Work behavior feature driven rehabilitation training effect dynamic evaluation method

Abstract

The application belongs to the technical field of artificial intelligence, and discloses a rehabilitation training effect dynamic evaluation method driven by work behavior characteristics, multi-dimensional operation behavior data and physiological sensor data of a patient on a simulated driving device are acquired, compensatory behavior characteristics and implicit behavior mode characteristics are extracted, and a behavior mode migration characteristic set is constructed. Based on the compensatory behavior characteristics in the characteristic set, a behavior mode migration index is calculated, and the gradual conversion process of the patient from compensatory behavior to normal behavior is accurately identified. Further, a rehabilitation ability dynamic evaluation model is constructed, real-time rehabilitation ability scores are generated, the complexity of the simulated driving scene is dynamically adjusted, adaptive behavior data are collected, a behavior mode migration trend curve and a rehabilitation effect prediction model are constructed, and closed-loop optimization of a personalized training scheme is realized. The method breaks through the subjectivity and staticity limitation of traditional rehabilitation evaluation, and significantly improves the accuracy and effectiveness of neural rehabilitation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, more specifically, the present application relates to a rehabilitation training effect dynamic evaluation method driven by work behavior characteristics. BACKGROUND

[0002] With the development of neurorehabilitation medicine, simulated driving environment as a means of rehabilitation training has been more and more widely used in clinical practice. For patients with systemic diseases, regaining the ability to drive not only relates to their life independence, but also is an important guarantee for social participation and mental health. Driving, as a complex work behavior, involves the integration of visual perception, attention allocation, cognitive judgment, and motor coordination, and therefore becomes an ideal carrier for evaluating the recovery degree of patients' neural function.

[0003] At present, various simulated driving devices have been used in clinical rehabilitation for training, ranging from simple desktop controllers to highly immersive virtual cockpits. These devices can provide patients with a driving situation close to reality, allowing them to practice driving skills in a safe environment. Traditional evaluation methods mainly rely on therapists' on-site observation combined with simple driving performance indicators (such as task completion time, error frequency, etc.), and some advanced centers will combine standardized neuropsychological assessment scales for auxiliary judgment.

[0004] However, the existing rehabilitation training evaluation system cannot effectively identify the gradual transition process of patients from compensatory behavior to normal behavior. In clinical practice, patients with nervous system damage often develop a series of compensatory behavior strategies to complete daily tasks, such as hemiplegic patients using the non-damaged side of the body to overcompensate, brain injury patients using non-standard action sequences to complete operation tasks, or cervical spondylosis patients rotating their trunks instead of their necks during driving, etc. Although these compensatory behaviors can help patients regain functional independence in the short term, they hinder neural remodeling and the reconstruction of normal motor patterns in the long term. Traditional evaluation methods mainly rely on doctors' intermittent observation and subjective scoring scales, and cannot capture the subtle changes and gradual decline of patients' compensatory behaviors during rehabilitation. More importantly, existing technologies lack the ability to correlate physiological sensor data with operational behavior, making it difficult to reveal changes in cognitive load and neural control strategy changes hidden beneath surface behavior. For example, in simulated driving rehabilitation training, patients may have completed the operation task on the surface, but the system cannot detect changes in implicit indicators such as heart rate variability, skin electrical response, and muscle electrical activity when completing the task, making it difficult to accurately determine whether the patient continues to rely on compensatory strategies or has begun to restore normal neural control mechanisms. The lack of this fine-grained behavior pattern migration evaluation mechanism makes it difficult for clinicians to determine the optimal intervention time and adjust rehabilitation strategies, often leading to the solidification of compensatory behaviors or low training efficiency, which seriously affects the rehabilitation effect and the quality of life of patients.

[0005] In view of this, the application proposes a rehabilitation training effect dynamic evaluation method driven by work behavior characteristics to solve the above problems. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the application provides the following technical scheme: a rehabilitation training effect dynamic evaluation method driven by work behavior characteristics, comprising:

[0007] Obtain multi-dimensional operation behavior data of a patient on a simulated driving device and synchronously collected physiological sensor data;

[0008] According to the time sequence distribution and force variation of the operation action in the multi-dimensional operation behavior data, extract compensatory behavior characteristics; according to the fluctuation characteristics of the physiological indicators in the physiological sensor data, extract implicit behavior mode characteristics; based on the compensatory behavior characteristics and the implicit behavior mode characteristics, construct a behavior mode migration feature set;

[0009] According to the attenuation trend of the compensatory behavior characteristics and the enhancement trend of the implicit behavior mode characteristics in the behavior mode migration feature set, calculate a behavior mode migration index; based on the behavior mode migration index, identify the gradual transition process of the patient from compensatory behavior to normal behavior;

[0010] According to the behavior mode migration index and a preset health population behavior mode benchmark, construct a patient rehabilitation ability dynamic evaluation model; based on the patient rehabilitation ability dynamic evaluation model, generate a real-time rehabilitation ability score;

[0011] According to the real-time rehabilitation ability score, design an adaptive adjustment strategy of a virtual driving scene, dynamically adjust the complexity parameters of the simulated driving scene, and simultaneously collect adaptive behavior data of the patient in the adjusted simulated driving scene;

[0012] According to the adaptive behavior data and the behavior mode migration index, construct a behavior mode migration trend curve; based on the behavior mode migration trend curve, generate a rehabilitation effect prediction model;

[0013] According to the rehabilitation effect prediction model and the behavior mode migration trend curve, generate an individualized rehabilitation training report, and optimize the training scheme parameters of the next stage, to realize closed-loop dynamic optimization of rehabilitation training.

[0014] Preferably, the extraction of compensatory behavior characteristics comprises:

[0015] Segment the time sequence distribution of the operation action in the multi-dimensional operation behavior data, and obtain the frequency, force and duration of the operation action in each time period;

[0016] According to the frequency, intensity and duration of the operation action in each time period, a smoothness index of the operation action is calculated; the smoothness index is used to represent the coordination of the operation action;

[0017] According to the deviation of the smoothness index from a preset health population operation action smoothness threshold, an abnormal fluctuation feature of the operation action is extracted;

[0018] According to the distribution density and duration of the abnormal fluctuation feature, a compensatory behavior feature is determined; the compensatory behavior feature includes the unnecessary repetition number and intensity overload ratio of the operation action.

[0019] Preferably, the extraction of the implicit behavior pattern feature includes:

[0020] The fluctuation characteristics of the physiological indicators in the physiological sensor data are subjected to time-frequency analysis, and the frequency energy distribution and time domain fluctuation amplitude of the physiological indicators in each time period are obtained;

[0021] According to the concentration of the frequency energy distribution and the stability of the time domain fluctuation amplitude, a dynamic balance index of the physiological indicators is calculated; the dynamic balance index is used to represent the potential cognitive load of the patient in the operation behavior;

[0022] According to the difference between the dynamic balance index and a preset health population dynamic balance benchmark, an implicit behavior pattern feature is extracted; the implicit behavior pattern feature includes a cognitive load overload feature and a physiological stress response feature.

[0023] Preferably, the calculation of the behavior pattern migration index includes:

[0024] The attenuation trend of the compensatory behavior feature is quantified, and an attenuation rate of the compensatory behavior feature is obtained; the attenuation rate is obtained by calculating the change slope of the compensatory behavior feature in consecutive time periods;

[0025] The enhancement trend of the implicit behavior pattern feature is quantified, and an enhancement rate of the implicit behavior pattern feature is obtained; the enhancement rate is obtained by calculating the change slope of the implicit behavior pattern feature in consecutive time periods;

[0026] According to the attenuation rate of the compensatory behavior feature and the enhancement rate of the implicit behavior pattern feature, a behavior pattern migration index is calculated; the behavior pattern migration index is a weighted ratio of the attenuation rate and the enhancement rate, wherein the weight is determined by the influence factor of the compensatory behavior feature and the implicit behavior pattern feature on the rehabilitation ability.

[0027] Preferably, the construction of the patient rehabilitation ability dynamic evaluation model includes:

[0028] According to the behavior pattern migration index, a behavior pattern migration state space is constructed; the behavior pattern migration state space includes a compensatory behavior dominant state, a transition state and a normal behavior dominant state;

[0029] According to the distance between the state currently occupied by the patient in the behavior pattern migration state space and the health population behavior pattern benchmark, a rehabilitation ability dynamic score is calculated; the distance is calculated by Euclidean distance;

[0030] According to the time series change of the rehabilitation ability dynamic score, a patient rehabilitation ability dynamic evaluation model is constructed; the patient rehabilitation ability dynamic evaluation model uses a long short-term memory network to model the time series change.

[0031] Preferably, the adaptive adjustment strategy for designing a virtual driving scene includes:

[0032] According to the real-time rehabilitation ability score, the adjustment direction of the complexity parameter of the simulated driving scene is determined; the complexity parameter includes road curvature, obstacle density and traffic flow;

[0033] According to the behavior pattern migration index, the adjustment amplitude of the complexity parameter of the simulated driving scene is determined; the adjustment amplitude is positively correlated with the absolute value of the behavior pattern migration index;

[0034] According to the adjustment direction and the adjustment amplitude, an adaptive adjustment strategy for a virtual driving scene is generated; the adaptive adjustment strategy includes preferentially reducing road curvature and obstacle density when the behavior pattern migration index is lower than a preset migration threshold.

[0035] Preferably, the behavior pattern migration trend curve is constructed, including:

[0036] The adaptive behavior data is subjected to cluster analysis to obtain a behavior pattern migration subset of the patient under different complexity parameters; the behavior pattern migration subset includes a decay subset of compensatory behavior characteristics and an enhancement subset of implicit behavior pattern characteristics;

[0037] According to the time series change of the behavior pattern migration subset, a behavior pattern migration trend curve is fitted; the behavior pattern migration trend curve is fitted using a polynomial regression model;

[0038] According to the curvature change of the behavior pattern migration trend curve, an inflection point of behavior pattern migration is determined; the inflection point is used to represent a significant transition stage of the patient from compensatory behavior to normal behavior.

[0039] Preferably, the rehabilitation effect prediction model is generated, including:

[0040] According to the behavior mode migration trend curve, long-term trend characteristics and short-term fluctuation characteristics of the trend curve are extracted; the long-term trend characteristics are obtained by low-pass filtering, and the short-term fluctuation characteristics are obtained by high-pass filtering;

[0041] According to the long-term trend characteristics and the short-term fluctuation characteristics, a rehabilitation effect prediction model is constructed; the rehabilitation effect prediction model adopts an autoregressive integrated moving average model to jointly model the long-term trend characteristics and the short-term fluctuation characteristics;

[0042] According to the output of the rehabilitation effect prediction model, a behavior mode migration index and a rehabilitation ability score of the patient in a future training period are predicted.

[0043] Preferably, the training scheme parameters of the next stage are optimized, including:

[0044] According to the insufficient items of the behavior mode migration index in the individualized rehabilitation training report, an optimization direction of the training scheme parameters is determined; the insufficient items include insufficient attenuation of compensatory behavior characteristics and insufficient enhancement of implicit behavior mode characteristics;

[0045] According to the prediction result of the rehabilitation effect prediction model, an optimization amplitude of the training scheme parameters is determined; the optimization amplitude is positively correlated with the promotion space of the predicted behavior mode migration index;

[0046] According to the optimization direction and the optimization amplitude, the training scheme parameters of the next stage are adjusted; the training scheme parameters include complexity parameters of a simulated driving scene, training duration and training frequency.

[0047] Preferably, the compensatory behavior characteristics are determined, including:

[0048] The distribution density of the abnormal fluctuation characteristics is kernel density estimation, and the probability density function of the abnormal fluctuation characteristics is obtained;

[0049] According to the number of peak values and the peak value spacing of the probability density function, the mode type of the compensatory behavior characteristics is determined; the mode type includes single mode compensation and multi-mode compensation;

[0050] According to the mode type and the duration of the abnormal fluctuation characteristics, the saliency score of the compensatory behavior characteristics is calculated; the saliency score is used to represent the contribution degree of the compensatory behavior characteristics to the rehabilitation ability evaluation.

[0051] The technical effects and advantages of the work behavior characteristic driven rehabilitation training effect dynamic evaluation method of the application are:

[0052] The present application can accurately identify the gradual transition process of the patient from the compensatory strategy to the normal behavior mode, so that the clinician can grasp the golden intervention opportunity, avoid the solidification of the compensatory behavior and the secondary damage caused thereby, such as muscle and bone imbalance and abnormal posture control, and the like. This evaluation method breaks through the limitations of traditional subjective evaluation, realizes the visualization and quantification of the rehabilitation process, and significantly enhances the confidence and compliance of the patient to the rehabilitation treatment. In addition, the present application dynamically adjusts the training difficulty to keep the patient always in the "zone of proximal development", which not only avoids the rehabilitation stagnation caused by insufficient training intensity, but also prevents the frustration and secondary damage caused by excessive training, and optimizes the rehabilitation efficiency and patient experience. Most importantly, the present application accurately guides the neural remodeling process to help the patient recover to a functional mode closer to normal rather than relying on compensatory strategies, which significantly improves the long-term quality of life and social participation ability of the patient. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 A step flowchart of the work behavior feature driven rehabilitation training effect dynamic evaluation method of the present application. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0055] The examples of the present application provide a work behavior feature driven rehabilitation training effect dynamic evaluation method. The execution subjects of the method include, but are not limited to: a simulation driving device, a rehabilitation training system, a physiological data acquisition device, an evaluation server, and the like, which can be regarded as general computing nodes of the present application, and the rehabilitation training system includes, but is not limited to: a virtual reality training device, an interactive rehabilitation device, and an intelligent feedback system at least one of which.

[0056] The present application provides a work behavior feature driven rehabilitation training effect dynamic evaluation method, which collects the operation behavior data and physiological sensor data of the patient on the simulation driving device, extracts compensatory behavior features and implicit behavior mode features, constructs a behavior mode migration feature set, calculates a behavior mode migration index, constructs a rehabilitation ability dynamic evaluation model, and dynamically adjusts the rehabilitation training scheme based on the dynamic evaluation model, to realize the closed-loop dynamic optimization of the rehabilitation training.

[0057] The application realizes the consistency of feature extraction and analysis through the standardized data acquisition interface and processing flow, adopts the feature-driven evaluation method to improve the objectivity and accuracy of evaluation, monitors the gradual transition process of patients from compensatory behavior to normal behavior through the built-in intelligent tracking mechanism, supports the accurate evaluation of rehabilitation progress, realizes the separation and cooperation of evaluation dimensions through the multi-layer evaluation architecture (compensatory behavior evaluation and implicit behavior pattern evaluation), and meets the strict requirements of objectivity, adaptability and individualization in the rehabilitation training process.

[0058] Please refer to Figure 1 In the embodiment of the application, the detailed implementation steps of the work behavior feature-driven rehabilitation training effect dynamic evaluation method include:

[0059] Multi-dimensional operation behavior data of the patient on the simulation driving device and synchronous physiological sensor data are acquired. The multi-dimensional operation behavior data includes steering wheel steering angle, steering force, pedal pressure distribution, operation timing and other driving operation parameters, reflecting the explicit behavior performance of the patient in the simulation driving task. The physiological sensor data includes heart rate variability, skin electricity reaction, muscle electricity activity, eye movement tracking and other physiological indexes, reflecting the internal physiological state changes of the patient during the task execution. These data are collected through a standardized interface and are time-stamped synchronously, ensuring the consistency and integrity of the data.

[0060] According to the timing distribution and force change of the operation action in the multi-dimensional operation behavior data, compensatory behavior features are extracted. Compensatory behavior refers to the non-standard operation mode adopted by the patient to compensate for functional defects, which is usually manifested as uncoordinated operation action, excessive force or unnecessary repeated action. The extraction process identifies the compensatory strategy of the patient by analyzing the volatility of the operation timing, the uneven distribution of the force and the repeated mode of the action.

[0061] According to the fluctuation characteristics of the physiological indexes in the physiological sensor data, implicit behavior pattern features are extracted. The implicit behavior pattern reflects the cognitive load and physiological stress state of the patient during the task execution, which is not directly shown in the operation behavior, and is an important supplementary index for evaluating the rehabilitation state. The extraction process identifies the physiological response mode related to cognitive effort and emotional state by analyzing the frequency domain and time domain characteristics of the physiological signal.

[0062] Based on the compensatory behavior features and the implicit behavior pattern features, a behavior pattern migration feature set is constructed. The feature set comprehensively reflects the feature performance of the patient in the rehabilitation process from compensatory behavior to normal behavior, providing a data basis for evaluating the rehabilitation progress. The feature set is standardized and fused to form a multi-dimensional evaluation index system.

[0063] According to the attenuation trend of compensatory behavior features and the enhancement trend of implicit behavior pattern features in the behavior pattern migration feature set, a behavior pattern migration index is calculated. The index quantifies the degree of change in behavior patterns during the patient's rehabilitation process and is a core indicator for evaluating rehabilitation effectiveness. The calculation process takes into account the rate and direction of feature changes, and generates a comprehensive score through weighted fusion.

[0064] Based on the behavior pattern migration index, the gradual transition process of patients from compensatory behavior to normal behavior is identified. This identification process tracks the time series changes of the behavior pattern migration index, identifies key transition stages and rehabilitation milestones, and provides objective evidence for clinical decision-making.

[0065] According to the behavior pattern migration index and the preset behavior pattern benchmark of healthy people, a dynamic assessment model of patient rehabilitation ability is constructed. This model compares the patient's current behavior pattern with the health standard, assesses the rehabilitation progress and potential, and provides quantitative rehabilitation ability assessment results. The model uses machine learning algorithms to establish the mapping relationship between behavior patterns and rehabilitation ability, enabling dynamic assessment.

[0066] Based on the dynamic assessment model of patient rehabilitation ability, real-time rehabilitation ability scores are generated. This score directly reflects the patient's current rehabilitation status, provides a basis for training program adjustment, and also serves as a feedback indicator for patient rehabilitation progress, enhancing the patient's rehabilitation confidence and motivation.

[0067] According to the real-time rehabilitation ability score, an adaptive adjustment strategy for virtual driving scenarios is designed, dynamically adjusting the complexity parameters of the simulated driving scenario, while collecting adaptive behavior data of patients in the adjusted simulated driving scenario. Adaptive adjustment ensures that the training difficulty is always in the "zone of proximal development" of the patient's ability, challenging but not exceeding the patient's ability range, maximizing training effectiveness.

[0068] According to the adaptive behavior data and the behavior pattern migration index, a behavior pattern migration trend curve is constructed. This curve describes the long-term trend of behavior pattern changes during the patient's rehabilitation process, providing a data basis for predicting rehabilitation effectiveness. The trend curve is constructed through data fitting and trend analysis methods, reflecting the dynamic change characteristics of the rehabilitation process.

[0069] Based on the behavior pattern migration trend curve, a rehabilitation effectiveness prediction model is generated. This model is based on historical rehabilitation data to predict future rehabilitation progress, providing a scientific basis for long-term rehabilitation planning. The prediction model uses time series analysis methods, taking into account trends and volatility characteristics, to provide reliable rehabilitation predictions.

[0070] According to the rehabilitation effect prediction model and the behavior pattern migration trend curve, a personalized rehabilitation training report is generated, and the training scheme parameters of the next stage are optimized to realize the closed-loop dynamic optimization of rehabilitation training. The personalized report provides detailed rehabilitation evaluation results and suggestions, and the training scheme optimization is based on the evaluation results and the prediction model to adjust the training content, intensity and frequency to form a continuous improvement rehabilitation closed loop.

[0071] In the embodiment of the present application, the detailed implementation steps of extracting compensatory behavior characteristics include:

[0072] The time sequence distribution of operation actions in multi-dimensional operation behavior data is segmented and processed to obtain the frequency, intensity and duration of operation actions in each time period. The sliding window technology is used for segmentation processing, and the window length is determined according to the task type and operation characteristics, usually 5-30 seconds. For operation data in each window, the occurrence frequency of operation actions is counted, the mean and standard deviation of intensity are calculated, and the distribution characteristics of duration are measured. These indicators comprehensively reflect the space-time characteristics of operation actions.

[0073] According to the frequency, intensity and duration of operation actions in each time period, the smoothness index of operation actions is calculated.

[0074] The smoothness index of operation actions The specific calculation formula is:

[0075] ;

[0076] Wherein:

[0077] , indicates the frequency stability, is the frequency standard deviation, is the frequency mean;

[0078] , indicates the intensity uniformity, is the intensity standard deviation, is the intensity mean;

[0079] , indicates the duration consistency, is the duration standard deviation, is the duration mean; , , are the weight coefficients of frequency, intensity and duration respectively;

[0080] The default value is , , .

[0081] To ensure that the calculation result is between 0-1, normalization processing is adopted, and then the more the smoothness index approaches 1, the more stable and coordinated the operation action is; the more the smoothness index approaches 0, the more unstable the operation action is.

[0082] The smoothness index is a comprehensive index for quantifying the coordination of operation action, reflecting the stability and fluency of the operation behavior of the patient. The smoothness index is used to represent the coordination of operation action, and the calculation formula considers the stability of operation frequency, the uniformity of force, and the consistency of action duration, to generate a normalized score between 0-1, and the higher the score, the more stable the operation is. In the calculation of the smoothness index, the outliers of the operation data are filtered to avoid evaluation deviation caused by temporary interference.

[0083] According to the deviation of the smoothness index from the preset health population operation action smoothness threshold, the abnormal fluctuation feature of the operation action is extracted. The smoothness threshold of the health population is statistically derived based on a large amount of normal driving data, and is used as an evaluation benchmark. By calculating the difference between the smoothness index of the patient and the health threshold, abnormal fluctuations in the operation behavior are identified, which are often direct manifestations of compensatory behavior. The abnormal fluctuation feature is determined by the fluctuation of the smoothness index in the time sequence and the deviation from the health threshold, considering both the amplitude and the pattern of the fluctuation.

[0084] According to the distribution density and duration of the abnormal fluctuation feature, the compensatory behavior feature is determined. The distribution density reflects the universality of the abnormal fluctuation in the operation behavior, and the duration reflects the stability and persistence of the compensatory behavior. The compensatory behavior feature includes the number of unnecessary repetitions of the operation action and the proportion of force overload, which directly reflect the main forms of compensatory behavior. The number of unnecessary repetitions counts the number of redundant actions to complete the same operation goal, and the proportion of force overload quantifies the proportion of operation that exceeds the necessary force level, which together constitute the quantitative description of compensatory behavior.

[0085] In the embodiments of the present application, the detailed implementation steps for extracting the implicit behavior pattern feature include:

[0086] The fluctuation characteristics of the physiological indicators in the physiological sensor data are analyzed in time and frequency to obtain the frequency energy distribution and time domain fluctuation amplitude of the physiological indicators in each time period. Time-frequency analysis uses methods such as short-time Fourier transform or wavelet transform to decompose the physiological signal into different frequency components and analyze their time variation characteristics. The frequency energy distribution reflects the intensity of physiological activities at different frequencies, and the time domain fluctuation amplitude reflects the degree of change of the physiological indicators over time, which together provide a comprehensive description of the physiological state.

[0087] According to the concentration of the frequency energy distribution and the stability of the time domain fluctuation amplitude, the dynamic balance index of the physiological indicators is calculated.

[0088] Dynamic balance index The specific calculation formula is as follows:

[0089] ;

[0090] in, The frequency domain energy entropy is calculated using the following formula: , It is the first Normalized energy values ​​for each frequency band; The time-domain variation coefficient is calculated using the following formula: , It is the standard deviation of the time-domain fluctuation. It is the mean of the time-domain fluctuation; is a weighting coefficient used to balance the importance of frequency domain and time domain features. Its value ranges from [0,1], and it is usually taken as 0.5.

[0091] Frequency domain energy entropy The smaller the value, the more concentrated the energy distribution; the time-domain variation coefficient The smaller the value, the more stable the fluctuation range. Both are subtracted from 1, so a higher exponent value indicates better dynamic equilibrium. The value range is [0,1].

[0092] The dynamic balance index is a comprehensive indicator for assessing the stability of a physiological system, reflecting a patient's physiological regulatory capacity during task performance. The dynamic balance index characterizes a patient's potential cognitive load during performance activities. The calculation formula comprehensively considers the entropy value of the frequency domain energy distribution and the coefficient of variation of time domain fluctuations to generate a standardized index value. A higher dynamic balance index indicates a more stable physiological system and a more moderate cognitive load; a lower index indicates significant physiological fluctuations and a heavier cognitive load.

[0093] Implicit behavioral pattern characteristics were extracted based on the difference between the dynamic balance index and the pre-defined dynamic balance benchmark for healthy individuals. The dynamic balance benchmark for healthy individuals was derived from statistical analysis of physiological data of normal individuals under the same task conditions and served as an assessment reference standard. By comparing the dynamic balance index of patients with the healthy benchmark, implicit behavioral pattern characteristics were identified. These characteristics reflect the patients' adaptive status at the cognitive and emotional levels. Implicit behavioral pattern characteristics include cognitive overload characteristics and physiological stress response characteristics. The former reflects the consumption of cognitive resources when performing tasks, while the latter reflects the degree of stress response when facing task challenges. Together, they constitute the core indicators of patients' implicit behavioral patterns.

[0094] In this embodiment of the invention, the detailed implementation steps for calculating the behavioral pattern migration index include:

[0095] The attenuation trend of compensatory behavior characteristics is quantified to obtain an attenuation rate of compensatory behavior characteristics. The attenuation rate reflects how fast compensatory behavior decreases with rehabilitation progress and is a key indicator for evaluating rehabilitation effect. The attenuation rate is obtained by calculating the slope of the change of compensatory behavior characteristics in a continuous time period. A linear regression method is used to fit the change trend of the characteristic values over time, and the negative value of the slope is the attenuation rate. The greater the attenuation rate, the faster the compensatory behavior decreases, and the more obvious the rehabilitation progress. If the attenuation rate is close to zero or positive, it indicates that compensatory behavior persists or has an increasing trend, and the rehabilitation effect is poor.

[0096] The enhancement trend of implicit behavior pattern characteristics is quantified to obtain an enhancement rate of implicit behavior pattern characteristics. The enhancement rate reflects the speed of establishing normal behavior patterns and is an important indicator for evaluating rehabilitation quality. The enhancement rate is obtained by calculating the slope of the change of implicit behavior pattern characteristics in a continuous time period. Similarly, a linear regression method is used to fit the change trend of the characteristic values over time, and the positive value of the slope is the enhancement rate. The greater the enhancement rate, the faster the normal behavior patterns are established, and the higher the rehabilitation quality. If the enhancement rate is close to zero or negative, it indicates that the normal behavior patterns are established slowly or degenerate, and the rehabilitation quality needs to be improved.

[0097] According to the attenuation rate of compensatory behavior characteristics and the enhancement rate of implicit behavior pattern characteristics, a behavior pattern migration index is calculated. The behavior pattern migration index is a weighted ratio of the attenuation rate and the enhancement rate, wherein the weights are determined by the influence factors of compensatory behavior characteristics and implicit behavior pattern characteristics on rehabilitation ability. The calculation formula is:

[0098] Behavior pattern migration index = (w1 x attenuation rate) / (w2 x enhancement rate);

[0099] wherein w1 and w2 are weight coefficients determined according to the importance of the characteristics for rehabilitation evaluation. In this calculation logic, the ratio of the attenuation rate and the enhancement rate reflects the balance between the reduction of compensatory behavior and the establishment of normal behavior, and the weighted processing considers the relative importance of different characteristics to generate a more meaningful evaluation index.

[0100] It should be noted that the weights are determined by the influence factors of compensatory behavior characteristics and implicit behavior pattern characteristics on rehabilitation ability, and the specific determination method adopts a combination of data-driven and expert evaluation:

[0101] w1 = β1 x r1 + (1 - β1) x e1;

[0102] w2 = β1 x r2 + (1 - β1) x e2;

[0103] wherein:

[0104] r1, r2 are absolute values of Pearson correlation coefficients between compensatory behavior characteristics and implicit behavior pattern characteristics and rehabilitation results respectively;

[0105] e1, e2 are importance scores of compensatory behavior characteristics and implicit behavior pattern characteristics evaluated by experts, with a value range of [0, 10]; r1, r2, e1 and e2 are the above influence factors;

[0106] β1 is a coefficient for balancing data-driven and expert evaluation, with a value range of [0, 1], and usually takes 0.7.

[0107] In practical applications, the initial weights can be set as w1=0.6 and w2=0.4, and with the accumulation of more rehabilitation data, the weight coefficients are recalculated regularly to realize adaptive optimization of the weights.

[0108] In the embodiment of the application, the detailed implementation steps for constructing the patient rehabilitation ability dynamic evaluation model include:

[0109] According to the behavior pattern migration index, a behavior pattern migration state space is constructed. The state space is a multi-dimensional representation of the change of behavior patterns in the rehabilitation process of the patient, and provides a theoretical framework for rehabilitation ability evaluation. The behavior pattern migration state space includes a compensatory behavior dominant state, a transition state and a normal behavior dominant state, which reflect the main stages in the rehabilitation process. The compensatory behavior dominant state indicates that the patient mainly relies on compensatory strategies to complete tasks, the transition state indicates that compensatory behavior and normal behavior coexist, and the normal behavior dominant state indicates that the patient mainly uses normal behavior patterns to perform tasks.

[0110] According to the distance between the current state of the patient in the behavior pattern migration state space and the behavior pattern benchmark of the healthy population, the rehabilitation ability dynamic score is calculated. The distance is calculated by the Euclidean distance, which reflects the difference between the current behavior pattern of the patient and the health standard. The smaller the distance, the closer the behavior pattern of the patient to the health standard, and the higher the rehabilitation degree; the larger the distance, the more significant the difference, and the greater the space for rehabilitation. The rehabilitation ability dynamic score is nonlinearly mapped based on the distance value to generate a standardized score of 0-100, which intuitively reflects the rehabilitation state.

[0111] According to the time series change of the rehabilitation ability dynamic score, a patient rehabilitation ability dynamic evaluation model is constructed. The patient rehabilitation ability dynamic evaluation model uses a long short-term memory network to model the time series change, capturing long-term trends and short-term fluctuations in the rehabilitation process. The long short-term memory network is a special recurrent neural network with long-term memory capability, which is suitable for processing time series data in rehabilitation evaluation. The model training uses historical rehabilitation data to optimize the network parameters through a supervised learning method, establishing a mapping relationship between the score sequence and the rehabilitation ability. The model output includes current rehabilitation ability evaluation and short-term prediction, providing real-time guidance for rehabilitation training.

[0112] In the embodiment of the present application, the detailed implementation steps of the adaptive adjustment strategy of the virtual driving scene design include:

[0113] According to the real-time rehabilitation ability score, the complexity parameter adjustment direction of the simulation driving scene is determined. The complexity parameter is a key factor to control the difficulty of the driving task, and appropriate difficulty adjustment is an important means to optimize the training effect. The complexity parameters include road curvature, obstacle density and traffic flow, which directly affect the cognitive and operation difficulty of the driving task. According to the level of the rehabilitation ability score, the system determines whether to increase or decrease the task difficulty, and when the score is low, the complexity is reduced to avoid frustration; when the score is high, the complexity is appropriately increased to create a moderate challenge.

[0114] According to the behavior pattern migration index, the adjustment amplitude of the simulation driving scene is determined. The adjustment amplitude is positively correlated with the absolute value of the behavior pattern migration index, reflecting the intensity and speed of adjustment. When the absolute value of the behavior pattern migration index is large, it indicates that the patient's behavior pattern is changing rapidly, and the system should make a large amplitude adjustment to match the change speed; when the absolute value is small, it indicates that the behavior pattern changes slowly, and the system should use small amplitude progressive adjustment to avoid excessive intervention.

[0115] According to the adjustment direction and adjustment amplitude, the adaptive adjustment strategy of the virtual driving scene is generated. The adaptive adjustment strategy includes preferentially reducing the road curvature and obstacle density when the behavior pattern migration index is lower than the preset migration threshold. This strategy takes into account the influence characteristics of different complexity parameters on the driving task, and the road curvature and obstacle density have higher requirements for operation precision, so these parameters are preferentially reduced when the patient's behavior pattern is not yet stable, reducing the operation burden; while the traffic flow mainly affects the cognitive load, which can be adjusted gradually after the basic operation ability is established. The adaptive adjustment adopts a progressive method to avoid maladjustment caused by drastic changes, and records the adjustment effect to form a feedback optimization mechanism.

[0116] In the embodiment of the present application, the detailed implementation steps of constructing the behavior pattern migration trend curve include:

[0117] The adaptive behavior data is clustered and analyzed to obtain the behavior pattern migration subsets of the patient under different complexity parameters. The clustering analysis uses K-means or hierarchical clustering algorithm to group the behavior data according to similarity, and identifies the behavior pattern characteristics under different task conditions. The behavior pattern migration subsets include a decay subset of compensatory behavior characteristics and an enhancement subset of implicit behavior pattern characteristics, which respectively reflect the reduction process of compensatory behavior and the establishment process of normal behavior, providing detailed data for trend analysis.

[0118] According to the time series change of the behavior pattern migration subset, a behavior pattern migration trend curve is fitted. The behavior pattern migration trend curve is fitted by a polynomial regression model to capture the nonlinear change characteristics in the rehabilitation process. The order of the polynomial regression model is dynamically determined according to the data complexity, and a 3-5 order polynomial is usually selected to balance the fitting accuracy and model complexity. The least squares method is used to optimize the polynomial coefficients in the fitting process, and a regularization term is introduced to prevent overfitting, ensuring the smoothness and prediction ability of the trend curve.

[0119] According to the curvature change of the behavior pattern migration trend curve, the inflection point of the behavior pattern migration is determined. The inflection point is used to represent the significant transition stage of the patient from compensatory behavior to normal behavior, which is a key milestone in the rehabilitation process. The mathematical definition of the inflection point is the point where the second derivative of the trend curve changes sign, indicating the position where the curve changes from concave to convex or from convex to concave, which corresponds to the turning point of the behavior pattern change rate in rehabilitation assessment. The inflection point analysis uses numerical differentiation and extreme value detection methods to identify the key transition position on the trend curve, providing an objective basis for rehabilitation stage division and progress assessment.

[0120] In the embodiment of the present application, the detailed implementation steps of generating the rehabilitation effect prediction model include:

[0121] According to the behavior pattern migration trend curve, the long-term trend characteristics and short-term fluctuation characteristics of the trend curve are extracted. The long-term trend characteristics are obtained by low-pass filtering, reflecting the basic direction and speed of the rehabilitation process; the short-term fluctuation characteristics are obtained by high-pass filtering, reflecting the fluctuation and adaptation characteristics in the rehabilitation process. The filtering process uses Butterworth filter or wavelet decomposition method to decompose the trend curve into different frequency components, effectively separating the long-term trend and short-term fluctuation. The long-term trend characteristics are used to predict the overall rehabilitation trend, and the short-term fluctuation characteristics are used to evaluate the rehabilitation stability and the ability to respond to changes.

[0122] According to the long-term trend characteristics and short-term fluctuation characteristics, a rehabilitation effect prediction model is constructed. The rehabilitation effect prediction model uses an autoregressive integrated moving average model to jointly model the long-term trend characteristics and short-term fluctuation characteristics to predict future rehabilitation progress. The autoregressive integrated moving average model (ARIMA) is a classic time series analysis method suitable for handling data with trends and seasonality, which can effectively capture the regularity and volatility of the progress in rehabilitation prediction. The model parameters (p, d, q) are determined by autocorrelation function and partial autocorrelation function analysis, and the maximum likelihood estimation is used to optimize the parameter values to establish the mapping relationship between data and predicted values.

[0123] According to the output of the rehabilitation effect prediction model, the behavior pattern transfer index and the rehabilitation ability score of the patient in the future training period are predicted. The prediction adopts a rolling prediction strategy, first predicts the near-term value, then takes the prediction result as the new input to predict the more distant future, and gradually extends the prediction time range. The prediction result includes a point prediction value and a prediction interval, the point prediction value provides the most likely rehabilitation trajectory, and the prediction interval considers the uncertainty of the prediction and provides the range of possible changes. The prediction result is used to develop a long-term rehabilitation plan and set reasonable rehabilitation goals, guide clinical decision-making and patient expectation management.

[0124] In the embodiment of the present application, the detailed implementation steps of optimizing the training scheme parameters of the next stage include:

[0125] According to the deficiency items of the behavior pattern transfer index in the individualized rehabilitation training report, the optimization direction of the training scheme parameters is determined. The deficiency items include insufficient attenuation of compensatory behavior characteristics and insufficient enhancement of implicit behavior pattern characteristics, which reflect the deficiencies of compensatory behavior elimination and normal behavior establishment respectively. By analyzing the specific performance of the deficiency items, such as which compensatory actions persist and which normal behavior patterns fail to establish, the system determines the optimization direction of the training scheme and specifically strengthens the training of specific abilities.

[0126] According to the prediction result of the rehabilitation effect prediction model, the optimization amplitude of the training scheme parameters is determined. The optimization amplitude is positively correlated with the improvement space of the predicted behavior pattern transfer index, reflecting the intensity and speed of adjustment. The prediction result provides an estimate of future rehabilitation progress under the current training scheme, and by comparing the gap between the prediction value and the rehabilitation goal, the required optimization intensity is determined. When the prediction value is much lower than the goal, a large amplitude of scheme adjustment is needed; when the prediction value is close to the goal, small amplitude fine-tuning can be used to consolidate the existing achievements.

[0127] According to the optimization direction and optimization amplitude, the training scheme parameters of the next stage are adjusted. The training scheme parameters include the complexity parameter of the simulated driving scene, the training duration and the training frequency, which together determine the intensity and content of the training. The complexity parameter adjustment targets specific ability requirements, such as increasing road curvature to train fine steering control and increasing traffic flow to train multi-task processing ability; the training duration and frequency adjustment considers the need for fatigue management and learning consolidation, and is personalized according to the patient's tolerance and learning characteristics. The adjustment process adopts a gradual method to avoid maladjustment caused by drastic changes, and establishes a tracking evaluation mechanism for the adjustment effect to form a closed-loop management of continuous optimization.

[0128] In the embodiment of the present application, the detailed implementation steps of determining the compensatory behavior characteristics include:

[0129] The distribution density of the abnormal fluctuation features is kernel density estimation, and the probability density function of the abnormal fluctuation features is obtained. Kernel density estimation is a non-parametric estimation method, which can effectively describe the distribution characteristics of data and is not limited by specific distribution assumptions. The estimation process uses a Gaussian kernel function to smooth the abnormal fluctuation data, generates a continuous probability density function, and intuitively reflects the distribution pattern of the abnormal fluctuation. The peak value of the probability density function corresponds to the high-frequency abnormal fluctuation mode, which is the key information for identifying the main compensatory behavior.

[0130] According to the number of peak values and the peak value interval of the probability density function, the mode type of the compensatory behavior features is determined. The mode type includes single mode compensation and multi-mode compensation, which reflects the complexity and stability of the patient's compensatory strategy. Single mode compensation is characterized by a single main peak in the probability density function, indicating that the patient adopts a relatively consistent compensatory strategy; multi-mode compensation is characterized by multiple obvious peaks, indicating that the patient switches different compensatory strategies according to task requirements, usually representing more complex dysfunction. The peak value interval reflects the distinction between different compensation modes, and the larger the interval, the more obvious the mode difference, which helps to accurately identify and targeted training.

[0131] According to the mode type and the duration of the abnormal fluctuation features, the significance score of the compensatory behavior features is calculated.

[0132] The specific calculation formula of the significance score SU of the compensatory behavior features is:

[0133] ;

[0134] Wherein:

[0135] is the mode complexity coefficient, which is calculated differently according to the mode type:

[0136] Single mode compensation: ;

[0137] Multi-mode compensation: ;

[0138] Wherein, is the peak height, is the maximum possible peak height, N is the number of peaks, is the minimum distance between peaks, is the maximum possible distance;

[0139] is the duration weight, and the calculation formula is:

[0140] ;

[0141] Wherein, T is the duration of abnormal fluctuation, and Ttotal is the total observation time.

[0142] Finally, the significance score is standardized to the range of 0-10, and the higher the significance score, the greater the impact of the compensatory behavior feature on the rehabilitation ability evaluation, and the higher priority it needs to be given in the training program.

[0143] The significance score is used to represent the contribution of the compensatory behavior feature to the rehabilitation ability evaluation, and is an important basis for determining the feature weight. The calculation process comprehensively considers the stability and persistence of the mode, and the compensatory behavior of a single mode and long duration usually represents a deep-rooted alternative strategy, which needs to be focused on; the compensatory behavior of multiple modes and short duration is likely to be a temporary adaptive strategy, and the priority is relatively low. The significance score adopts a standardized scale of 0-10, which directly reflects the importance of the feature and guides the subsequent evaluation and training program design.

[0144] Through the description of the above detailed implementation manner, the present application realizes the objective evaluation of the rehabilitation training effect and the dynamic optimization of the training program by comprehensively analyzing the explicit operation behavior and the implicit physiological reaction of the patient, constructing the behavior mode migration feature set and the rehabilitation ability evaluation model.

[0145] The above only describes the preferred embodiments of the present application and is not used to limit the present application, although the foregoing embodiments of the present application have been described in detail, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments, or equivalently replace some technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0146] It should be noted that the preset parameters and threshold values are set by the person skilled in the art according to the actual situation.

[0147] Although the embodiments of the present application have been shown and described, those skilled in the art can understand that various changes, modifications, replacements and variations can be made to these embodiments without departing from the principles and purposes of the present application, and the scope of the present application is defined by the claims and their equivalents.

Claims

1. A method for dynamically evaluating the effect of rehabilitation training driven by work behavior characteristics, characterized in that, The method comprises the following steps: acquiring multi-dimensional operation behavior data of a patient on a simulated driving device and synchronously collected physiological sensor data; extracting compensatory behavior features according to the time sequence distribution and force variation of operation actions in the multi-dimensional operation behavior data; extracting implicit behavior pattern features according to the fluctuation characteristics of physiological indexes in the physiological sensor data; constructing a behavior pattern migration feature set based on the compensatory behavior features and the implicit behavior pattern features; calculating a behavior pattern migration index according to the attenuation trend of the compensatory behavior features and the enhancement trend of the implicit behavior pattern features in the behavior pattern migration feature set; constructing a patient rehabilitation ability dynamic evaluation model according to the behavior pattern migration index and a preset health population behavior pattern benchmark; generating a real-time rehabilitation ability score based on the patient rehabilitation ability dynamic evaluation model; designing an adaptive adjustment strategy for a virtual driving scene according to the real-time rehabilitation ability score, dynamically adjusting the complexity parameters of the simulated driving scene, and simultaneously collecting adaptive behavior data of the patient in the adjusted simulated driving scene; constructing a behavior pattern migration trend curve according to the adaptive behavior data and the behavior pattern migration index; generating a rehabilitation effect prediction model based on the behavior pattern migration trend curve; generating an individualized rehabilitation training report and optimizing the training scheme parameters of the next stage according to the rehabilitation effect prediction model and the behavior pattern migration trend curve. 2.The job behavior feature driven rehabilitation training effect dynamic evaluation method according to claim 1, characterized in that, The extraction of compensatory behavior features comprises the following steps: segmenting the time sequence distribution of operation actions in the multi-dimensional operation behavior data to obtain the frequency, force and duration of operation actions in each time period; calculating the smoothness index of operation actions according to the frequency, force and duration of operation actions in each time period; extracting the abnormal fluctuation features of operation actions according to the deviation of the smoothness index from the preset health population operation action smoothness threshold; determining the compensatory behavior features according to the distribution density and duration of the abnormal fluctuation features; the compensatory behavior features include the unnecessary repetition times and force overload proportion of operation actions. 3.The job behavior feature driven rehabilitation training effect dynamic evaluation method according to claim 1, characterized in that, The extraction of implicit behavior pattern features comprises the following steps: performing time-frequency analysis on the fluctuation characteristics of physiological indexes in the physiological sensor data to obtain the frequency energy distribution and time domain fluctuation amplitude of physiological indexes in each time period; calculating the dynamic balance index of physiological indexes according to the concentration of the frequency energy distribution and the stability of the time domain fluctuation amplitude; extracting the implicit behavior pattern features according to the difference between the dynamic balance index and the preset health population dynamic balance benchmark; the implicit behavior pattern features include cognitive load overload features and physiological stress response features. 4.The job behavior feature driven rehabilitation training effect dynamic evaluation method according to claim 1, characterized in that, The calculation of the behavior pattern migration index comprises the following steps: quantifying the attenuation trend of the compensatory behavior features to obtain the attenuation rate of the compensatory behavior features; quantifying the enhancement trend of the implicit behavior pattern features to obtain the enhancement rate of the implicit behavior pattern features; According to the attenuation rate of the compensatory behavior feature and the enhancement rate of the implicit behavior pattern feature, a behavior pattern migration index is calculated; the behavior pattern migration index is a weighted ratio of the attenuation rate and the enhancement rate. 5.The job behavior feature driven rehabilitation training effect dynamic evaluation method according to claim 1, characterized in that, The model for dynamically evaluating the rehabilitation ability of the patient comprises: According to the behavior pattern migration index, a behavior pattern migration state space is constructed; the behavior pattern migration state space comprises a compensatory behavior dominant state, a transition state and a normal behavior dominant state; According to the distance between the state currently occupied by the patient in the behavior pattern migration state space and the behavior pattern benchmark of the healthy population, a dynamic score of the rehabilitation ability is calculated; the distance is calculated by the Euclidean distance; According to the time series change of the dynamic score of the rehabilitation ability, a model for dynamically evaluating the rehabilitation ability of the patient is constructed; the model for dynamically evaluating the rehabilitation ability of the patient uses a long short-term memory network to model the time series change. 6.The job behavior feature driven rehabilitation training effect dynamic evaluation method according to claim 1, characterized in that, The adaptive adjustment strategy for designing the virtual driving scene comprises: According to the real-time rehabilitation ability score, a complexity parameter adjustment direction of the simulated driving scene is determined; the complexity parameter comprises road curvature, obstacle density and traffic flow; According to the behavior pattern migration index, an adjustment amplitude of the complexity parameter of the simulated driving scene is determined; the adjustment amplitude is positively correlated with the absolute value of the behavior pattern migration index; According to the adjustment direction and the adjustment amplitude, an adaptive adjustment strategy of the virtual driving scene is generated; the adaptive adjustment strategy comprises preferentially reducing the road curvature and the obstacle density when the behavior pattern migration index is lower than a preset migration threshold. 7.The job behavior feature driven rehabilitation training effect dynamic evaluation method according to claim 1, characterized in that, The behavior pattern migration trend curve is constructed, comprising: The adaptive behavior data is subjected to cluster analysis to obtain a behavior pattern migration subset of the patient under different complexity parameters; the behavior pattern migration subset comprises an attenuation subset of the compensatory behavior feature and an enhancement subset of the implicit behavior pattern feature; According to the time series change of the behavior pattern migration subset, a behavior pattern migration trend curve is fitted; the behavior pattern migration trend curve is fitted by using a polynomial regression model; According to the curvature change of the behavior pattern migration trend curve, an inflection point of the behavior pattern migration is determined. 8.The job behavior feature driven rehabilitation training effect dynamic evaluation method according to claim 1, characterized in that, The rehabilitation effect prediction model is generated, comprising: According to the behavior pattern migration trend curve, long-term trend features and short-term fluctuation features of the trend curve are extracted; According to the long-term trend features and the short-term fluctuation features, a rehabilitation effect prediction model is constructed; the rehabilitation effect prediction model uses an autoregressive integrated moving average model to jointly model the long-term trend features and the short-term fluctuation features; According to the output of the rehabilitation effect prediction model, behavior pattern migration indexes and rehabilitation ability scores of the patient in future training periods are predicted. 9.The job behavior feature driven rehabilitation training effect dynamic evaluation method according to claim 1, characterized in that, The training scheme parameters of the next stage are optimized, comprising: According to the deficiency items of the behavior pattern migration index in the individualized rehabilitation training report, an optimization direction of the training scheme parameters is determined; the deficiency items comprise insufficient attenuation of the compensatory behavior feature and insufficient enhancement of the implicit behavior pattern feature; According to the prediction result of the rehabilitation effect prediction model, an optimization amplitude of the training scheme parameter is determined; According to the optimization direction and the optimization amplitude, a training scheme parameter of a next stage is adjusted; the training scheme parameter includes a complexity parameter of a simulated driving scene, a training duration, and a training frequency. 10.The job behavior feature driven rehabilitation training effect dynamic evaluation method according to claim 2, characterized in that, The determination of the compensatory behavior feature includes: Kernel density estimation is performed on the distribution density of the abnormal fluctuation feature to obtain a probability density function of the abnormal fluctuation feature; According to the number of peak values and the peak value interval of the probability density function, a mode type of the compensatory behavior feature is determined; the mode type includes single mode compensation and multi-mode compensation; According to the mode type and the duration of the abnormal fluctuation feature, a significance score of the compensatory behavior feature is calculated.

Images