State perception and adaptation method and system based on multi-source heterogeneous data fusion

Abstract

The application belongs to the technical field of data analysis, and discloses a state perception and adaptation method and system based on multi-source heterogeneous data fusion, which comprises the following steps: collecting heart rate data, state data and cognitive data of a subject, generating baseline heart rate, state characteristics and initial cognitive characteristics respectively; generating a state perception result; judging whether the initial cognitive characteristics meet the adaptation input condition according to the state perception result; if not, performing recovery mode training, and re-collecting cognitive data and state data after the recovery mode training is completed, generating updated cognitive characteristics and comprehensive recovery confidence; then judging whether the adaptation input condition is met, if yes, taking the initial cognitive characteristics or the updated cognitive characteristics that meet the adaptation input condition as effective cognitive characteristics, and calling the effective cognitive characteristics to perform recommended scheme adaptation, otherwise, repeatedly performing the recovery mode training; the application improves the reliability and applicability of the ability evaluation and rehabilitation system.

Description

Technical Field

[0001] This application relates to the field of data analysis technology, and more specifically, to a state perception and adaptation method and system based on multi-source heterogeneous data fusion. Background Technology

[0002] In ability assessment and rehabilitation training systems, existing technologies typically employ a fixed process of assessment, fitting, and training. Specifically, the system directly recommends a fitting plan or generates a training plan based on the assessment results, and the subject then enters the fitting trial or training phase. This process implicitly assumes that the subject's state after the assessment is consistent with their average state during the assessment process and that they are able to effectively perform subsequent fitting or training operations.

[0003] However, in practical applications, when the assessment process involves multiple assessment modules and the continuous execution time is long, subjects may experience increased heart rate, increased task time, or an increase in unplanned interruptions. When the assessment content involves multiple functional modules and many test items, subjects often experience changes in their state such as increased heart rate, decreased attention, and accumulated fatigue in the later stages of the assessment. In this state, if the adaptation trial is immediately carried out according to the fixed process of existing technology, subjects may not be able to truly reflect the actual usability of the adaptation plan due to fatigue, lack of concentration, or subjective resistance. This may lead to a plan that should be well-adapted being misjudged as unsuitable, or an unsuitable plan being incorrectly adopted because the subject has not fully performed the trial task. Similarly, if training is started immediately, subjects may experience frustration due to overload, reducing compliance with subsequent training.

[0004] While existing technologies can collect some physiological data during the assessment process, they are only used as independent indicators for health monitoring or abnormal alarms. This results in subsequent adaptation and training being carried out in a fixed order even when the subject is in poor condition at the end of the assessment. This leads to distorted adaptation results, unsuitable training loads, low overall intervention efficiency, and may even cause the subject to resist the rehabilitation process.

[0005] In view of this, this application proposes a state awareness and adaptation method and system based on multi-source heterogeneous data fusion to solve the above problems. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this application aims to provide a state perception and adaptation method and system based on multi-source heterogeneous data fusion. This method collects baseline heart rate, state characteristics, and initial cognitive characteristics. It generates state perception results by comparing the baseline heart rate with the state characteristics, and determines whether the initial cognitive characteristics meet the adaptation input conditions. If not, it executes recovery mode training. After training, it re-collects data to generate updated cognitive characteristics and a comprehensive recovery confidence score. Once the conditions are verified, the initial or updated cognitive characteristics are used as valid cognitive characteristics to execute the recommended adaptation scheme. This solves the problems in existing technologies, such as mismatched adaptation timing, unsuitable training load, and disconnect between assessment and training, caused by neglecting the subject's real-time state. It achieves the technical effect of improving the authenticity of adaptation results, the matching degree of training load, and the closed-loop nature and reliability of the overall intervention process.

[0007] This application provides the following technical solution: a state awareness and adaptation method based on multi-source heterogeneous data fusion, including:

[0008] Heart rate data of subjects were collected before the execution of a preset standardized procedure, and state data and cognitive data were collected during the execution of the preset standardized procedure to generate baseline heart rate, state characteristics and initial cognitive characteristics, respectively.

[0009] A state perception result is generated by comparing and analyzing the baseline heart rate with the state characteristics.

[0010] Based on the state perception results, determine whether the initial cognitive features meet the adaptation input conditions;

[0011] If the initial cognitive features do not meet the input adaptation conditions, recovery mode training is performed, and cognitive and state data are re-collected after the recovery mode training is completed to generate updated cognitive features and comprehensive recovery confidence. The comprehensive recovery confidence is used to determine whether the updated cognitive features meet the input adaptation conditions. If the updated cognitive features do not meet the input adaptation conditions, recovery mode training is repeated.

[0012] The initial or updated cognitive features that meet the adaptation input conditions are taken as valid cognitive features, and the recommendation scheme is adapted by calling the valid cognitive features.

[0013] The state awareness and adaptation system based on multi-source heterogeneous data fusion, implementing the state awareness and adaptation method based on multi-source heterogeneous data fusion, includes:

[0014] The data acquisition and analysis module is used to collect heart rate data of the subjects before the execution of the preset standardized procedure, and to collect state data and cognitive data during the execution of the preset standardized procedure, and to generate baseline heart rate, state characteristics and initial cognitive characteristics, respectively.

[0015] The state analysis module is used to compare and analyze the baseline heart rate with the state characteristics to generate state perception results.

[0016] The first decision module is used to determine whether the initial cognitive features meet the adaptation input conditions based on the state perception results.

[0017] The second decision module is used to perform recovery mode training if the initial cognitive features do not meet the adaptation input conditions, and to re-collect cognitive data and state data after the recovery mode training is completed, and generate updated cognitive features and comprehensive recovery confidence. Based on the comprehensive recovery confidence, it is determined whether the updated cognitive features meet the adaptation input conditions. If the updated cognitive features do not meet the adaptation input conditions, the recovery mode training is repeated.

[0018] The scheme adaptation module is used to take the initial or updated cognitive features that meet the adaptation input conditions as effective cognitive features and call the effective cognitive features to perform the recommended scheme adaptation.

[0019] The technical effects and advantages of this application regarding the state awareness and adaptation method and system based on multi-source heterogeneous data fusion are as follows:

[0020] 1. This application uses the state perception results at the end of the assessment as the basis for judging the adaptation input conditions, and only allows the initial cognitive features to enter the adaptation process when the subject is in a good state; it effectively avoids the drawback of the prior art that directly uses the assessment results for adaptation regardless of whether the subject is tired or focused, and avoids the problem of misjudging a well-fitted solution as unfit or incorrectly adopting an unfit solution due to the subject's poor state.

[0021] 2. When the initial cognitive features are determined not to meet the adaptation input conditions, this application does not directly abandon the adaptation, but actively performs recovery mode training to alleviate the fatigue and load of the subjects, and re-collects data to generate updated cognitive features. This mechanism transforms the state recovery process, which is ignored in the prior art, into a necessary correction step before adaptation, so that the cognitive data that was originally distorted due to poor state can be reconstructed, providing reliable input for subsequent adaptation.

[0022] 3. After the recovery mode training is completed, this application further calculates the comprehensive recovery confidence and uses it as a quantitative basis for judging whether the updated cognitive features are effective. Only when the confidence reaches the preset threshold is the updated cognitive feature allowed to enter the adaptation stage. If it does not reach the threshold, the recovery training is repeated until the state is fully recovered. This closed-loop verification mechanism avoids the unreliability of directly entering the adaptation stage based on a single recovery training and further ensures the authenticity of the adaptation input.

[0023] 4. The cognitive features used for adaptation in this application have all been screened and verified by the adaptation input conditions to ensure that they are valid data collected when the subject is in a suitable state. The recommended adaptation scheme implemented on this basis can truly reflect the degree of matching between the subject's functional needs and the assistive device, thereby reducing the probability of misjudgment caused by subject fatigue, perfunctory attitude or resistance, and improving the stability of the recommended adaptation scheme and the overall intervention effectiveness of the system.

[0024] In summary, this application organizes the steps of assessment, state perception, branch decision-making, adaptation, dual-mode training, state verification, and reassessment into a continuous data transmission chain. This ensures both refined adaptation and rehabilitation training when the subject is in a good state and progressive intervention of recovery followed by adaptation and rehabilitation when the subject is in a poor state. This improves the reliability and applicability of the ability assessment and rehabilitation system in complex real-world scenarios. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the state perception and adaptation method based on multi-source heterogeneous data fusion in this application.

[0026] Figure 2 A schematic diagram of the method flow for adapting the recommended scheme in this application;

[0027] Figure 3 This is a schematic diagram of the method flow for implementing the rehabilitation mode in this application;

[0028] Figure 4 This is a schematic diagram of the state perception and adaptation system structure based on multi-source heterogeneous data fusion in this application. Detailed Implementation

[0029] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0030] Example 1:

[0031] Please see Figure 1 As shown, this embodiment provides a state awareness and adaptation method based on multi-source heterogeneous data fusion, including:

[0032] Heart rate data of subjects were collected before the execution of the pre-set standardized procedure, and state data and cognitive data were collected during the execution of the pre-set standardized procedure to generate baseline heart rate, state characteristics and initial cognitive characteristics, respectively.

[0033] In specific implementations, methods for constructing a baseline heart rate include:

[0034] After initiating the first data acquisition process, the control unit automatically detects whether the heart rate monitoring terminal is connected. If connected, the control unit generates a first prompt command to guide the subject into a resting state. After the resting time reaches a preset resting waiting threshold, the control unit continuously collects a preset number of instantaneous heart rate values ​​through the heart rate monitoring terminal. The arithmetic mean of the instantaneous heart rate values ​​is taken as the baseline heart rate to characterize the subject's resting physiological baseline. When the number of valid instantaneous heart rate values ​​is less than the preset number of valid samples, no baseline heart rate is generated, and a heart rate acquisition invalidation flag is output. When the heart rate acquisition invalidation flag exists, the state perception result calculation based on the baseline heart rate is not performed. The heart rate monitoring terminal includes a wearable heart rate acquisition terminal and an optoelectronic transmission device integrated into the first test terminal. Sensing unit; the heart rate monitoring terminal referred to below refers to any of the above-mentioned acquisition terminals capable of outputting instantaneous heart rate values; the method for setting the duration of the resting state is: determined based on the time required for the heart rate to stabilize after the human body transitions from an active state to a resting state; the method for setting the resting wait threshold is: taking the average time required for a healthy subject population to sit still from a state of light activity until the heart rate reaches a stable level, and considering the relatively slow physiological regulation of the subject population, increasing the preset safety margin as the resting wait threshold; the method for setting the number of samplings is: determined based on the fluctuation characteristics of the heart rate signal and the requirements of measurement accuracy, taking the number of samplings that can achieve a balance between measurement stability and measurement efficiency, usually the arithmetic mean of multiple samplings to reduce random errors.

[0035] If the heart rate monitoring terminal is not connected, or if no valid instantaneous heart rate value is obtained within the preset invalid sampling time, the system will switch to the backup acquisition channel. The valid instantaneous heart rate value is the heart rate value that is within the preset physiological heart rate range and carries a valid sampling identifier.

[0036] In specific implementations, the methods for extracting and obtaining state features include:

[0037] Subjects followed a pre-defined standardized procedure, sequentially retrieving assessment tools from the first functional unit of the integrated resource platform to complete data collection for each assessment module. The pre-defined standardized procedure was determined based on the execution order of the assessment modules, the order in which assessment tools were retrieved, and the standard time consumption for each assessment task. The standard time consumption was set by having multiple experienced assessors conduct multiple tests on healthy subjects in a standard environment during the pre-experiment phase, taking the average completion time of each task as the initial standard time consumption, which could then be dynamically calibrated based on the median or mode of the actual operational data. The integrated resource platform was adaptively adjusted according to the subjects. In an optional embodiment, taking elderly capacity assessment as an example, the integrated resource platform could consist of three functional cabinets (A, B, and C), with a total footprint of approximately 1.8m².2 Cabinet A has a seven-layer structure with specific partitions and slots, used to centrally store 70 small and micro assessment tools and matching assistive devices for the four modules of information collection, self-care ability, mental cognition, and perception, such as blood pressure monitors, hand strength testers, cognitive cards, and visual aids; Cabinets B and C are used to centrally store 10 medium-sized assistive devices for the mobility module, such as wheelchairs, walkers, and transfer boards; the layout of all tools corresponds to the standardized process modules of the subject assessment guidelines, and assessors can operate along the preset flow, greatly improving efficiency and standardization.

[0038] During the subjects' execution of the pre-set standardized procedure, heart rate was continuously monitored through a heart rate monitoring terminal. The heart rate collection intervals for each task were divided according to the start and end times of the assessment task. The arithmetic mean of the instantaneous heart rate values ​​within each heart rate collection interval was calculated to obtain the average heart rate during each assessment task.

[0039] Record the start and end times of each assessment task, calculate the difference between the end time and the start time to obtain the actual time consumed, and calculate the ratio of the actual time consumed to the standard time consumed after each assessment task is completed to obtain the task time ratio corresponding to the assessment task.

[0040] Between two adjacent evaluation tasks, record the end time of the previous evaluation task and the start time of the next evaluation task, calculate the difference between the start time of the next evaluation task and the end time of the previous evaluation task, and obtain the task switching duration between the corresponding adjacent tasks; combine the task switching durations according to the execution order of the evaluation tasks to generate a task switching duration sequence.

[0041] When the evaluator triggers a pause command on the control unit, the interruption time is recorded as the time when the pause command is triggered; or when an operation interval exceeding the preset interruption judgment duration is detected, an interruption is automatically counted once, and the start time of the operation interval is recorded as the interruption time; the preset interruption judgment duration is set by determining it based on the reasonable interval time of task switching during normal evaluation, and exceeding this interval is judged as an unplanned interruption; an operation interval refers to the absence of task confirmation operation, key operation, touch operation, tool retrieval record, or evaluator's continuation process confirmation record within the corresponding time period.

[0042] After the evaluation is completed, the system generates status features, which include an average heart rate sequence, a task time ratio sequence, an unplanned interruption count sequence, and a task switching duration sequence. The average heart rate sequence is formed by arranging the average heart rate during each evaluation task in the order of task execution. The task time ratio sequence is formed by arranging the task time ratio corresponding to each evaluation task in the order of task execution. The task switching duration sequence is formed by arranging the task switching duration between adjacent evaluation tasks in the order of task execution. The unplanned interruption count is used to characterize the cumulative result of unplanned interruptions during the execution of the preset standardized process.

[0043] In specific implementations, methods for obtaining cognitive features include:

[0044] During the execution of the pre-defined standardized process, performance data for four cognitive dimensions were collected, and scores for each cognitive dimension were calculated. The first cognitive task type corresponds to the memory ability dimension, the second cognitive task type corresponds to the attention ability dimension, the third cognitive task type corresponds to the executive function dimension, and the fourth cognitive task type corresponds to the language ability dimension. Specifically, for the first cognitive task type, accuracy and reaction time were recorded. Accuracy was the number of correctly answered tasks divided by the total number of tasks, with a value ranging from 0 to 1. Reaction time was the time interval from task presentation to the subject's response. A baseline reaction time for the first cognitive dimension was pre-defined, which was set by taking the average reaction time of a healthy subject population on the first cognitive dimension task, obtained through literature review or pre-experimentation. Accuracy was directly used as the accuracy contribution value. The ratio of actual reaction time to baseline reaction time was normalized. The normalization method was: when the actual reaction time was less than or equal to the baseline reaction time, the reaction time was normalized. The contribution value is 1; when the actual reaction time is longer than the baseline reaction time, the reaction time contribution value is equal to the baseline reaction time divided by the actual reaction time; the accuracy contribution value and the reaction time contribution value are weighted and summed according to preset weights, and then multiplied by the upper limit of the score to obtain the first cognitive dimension score within the range of 0 to the upper limit of the score. Each weight is a value greater than or equal to 0, and the sum of each weight in the same weighting calculation is 1; the accuracy weight is set according to the consensus in the field of cognitive rehabilitation that accuracy is more clinically significant than speed, and the accuracy weight is usually greater than the reaction time weight; for example, data on memory ability dimension can be collected and analyzed through picture matching tasks or digit span tasks to obtain the memory ability dimension score; the upper limit of the score is set according to the scoring range commonly used in clinical assessment scales, taking an integer that is easy to understand and calculate, while ensuring that the score has sufficient discrimination to distinguish subtle differences in ability; in this embodiment, all upper limits of the score use the same value.

[0045] For the second cognitive task type, the omission rate and false alarm rate are recorded. The omission rate is the number of missed targets divided by the total number of targets; the false alarm rate is the number of incorrectly reported non-targets divided by the total number of non-targets. Both rates range from 0 to 1. The omission rate and false alarm rate are weighted and summed according to preset weights to obtain a comprehensive error value. This value is then subtracted from 1 and multiplied by the upper limit of the score to obtain the second cognitive dimension score within the range of 0 to the upper limit of the score. Each weight is a value greater than or equal to 0, and the sum of all weights in the same weighting calculation is 1. The preset weights are determined based on the relative importance of omission and false alarm to attentional ability assessment in clinical practice. Usually, the weight of the omission rate is equal to the weight of the false alarm rate. Attentional ability dimension data can be collected and analyzed through continuous attention tracking tasks or selective attention screening tasks to obtain the attentional ability dimension score.

[0046] For the third cognitive task type, the switching cost is recorded. The switching cost is the average reaction time under the switching condition minus the average reaction time under the non-switching condition. A maximum acceptable switching cost is preset. The method for setting the maximum acceptable switching cost is: determined based on the upper limit of the switching cost distribution of healthy subjects on this task, usually twice the mean of the normal population is taken as the maximum acceptable value. The cost ratio of the actual switching cost to the maximum acceptable switching cost is clamped to no more than 1. If the value exceeds 1, the cost ratio is forcibly set to 1. Then, 1 is subtracted from the cost ratio and multiplied by the upper limit of the score to obtain the score of the third cognitive dimension within the range of 0 to the upper limit of the score. If data of the executive function dimension can be collected through task switching tasks for analysis, the executive function dimension score can be obtained.

[0047] For the fourth cognitive task type, completion and accuracy are recorded. Completion is the number of correctly named or repeated items divided by the total number of items, ranging from 0 to 1. Accuracy is the semantic correctness score, obtained by the system automatically comparing with preset answers, ranging from 0 to 1. The completion and accuracy are weighted and summed according to preset weights, and then multiplied by the upper limit of the score to obtain the fourth cognitive dimension score within the range of 0 to the upper limit of the score. Each preset weight is a value greater than or equal to 0, and the sum of all weights in the same weighted calculation is 1. The preset weights are set according to the relative importance of the quantity and quality of completion to the ability reflection in the language rehabilitation assessment. The language ability dimension score can be obtained by collecting and analyzing data from naming or repeating tasks.

[0048] The scores of the first cognitive dimension, the second cognitive dimension, the third cognitive dimension, and the fourth cognitive dimension are concatenated to obtain cognitive features.

[0049] A state perception result is generated by comparing and analyzing the baseline heart rate with state characteristics.

[0050] In specific implementations, methods for obtaining state-aware results include:

[0051] When the state features include average heart rate sequences from multiple assessment tasks, the physiological stability dimension sub-scores for each assessment task are calculated separately, and the arithmetic mean of the sub-scores within the most recent preset window number is taken as the physiological stability dimension score. The physiological stability dimension score is calculated based on the average heart rate and baseline heart rate during the current assessment task in the state features. Specifically, the absolute value of the difference between the average heart rate and the baseline heart rate is calculated and denoted as the absolute deviation; the product of the baseline heart rate and a preset physiological deviation ratio is calculated and denoted as the resting-state deviation; the ratio of the absolute deviation value to the resting-state deviation is calculated to obtain the initial deviation level coefficient. The smaller of the initial deviation coefficient and value 1 is obtained as the deviation coefficient. The difference between value 1 and the deviation coefficient is calculated and recorded as the stability. The product of the stability and the upper limit of the score is calculated to obtain the physiological stability dimension score. When the average heart rate is equal to the baseline heart rate, the physiological stability dimension score is the upper limit of the score. When the average heart rate deviates to or exceeds the preset physiological deviation ratio, the physiological stability dimension score is 0. The preset physiological deviation ratio is set by determining it according to the allowable range of physiological fluctuations of the subject and the clinical fatigue judgment criteria, and taking a critical value that can distinguish between normal physiological fluctuations and significant fatigue state.

[0052] The operational load dimension score is calculated based on the task time ratio and the number of unplanned interruptions. Specifically, the task time ratio score and the score for the number of unplanned interruptions are calculated separately, and a weighted sum of these scores is obtained to determine the operational load dimension score. The task time ratio score is calculated as follows: the ratio of actual time consumption to standard time consumption is used as the task time ratio; a preset allowable fluctuation coefficient and an allowable deviation upper limit for the time ratio are used; 1 minus the allowable fluctuation coefficient is used as the lower limit of the ideal interval, and 1 plus the allowable fluctuation coefficient is used as the upper limit of the ideal interval, thus obtaining the ideal interval for the task time ratio; when the task time ratio is within the ideal interval, the task time ratio score is the upper limit; when the task time ratio is greater than the upper limit of the ideal interval, the difference between the task time ratio and the upper limit of the ideal interval is calculated, and the corresponding difference is divided by the allowable deviation upper limit of the time ratio to obtain the initial exceedance level; when the task time ratio is less than the lower limit of the ideal interval, the lower limit of the ideal interval and the task time ratio are calculated separately. The difference between the time consumption ratio and the value 1 is used to obtain the initial excess degree. The smaller value between the initial excess degree and the value 1 is taken as the excess degree coefficient. The difference between the value 1 and the excess degree coefficient is calculated to obtain the time consumption coefficient. The product of the score limit and the time consumption coefficient is calculated to obtain the task time consumption ratio score. The method for setting the time consumption ratio allowable fluctuation coefficient is as follows: Collect historical task time consumption ratio samples that have completed the preset standardized process normally, calculate the absolute value of the deviation between each task time consumption ratio and the value 1, and take the 75th percentile of the absolute value of the deviation as the time consumption ratio allowable fluctuation coefficient. The method for setting the time consumption ratio allowable deviation limit is as follows: In the historical task time consumption ratio samples that have completed the preset standardized process normally, select task time consumption ratio samples that exceed the ideal range, calculate the deviation value of the deviation value that exceeds the corresponding ideal range boundary, and take the 90th percentile of the deviation value as the time consumption ratio allowable deviation limit. If the historical samples are insufficient, the time consumption ratio allowable fluctuation coefficient is used as the time consumption ratio allowable deviation limit.

[0053] The calculation method for the score of unplanned interruptions is as follows: a maximum allowed number of interruptions is preset, the ratio of the actual number of interruptions to the maximum allowed number of interruptions is calculated and recorded as the interruption ratio, and the smaller value between the interruption ratio and value 1 is taken to obtain the interruption coefficient; the difference between value 1 and the interruption coefficient is calculated and recorded as the unplanned interruption coefficient, and the product of the upper limit of the score and the unplanned interruption coefficient is calculated to obtain the score of unplanned interruptions; the method for setting the maximum allowed number of interruptions is: determined according to the reasonable fluctuation range of the subject's cooperation during the normal assessment process, and the upper limit of the number of interruptions that most subjects can successfully complete the assessment is taken.

[0054] The weight of the task duration score is greater than the weight of the score for the number of unplanned interruptions. The weight is determined by the relative importance of the degree of influence of task duration deviation on status assessment and the degree of influence of the number of interruptions on status assessment in clinical practice.

[0055] The task continuity dimension score is calculated based on the task switching duration. Specifically, the task switching duration is the time difference between the end time of the previous evaluation task and the start time of the next evaluation task. An ideal duration range and an allowable deviation duration are preset. When the task switching duration is within the ideal duration range, the task continuity dimension score is the upper limit. When the task switching duration exceeds the ideal duration range, the difference between the task switching duration and the nearest boundary value of the ideal duration range is calculated and recorded as the duration difference. The duration difference is divided by the allowable deviation duration to obtain the initial switching degree, which is then taken as the initial switching degree. The smaller value in value 1 is used as the switching degree coefficient. The difference between value 1 and the switching degree coefficient is calculated and multiplied by the upper limit of the score to obtain the task coherence dimension score. The ideal duration interval is set by determining the time required for the assessor to complete the normal transition such as tool preparation and subject posture adjustment. The allowable deviation duration is set by collecting historical samples of task switching time when the preset standardized process is completed normally, calculating the sample standard deviation, and using the sample standard deviation as the allowable deviation duration. If there are no historical samples, the difference between the upper and lower bounds of the ideal duration interval is used as the allowable deviation duration.

[0056] The comprehensive state credibility index is obtained by weighting and summing the scores of physiological stability dimension, operational load dimension, and task coherence dimension. The comprehensive state credibility index is used as the state perception result. The weighting weights are set according to the significance of each indicator's influence on fatigue state in rehabilitation medicine, obtained through expert scoring or clinical data statistical analysis, and the sum of each weighting weight is 1.

[0057] The overall state credibility index is updated once after each evaluation task is completed. A sliding window method is used, and the average value of the most recent tasks within a preset window is taken as the current overall state credibility index, i.e., the current state perception result. The preset window number is determined based on the total number of evaluation tasks and the time resolution requirements of state changes, ensuring that the index can reflect the state trend without causing drastic fluctuations due to a single outlier.

[0058] Based on the state perception results, determine whether the initial cognitive features meet the adaptation input conditions.

[0059] In specific implementations, methods for determining whether initial cognitive features meet the adaptation input conditions include:

[0060] If the overall state credibility index is greater than or equal to the preset state threshold, the initial cognitive features are deemed to meet the adaptation input conditions, and the recommended scheme is implemented for adaptation; otherwise, the initial cognitive features are deemed not to meet the adaptation input conditions, and recovery mode training is implemented for data correction. The preset state threshold is set as follows: historical evaluation samples are collected, each historical evaluation sample includes the overall state credibility index and the corresponding adaptation trial result; samples that can complete the adaptation trial and whose trial data is confirmed as valid by the evaluator are taken as valid trial samples, and the lower quartile of the overall state credibility index corresponding to the valid trial samples is calculated, and the lower quartile is taken as the preset state threshold; if the number of historical evaluation samples does not reach the preset sample number requirement, the initial state threshold set by the expert is adopted.

[0061] If the initial cognitive features do not meet the input adaptation conditions, recovery mode training is performed. After the recovery mode training is completed, cognitive data and state data are collected again to generate updated cognitive features and comprehensive recovery confidence. The comprehensive recovery confidence is used to determine whether the updated cognitive features meet the input adaptation conditions. If the updated cognitive features do not meet the input adaptation conditions, recovery mode training is repeated.

[0062] In a specific implementation, the method for performing recovery mode training includes:

[0063] The recovery mode training task is generated based on the current state awareness result. The training parameters of the recovery mode training task include training duration, training difficulty, and training content; as detailed below:

[0064] Training duration setting: A mapping relationship between preset state perception results and training duration is established, and the corresponding training duration is queried based on the current state perception result. The mapping relationship is set according to the optimal duration range for promoting state recovery through light activity in rehabilitation medicine. The lower the overall state reliability index, the worse the subject's current state, and the shorter the corresponding recovery training duration, to avoid adding extra burden in a fatigued state. The specific mapping relationship is implemented through a preset piecewise function or lookup table. The difference between the overall state reliability index and the preset state threshold is used as the state deviation, which is divided into multiple state deviation intervals in ascending order. Each state deviation interval corresponds to a recovery training duration. The interval containing the state deviation is obtained by the control unit, and the retrieved recovery training duration is used as the training duration for the recovery mode training task. The mapping relationship is stored in the form of a mapping table, which includes at least the state deviation interval, training duration, training content identifier, and training difficulty level. The control unit reads the corresponding training duration based on the state deviation interval containing the state deviation.

[0065] Training difficulty setting: Recovery mode is uniformly set to the lowest difficulty level.

[0066] Training content selection: Preset content selection rules are used to select corresponding training content based on the state perception results and the preset content selection rules. The method for setting the preset content selection rules is as follows: determined according to the recovery strategies corresponding to different types of fatigue in rehabilitation medicine. When the score of the physiological stability dimension in the comprehensive state credibility index is less than the preset physiological fatigue threshold, it is determined that physiological fatigue is the main factor, and relaxation and soothing training tasks are selected; otherwise, training tasks with fun and low cognitive load are selected.

[0067] After recovery mode training is completed, state and cognitive data are re-collected and analyzed to obtain the overall recovery confidence score. Methods for obtaining the overall recovery confidence score include:

[0068] After the recovery mode training is completed, state data and cognitive data are collected in real time. Recovery verification indicators are calculated based on the real-time collected state data and cognitive data, and a comprehensive recovery confidence score is obtained based on the recovery verification indicators. The recovery verification indicators include physiological recovery, training performance recovery, and subjective state recovery.

[0069] The method for calculating physiological recovery includes: after the recovery mode training is completed, an average heart rate sequence is collected, and the degree of heart rate decline is calculated, that is, the ratio of the average heart rate sequence to the baseline heart rate; a preset segmented mapping rule is used to map the degree of heart rate decline to a physiological recovery range of 0 to 1; the segment boundary in the segmented mapping rule is set according to the evaluation criteria of physiological heart rate recovery, for example, the degree of decline is divided into four levels: complete recovery, basic recovery, partial recovery, and no recovery, and each level corresponds to a different heart rate decline ratio range.

[0070] The method for calculating training performance recovery includes: comparing the performance scores of the final preset number of training tasks with the initial preset number of training tasks; the preset number of training tasks is determined based on the total number of training tasks and the minimum sample size that can reflect stable performance; calculating the difference between the average performance score of the final preset number of training tasks and the average performance score of the initial preset number of training tasks, dividing the score difference by the upper limit of the performance score to obtain the improvement coefficient, and clamping the improvement coefficient to the range of 0 to 1, that is, when the improvement coefficient is less than 0, the improvement coefficient is forced to 0, and when the improvement coefficient is greater than 1, the improvement coefficient is forced to 1; simultaneously calculating the average accuracy of the final preset number of training tasks and comparing it with the preset ideal accuracy threshold, when the average accuracy is greater than or equal to the ideal accuracy threshold, the accuracy achievement coefficient is 1; otherwise... The average accuracy rate is divided by the ideal accuracy rate threshold to obtain the accuracy rate achievement coefficient. The method for setting the preset ideal accuracy rate threshold is as follows: for each difficulty level training task, the task is completed by healthy subjects in the pre-experiment phase, and the lower quartile or 30th percentile of the accuracy rate distribution is taken as the ideal accuracy rate threshold. If there is no pre-experiment data, it can be set to 0.80 based on experience, indicating that when the average accuracy rate reaches above 80%, the training performance is considered to have recovered to an effective level. The improvement coefficient and the accuracy rate achievement coefficient are weighted and summed to obtain the training performance recovery degree. Each weight is a value greater than or equal to 0, and the sum of all weights in the same weighting calculation is 1. The method for setting the weight is that the ability improvement reflects the state recovery better than the current accuracy rate, so the improvement coefficient has a higher weight.

[0071] The method for calculating subjective state recovery includes: after recovery mode training, guiding subjects to conduct self-assessment of subjective load after training; calculating subjective recovery based on the upper limit of the score and the self-assessment score. Specifically, the difference between the upper limit of the score and the self-assessment score is calculated, and the ratio of the score difference to the upper limit of the score is calculated to obtain the subjective state recovery. The higher the self-assessment score, the greater the subjective load. The lower limit of the score is 0, and the upper limit of the score is the upper limit of the score.

[0072] The overall recovery confidence score is obtained by weighting physiological recovery, training performance recovery, and subjective state recovery. The weighting method is as follows: physiological recovery and training performance recovery are more objective and have higher weights; subjective state recovery is more affected by expressive ability and emotions and has lower weights, and the sum of all weights is 1.

[0073] Simultaneously, after the recovery mode training is completed, cognitive data is re-collected to generate updated cognitive features. The method for generating updated cognitive features is the same as that for the initial cognitive features, that is, the cognitive tasks corresponding to the first, second, third, and fourth cognitive dimensions are re-executed to obtain the scores of each cognitive dimension and concatenate them into updated cognitive features. When the preset standardized process has a shortened collection requirement, the shortened collection requirement retains at least the core tasks corresponding to the first, second, third, and fourth cognitive task types, and obtains updated cognitive features according to the same scoring calculation method as the initial cognitive features.

[0074] In specific implementations, the method for determining whether the updated cognitive features meet the adaptation input conditions based on the comprehensive recovery confidence score includes:

[0075] When the overall recovery confidence score is greater than or equal to the preset recovery verification threshold, the subject's state is determined to be recovered, the updated cognitive features are effective, the adaptation input conditions are met, and the recommended scheme adaptation is executed by calling the updated cognitive features, i.e., proceeding to the next step; when the overall recovery confidence score is less than the preset recovery verification threshold, the subject's state is determined to be unrecovered, and a prompt to continue resting is output. If the cumulative number of times the recovery mode training is executed has not reached the preset repetition limit, the recovery mode training is executed again after the preset rest end condition is reached, and the updated cognitive features and overall recovery confidence score are regenerated after the retraining is completed; if ..., the recovery mode training is executed again after the preset rest end condition is reached, and the updated cognitive features and overall recovery confidence score are regenerated after the retraining is completed. If the number of training iterations reaches the preset repetition limit and the overall recovery confidence is still less than the preset recovery verification threshold, a temporary adaptation delay prompt and a manual review flag are generated, and the currently updated cognitive features are not considered valid cognitive features. The preset repetition limit is set as follows: first, determine the total upper limit of the available training time for recovery patterns based on the allowed extension time of the preset standardized process and the subject's tolerance time; then, determine the duration of a single recovery training session based on the training time of the current recovery pattern training task; and use the integer part of the result obtained by dividing the total upper limit by the duration of a single recovery training session as the candidate repetition limit. When historical recovery samples exist, the statistical... The 90th percentile of the number of recovery mode training sessions required to reach the adaptation input conditions in the historical recovery samples is used, and the smaller value between the 90th percentile and the candidate repetition upper limit is taken as the preset repetition upper limit. When there are no historical recovery samples, the candidate repetition upper limit is used as the preset repetition upper limit, and the preset repetition upper limit is not less than 1. The method for setting the preset recovery verification threshold is as follows: collect historical recovery samples, each of which includes a comprehensive recovery confidence score and a corresponding adaptation trial validity label; samples that can complete the adaptation trial after recovery and whose trial data is confirmed as valid by the evaluator are taken as valid recovery samples, and the lower quartile of the comprehensive recovery confidence score corresponding to the valid recovery samples is calculated, and the lower quartile is used as the preset recovery verification threshold; if the number of historical recovery samples does not reach the preset sample number requirement, the initial recovery verification threshold set by the expert is adopted; the preset rest end condition is: the rest time reaches the preset rest time, or the deviation between the re-collected average heart rate and the baseline heart rate is less than the preset rest recovery deviation threshold; the preset rest time is determined according to the minimum interval requirement after recovery mode training, and the preset rest recovery deviation threshold is determined according to the heart rate deviation distribution when the heart rate in the historical recovery samples falls back to the state where training can continue.

[0076] The initial or updated cognitive features that meet the adaptation input conditions are taken as valid cognitive features, and the recommendation scheme is adapted by calling the valid cognitive features.

[0077] In a specific implementation, refer to Figure 2 The methods for implementing recommended scheme adaptation include:

[0078] Based on the comparison results of the scores of each cognitive dimension in the effective cognitive features with the corresponding cognitive dimension thresholds, the corresponding assistive device type is matched from the integrated resource platform. Specifically, based on mapping rules, the scores of each cognitive dimension in the effective cognitive features are compared with the corresponding cognitive dimension thresholds, and the corresponding recommended assistive devices are obtained from the integrated resource platform according to the comparison results, generating a recommended assistive device list. The control unit is configured with mapping rules between cognitive features and recommended assistive devices. The integrated resource platform provides the corresponding physical assistive devices according to the assistive device identifiers returned by the mapping rules. The control unit and the integrated resource platform are set up independently but are communicatively connected. The control unit is used to perform data acquisition, feature analysis, status judgment, and command sending, etc. The integrated resource platform is used to store assessment tools and fitting assistive devices, and provides the corresponding physical assistive devices according to the instructions of the control unit. The mapping rule setting method is as follows: based on the correspondence principle between functional defects and compensatory assistive devices in rehabilitation engineering, combined with the clinical assistive device fitting guidelines, the status of each cognitive dimension score being less than the corresponding cognitive dimension threshold is associated with the corresponding assistive device type. Specifically:

[0079] When the score of the first cognitive dimension is less than the preset threshold of the first cognitive dimension, the corresponding assistive device is recommended. For example, when the memory ability score is less than the preset threshold of memory function, assistive devices with reminder functions, such as smart pillboxes, voice memos and schedule reminders, are recommended, with the reason being that memory impairment requires external compensation.

[0080] When the score of the second cognitive dimension is less than the preset threshold of the second cognitive dimension, the corresponding assistive device is recommended. For example, when the score of attention ability is less than the preset threshold of attention function, assistive devices with environmental simplification or interference shielding functions are recommended, such as noise-canceling headphones and single-task operation panels.

[0081] When the score of the third cognitive dimension is less than the preset threshold of the third cognitive dimension, the corresponding auxiliary tool is recommended. For example, when the score of the execution function is less than the preset threshold of the execution function, the auxiliary tool with step guidance function is recommended, such as the task decomposition prompter and the sequential operation guidance device.

[0082] When the score of the fourth cognitive dimension is less than the preset threshold of the fourth cognitive dimension, the corresponding assistive device is recommended. For example, when the language ability score is less than the preset language function threshold, assistive devices with alternative communication functions are recommended, such as picture communication boards and voice output devices.

[0083] The thresholds for each cognitive dimension are set as follows: the threshold for each cognitive dimension is determined based on the upper limit of the corresponding cognitive dimension score, the functional impairment threshold requirements in the clinical assessment scale, and the changes in the success rate of fitting in historical fitting data; when the score of a certain cognitive dimension is less than the corresponding cognitive dimension threshold, it is determined that there is a functional gap in that cognitive dimension that requires assistive device compensation; in addition, collaborative filtering recommendations can be made based on historical fitting data. When the similarity between the cognitive characteristics of a new subject and a case in the historical case library exceeds the preset similarity threshold, the successful fitting scheme of that case can be referred to for recommendation; the similarity calculation uses cosine similarity or Euclidean distance, and the preset similarity threshold is set by taking an empirical value that makes the recommendation accuracy reach a clinically acceptable level.

[0084] The assessor provides the assistive device for the subject to try out, and data on the success rate, efficiency, and subjective acceptance are collected during the trial. Specifically, objective data are collected during the subject's trial of the assistive device through the following methods.

[0085] If the assistive device is equipped with sensors, the system automatically records the success rate and the time taken for each operation; if the assistive device is a sensorless ordinary device, the assessor manually records the following through the timer and counter on the control unit: the number of times the subject independently completes the core operation without prompting, the time taken for each operation, and whether prompting or limb assistance is required.

[0086] Simultaneously, nonverbal behaviors of the subjects during the trial were recorded. Facial expressions were used to generate facial expression scores, while whether the subjects actively expressed their willingness to use the product, whether they refused to continue using it, and whether they needed repeated prompts were used to generate operational observation scores.

[0087] Based on data on operation success rate, operation efficiency, and subjective acceptance, the system automatically calculates the adaptation level using preset judgment rules. These rules are weighted according to the following three dimensions:

[0088] The number of operations successfully completed independently by the subject within a preset number of trial tasks and the total number of attempts are obtained. The ratio of the number of independently successful operations to the total number of attempts is calculated to obtain the raw success rate. A minimum acceptable success rate threshold and an ideal success rate threshold are preset, with the minimum acceptable success rate threshold being less than the ideal success rate threshold. The method for setting the preset number of trial tasks is as follows: the core operation set is determined according to the standard usage process of the assistive device to be tested, and each independent operation step necessary to complete the main compensation function of the assistive device is defined as a trial task. When the same core operation needs to be repeatedly verified, the number of repetitions is determined according to the minimum number of repetitions required for the success rate of the core operation to stabilize in the historical trial records. Specifically, the independent completion success rate of the core operation is calculated cumulatively according to the trial sequence. When the difference between two consecutive cumulative success rates is less than the success rate stabilization judgment value, the corresponding cumulative number is determined as the number of repetitions of the core operation. The sum of the number of repetitions corresponding to each core operation is the preset number of trial tasks. If there are no historical trial records, the number of repetitions is determined according to the assistive device instruction manual or... The core operation count specified in the clinical trial process determines the preset trial task count. The method for setting the minimum acceptable success rate threshold is as follows: in historical trial samples, samples with safety interruptions, requiring full prompts, or unable to complete the core operation are considered invalid trial samples. The upper quartile of the original success rate of invalid trial samples is used as the minimum acceptable success rate threshold. The method for setting the ideal success rate threshold is as follows: in historical trial samples, samples that can complete the core operation independently without prompts are considered valid trial samples. The lower quartile of the original success rate of valid trial samples is used as the ideal success rate threshold. When the original success rate is not less than the ideal success rate threshold, the operation success rate score is 1; when the original success rate is not greater than the minimum acceptable success rate threshold, the operation success rate score is 0. In other cases, the difference between the original success rate and the minimum acceptable success rate threshold is calculated and recorded as the first difference; the difference between the ideal success rate threshold and the minimum acceptable success rate threshold is calculated and recorded as the second difference; the ratio of the first difference to the second difference is calculated to obtain the operation success rate score.

[0089] The average time taken by a subject for a single operation was obtained, and a baseline time for the assistive device was preset. The baseline time for the assistive device was determined by taking the average time taken by healthy subjects to complete the core operation for the first time using the corresponding assistive device in a standard environment. Simultaneously, a maximum acceptable time was preset. The maximum acceptable time was determined by multiplying the baseline time of the assistive device by a preset allowable extension factor. The preset allowable extension factor was determined based on the 90th percentile of the ratio of the single operation time to the baseline time in historical valid trial samples. When the average time taken for a single operation is less than or equal to the assistive device's baseline time... When the baseline time is reached, the operational efficiency score is 1; when the average time of a single operation is greater than or equal to the maximum acceptable time, the operational efficiency score is 0; when the average time of a single operation is greater than the baseline time of the assistive device but less than the maximum acceptable time, the difference between the average time of a single operation and the baseline time of the assistive device is calculated and recorded as the third difference; the difference between the maximum acceptable time and the baseline time of the assistive device is calculated and recorded as the fourth difference; the ratio of the third difference to the fourth difference is calculated and recorded as the efficiency ratio; the difference between the value 1 and the efficiency ratio is calculated to obtain the operational efficiency score.

[0090] Multiple frames of facial images were captured during the trial using a camera. A pre-trained expression recognition algorithm categorized these images as positive, neutral, or negative, and assigned scores accordingly. A positive expression received a higher score than a neutral one, and vice versa. All scores ranged from 0 to 1. The corresponding expression value was then calculated based on the expression category of each frame, and the arithmetic mean of these values ​​was obtained, resulting in a facial expression score normalized to 0-1. Evaluators selected three levels of acceptance—high, moderate, and low—on the control unit based on whether the user actively expressed liking or refused to continue using the product. Scoring is performed, with high acceptance values ​​higher than average acceptance values, and average acceptance values ​​higher than low acceptance values. High, average, and low acceptance values ​​are all within the range of 0 to 1. Based on the subject's actual reaction, corresponding acceptance values ​​are read to obtain an operational observation score normalized to the 0-1 range. The facial expression score and the operational observation score are weighted and summed to obtain the subjective acceptance score. Each weight is a value greater than or equal to 0, and the sum of all weights in the same weighting calculation is 1. The weights can be set and adjusted based on clinical experience. The expression recognition algorithm can use an existing trained classification model. This application does not improve the expression recognition algorithm itself, but only uses its output expression category as the input to the subjective acceptance score.

[0091] The overall fitting score is obtained by weighting and summing the operation success rate score, operation efficiency score, and subjective acceptance score. Each weight is a value greater than or equal to 0, and the sum of all weights in the same weighted calculation is 1. The weights are set according to the importance of each indicator in the clinical assistive device fitting assessment. The operation success rate directly reflects the functional compensation effect and has the highest weight; the operation efficiency reflects the ease of use and has the second highest weight; the subjective acceptance is greatly affected by individual preferences and has the lowest weight.

[0092] The first fitting threshold is greater than the second fitting threshold; when the overall fitting score is greater than or equal to the first fitting threshold, it is judged as a good fit; when the overall fitting score is less than the second fitting threshold, it is judged as an unsuitable fit; when the overall fitting score is greater than or equal to the second fitting threshold and less than the first fitting threshold, it is judged as a moderate fit; the method for setting the first fitting threshold and the second fitting threshold is: determined according to the commonly used dividing points for distinguishing effective fit, partial fit and ineffective fit in clinical practice.

[0093] The automatically generated fitting results are displayed on the control unit, where assessors can review and correct them based on clinical experience. If an assessor determines that the fitting results do not match the actual situation, they can modify the fitting level and fill in the reason for the correction. When the assessor corrects the fitting level, the original collected data and the fitting comprehensive score automatically calculated by the system are not modified; only the corrected fitting level, the reason for the correction, and the correction time are generated. The correction record is stored in the database for subsequent optimization of the judgment rules. The database stores at least the subject identifier, collection time, effective cognitive characteristics, state perception results, comprehensive recovery confidence, recommended assistive device identifier, trial data, fitting results, and correction records.

[0094] After the assessor confirms the fitting result, corresponding follow-up processing is performed based on different fitting results. Specifically, when the fitting result is good: the corresponding assistive device is marked as successfully fitted, and fitting success information, such as assistive device type, fitting time, trial duration, and overall fitting score, is recorded. The fitting result is used as input for determining rehabilitation mode training parameters. This successful fitting case is also stored in the historical case database for subsequent optimization of collaborative filtering recommendations. When the fitting result is average: it is determined that the corresponding assistive device partially meets the requirements but has deficiencies. At this time, the system performs the following operations: automatically classifying... The system analyzes the shortcomings in each dimension of the overall adaptation score: if the success rate is low, it records the difficulty of operation; if the efficiency is low, it records the slowness of operation; if the subjective acceptance is low, it records the low acceptance. Based on the cause of the shortcomings, the system automatically searches for alternative solutions in the list of recommended assistive devices. For example, if the cause is difficulty or slowness of operation, it recommends an assistive device of the same function but with simpler operation; if the cause is low acceptance, it recommends an alternative model of the same function but with a different appearance or interaction method. The evaluator takes out the alternative assistive device provided by the system, tries it again, and returns for re-adaptation.

[0095] If the results of the alternative solutions are all only moderately suitable after a series of preset attempts, the subject will be marked as needing expert evaluation, and a recommendation will be given to refer the subject to a rehabilitation expert for personalized adaptation. The method for setting the number of consecutive preset attempts is based on the reasonable number of attempts to try alternative assistive devices of the same functional type in clinical practice.

[0096] When the adaptation result is "incompatible," it is determined that the corresponding assistive device does not meet the requirements at all; at this time, the system performs the following operations:

[0097] The system automatically analyzes the main reasons for mismatch, including at least one of the following: operation success rate score of 0, operation efficiency score of 0, and subject's explicit refusal to continue use.

[0098] The corresponding assistive device is added to the subject's prohibited recommendation list and will not be recommended again in subsequent recommendations. Other assistive devices with the same functional type but significantly different designs are automatically retrieved from the recommended assistive device list as alternatives. If an alternative exists, the assessor will select it for trial. If all available assistive devices for the corresponding functional type have been tried and all results are unsuitable, the system determines that the adaptation requirements for the corresponding functional dimension cannot be met by the existing resource platform, outputs a prompt that no assistive devices are available for the corresponding functional dimension, and suggests referral to customized assistive device services. Regardless of whether the adaptation is moderate or unsuitable, the adaptation results, overall adaptation score, and scores for each dimension are stored in the database for optimization of the integrated resource platform.

[0099] The system calculates the success rate of each assistive device, which is the number of successful fittings divided by the total number of trials. When the success rate is less than the preset success rate threshold, the system lowers the priority of the corresponding assistive device in the recommendation list or marks the corresponding assistive device as needing review. The preset success rate threshold is set by taking the lowest success rate standard that is considered to have clinical value in clinical practice. The specific threshold can be adaptively calibrated based on the statistical analysis of historical fitting data.

[0100] The distribution of the main reasons for poor or inadequate fit under different cognitive characteristics is analyzed to improve assistive device recommendation rules, such as adjusting the value of preset function thresholds and adding multi-dimensional feature joint recommendation.

[0101] For subjects marked as requiring expert evaluation, their complete evaluation data and fitting attempt records are stored in the database for subsequent optimization of assistive device recommendation rules.

[0102] After the recommended treatment plan is adapted, the system can train the subject in rehabilitation mode; refer to Figure 3During rehabilitation training, the training parameters include the type of training task, difficulty level, training frequency, and duration of each training session. Methods for obtaining these training parameters include:

[0103] The training task type is determined based on the cognitive dimension with the lowest score among the effective cognitive features. If multiple dimensions are tied for the lowest score, the training task type is selected according to the preset priority. The preset priority is determined based on the degree of influence of each cognitive domain on daily living abilities in cognitive rehabilitation theory, with memory ability usually having the highest priority.

[0104] Based on the lowest score of the cognitive dimension among the effective cognitive features and the functional requirement level in the adaptation results, the corresponding difficulty level is determined through a preset difficulty mapping rule. The method for obtaining the functional requirement level in the adaptation results includes: determining the functional requirement level based on the overall adaptation score, the adaptation level, and the number of cognitive dimensions below the cognitive dimension threshold; when the overall adaptation score is below the second adaptation threshold or the number of cognitive dimensions below the cognitive dimension threshold reaches a preset number, the functional requirement level is determined to be fully dependent; when the overall adaptation score is greater than or equal to the second adaptation threshold and less than the first adaptation threshold, the functional requirement level is determined to be partially auxiliary; when the overall adaptation score is greater than or equal to the first adaptation threshold, the functional requirement level is determined to be partially auxiliary. The functional requirement level is set as independent completion. The method for setting the preset difficulty mapping rule is as follows: the lowest score in the cognitive dimension is divided into multiple score intervals, and each score interval corresponds to a basic difficulty level. The basic difficulty level is then adjusted according to the functional requirement level. Specifically, the degree of functional limitation decreases in the order of complete dependence, partial assistance, and independent completion. The initial difficulty corresponding to complete dependence is lower than that of partial assistance, and the initial difficulty corresponding to partial assistance is lower than that of independent completion. The control unit determines the basic difficulty level based on the score interval of the lowest score in the cognitive dimension, and adjusts, maintains, or lowers the basic difficulty level according to the functional requirement level to obtain the difficulty level of the rehabilitation mode training task.

[0105] The difficulty level is dynamically adjusted using the following methods: Real-time data collection of the subject's accuracy and reaction time for each training task is performed. The performance score is obtained by multiplying the weighted sum of the accuracy and reaction time contributions by a pre-defined weighted sum, where each weight is greater than or equal to 0, and the sum of all weights in the same weighted calculation is 1. When the performance score for a consecutive preset number of tasks is not lower than the upward adjustment threshold, the difficulty level is increased by a preset unit. When the performance score for a consecutive preset number of tasks is not higher than the downward adjustment threshold, the difficulty level is decreased by a preset unit. When the performance score for a consecutive preset number of tasks does not meet either the upward or downward adjustment conditions, the current difficulty level remains unchanged. The methods for setting the upward and downward adjustment thresholds, the consecutive preset number of tasks, and the preset units are determined based on the sensitivity requirements of difficulty adjustment, ensuring that the subject has truly mastered the current difficulty level before upgrading.

[0106] The training frequency and duration of each training session are preset for each rehabilitation cycle. The training frequency for each rehabilitation cycle is set according to the training frequency recommended in cognitive rehabilitation medicine, combined with the subject's tolerance and daily schedule. Initially, a general recommended value is used, and subsequent personalized adjustments can be made based on the subject's training compliance and effect feedback. The duration of each training session is set according to the state perception results of the first assessment. The better the state, the longer the training can be performed.

[0107] After completing a preset cycle of rehabilitation training, cognitive data is re-collected, and reassessment cognitive features are generated according to the method for generating effective cognitive features. The reassessment cognitive features are compared dimension by dimension with the effective cognitive features before entering the rehabilitation mode to obtain the rehabilitation effect, i.e., the change in each cognitive dimension. The method for obtaining the rehabilitation effect includes: calculating the change in each cognitive dimension between the reassessment cognitive features and the initial cognitive features; comparing each change with a preset effective improvement threshold, and counting the number of cognitive dimensions that reach or exceed the effective improvement threshold; when the number of dimensions reaches the preset number requirement, the rehabilitation is deemed effective; when the number of cognitive dimensions does not reach the preset number requirement, the rehabilitation is deemed average. The preset cycle is set according to the minimum intervention cycle usually required in cognitive rehabilitation training. The effective improvement threshold is set according to the minimum clinical significance difference of the cognitive ability assessment scale. The preset number requirement is set according to the total number of cognitive dimensions, and when the number of cognitive dimensions below the cognitive dimension threshold reaches more than half of the total number of cognitive dimensions, the preset number is determined to have been reached.

[0108] Example 2:

[0109] Please see Figure 4 As shown, this embodiment provides a state awareness and adaptation system based on multi-source heterogeneous data fusion, including:

[0110] The data acquisition and analysis module is used to collect the heart rate data of the subjects before the execution of the preset standardized procedure, and to collect state data and cognitive data during the execution of the preset standardized procedure, and generate baseline heart rate, state characteristics and initial cognitive characteristics respectively.

[0111] The status analysis module is used to compare and analyze baseline heart rate with status characteristics to generate status perception results.

[0112] The first decision module is used to determine whether the initial cognitive features meet the adaptation input conditions based on the state perception results.

[0113] The second decision module is used to perform recovery mode training if the initial cognitive features do not meet the adaptation input conditions. After the recovery mode training is completed, cognitive data and state data are collected again to generate updated cognitive features and comprehensive recovery confidence. The module determines whether the updated cognitive features meet the adaptation input conditions based on the comprehensive recovery confidence. If the updated cognitive features do not meet the adaptation input conditions, the recovery mode training is repeated.

[0114] The scheme adaptation module is used to take the initial or updated cognitive features that meet the adaptation input conditions as effective cognitive features and call the effective cognitive features to perform the recommended scheme adaptation.

[0115] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A state awareness and adaptation method based on multi-source heterogeneous data fusion, characterized in that, include: Heart rate data of subjects were collected before the execution of a preset standardized procedure, and state data and cognitive data were collected during the execution of the preset standardized procedure to generate baseline heart rate, state characteristics and initial cognitive characteristics, respectively. A state perception result is generated by comparing and analyzing the baseline heart rate with the state characteristics. Based on the state perception results, determine whether the initial cognitive features meet the adaptation input conditions; If the initial cognitive features do not meet the input adaptation conditions, recovery mode training is performed, and cognitive and state data are re-collected after the recovery mode training is completed to generate updated cognitive features and comprehensive recovery confidence. The comprehensive recovery confidence is used to determine whether the updated cognitive features meet the input adaptation conditions. If the updated cognitive features do not meet the input adaptation conditions, recovery mode training is repeated. The initial or updated cognitive features that meet the adaptation input conditions are taken as valid cognitive features, and the recommendation scheme is adapted by calling the valid cognitive features.

2. The state awareness and adaptation method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, Methods for obtaining state-aware results include: The physiological stability dimension score is obtained by calculating and mapping the average heart rate sequence in the state characteristics to the baseline heart rate. The task time ratio and the number of unplanned interruptions in the status characteristics are calculated and mapped to obtain the operational load dimension score; Analyze and map the task switching duration in the state features to obtain the task coherence dimension score; The scores of physiological stability, operational load, and task coherence are weighted and summed according to preset weights to obtain a comprehensive state credibility index, which is then used as the state perception result.

3. The state awareness and adaptation method based on multi-source heterogeneous data fusion according to claim 2, characterized in that, The method for determining whether the initial cognitive features meet the adaptation input conditions includes: If the overall state credibility index is greater than or equal to the preset state threshold, the initial cognitive features are determined to meet the adaptation input conditions; otherwise, the initial cognitive features are determined not to meet the adaptation input conditions.

4. The state awareness and adaptation method based on multi-source heterogeneous data fusion according to claim 3, characterized in that, Methods for performing recovery mode training include: Based on the current state perception results, training tasks with different training parameters in the recovery mode are generated; the training parameters include training duration, training difficulty, and training content. Methods for obtaining different training parameters in recovery mode include: Based on the state perception results, the training duration, training difficulty, and training content of the recovery mode are set; the physiological stability dimension score obtained during the generation of the comprehensive state credibility index is read, the physiological stability dimension score is compared with the preset physiological fatigue threshold, and the training content is set according to the comparison results.

5. The state awareness and adaptation method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, Methods for obtaining the overall recovery confidence include: The ratio of the average heart rate sequence after recovery mode training to the baseline heart rate is mapped to the physiological recovery degree. The accuracy achievement coefficient is calculated based on the average accuracy of the last preset number of tasks before the end of recovery mode training and the preset ideal accuracy threshold; the improvement coefficient is calculated based on the difference between the average performance score of the last preset number of tasks before the end of recovery mode training and the average performance score of the first preset number of tasks after the start of recovery mode training; the training performance recovery degree is obtained based on the accuracy achievement coefficient and the improvement coefficient. The self-assessment score after recovery mode training is linearly mapped to the subjective state recovery degree. The comprehensive recovery confidence score is obtained by weighting and summing the physiological recovery rate, training performance recovery rate, and subjective state recovery rate.

6. The state awareness and adaptation method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, Methods for determining whether updated cognitive features meet the input adaptation conditions based on comprehensive recovery confidence include: When the overall recovery confidence is greater than or equal to the preset recovery verification threshold, it is determined that the updated cognitive features meet the adaptation input conditions; when the overall recovery confidence is less than the preset recovery verification threshold, it is determined that the updated cognitive features do not meet the adaptation input conditions.

7. The state awareness and adaptation method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The methods for implementing recommended scheme adaptation include: Effective cognitive features include scores for the first cognitive dimension, the second cognitive dimension, the third cognitive dimension, and the fourth cognitive dimension; based on the comparison results of each cognitive dimension score with the corresponding cognitive dimension threshold, assistive device types for the corresponding dimensions are matched from the integrated resource platform; Assistive devices corresponding to the type of assistive device were provided to the subjects for trial use, and data on operation success rate, operation efficiency, and subjective acceptance were collected during the trial. The adaptation results are obtained by analyzing data on success rate, efficiency, and subjective acceptance, combined with preset judgment rules.

8. The state awareness and adaptation method based on multi-source heterogeneous data fusion according to claim 7, characterized in that, After adapting the recommended program, the subjects were also trained in a rehabilitation mode. The training parameters of the rehabilitation mode included the type of training task, difficulty level, training frequency, and duration of a single training session. Methods for obtaining training parameters for rehabilitation patterns include: The training task type is determined based on the cognitive dimension with the lowest score among the effective cognitive features; the corresponding difficulty level is determined based on the score of the cognitive dimension with the lowest score and the functional requirement level in the adaptation results through a preset difficulty mapping rule. Preset training frequency and training duration.

9. The state awareness and adaptation method based on multi-source heterogeneous data fusion according to claim 8, characterized in that, The methods for dynamically adjusting the difficulty level include: The accuracy and reaction time of each subject performing a training task are collected in real time, and a performance score is calculated. When the performance score of a series of tasks is greater than or equal to the threshold for adjustment, the difficulty level will be increased by a preset unit. When the performance score of a task is less than or equal to the reduction threshold for a consecutive preset number of tasks, the difficulty level will be reduced by a preset unit. If the performance score of a task for a consecutive preset number of times is less than the upward adjustment threshold, or if the performance score of a task for a consecutive preset number of times is greater than the downward adjustment threshold, the current difficulty level remains unchanged.

10. A state awareness and adaptation system based on multi-source heterogeneous data fusion, implementing the state awareness and adaptation method based on multi-source heterogeneous data fusion as described in any one of claims 1-9, characterized in that, include: The data acquisition and analysis module is used to collect heart rate data of the subjects before the execution of the preset standardized procedure, and to collect state data and cognitive data during the execution of the preset standardized procedure, and to generate baseline heart rate, state characteristics and initial cognitive characteristics, respectively. The state analysis module is used to compare and analyze the baseline heart rate with the state characteristics to generate state perception results. The first decision module is used to determine whether the initial cognitive features meet the adaptation input conditions based on the state perception results. The second decision module is used to perform recovery mode training if the initial cognitive features do not meet the adaptation input conditions, and to re-collect cognitive data and state data after the recovery mode training is completed, and generate updated cognitive features and comprehensive recovery confidence. Based on the comprehensive recovery confidence, it is determined whether the updated cognitive features meet the adaptation input conditions. If the updated cognitive features do not meet the adaptation input conditions, the recovery mode training is repeated. The scheme adaptation module is used to take the initial or updated cognitive features that meet the adaptation input conditions as effective cognitive features and call the effective cognitive features to perform the recommended scheme adaptation.

Images