If there is a standard behavior baseline in the third baseline subset with a comprehensive similarity greater than the preset high fit standard, then it is determined as the standard behavior baseline that matches the user. If none exist, then remove the abnormal standard behavior baselines in the third baseline subset whose comprehensive similarity is lower than the mean of the comprehensive similarity of the third baseline subset minus the standard deviation of the comprehensive similarity. Recalculate the comprehensive similarity of the remaining standard behavior baselines, and select the standard behavior baselines whose comprehensive similarity after recalculation is greater than the preset adaptation threshold to determine the standard behavior baselines that match the user.

4. The real-time data-driven oxygen therapy exercise training and rehabilitation method as described in claim 2, characterized in that, The steps for constructing a graded testing and rehabilitation exercise program include: Extract core parameters from the user's standard behavioral baseline, including exercise tolerance range parameters, physiological response threshold range parameters, and exercise adaptation range parameters; Based on the exercise tolerance interval parameters of the standard behavioral baseline, multiple continuous gradient intervals are divided in ascending order. The gradient intervals are divided according to the distribution density of exercise tolerance in the standard behavioral baseline. Combining the exercise adaptation interval parameters of the standard behavioral baseline, the total duration of a single test exercise is determined, and the duration is allocated according to the importance of each gradient interval. Referring to the physiological response threshold interval parameters of the standard behavioral baseline, the movement frequency change pattern is set for each gradient interval. A gradient test rehabilitation exercise program is thus formed.

5. The real-time data-driven oxygen therapy exercise training and rehabilitation method as described in claim 1, characterized in that, The steps of deviation standardization and anomaly detection threshold correction include: When a user triggers the execution of a personalized oxygen-enriched rehabilitation exercise program, real-time data of the user is collected simultaneously. The real-time data includes: dynamic exercise physiological data, brain-computer interface data, exercise execution data, and real-time oxygen supply parameters. Add a unified timestamp to the real-time data, divide the data into segments according to the training stage, and construct a real-time data sequence corresponding to each training stage; Based on the physiological state goals and exercise execution goals of each training stage, and combined with the current real-time oxygen supply parameters, a dynamic threshold system for stage goals is constructed to set standard threshold ranges for exercise physiological dynamic data and exercise execution data for each training stage. The real-time data sequences of exercise physiological dynamic data and exercise execution data are standardized for deviation. The absolute difference between the exercise physiological dynamic data and exercise execution data and the standard threshold interval of the corresponding training stage is calculated. The deviation value is converted into a corrected deviation value according to the proportion of the absolute difference to the total span of the standard threshold interval, thus forming the deviation sequence of exercise physiological dynamic data and exercise execution data. Based on the real-time data sequence of brain-computer interface data, the dynamic changing trends of the proportion of alpha wave energy and the proportion of beta wave energy are extracted, and the user's real-time fatigue and concentration values ​​are quantified. Based on fatigue and focus values, a two-dimensional error tolerance adjustment is performed to dynamically correct the user's preset error tolerance and generate the user's real-time error tolerance; and the adaptation deviation between the real-time oxygen supply parameters and the current stage preset oxygen supply standard is calculated to generate an oxygen supply adaptation value. Set anomaly detection threshold for each training stage, and adjust the anomaly detection threshold for each training stage based on the oxygen supply fit value and the user's real-time fault tolerance.

6. The real-time data-driven oxygen therapy exercise training and rehabilitation method as described in claim 5, characterized in that, The steps of performing AI anomaly detection, identifying deviation data points, quantifying the degree of anomaly, classifying anomaly types, and generating AI wake-up prompts include: The deviation sequence of dynamic exercise physiological data and exercise execution data is compared one by one with the anomaly judgment threshold after correction in the corresponding training stage. Deviation data points that exceed the anomaly judgment threshold are marked. The proportion of deviation data points within the preset detection period is calculated. If it is greater than the preset proportion threshold, it is determined that the current user has an execution abnormality. The degree of anomaly is classified based on the degree of deviation of each deviation data point from the anomaly judgment threshold and the proportion of deviation data points within the detection period. An AI algorithm combining decision tree and gradient boosting tree was used to extract anomaly features from real-time data that deviated from the time window of the data point and to classify the anomaly types. The anomaly features included: the deviation between the dynamic data of exercise physiology and the data of exercise execution, the real-time fault tolerance, and the oxygen supply adaptation value. Establish a dynamic mapping library of anomaly types, anomaly severity, user status, and wake-up strategies. This library is used to match wake-up strategies based on the user's current anomaly type, anomaly severity, fatigue level, and focus level, and to generate AI wake-up prompts through multimodal interaction components.

7. The real-time data-driven oxygen therapy exercise training and rehabilitation method as described in claim 1, characterized in that, The step of classifying and correcting the user's personalized behavioral baseline based on the anomaly type includes: When an abnormality is detected in the user's execution, the user's personalized behavior baseline correction operation is triggered to construct an abnormal dataset of the current personalized oxygen-enriched rehabilitation exercise program execution. Based on the preset association mapping rules between anomaly types and baseline parameters, the target correction dimension parameters that do not fit in the personalized behavior baseline are located. Feature extraction is performed on various time series data in the abnormal dataset. Combined with the functional attributes of the target correction dimension parameters, feature indicators of the target correction dimension parameters are selected to construct a feature set for correction. Based on the characteristic indicators of the target correction dimension parameters, the degree of adaptation deviation of the target correction dimension parameters is quantified by multi-dimensional feature weighted deviation calculation. This is used to determine the adjustment direction and adjustment magnitude of the target correction dimension parameters and to make targeted adjustments to the target correction dimension parameters. The corrected target dimension parameters are subjected to safety boundary verification and correction. The corrected target dimension parameters are then integrated with the remaining parameters of the personalized behavior baseline to form the corrected personalized behavior baseline.

8. A real-time data-driven oxygen therapy exercise training and rehabilitation device, used to implement the real-time data-driven oxygen therapy exercise training and rehabilitation method according to any one of claims 1-7, characterized in that, It consists of a rehabilitation intelligent monitoring and assessment platform, a motion carrier, an oxygen concentrator, physiological monitoring components, and a brain-computer interface component; The intelligent rehabilitation monitoring and evaluation platform is used to construct a personalized behavioral baseline for the user, generate a personalized oxygen-enriched rehabilitation exercise plan, and perform abnormal execution diagnosis to provide feedback and correct the personalized behavioral baseline; the exercise carrier is used to provide an exercise execution environment adapted to the personalized oxygen-enriched rehabilitation exercise plan; the oxygen concentrator is used to dynamically output adapted oxygen supply parameters based on the personalized oxygen-enriched rehabilitation exercise plan and real-time monitoring data; the physiological monitoring component is used to collect the user's dynamic exercise physiological data. The brain-computer interface component is used to collect electroencephalogram (EEG) characteristic data related to fatigue and concentration.

Images