A cerebral palsy child rehabilitation game task generation method based on adaptive adjustment

Abstract

This invention discloses a method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment, belonging to the field of rehabilitation training technology. The method includes the following steps: S1, acquiring data to construct an initial ability map and performing weighted correction to generate a dynamic ability weight vector; S2, pre-setting a task library, adjustment rules, and matching rules, configuring adjustable parameters, and establishing a mapping relationship table; S3, constructing a model, calculating game parameter combinations, and generating personalized rehabilitation game tasks; S4, executing the task, collecting data in real time, and fusing the data to obtain real-time rehabilitation status data; S5, matching the data with the mapping relationship table, generating real-time stimulus signals, and then combining the adjustment rules to perform real-time adaptive adjustment of the task. This invention, by constructing an integrated method for generating rehabilitation game tasks for children with cerebral palsy, realizes the generation of personalized rehabilitation game tasks that match the individual abilities of children with cerebral palsy, simplifying the personalized customization process of rehabilitation tasks.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation training technology, and more specifically, to a method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment. Background Technology

[0002] The adaptive adjustment-based method for generating rehabilitation game tasks for children with cerebral palsy serves as a core support for the construction of a digital and intelligent medical rehabilitation system for children's rehabilitation medicine. Its task adaptation accuracy, training fun, sustainable rehabilitation effect, and ease of operation directly determine the quality of rehabilitation training for children with cerebral palsy, the work efficiency of rehabilitation therapists, and the accessibility of home rehabilitation. It has become a key technology for promoting the intelligent transformation of children's rehabilitation and realizing the precise implementation of personalized rehabilitation.

[0003] The significant individual differences among children with cerebral palsy, the complexity and diversity of clinical classifications, the dynamic changes in rehabilitation goals, the low cognitive acceptance among children, and the poor long-term training compliance are the core constraints affecting the efficient application and stable implementation of rehabilitation games for children with cerebral palsy. These factors are dynamically combined by the differences in different clinical classifications such as spastic, athetoid, and ataxic types, the uneven development of motor and cognitive functions, large fluctuations in emotional state, and the long and slow rehabilitation training cycle. Meanwhile, adaptive game task generation technology for rehabilitation scenarios specifically for children with cerebral palsy is a key means to solve the problems of poor adaptability, insufficient fun, and low personalization of existing rehabilitation games, directly ensuring that rehabilitation therapists can quickly complete personalized rehabilitation task customization and match the child's abilities.

[0004] However, existing methods and systems for generating rehabilitation game tasks for children with cerebral palsy often rely on generic children's game templates in practical applications. They lack specific adaptation mechanisms for clinical classification of cerebral palsy, ignore the dynamic changes in children's abilities, and fail to integrate and evaluate the value of multimodal rehabilitation status data and personalize matching logic for positive incentives. They cannot cover the complex needs of children with cerebral palsy with different clinical classifications, ability levels, rehabilitation stages, and interests. This results in a one-sided approach to rehabilitation games that prioritizes entertainment over rehabilitation professionalism. Furthermore, existing methods often rely on manual adjustment, failing to address issues such as a single dimension of ability assessment, delayed parameter adjustment, homogenized incentive mechanisms, and insufficient task adaptability. Consequently, the matching degree between tasks and children's abilities is low, training effects are poor, children are prone to resistance, and rehabilitation therapists face heavy workloads.

[0005] There are currently no effective solutions to the problems in the relevant technologies. Summary of the Invention

[0006] To address the problems in related technologies, this invention proposes a method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment, in order to overcome the aforementioned technical problems existing in the existing related technologies.

[0007] To achieve the above objectives, the specific technical solution adopted by the present invention is as follows: A method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment, comprising the following steps: S1. Obtain motor function data, cognitive function data, emotional state data, and preference characteristic data of children with cerebral palsy to be trained to construct an initial ability map, and then make weighted corrections to the initial ability map according to clinical classification, and then generate a dynamic ability weight vector according to clinical rehabilitation goals. As a preferred embodiment, the process of acquiring motor function data, cognitive function data, emotional state data, and preference characteristic data of children with cerebral palsy to be trained to construct an initial ability map, and then weighting and correcting the initial ability map according to clinical classification, and finally generating a dynamic ability weight vector according to clinical rehabilitation goals, includes the following steps: S11. Preset the ability dimension framework, and collect corresponding motor function data, cognitive function data, emotional state data and preference feature data of children with cerebral palsy to be trained, and map them to the ability dimension framework to form an initial ability map. S12. Preset the focus of clinical rehabilitation goals, and based on the clinical classification of children with cerebral palsy, set the weighted correction coefficients corresponding to the ability dimensions to correct the dimensional weights of the initial ability map. S13. Combine the initial capability map after dimension weight correction with the focus of clinical rehabilitation goals, assign weights to each rehabilitation dimension, and generate a dynamic capability weight vector.

[0008] S2. Pre-set standardized game task library, non-linear strategy adjustment rules and positive incentive matching rules, configure adjustable parameters for each standardized game task, and establish a mapping relationship table between positive incentive parameters and game task execution status; As a preferred embodiment, the preset standardized game task library, nonlinear strategy adjustment rules, and positive incentive matching rules, which configure adjustable parameters for each standardized game task and establish a mapping relationship table between positive incentive parameters and game task execution status, include the following steps: S21. Categorize and organize rehabilitation game units of different difficulty according to the dimensions of rehabilitation training, and compile them into a standardized game task library; S22. Based on the pattern of ability changes in children with cerebral palsy, set nonlinear strategy adjustment rules and define the task escalation and demotion adjustment boundaries; S23. Develop positive incentive matching rules based on the cognitive acceptance level of children with cerebral palsy, and set incentive thresholds at different levels; S24. Configure adjustable parameters for motion difficulty, cognitive load, and interaction rhythm for standardized game tasks. S25. Divide the standardized game tasks into different execution state categories, associate and match the corresponding positive incentive parameters, and organize them into a mapping relationship table.

[0009] As a preferred embodiment, the process of classifying standardized game tasks into different execution state categories, associating and matching corresponding positive incentive parameters, and organizing them into a mapping relationship table includes the following steps: S251. Extract the execution features of standardized game tasks during the training process, and classify various task execution state categories based on the execution features; S252. Associate and match the various task execution state categories with the positive stimulus parameters one by one, and solidify the correspondence between the execution state categories and the positive stimulus parameters to generate a mapping table.

[0010] S3. Construct a game parameter adjustment model based on fuzzy inference algorithm, input the weighted and corrected initial ability map and dynamic ability weight vector into the game parameter adjustment model, calculate the game parameter combination, and generate personalized rehabilitation game tasks by combining preference feature data. As a preferred embodiment, the step of constructing a game parameter adjustment model based on a fuzzy inference algorithm, inputting the weighted and corrected initial ability map and dynamic ability weight vector into the game parameter adjustment model, calculating the game parameter combination, and generating a personalized rehabilitation game task by combining preference feature data includes the following steps: S31. Based on the fuzzy inference algorithm, a fuzzy input subset, a fuzzy rule base and a fuzzy decision criterion are set to build a game parameter adjustment model that is adapted to the cerebral palsy rehabilitation scenario. S32. Import the weighted and corrected initial ability map and dynamic ability weight vector into the game parameter adjustment model, and use fuzzy reasoning rules to perform logical deduction to solve the game parameter combination that is suitable for the current child's ability. As a preferred embodiment, the step of importing the weighted and corrected initial ability map and dynamic ability weight vector into the game parameter adjustment model, and solving for the game parameter combination that adapts to the current child's ability through logical deduction using fuzzy inference rules, includes the following steps: S321. Perform data quantization processing on the weighted and corrected initial capability map and dynamic capability weight vector, and convert them into standardized input quantities that the model can recognize. S322. Based on the fuzzy rule base, perform condition matching on the standardized input quantities, and import the condition-matched standardized input quantities into the game parameter adjustment model to generate multiple sets of candidate game parameter combinations through step-by-step logical deduction. S323. Determine the degree of fit between each group of candidate game parameter combinations and the ability map, as well as the degree of fit with the dynamic ability weight vector. Use the degree of fit and the degree of fit as the criteria for selection to obtain game parameter combinations that are suitable for the current rehabilitation ability of children.

[0011] S33. Match the game parameter combination with the standardized game task library, and integrate the matched standardized game tasks with preference feature data to generate personalized rehabilitation game tasks.

[0012] As a preferred embodiment, the process of matching game parameter combinations with a standardized game task library, and then integrating the matched standardized game tasks with preference feature data to generate personalized rehabilitation game tasks includes the following steps: S331. Using the selected game parameter combinations as a matching benchmark, compare the adjustable parameters of standardized game tasks in the standardized game task library, and select the standardized game task with the highest parameter matching degree as the basic rehabilitation task. S332. Embed preference feature data into the game structure of basic rehabilitation tasks, and adapt and adjust the basic rehabilitation tasks based on preference features to generate personalized rehabilitation game tasks.

[0013] S4. Perform personalized rehabilitation game tasks, collect modal physiological data and task performance data in real time during the execution process, and fuse them to obtain real-time rehabilitation status data; As a preferred embodiment, the execution of personalized rehabilitation game tasks, the real-time collection of modal physiological data and task performance data during the execution process, and the fusion of these data to obtain real-time rehabilitation status data include the following steps: S41. Implement personalized rehabilitation game tasks to guide children with cerebral palsy to carry out rehabilitation interactive training, and collect multiple types of modal physiological data simultaneously during the task execution process, and then record the children's task completion status as task performance data. S42. The collected multi-modal physiological data are preprocessed, and the preprocessed multi-modal physiological data and task performance data are subjected to feature alignment and dimension normalization. Then, the multi-source data information is integrated by feature fusion, and the real-time rehabilitation status data is quantitatively output.

[0014] As a preferred embodiment, the steps of preprocessing the collected multi-modal physiological data, performing feature alignment and dimensionality normalization on the preprocessed multi-modal physiological data and task performance data, and then integrating multi-source data information using a feature fusion method to quantify and output real-time rehabilitation status data include the following steps: S421. Organize and clean the collected multi-modal physiological data, align the processed modal physiological data with the task performance data in time series, eliminate the differences in the units of measurement between different data, and complete the dimension normalization process. S422. Extract the feature components of the normalized modal physiological data and task performance data, assign weights, and integrate the weighted feature components to generate comprehensive rehabilitation feature information. Then, quantify and rate the comprehensive rehabilitation feature information to obtain real-time rehabilitation status data.

[0015] S5. Match the real-time rehabilitation status data with the mapping relationship table, generate real-time incentive signals according to the positive incentive matching rules, and then combine the nonlinear strategy adjustment rules to make real-time adaptive adjustments to the personalized rehabilitation game tasks.

[0016] As a preferred embodiment, the process of matching real-time rehabilitation status data with a mapping table, generating real-time incentive signals based on positive incentive matching rules, and then combining nonlinear strategy adjustment rules to adaptively adjust personalized rehabilitation game tasks in real time includes the following steps: S51. The real-time rehabilitation status data is compared and matched with the mapping relationship table one by one, and the current state is determined according to the positive incentive matching rule. The corresponding real-time incentive signal is generated, and the nonlinear strategy adjustment rule is retrieved simultaneously to determine the adjustment direction and adjustment range of the current rehabilitation task. S52. Adjust the adjustable parameters corresponding to the movement difficulty, cognitive load, and interaction rhythm of the personalized rehabilitation game task according to the adjustment direction and adjustment range, and perform real-time adaptive adjustment of the rehabilitation game task.

[0017] The beneficial effects of this invention are as follows: 1. This invention constructs a method for generating rehabilitation game tasks for children with cerebral palsy that integrates multi-dimensional ability assessment, clinical classification weighted correction, fuzzy inference parameter adjustment, real-time fusion of multi-source data, and adaptive task iteration. By combining initial ability map construction, dynamic ability weight vector generation, fuzzy inference of game parameters, and preference feature adaptation mechanism, it realizes the generation of personalized rehabilitation game tasks that match the individual abilities of children with cerebral palsy. At the same time, through standardized game task library configuration, positive incentive parameter mapping, and nonlinear strategy adjustment rules, it simplifies the personalized customization process of rehabilitation tasks, reduces the manual intervention burden and professional threshold of rehabilitation therapists, and improves the accuracy and personalization of rehabilitation training for children with cerebral palsy.

[0018] 2. This invention achieves continuous quantitative assessment of the rehabilitation status of children with cerebral palsy through synchronous acquisition of multimodal physiological data, real-time recording of task performance data, and multi-source feature fusion processing. At the same time, it dynamically adjusts the motor difficulty, cognitive load, and interaction rhythm of game tasks through a positive incentive grading matching mechanism and a nonlinear task adjustment strategy to ensure the suitability and fun of rehabilitation training. Furthermore, through clinical subtype-differentiated weighting, dynamic adaptation of rehabilitation goals, and real-time status closed-loop feedback, it improves the rehabilitation training compliance and long-term training effect of children with cerebral palsy, avoids the problems of insufficient training or excessive fatigue caused by fixed task difficulty, and realizes the personalized, intelligent, and sustainable development of rehabilitation training for children with cerebral palsy. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment, according to an embodiment of the present invention. Detailed Implementation

[0021] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0022] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0023] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, according to an embodiment of the present invention, a method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment includes the following steps: S1. Obtain motor function data, cognitive function data, emotional state data, and preference characteristic data of children with cerebral palsy to be trained to construct an initial ability map, and then make weighted corrections to the initial ability map according to clinical classification, and then generate a dynamic ability weight vector according to clinical rehabilitation goals. In this embodiment of the application, the steps of acquiring motor function data, cognitive function data, emotional state data, and preference feature data of children with cerebral palsy to be trained to construct an initial ability map, weighting and correcting the initial ability map according to clinical classification, and then generating a dynamic ability weight vector according to clinical rehabilitation goals include the following steps: S11. Preset the ability dimension framework, and collect corresponding motor function data, cognitive function data, emotional state data and preference feature data of children with cerebral palsy to be trained, and map them to the ability dimension framework to form an initial ability map. The method for constructing the initial capability map is as follows: The system pre-defines a capability dimension framework comprising four core dimensions: motor function, cognitive function, emotional state, and preference characteristics. Each dimension is further divided into three sub-dimensions, and standardized quantitative scoring criteria are defined. Multi-source raw data corresponding to children with cerebral palsy undergoing training is collected, including clinical rehabilitation scale assessment data, multimodal physiological monitoring data, rehabilitation therapist behavioral observation data, and preference data jointly confirmed by parents and children. The collected multi-source raw data undergoes format normalization and outlier removal to unify data interaction formats and scoring benchmarks. The pre-processed data is mapped to the corresponding three-level sub-dimensions of the capability dimension framework, completing the quantitative assignment of single-dimensional capabilities. The quantitative scoring results of each dimension are integrated to establish inter-dimensional correlation mapping relationships, generating a structured and visualized initial capability map.

[0024] It should be explained that the compliance standards for raw data include: Clinical scale data validity: Assessment was conducted using internationally recognized standardized scales such as GMFM-88 and Peabody-2, performed by certified rehabilitation therapists, with a scoring error of ≤5%; Physiological monitoring data completeness: Physiological data such as heart rate, skin conductance, and eye movement were continuously collected for ≥10 minutes, with a valid data segment ratio of ≥90%; Behavioral observation data objectivity: Observed and scored independently by two or more rehabilitation therapists, with an intragroup correlation coefficient of ≥0.85; Preference characteristic data authenticity: Obtained through a combination of children's self-selection and parental verification, without leading questions or guided choices; Data timeliness: All collected data are valid data within the last 7 days, accurately reflecting the child's current true ability and status.

[0025] The range parameters for constructing the initial capability map are as follows: Ability Dimension Hierarchy: A three-level hierarchical structure is adopted, with the first level being the core dimension, the second level being the ability category, and the third level being the specific assessment indicator, ensuring that the assessment granularity is fine and controllable; Single indicator scoring range: All three-level assessment indicator scores are uniformly quantified to the range of [0, 10], where 0 points indicate no corresponding ability and 10 points indicate that the ability reaches the level of normal children of the same age; Single collection time: The total time for simultaneous collection of multi-source data is controlled within 20-40 minutes to avoid data distortion caused by children's fatigue; Scale retest interval: The interval between two assessments of the same clinical scale is ≥3 days to eliminate scoring bias caused by the practice effect; Atlas validity period: The initial ability atlas is valid for 14 days, and the data recollection and atlas update process is automatically triggered after the expiration.

[0026] An isolated forest anomaly detection algorithm was used to remove outlier data points caused by sudden emotional fluctuations in children or interference from the external environment. Weighted mean smoothing was applied to multiple batches of data collected in the same dimension to improve the stability of single-dimensional ability scores. Assessment data from different sources and with different scales were uniformly normalized to the standard interval of [0, 10] to eliminate systematic bias caused by differences in assessment methods. Scoring was calibrated based on a standard ability norm library for children with cerebral palsy aged 0-12 years to ensure the objectivity and horizontal comparability of the initial ability map.

[0027] S12. Preset the focus of clinical rehabilitation goals, and based on the clinical classification of children with cerebral palsy, set the weighted correction coefficients corresponding to the ability dimensions to correct the dimensional weights of the initial ability map. Specifically, the method for adjusting dimension weights is as follows: The study prioritizes three clinical rehabilitation goals: motor function, cognitive function, and comprehensive ability improvement. It establishes a mapping relationship between clinical classifications such as spastic, athetoid, and ataxia and the basic weights of ability dimensions. The basic weights are then adjusted in conjunction with the rehabilitation goal priorities to generate the final weighted correction coefficients for each dimension. Finally, the weighted correction coefficients are multiplied by the corresponding dimension scores in the initial ability map to complete the dimension weight correction.

[0028] It should be explained that the principle for setting the basic weights for clinical classification is as follows: Spastic type: Motor function weight ≥ 0.6, cognitive function weight ≤ 0.2; Athetoid type: Motor function weight 0.4~0.5, emotional state weight 0.3~0.4; Ataxia type: Cognitive function weight 0.4~0.5, motor coordination sub-dimension weight ≥ 0.3.

[0029] The range parameters for dimension weight correction are as follows: The weighted correction coefficient ranges from 0.1 to 2.0; the corrected single-dimensional score ranges from 0 to 10, with automatic truncation outside this range; the weight adjustment step size is 0.05 to ensure adjustment accuracy. A dynamic weight fine-tuning mechanism is adopted, making small adjustments to the correction coefficient within ±0.1 based on the child's weekly rehabilitation progress to avoid sudden weight changes affecting training continuity.

[0030] S13. Combine the initial capability map after dimension weight correction with the focus of clinical rehabilitation goals, assign weights to each rehabilitation dimension, and generate a dynamic capability weight vector.

[0031] Specifically, the method for generating the dynamic capability weight vector is as follows: Extract the capability deficiency features of each dimension of the initial capability map after dimension weight correction; determine the priority of weight allocation for each rehabilitation dimension based on the focus of clinical rehabilitation goals and capability deficiency features; calculate the initial dynamic weight value of each dimension according to the priority; normalize the initial dynamic weight value to generate a standardized dynamic capability weight vector.

[0032] It should be explained that the weight allocation follows the following core principles: The principle of prioritizing weaknesses: the lower the score of a dimension, the higher the priority of its weight allocation; the principle of goal orientation: the weight of dimensions that are emphasized in the rehabilitation goal is increased by 20% to 30%; the principle of balance: the minimum weight of a single dimension is not less than 0.1, and non-core dimensions should not be completely ignored.

[0033] The range parameters of the dynamic capability weight vector are as follows: Vector dimensions: fixed at 4 dimensions, corresponding to the four core dimensions of motion, cognition, emotion, and preference; single-dimensional weight range: 0.1 to 0.6; weights and constraints: the sum of the weights of all dimensions is always equal to 1; vector update cycle: automatically updated every 7 days; weight adjustment step size: 0.05.

[0034] A capability change trend correction factor is introduced, and the dynamic weights are fine-tuned within a range of ±0.1 based on the rate of improvement of the child's capabilities in each dimension over the past two weeks. When a child is detected to have obvious emotional resistance, the weight of the emotional state dimension is temporarily increased to 0.4 to prioritize training compliance.

[0035] S2. Pre-set standardized game task library, non-linear strategy adjustment rules and positive incentive matching rules, configure adjustable parameters for each standardized game task, and establish a mapping relationship table between positive incentive parameters and game task execution status; In this embodiment of the application, the preset standardized game task library, nonlinear strategy adjustment rules, and positive incentive matching rules, which configure adjustable parameters for each standardized game task and establish a mapping relationship table between positive incentive parameters and game task execution status, include the following steps: S21. Categorize and organize rehabilitation game units of different difficulty according to the dimensions of rehabilitation training, and compile them into a standardized game task library; Specifically, the method for constructing a standardized game task library is as follows: The system includes a pre-defined classification system for rehabilitation training dimensions and difficulty grading standards; selection of game units that meet the rehabilitation guidelines for children with cerebral palsy and their standardized packaging; configuration of adjustable parameter templates and rehabilitation effect evaluation indicators for each game unit; establishment of an association index between game units and training dimensions and difficulty levels; and compilation of all standardized game units to form a structured standardized game task library.

[0036] It should be noted that the game quest library is built following these core principles: Rehabilitation-oriented principle: Each game unit corresponds to at least one clear rehabilitation training goal; Progressive difficulty principle: The increase in ability requirements between adjacent difficulty levels is controlled within 15% to 20%; Adjustable principle: Each game unit supports continuous adjustment of at least 3 core parameters; Safety principle: All game movements have been clinically validated and have no potential risk of sports injury.

[0037] The scope parameters of the standardized game task library are as follows: Number of training dimensions: no less than 6, covering upper limb movement, lower limb movement, hand-eye coordination, attention, memory, and reaction time; Number of difficulty levels: fixed at 5 levels, corresponding to children with cerebral palsy at different ability stages; Number of adjustable parameters per game: 3 to 8, covering movement difficulty, cognitive load, and interaction rhythm; Range of game duration per game: 5 to 20 minutes, adapting to the duration of children's attention; Minimum library capacity: no less than 200 standardized game units, covering the needs of the entire rehabilitation cycle.

[0038] An iterative feedback mechanism for game performance is introduced, and game units with poor performance are periodically eliminated based on children's rehabilitation training data; a dynamic adaptation mechanism for children's preferences is established, and children's game participation is counted every 14 days to adjust the game recommendation weight, prioritizing the fun and participation of training.

[0039] S22. Based on the pattern of ability changes in children with cerebral palsy, set nonlinear strategy adjustment rules and define the task escalation and demotion adjustment boundaries; Specifically, the method for setting the nonlinear strategy adjustment rules and upgrade / degrade boundaries is as follows: The study analyzed the distribution characteristics of the rate of ability improvement in children with cerebral palsy under different clinical classifications; constructed a nonlinear adjustment function with the magnitude of ability change as the independent variable; set the basic adjustment coefficients for different rehabilitation stages; and delineated the critical threshold ranges for task escalation and de-escalation based on clinical rehabilitation data.

[0040] It should be explained that the adjustment of the nonlinear strategy follows the following core principles: The principle of matching ability: the adjustment range is positively correlated with the change range of the child's ability, and the maximum adjustment range in a single instance is ≤30%; the principle of gradual progress: an upgrade can be triggered only after three consecutive successful tasks, and a downgrade is triggered if the error rate in a single task is ≥60%; the principle of stability: the adjustment coefficient is reduced by 50% during periods of ability fluctuation to avoid frequent changes in task difficulty.

[0041] The range parameters for the nonlinear strategy adjustment rules and the upgrade / degrade boundaries are as follows: Adjustment function type: adopts power function model, with an exponent value of 0.6 to 0.8; single difficulty adjustment step size: 0.1 to 0.3 difficulty levels; upgrade completion rate threshold: average completion rate of 3 consecutive tasks ≥ 80%; downgrade error rate threshold: single task completion rate ≤ 40%; boundary buffer: ±5%, used to filter out erroneous adjustments caused by accidental fluctuations.

[0042] An emotional state correction factor is introduced. When a child's emotional fluctuations are detected to be large, the adjustment range is reduced to 50% of the original range. When a child's ability does not show significant improvement for two consecutive weeks, the system automatically switches to a step-by-step linear adjustment mode to prioritize training stability.

[0043] S23. Develop positive incentive matching rules based on the cognitive acceptance level of children with cerebral palsy, and set incentive thresholds at different levels; Specifically, the method for setting the positive incentive matching rules and hierarchical thresholds is as follows: The system is designed with four main categories of incentives: visual, auditory, interactive, and achievement-based. It is divided into five incentive suitability levels based on the cognitive development level of children with cerebral palsy. A corresponding mapping relationship is established between cognitive acceptance and the complexity of incentive types. Incentive trigger thresholds are set for each level based on the quality of task completion and the extent of rehabilitation progress. The system is integrated to form a hierarchical and categorized positive incentive matching rule.

[0044] It should be explained that positive incentive matching follows the following core principles: Cognitive Adaptation Principle: Children under 3 years old only use basic sound and light-based incentives, while those over 6 years old can be introduced with achievement badges and leaderboard incentives; Gradual Progression Principle: Higher-level incentives are only unlocked after completing lower-level tasks consecutively; Personalization Principle: Prioritize matching the top 3 incentive types in the child's preference feature data; Positive Reinforcement Principle: All incentives are instantaneous feedback with a delay time of ≤2 seconds.

[0045] The range parameters for the positive stimulus matching rule and the hierarchical threshold are as follows: Number of incentive levels: fixed at 5 levels, corresponding to different task completion qualities; Number of incentive types per level: 3 to 6 types to avoid incentive monotony; Basic incentive trigger threshold: task completion rate ≥ 60%; Highest incentive trigger threshold: 5 consecutive task completion rates ≥ 90% with improved ability; Incentive display duration: 1 to 3 seconds to avoid interrupting training continuity.

[0046] A dynamic incentive preference update mechanism is introduced, which counts children's response time and attention to different incentives every 7 days and automatically adjusts the trigger priority of each type of incentive. When a decline in children's training enthusiasm is detected, the incentive level is temporarily increased by one level, and the original rules are restored after 1 to 2 task cycles.

[0047] S24. Configure adjustable parameters for motion difficulty, cognitive load, and interaction rhythm for standardized game tasks. Specifically, the configuration method for the adjustable parameters of the game tasks is as follows: It pre-sets three core parameter systems and grading standards for movement difficulty, cognitive load, and interaction rhythm; configures the basic adjustment range of corresponding parameters for each standardized game unit; establishes a linear mapping relationship between parameter values ​​and children's ability scores; and generates a unified format game parameter configuration template and embeds it into the task library.

[0048] It should be noted that the parameter configuration follows these core principles: Three-dimensional decoupling principle: Exercise difficulty, cognitive load, and interaction rhythm can be adjusted independently without interfering with each other; Ability adaptation principle: The parameter adjustment range corresponds one-to-one with the corresponding dimension ability score range; Continuous adjustability principle: All core parameters support continuous adjustment in 0.1 steps, without level gaps.

[0049] The adjustable parameters for the game tasks are as follows: The range of adjustment for exercise difficulty is 0.1 to 1.0, corresponding to the full range of abilities from passive assistance to autonomous completion; the range of adjustment for cognitive load is 0.1 to 1.0, corresponding to cognitive levels from single command to multi-task parallelism; the range of adjustment for interaction rhythm is 0.5 to 3.0 seconds / time, corresponding to response requirements from slow to fast; the parameter adjustment step size is fixed at 0.1 to ensure adjustment accuracy; the parameter update cycle is synchronized with the dynamic ability weight vector and updated once every 7 days.

[0050] An adaptive parameter calibration mechanism is introduced to automatically fine-tune the upper and lower limits of corresponding parameters based on the child's actual performance in three consecutive tasks; a safety parameter threshold interception mechanism is set to automatically trigger protection when the difficulty parameter of the exercise exceeds the child's current ability limit to avoid sports injuries.

[0051] S25. Divide the standardized game tasks into different execution state categories, associate and match the corresponding positive incentive parameters, and organize them into a mapping relationship table.

[0052] In this embodiment of the application, the steps of classifying the standardized game task into different execution state categories, associating and matching the corresponding positive incentive parameters, and organizing them into a mapping relationship table include the following steps: S251. Extract the execution features of standardized game tasks during the training process, and classify various task execution state categories based on the execution features; Specifically, the method for dividing task execution states is as follows: Four core execution features are extracted: task completion rate, average response time, error type distribution, and operation consistency. The execution features collected in real time are smoothed using a sliding window. The feature vectors are classified in an unsupervised manner using the K-means clustering algorithm. Standardized task execution status labels and feature threshold ranges are defined for each cluster.

[0053] It should be explained that the division of task execution states follows the following core principles: Clinical relevance principle: Each state category corresponds to a clear rehabilitation training effect and the child's degree of adaptation; Feature independence principle: There is no significant linear correlation between the selected executive features; State stability principle: The duration of a single state is not less than one complete game task cycle.

[0054] The parameters for dividing the task execution state are as follows: Number of core execution features: fixed at 4 categories, covering the main dimensions of task performance; Number of task execution status categories: 5 categories, namely excellent, good, average, difficult, and resistant; Feature sampling frequency: 1Hz, to ensure real-time status determination; Sliding window size: 30 seconds, to filter out instantaneous fluctuation interference; State switching delay: 10 seconds, to avoid frequent state jumps.

[0055] A physiological data-assisted correction mechanism is introduced, which combines physiological indicators such as heart rate and skin conductance to calibrate the state judgment results; an abnormal state warning threshold is set, and when a resistant state is detected to last for more than 2 task cycles, the task difficulty reduction process is automatically triggered.

[0056] S252. Associate and match the various task execution state categories with the positive stimulus parameters one by one, and solidify the correspondence between the execution state categories and the positive stimulus parameters to generate a mapping table.

[0057] Specifically, the method for generating the mapping table is as follows: Establish a basic one-to-one correspondence between 5 types of task execution states and 5 levels of positive incentive parameters; adjust the priority of incentive types in each state based on children's preference characteristics data; set incentive superposition rules when switching between consecutive states; solidify all correspondences and generate a structured, callable mapping table.

[0058] It should be explained that the mapping relationship is established following core principles: Intensity matching principle: The intensity of the incentive is positively correlated with the task performance status, and the incentive intensity for an excellent state is 2 to 3 times that for a normal state; Type matching principle: Under the same state, the top 2 incentive types preferred by children are prioritized; Superimposed incentive principle: When maintaining an excellent state for 2 consecutive times or more, an additional random surprise incentive is triggered; No punishment principle: Difficult and resistant states do not trigger negative incentives, and only basic encouragement incentives are provided.

[0059] The range parameters of the mapping table are as follows: Number of basic mapping relationships: 5 groups, corresponding to 5 types of task execution states; Dimension of single-state incentive parameters: 3 dimensions, including incentive type, intensity, and display duration; Number of special state mappings: 3 groups, covering scenarios of continuous excellence, state decline, and first achievement; Table update cycle: incentive type priority is updated synchronously every 14 days; Matching response time: ≤100ms, ensuring that incentives are triggered instantly.

[0060] A dynamic iteration mechanism for mapping relationships is introduced, and the incentive intensity coefficients of each state are fine-tuned every 7 days based on the children's response to different incentives. An abnormal state fallback mapping rule is established, and when an undefined transition state is detected, the incentive parameters of the adjacent lower-level state are automatically matched.

[0061] S3. Construct a game parameter adjustment model based on fuzzy inference algorithm, input the weighted and corrected initial ability map and dynamic ability weight vector into the game parameter adjustment model, calculate the game parameter combination, and generate personalized rehabilitation game tasks by combining preference feature data. In this embodiment of the application, the step of constructing a game parameter adjustment model based on a fuzzy inference algorithm, inputting the weighted and corrected initial ability map and dynamic ability weight vector into the game parameter adjustment model, calculating the game parameter combination, and generating a personalized rehabilitation game task by combining preference feature data includes the following steps: S31. Based on the fuzzy inference algorithm, a fuzzy input subset, a fuzzy rule base and a fuzzy decision criterion are set to build a game parameter adjustment model that is adapted to the cerebral palsy rehabilitation scenario. Specifically, the method for building the game parameter adjustment model is as follows: The weighted and corrected ability score and dynamic ability weight vector are used as model inputs. The corresponding fuzzy input subsets are divided and membership functions are defined. A fuzzy rule base of input features and output parameters is constructed based on the experience of clinical rehabilitation experts. The centroid method is selected as the fuzzy decision criterion to achieve precise conversion of fuzzy quantities. After completing model training and clinical validation, a special game parameter adjustment model adapted to cerebral palsy rehabilitation scenarios is generated.

[0062] It should be noted that the model was built following these core principles: Clinically-oriented principle: All fuzzy rules are jointly formulated by rehabilitation therapists with more than 3 years of experience; Input decoupling principle: Input features of each dimension are independently mapped to corresponding output parameters without cross interference; Reasoning transparency principle: Complete reasoning process logs are retained, supporting manual review by rehabilitation therapists; Boundary constraint principle: Output parameters are automatically limited to a preset safe range to avoid outliers.

[0063] The range of parameters for adjusting game parameters is as follows: Number of fuzzy input subsets: 8, corresponding to low, medium, and high levels of the four capability dimensions; Size of fuzzy rule base: no less than 120 rules, covering all common capability combination scenarios; Membership function type: triangular membership function, which is computationally efficient and adaptable; Fuzzy decision method: centroid method, with smooth and stable output results; Single inference time: ≤50ms, meeting real-time adjustment requirements.

[0064] A dynamic rule base update mechanism is introduced, and fuzzy rules are supplemented and optimized every 3 months based on clinical rehabilitation data; an input anomaly detection module is set up, and when the input data is detected to be outside the reasonable range, the previous valid input is automatically used for inference to ensure the stability of model operation.

[0065] S32. Import the weighted and corrected initial ability map and dynamic ability weight vector into the game parameter adjustment model, and use fuzzy reasoning rules to perform logical deduction to solve the game parameter combination that is suitable for the current child's ability. In this embodiment of the application, the step of importing the weighted and corrected initial ability map and dynamic ability weight vector into the game parameter adjustment model, and solving the game parameter combination that adapts to the current child's ability through logical deduction using fuzzy inference rules, includes the following steps: S321. Perform data quantization processing on the weighted and corrected initial capability map and dynamic capability weight vector, and convert them into standardized input quantities that the model can recognize. Specifically, the method for converting standardized input quantities is as follows: Extract the four-dimensional quantitative scores of the weighted and corrected capability map and the four-dimensional weight values ​​of the dynamic capability weight vector; unify the data format and align the dimensions to generate an 8-dimensional original input vector; map all values ​​to the standard input interval [0, 1] required by the model; add data validation markers to generate standardized input quantities that the model can recognize.

[0066] It should be explained that data quantification processing follows these core principles: Completeness principle: Retain all core dimension data without information loss; Consistency principle: All input data adopt the same quantization benchmark and precision; Stability principle: When the fluctuation range of input data is ≤5%, the output results remain stable; Traceability principle: Retain the complete log of the original data and quantization process.

[0067] The range parameters for standardized input quantities are as follows: Input vector dimension: fixed at 8 dimensions, corresponding to 4 ability scores + 4 weight values; quantization precision: retain 2 decimal places; standard input range: [0, 1]; single data processing time: ≤10ms; data validation pass rate: ≥99.9%.

[0068] A sliding window smoothing mechanism is introduced to perform weighted averaging on the input data collected three times consecutively, reducing the impact of instantaneous fluctuations; an outlier automatic replacement rule is set, which automatically replaces invalid data with the previous valid input value to ensure continuous model operation.

[0069] S322. Based on the fuzzy rule base, perform condition matching on the standardized input quantities, and import the condition-matched standardized input quantities into the game parameter adjustment model to generate multiple sets of candidate game parameter combinations through step-by-step logical deduction. Specifically, the method for generating the candidate game parameter combinations is as follows: The similarity between the standardized input and the preconditions of all rules in the fuzzy rule base is calculated; effective matching rules with similarity ≥ a set threshold are selected; after sorting the rules according to their priority, they are imported into the game parameter adjustment model; step-by-step fuzzy inference from input features to output parameters is performed; and 3 to 5 sets of candidate game parameter combinations that meet different rehabilitation focuses are generated.

[0070] It should be explained that the fuzzy reasoning process follows the following core principles: Priority matching principle: Rules corresponding to high-weight dimensions are matched and executed first; Multiple solutions principle: Multiple sets of parameter combinations with different emphases are generated for the same input condition; Inference consistency principle: The deviation of inference results is ≤5% under the same input condition; Safety fallback principle: All candidate combinations are automatically filtered to remove parameters that exceed the safety threshold.

[0071] The range parameters for fuzzy inference and parameter generation are as follows: Rule matching similarity threshold: 0.7; Number of reasoning levels: 3, corresponding to input layer, rule layer, and output layer; Number of candidate combinations: 3 to 5, focusing on rehabilitation effect, fun, and safety respectively; Single reasoning time: ≤30ms; Rule matching accuracy: ≥95%.

[0072] A rule conflict resolution mechanism is introduced, and when multiple rules match simultaneously, a weighted voting method is used to determine the final execution rule; a candidate combination pre-evaluation mechanism is established, and the expected effects of each group of parameters are scored and ranked based on historical rehabilitation data; an inference anomaly interception module is set up, and when the inference results show abnormal fluctuations, the default safe parameter combination is automatically adopted.

[0073] S323. Determine the degree of fit between each group of candidate game parameter combinations and the ability map, as well as the degree of fit with the dynamic ability weight vector. Use the degree of fit and the degree of fit as the criteria for selection to obtain game parameter combinations that are suitable for the current rehabilitation ability of children.

[0074] Specifically, the method for selecting the optimal combination of game parameters is as follows: Define the formulas for calculating the fit between parameter combinations and ability maps, and the formulas for calculating the fit between dynamic weight vectors; calculate the fit score and fit score for each candidate combination; calculate the comprehensive score by weighting according to preset weights; sort all candidate combinations in descending order of comprehensive score; select the combination with the highest comprehensive score that meets the safety threshold requirement as the final game parameter combination.

[0075] It should be explained that the parameter combination selection follows the following core principles: Comprehensive judgment principles: The weighting ratio of fit and suitability is 6:4, taking into account both ability matching and goal orientation; Safety veto principle: Any combination with parameters exceeding the safety threshold is directly excluded; Gradient retention principle: The top 3 combinations in the comprehensive score are all retained as alternatives; Explainability principle: Output the scoring basis of the final combination and the detailed scores of each dimension.

[0076] The range parameters for parameter combination filtering are as follows: Fit weighting coefficient: 0.6; Fit weighting coefficient: 0.4; Overall score passing threshold: ≥0.75; Number of candidate combinations to retain: 3 groups; Calculation time per screening: ≤20ms.

[0077] A child preference correction factor is introduced to increase the overall score of parameter combinations that include children's preferred game types by 5% to 10%; a historical effect backtracking mechanism is established to fine-tune the overall score within ±5% by referring to the training effect of the child's last three training sessions with the same parameter combinations; when the overall score of all candidate combinations is lower than the qualified threshold, the model parameters are automatically recalibrated and candidate combinations are regenerated.

[0078] S33. Match the game parameter combination with the standardized game task library, and integrate the matched standardized game tasks with preference feature data to generate personalized rehabilitation game tasks.

[0079] In this embodiment of the application, the step of matching the game parameter combination with a standardized game task library, and integrating the matched standardized game tasks with preference feature data to adapt and generate personalized rehabilitation game tasks includes the following steps: S331. Using the selected game parameter combinations as a matching benchmark, compare the adjustable parameters of standardized game tasks in the standardized game task library, and select the standardized game task with the highest parameter matching degree as the basic rehabilitation task. Specifically, the matching method for basic rehabilitation tasks is as follows: Extract the core features of movement difficulty, cognitive load, and interaction rhythm of the selected game parameter combinations; traverse the adjustable parameter range of all tasks in the standardized game task library; calculate the multidimensional matching score of each task with the baseline parameters; and select the task with the highest matching degree and that meets the rehabilitation goal requirements as the basic rehabilitation task.

[0080] It should be explained that task matching follows these core principles: Parameter matching priority principle: the matching degree of core parameters accounts for ≥70%; rehabilitation goal consistency principle: the task must cover the training dimension corresponding to the current rehabilitation goal; low repetition principle: the same task should not be used more than twice in a row to avoid training fatigue.

[0081] The range parameters for matching basic rehabilitation tasks are as follows: Matching degree calculation dimensions: 3 dimensions, corresponding to three core parameters; single-dimensional matching weight: exercise difficulty 0.4, cognitive load 0.3, interaction rhythm 0.3; minimum matching degree threshold: ≥0.7; number of candidate tasks: 5, for rehabilitation therapists to manually review and select; single matching time: ≤20ms.

[0082] A historical performance correction mechanism is introduced, which increases the matching score by 5% for similar tasks with a historical training performance score of ≥85 for children; a task rotation mechanism is set up, which prioritizes tasks that have not been used in the past 30 days when the matching difference is ≤0.05, thereby increasing the freshness of training.

[0083] S332. Embed preference feature data into the game structure of basic rehabilitation tasks, and adapt and adjust the basic rehabilitation tasks based on preference features to generate personalized rehabilitation game tasks.

[0084] Specifically, the method for generating personalized rehabilitation game tasks is as follows: Extract four core preference dimensions—theme, role, color, and sound effect—from children's preference feature data; determine the embedding position and adjustment method of each preference dimension in the game structure; perform targeted adaptation of visual elements, interaction logic, and feedback forms for basic rehabilitation tasks; retain the core rehabilitation training movements unchanged to generate personalized rehabilitation game tasks.

[0085] It should be explained that preference embedding and adaptation follow these core principles: The core principle of rehabilitation remains unchanged: all adjustments should not change the core rehabilitation training goals and movement requirements of the task; the principle of appropriate adaptation: the adjustment of a single dimension preference should not exceed 30% to avoid excessive entertainment; the principle of multi-dimensional integration: at least two core preference features should be integrated at the same time to enhance the degree of personalization.

[0086] The parameters for the scope of personalized rehabilitation game tasks are as follows: Number of embeddable preference dimensions: 4, covering all aspects of visual, auditory, and interactive aspects; single-dimensional adjustment range: 10% to 30%; number of modifiable positions in the game structure: no less than 6, including background, characters, props, sound effects, etc.; single adaptation time: ≤100ms; preference feature coverage: ≥90% of core preferences can be reflected in the task.

[0087] A dynamic preference adaptation mechanism is introduced to adjust the display frequency and intensity of preferred elements based on children's real-time game responses; a rehabilitation effect priority mechanism is set up so that when preference adjustments are detected to cause a decline in training effect, the adjustment magnitude of the corresponding dimension is automatically reduced to prioritize the quality of rehabilitation training.

[0088] S4. Perform personalized rehabilitation game tasks, collect modal physiological data and task performance data in real time during the execution process, and fuse them to obtain real-time rehabilitation status data; In this embodiment of the application, the execution of personalized rehabilitation game tasks, the real-time collection of modal physiological data and task performance data during the execution process, and the fusion of these data to obtain real-time rehabilitation status data include the following steps: S41. Implement personalized rehabilitation game tasks to guide children with cerebral palsy to carry out rehabilitation interactive training, and collect multiple types of modal physiological data simultaneously during the task execution process, and then record the children's task completion status as task performance data. Specifically, the methods for task execution and data synchronization collection are as follows: Load personalized rehabilitation game tasks and display the guidance interface; guide children to complete standard rehabilitation interactive actions through voice and vision; simultaneously start multimodal physiological data acquisition equipment; record children's operation trajectory, response time and task completion status in real time; generate timestamp-aligned physiological data and task performance datasets.

[0089] It should be explained that task execution and data collection follow these core principles: Strict synchronization principle: the time stamp alignment error between physiological data and task performance data is ≤10ms; non-interference principle: all data collection devices are wearable, with a single device weight of ≤50g, so as not to affect the child's completion of the action; integrity principle: the proportion of effective data segments is ≥90%, and there is no missing data at key nodes.

[0090] The scope parameters for task execution and data acquisition are as follows: Physiological data types collected: 4 types, including heart rate, skin conductance, eye movement, and facial expression; physiological data sampling frequency: heart rate 1Hz, skin conductance 10Hz, eye movement 30Hz; task performance data dimensions: 4 dimensions, including completion rate, average response time, number of errors, and operation trajectory; data storage precision: retaining 2 decimal places; single task data volume: ≤5MB, facilitating fast transmission and processing.

[0091] A real-time data quality monitoring mechanism is introduced to automatically remove invalid data segments caused by equipment detachment or environmental interference; a physiological abnormality early warning mechanism is set up so that when a child's heart rate is detected to be outside the normal range of ±20%, the task is automatically paused and a rehabilitation therapist is notified to intervene.

[0092] S42. The collected multi-modal physiological data are preprocessed, and the preprocessed multi-modal physiological data and task performance data are subjected to feature alignment and dimension normalization. Then, the multi-source data information is integrated by feature fusion, and the real-time rehabilitation status data is quantitatively output.

[0093] In this embodiment of the application, the steps of preprocessing the collected multi-modal physiological data, performing feature alignment and dimensionality normalization on the preprocessed multi-modal physiological data and task performance data, and then integrating the multi-source data information using a feature fusion method to quantify and output real-time rehabilitation status data include the following steps: S421. Organize and clean the collected multi-modal physiological data, align the processed modal physiological data with the task performance data in time series, eliminate the differences in the units of measurement between different data, and complete the dimension normalization process. Specifically, the multi-source data fusion preprocessing method is as follows: The system performs missing value completion, outlier removal, and noise smoothing on multimodal physiological data; unifies the sampling time benchmark between physiological data and task performance data; performs frame-by-frame temporal matching and alignment of the two types of data based on timestamps; and performs interval mapping and conversion on data of different dimensions to unify and normalize them to a standard numerical range.

[0094] It should be explained that data preprocessing follows these core principles: Timing synchronization principle: The timing alignment deviation between the two types of data is controlled within a single frame interval; Lossless cleaning principle: Data cleaning only removes invalid noise and retains valid physiological and behavioral characteristics; Dimensional unification principle: All dimensions of data are normalized to the same interval without changing the original feature distribution.

[0095] The range parameters for data preprocessing are as follows: Timing alignment error: ≤10ms; Data normalization standard interval: uniformly mapped to [0,1]; Outlier judgment threshold: data that deviates from the mean by more than three times the standard deviation are directly removed; Noise smoothing window: mean filtering is performed on 5 consecutive frames of data; Preprocessing time per cycle: ≤15ms.

[0096] An adaptive timing calibration mechanism is introduced to automatically compensate for delay deviations from different acquisition devices; data integrity verification rules are set, and if the proportion of valid data is lower than a threshold, it is marked for re-acquisition to ensure the accuracy of subsequent rehabilitation status analysis.

[0097] S422. Extract the feature components of the normalized modal physiological data and task performance data, assign weights, and integrate the weighted feature components to generate comprehensive rehabilitation feature information. Then, quantify and rate the comprehensive rehabilitation feature information to obtain real-time rehabilitation status data.

[0098] Specifically, the method for generating real-time rehabilitation status data is as follows: Emotional features, physiological load features, and task completion features are extracted from normalized modal physiological data and task performance data, respectively. Dimensional weights are assigned based on the degree of influence of each feature on the rehabilitation status. All weighted feature components are weighted, summed, and integrated to construct comprehensive rehabilitation feature information. The comprehensive rehabilitation feature information is then hierarchically quantified according to preset rating standards to output standardized real-time rehabilitation status data.

[0099] It should be explained that feature weighting and state rating follow the following core principles: Objective weighting principle: The weights of physiological characteristics and task performance characteristics are fixed and not subject to subjective adjustment; Feature complementarity principle: Physiological abnormality characteristics and task behavior characteristics are cross-verified to improve the accuracy of status judgment; Equal hierarchical division principle: The feature score range corresponding to each rehabilitation status level is kept consistent.

[0100] The range parameters for generating real-time rehabilitation status are as follows: Extracting core feature components: 6 items, covering three major categories: physiological, emotional, and task performance; Physiological feature weight range: 0.4-0.5; Task performance feature weight range: 0.5-0.6; Rehabilitation status rating level: divided into 5 levels, corresponding to extremely poor, poor, average, good, and excellent; Comprehensive feature quantification range: uniformly set to [0,100]; Time taken for single feature integration and rating: ≤20ms.

[0101] A dynamic feature correction factor is introduced to fine-tune the weighting of physiological features based on the child's daily emotional baseline. A feature missing judgment rule is set up so that when a certain type of feature data is missing, the historical average value is used to supplement it in the weighted calculation to ensure that real-time rehabilitation status data can be continuously output.

[0102] S5. Match the real-time rehabilitation status data with the mapping relationship table, generate real-time incentive signals according to the positive incentive matching rules, and then combine the nonlinear strategy adjustment rules to make real-time adaptive adjustments to the personalized rehabilitation game tasks.

[0103] In this embodiment of the application, the steps of matching real-time rehabilitation status data with a mapping table, generating real-time incentive signals according to positive incentive matching rules, and then combining nonlinear strategy adjustment rules to perform real-time adaptive adjustment of personalized rehabilitation game tasks include the following steps: S51. The real-time rehabilitation status data is compared and matched with the mapping relationship table one by one, and the current state is determined according to the positive incentive matching rule. The corresponding real-time incentive signal is generated, and the nonlinear strategy adjustment rule is retrieved simultaneously to determine the adjustment direction and adjustment range of the current rehabilitation task. Specifically, the method for generating excitation signals and determining task adjustment is as follows: The system decomposes the multidimensional feature indicators of real-time rehabilitation status data; compares the similarity of each feature indicator with each execution status entry in the mapping relationship table; matches the corresponding incentive level according to the positive incentive matching rule and generates a real-time incentive signal; retrieves the preset nonlinear strategy adjustment rule, and determines the direction and quantitative adjustment range of task difficulty adjustment in combination with the status level.

[0104] It should be explained that feature matching and task tuning follow these core principles: The principle of precise feature matching: the execution status can be determined only when multiple features match simultaneously; the principle of immediate incentive: the determination of incentive level and the generation of incentive signal have no delay; the principle of prudent adjustment: only a single adjustment direction and fixed amplitude are determined at a time, and large adjustments across levels are prohibited.

[0105] The range parameters for determining the stimulus and regulation are as follows: Feature comparison dimensions: consistent with the execution state division dimensions, a total of 4 dimensions; Feature matching qualification threshold: similarity ≥ 0.75; Stimulus signal response delay: ≤ 50ms; Task adjustment direction: only supports three types: upgrade, downgrade, and maintain; Maximum adjustment range per time: not exceeding 0.3 difficulty levels.

[0106] A steady-state determination mechanism is introduced, requiring two consecutive determinations of the same rehabilitation state before incentive delivery and task adjustment are implemented; a boundary buffer protection mechanism is set up, which maintains the current task parameters unchanged by default when the child is in a critical state, avoiding frequent adjustments that may interfere with the training rhythm.

[0107] S52. Adjust the adjustable parameters corresponding to the movement difficulty, cognitive load, and interaction rhythm of the personalized rehabilitation game task according to the adjustment direction and adjustment range, and perform real-time adaptive adjustment of the rehabilitation game task.

[0108] Specifically, the adaptive adjustment method for rehabilitation game tasks is as follows: The baseline parameter values ​​for the original exercise difficulty, cognitive load, and interaction rhythm of the personalized rehabilitation game task are locked; the three core adjustable parameters are simultaneously increased or decreased according to the determined adjustment direction and quantitative adjustment range; it is verified whether each parameter after correction is within the preset safe value range; after boundary truncation correction of parameters exceeding the limit, the game task running configuration is updated to complete real-time adaptive adjustment.

[0109] It should be explained that the adaptive adjustment of task parameters follows the following core principles: Synchronous adjustment principle: Exercise difficulty, cognitive load, and interaction rhythm should be adjusted in the same direction and with the same amplitude; Boundary constraint principle: All parameters should not exceed the preset maximum and minimum value range after correction; Gradual adjustment principle: Large jumps across levels are strictly prohibited, and only small-amplitude step-by-step adjustments are supported; Experience priority principle: When children are in a state of emotional resistance, only the interaction rhythm should be slightly adjusted, without changing the difficulty parameters.

[0110] The adaptive adjustment range parameters for rehabilitation game tasks are as follows: Single parameter single adjustment range: 0.1 to 0.3 standard levels; exercise difficulty adjustment range: 0.1 to 1.0; cognitive load adjustment range: 0.1 to 1.0; interaction rhythm adjustment range: 0.5 to 3.0 seconds / time; parameter adjustment effective delay: ≤30ms.

[0111] A continuous adjustment cooling mechanism is introduced, allowing a maximum of two parameter adjustments within the same game task cycle to avoid frequent changes disrupting the training rhythm; a parameter history rollback mechanism is added, which automatically reverts to the previous version of effective parameter configuration if the child's training fit decreases significantly after adjustment.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment, characterized in that, Includes the following steps: S1. Obtain motor function data, cognitive function data, emotional state data, and preference characteristic data of children with cerebral palsy to be trained to construct an initial ability map, and then make weighted corrections to the initial ability map according to clinical classification, and then generate a dynamic ability weight vector according to clinical rehabilitation goals. S2. Pre-set standardized game task library, non-linear strategy adjustment rules and positive incentive matching rules, configure adjustable parameters for each standardized game task, and establish a mapping relationship table between positive incentive parameters and game task execution status; S3. Construct a game parameter adjustment model based on fuzzy inference algorithm, input the weighted and corrected initial ability map and dynamic ability weight vector into the game parameter adjustment model, calculate the game parameter combination, and generate personalized rehabilitation game tasks by combining preference feature data. S4. Perform personalized rehabilitation game tasks, collect modal physiological data and task performance data in real time during the execution process, and fuse them to obtain real-time rehabilitation status data; S5. Match the real-time rehabilitation status data with the mapping relationship table, generate real-time incentive signals according to the positive incentive matching rules, and then combine the nonlinear strategy adjustment rules to make real-time adaptive adjustments to the personalized rehabilitation game tasks.

2. The method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment according to claim 1, characterized in that, The process of acquiring motor function data, cognitive function data, emotional state data, and preference characteristic data of children with cerebral palsy to be trained, constructing an initial ability map, and then weighting and correcting the initial ability map according to clinical classification, and finally generating a dynamic ability weight vector according to clinical rehabilitation goals, includes the following steps: S11. Preset the ability dimension framework, and collect corresponding motor function data, cognitive function data, emotional state data and preference feature data of children with cerebral palsy to be trained, and map them to the ability dimension framework to form an initial ability map. S12. Preset the focus of clinical rehabilitation goals, and based on the clinical classification of children with cerebral palsy, set the weighted correction coefficients corresponding to the ability dimensions to correct the dimensional weights of the initial ability map. S13. Combine the initial capability map after dimension weight correction with the focus of clinical rehabilitation goals, assign weights to each rehabilitation dimension, and generate a dynamic capability weight vector.

3. The method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment according to claim 1, characterized in that, The preset standardized game task library, nonlinear strategy adjustment rules, and positive incentive matching rules, which configure adjustable parameters for each standardized game task and establish a mapping relationship table between positive incentive parameters and game task execution status, include the following steps: S21. Categorize and organize rehabilitation game units of different difficulty according to the dimensions of rehabilitation training, and compile them into a standardized game task library; S22. Based on the pattern of ability changes in children with cerebral palsy, set nonlinear strategy adjustment rules and define the task escalation and demotion adjustment boundaries; S23. Develop positive incentive matching rules based on the cognitive acceptance level of children with cerebral palsy, and set incentive thresholds at different levels; S24. Configure adjustable parameters for the movement difficulty, cognitive load, and interaction rhythm of standardized game tasks respectively; S25. Divide the standardized game tasks into different execution state categories, associate and match the corresponding positive incentive parameters, and organize them into a mapping relationship table.

4. The method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment according to claim 1, characterized in that, The process of constructing a game parameter adjustment model based on a fuzzy inference algorithm, inputting a weighted and corrected initial ability map and a dynamic ability weight vector into the game parameter adjustment model, calculating game parameter combinations, and generating personalized rehabilitation game tasks by combining preference feature data includes the following steps: S31. Based on the fuzzy inference algorithm, a fuzzy input subset, a fuzzy rule base and a fuzzy decision criterion are set to build a game parameter adjustment model that is adapted to the cerebral palsy rehabilitation scenario. S32. Import the weighted and corrected initial ability map and dynamic ability weight vector into the game parameter adjustment model, and use fuzzy reasoning rules to perform logical deduction to solve the game parameter combination that is suitable for the current child's ability. S33. Match the game parameter combination with the standardized game task library, and integrate the matched standardized game tasks with preference feature data to generate personalized rehabilitation game tasks.

5. The method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment according to claim 1, characterized in that, The process of executing personalized rehabilitation game tasks, collecting modal physiological data and task performance data in real time, and fusing them to obtain real-time rehabilitation status data includes the following steps: S41. Implement personalized rehabilitation game tasks to guide children with cerebral palsy to carry out rehabilitation interactive training, and collect multiple types of modal physiological data simultaneously during the task execution process, and then record the children's task completion status as task performance data. S42. The collected multi-modal physiological data are preprocessed, and the preprocessed multi-modal physiological data and task performance data are subjected to feature alignment and dimension normalization. Then, the multi-source data information is integrated by feature fusion, and the real-time rehabilitation status data is quantitatively output.

6. The method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment according to claim 1, characterized in that, The process of matching real-time rehabilitation status data with a mapping table, generating real-time incentive signals based on positive incentive matching rules, and then adaptively adjusting personalized rehabilitation game tasks in real time using nonlinear strategy adjustment rules includes the following steps: S51. The real-time rehabilitation status data is compared and matched with the mapping relationship table one by one, and the current state is determined according to the positive incentive matching rule. The corresponding real-time incentive signal is generated, and the nonlinear strategy adjustment rule is retrieved simultaneously to determine the adjustment direction and adjustment range of the current rehabilitation task. S52. Adjust the adjustable parameters corresponding to the movement difficulty, cognitive load, and interaction rhythm of the personalized rehabilitation game task according to the adjustment direction and adjustment range, and perform real-time adaptive adjustment of the rehabilitation game task.

7. The method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment according to claim 3, characterized in that, The process of classifying standardized game tasks into different execution state categories, associating and matching corresponding positive incentive parameters, and organizing them into a mapping table includes the following steps: S251. Extract the execution features of standardized game tasks during the training process, and classify various task execution state categories based on the execution features; S252. Associate and match the various task execution state categories with the positive stimulus parameters one by one, and solidify the correspondence between the execution state categories and the positive stimulus parameters to generate a mapping table.

8. The method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment according to claim 4, characterized in that, The process of importing the weighted and corrected initial ability map and dynamic ability weight vector into the game parameter adjustment model, and solving for the game parameter combination that adapts to the current child's ability through logical deduction using fuzzy inference rules, includes the following steps: S321. Perform data quantization processing on the weighted and corrected initial capability map and dynamic capability weight vector, and convert them into standardized input quantities that the model can recognize. S322. Based on the fuzzy rule base, perform condition matching on the standardized input quantities, and import the condition-matched standardized input quantities into the game parameter adjustment model to generate multiple sets of candidate game parameter combinations through step-by-step logical deduction. S323. Determine the degree of fit between each group of candidate game parameter combinations and the ability map, as well as the degree of fit with the dynamic ability weight vector. Use the degree of fit and the degree of fit as the criteria for selection to obtain game parameter combinations that are suitable for the current rehabilitation ability of children.

9. A method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment according to claim 4, characterized in that, The process of matching game parameter combinations with a standardized game task library, and then integrating the matched standardized game tasks with preference feature data to generate personalized rehabilitation game tasks includes the following steps: S331. Using the selected game parameter combinations as a matching benchmark, compare the adjustable parameters of standardized game tasks in the standardized game task library, and select the standardized game task with the highest parameter matching degree as the basic rehabilitation task. S332. Embed preference feature data into the game structure of basic rehabilitation tasks, and adapt and adjust the basic rehabilitation tasks based on preference features to generate personalized rehabilitation game tasks.

10. A method for generating rehabilitation game tasks for children with cerebral palsy based on adaptive adjustment according to claim 5, characterized in that, The process of preprocessing the collected multi-modal physiological data, performing feature alignment and dimensionality normalization on the preprocessed multi-modal physiological data and task performance data, and then integrating the multi-source data information using a feature fusion method to quantify and output real-time rehabilitation status data includes the following steps: S421. Organize and clean the collected multi-modal physiological data, align the processed modal physiological data with the task performance data in time series, eliminate the differences in the units of measurement between different data, and complete the dimension normalization process. S422. Extract the feature components of the normalized modal physiological data and task performance data, assign weights, and integrate the weighted feature components to generate comprehensive rehabilitation feature information. Then, quantify and rate the comprehensive rehabilitation feature information to obtain real-time rehabilitation status data.

Images