Rehabilitation exercise planning system based on computer assistance

Abstract

The invention relates to the technical field of rehabilitation exercise planning, in particular to a computer-aided rehabilitation exercise planning system, which comprises a data decomposition and feature extraction module, a data analysis and feature extraction module, a data analysis and feature extraction module, a data analysis and feature extraction module, a data analysis and feature extraction module, a data analysis and feature extraction module and a data analysis and feature extraction module, wherein the data analysis and feature extraction module collects rehabilitation data, sorts data according to a time sequence and decomposes data points to obtain single-group intrinsic mode functions and residual trends; key features are extracted for multiple parts of the body, and a rehabilitation data decomposition result is generated. According to the method, through careful collection and sorting of rehabilitation data, decomposition of data points, extraction of an intrinsic mode function and a residual trend, enhancement of capture of subtle changes in the data, and time sequence prediction, through reconstruction of the intrinsic mode function and the residual trend, accuracy and individuation of prediction are improved, a plan is adjusted more accurately, and the accuracy and the individuality of time sequence prediction are improved. The construction of the whole body motion relation graph and the extraction of the key motion mode optimize the understanding of the motion mode of the patient, and improve the pertinence and effectiveness of training.

Description

Technical Field

[0001] The present invention relates to the technical field of rehabilitation exercise planning, and in particular to a computer-assisted rehabilitation exercise planning system. Background Art

[0002] Rehabilitation motion planning is a technical field that integrates medicine, biomechanics, computer science, robotics, and data analysis to provide personalized and optimized rehabilitation treatment plans for injured or postoperative patients. By using sensor technology, motion capture, and machine learning methods, experts in this field are able to analyze patients' motor abilities and limitations and develop rehabilitation plans that are tailored to their current health conditions. These plans help patients improve joint flexibility, muscle strength, coordination, and restore motor function. Rehabilitation motion planning systems evaluate patients' motor performance, monitor rehabilitation progress, and adjust treatment plans based on real-time feedback to achieve the best treatment results.

[0003] Among them, the rehabilitation exercise planning system is used to help rehabilitation doctors and therapists develop, implement and monitor patients' rehabilitation training plans. This system generates personalized training programs by analyzing patients' needs and rehabilitation goals to support patients at all stages of the rehabilitation process. Its main uses include assessing patients' physical conditions, generating and adjusting personalized rehabilitation plans, monitoring rehabilitation progress, and providing interactive training environments through virtual reality technology. These systems improve the efficiency of rehabilitation training, ensure the safety and effectiveness of training, and help medical professionals optimize rehabilitation support and treatment strategies through detailed data analysis.

[0004] Existing technologies lack sensitivity and personalization in rehabilitation data processing, especially in multi-site analysis and prediction, which do not provide sufficient dimensions and accuracy, resulting in rehabilitation plans that cannot fully meet patient needs. Traditional time series prediction methods fail to fully capture the potential trends in complex data, affecting the comprehensiveness and accuracy of predictions. In terms of movement pattern analysis, there is a lack of effective methods to construct a whole-body movement relationship diagram, which affects the pertinence and effectiveness of training plans. Rehabilitation plan adjustments lack dynamism and real-time response, and do not fully consider the real-time changes of patients, which reduces rehabilitation efficiency and safety and delays the patient's recovery process. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a computer-assisted rehabilitation movement planning system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: A computer-assisted rehabilitation exercise planning system includes: The data decomposition and feature extraction module collects rehabilitation data, sorts them in chronological order, decomposes the data points to obtain a single set of intrinsic mode functions and residual trends, merges the results of all intrinsic mode functions, extracts key features for multiple parts of the body, and generates rehabilitation data decomposition results; The neural network time series prediction module reconstructs the time series of the intrinsic mode function and the residual trend based on the decomposition result of the rehabilitation data, calculates the predicted value of each component, combines the prediction results, predicts each part feature and movement pattern, and generates a multi-dimensional prediction result; The motion pattern and correlation analysis module extracts the prediction data of each part from the multi-dimensional prediction results, constructs a relationship diagram of the patient's whole body movement, uses connecting edges to represent the motion correlation between parts, aggregates neighbor node information, updates node features, extracts key motion patterns, and generates motion pattern analysis results; The rehabilitation plan and training optimization module identifies key movement patterns based on the movement pattern analysis results, adjusts the training content and sequence of multiple parts, automatically weights the information of neighbor nodes, and adjusts the local movement training intensity to generate an optimized rehabilitation plan.

[0007] As a further solution of the present invention, the steps of obtaining the intrinsic mode function and the residual trend are specifically as follows: Sort the rehabilitation data in chronological order to generate a sorted data sequence D; The sorted data sequence , decomposed by applying Hilbert-Huang transform, using the formula: ; ; Introducing dynamic adjustment coefficient and , calculate the natural mode components and residual trend , get the intrinsic modal components and residual trends before processing; in, and is a coefficient adjusted based on the characteristics of the data, is the first Points, is the total number of data points, is the number of components, Indicates frequency; For each of the natural mode components Perform energy normalization and retain the residual trend As part of the analysis, the formula was used: ; Get the normalized intrinsic mode function and residual trend; in, is the normalized The intrinsic mode function changes with time changes, is the integral, representing the sum of the energy over the entire data set.

[0008] As a further solution of the present invention, the steps of obtaining the rehabilitation data decomposition result are specifically as follows: The normalized intrinsic mode function is summarized as follows and residual trend To a data series, use the formula: ; Calculate the synthetic modal data series; in, is the comprehensive modal data sequence, are weights adjusted according to the variance of each mode function, is the standardized weight of the residual trend; The key feature extraction is performed on the comprehensive modal data sequence using the formula: ; Calculate the slope and curvature of the data to identify variation points and trend characteristics, and obtain key feature quantitative indicators; in, Quantitative indicators for key features, represents the first-order derivative of the comprehensive modal data sequence M, i.e., the slope, represents the second-order derivative of the integrated modal data sequence M, i.e., the curvature; Quantify the indicators based on the key characteristics Reconstruct the data points using the formula: ; Using functions Adjust the contribution of each data point to generate the rehabilitation data decomposition result; in, Decomposition results for rehabilitation data, is a threshold function, according to Adjust the value of contribution.

[0009] As a further solution of the present invention, the step of obtaining the multi-dimensional prediction result is specifically: Applying Hilbert-Huang transform, decomposing the rehabilitation data decomposition result into differentiated intrinsic mode functions to obtain an intrinsic mode time series; The future state of each intrinsic mode function time series is predicted using an autoregressive moving average model, and the formula is: ; Obtain the enhanced prediction results for each modal component; in, is the predicted value, is the autoregressive coefficient, It is before The observed value of the period, is the moving average coefficient, It is before The error term of the period, is the adjustment factor, represents the minimum weight of the time series, It is the maximum value of the error; The enhanced prediction results of each modal component are combined and a weighted linear superposition method is used to generate a prediction result time series; For differentiated rehabilitation parts and movement patterns, multivariate linear regression analysis is performed using the prediction result time series to determine the prediction results corresponding to each rehabilitation part and movement pattern, and obtain multi-dimensional prediction results for each part feature and movement pattern.

[0010] As a further solution of the present invention, the step of constructing a relationship diagram of the patient's whole body movement is specifically: From the multi-dimensional prediction results, filter the data of the target part by an index filtering algorithm to generate a part data sequence; Smoothing is applied to the site data series to reduce the impact of noise and outliers, using the formula: ; Generate smoothed data series; in, is the smoothed predicted value, is the original data sequence at time point The data value of is the neighborhood size, is the attenuation coefficient; Using the smoothed data sequence, the data is vectorized by vector representation, and the fit between the vectors is calculated using the formula: ; Generate goodness-of-fit results for the data vector; in, Indicates the part and parts The goodness of fit of the data vector between , Respectively represent the parts and The data vector, Representation vector and The elements, is the vector length; A relationship graph of whole-body movement is constructed based on the fitting results of the data vector, and the movement correlation between parts is represented by connecting edges. Each node represents a part, and the weight of each edge is calculated based on the fitting results of the data vector. The dynamic pattern of whole-body movement is represented by the node and edge relationship of the graph, and a relationship graph of the patient's whole-body movement is obtained.

[0011] As a further solution of the present invention, the step of obtaining the motion pattern analysis result is specifically: Using the whole body motion relationship graph, the neighbor node information is aggregated by calculating the weighted average of the adjacent node features, using the formula: ; Generate aggregated neighbor information; in, Is a node The aggregate eigenvalues ​​of yes The set of neighbor nodes of Neighbor node The characteristic value of Is a node and The weight of the edge between them; Use the aggregated neighbor node information to update the features of each node, combine the original features with the neighbor aggregate features, and use the formula: ; Determine the influence of original features and aggregated features, and generate updated node features; in, is the updated node feature, is the balance coefficient; According to the updated node features, the key motion patterns are identified and extracted using the formula: ; generating motion pattern analysis results; in, Is a node The critical rating of It is The weight of the feature, yes The update feature of Features.

[0012] As a further solution of the present invention, the steps of adjusting the training content and sequence of multiple parts are specifically as follows: The key patterns are identified from the motion pattern analysis results using the formula: ; Quantitatively analyze the influence of each mode and generate key movement mode scores; in, For Mode The critical rating of Is factor The influence weight of Yes Mode In the factor The performance indicators below, Is factor The attenuation coefficient of The impact analysis of the key motion modes is performed using the formula: ; Determine its impact on the training area and obtain the impact assessment of each key mode; in, Yes Mode The impact index, It is a feature The importance weight of Yes Mode In Features performance, It is a feature The enhancement factor of According to the impact assessment results of the key modes, the training content and sequence are optimized using the formula: ; Get adjustments to the training content and sequence of multiple parts; in, Indicates the training order, through the influence index Sort in descending order.

[0013] As a further solution of the present invention, the step of obtaining the optimized rehabilitation plan is specifically as follows: The whole body motion relationship graph is used to automatically calculate the neighbor node information of each node. The formula is: ; Aggregate the features of neighbor nodes to obtain weighted neighbor node information; in, Representation Node The aggregate eigenvalues ​​of yes The set of neighbor nodes of Neighbor node The characteristic value of Is a node and The connection weights between , is the nonlinear adjustment parameter, is a small constant to avoid division by zero errors; According to the weighted neighbor node information, the training intensity of each training part is adjusted using the formula: ; generating adjusted local exercise training intensity; in, is the adjusted training intensity, Is a node Basic training intensity, is the adjustment factor; The weighted neighbor node information and the adjusted local motion training intensity are integrated using the formula: ; Generate optimized rehabilitation plans; in, Indicates optimization of rehabilitation program.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, through the careful collection and sorting of rehabilitation data and the decomposition of data points, the new scheme extracts the intrinsic mode function and residual trend, and enhances the capture of subtle changes in the data. This improves the accuracy of the basic data and enables the rehabilitation plan to be more finely adapted to the needs of different parts. Time series prediction improves the accuracy and personalization of predictions by reconstructing the intrinsic mode function and residual trend, allowing plans to be adjusted more accurately. The construction of the whole-body motion relationship diagram and the extraction of key motion patterns optimize the understanding of the patient's motion pattern and improve the pertinence and effectiveness of training. The dynamic optimization mechanism ensures that the training content and intensity respond to patient feedback in real time, enhances the safety and personalization of training, and accelerates patient recovery. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a system flow chart of the present invention; Figure 2 A flow chart of the steps for obtaining the intrinsic mode function and residual trend of the present invention; Figure 3 A flowchart of the steps for obtaining the rehabilitation data decomposition results of the present invention; Figure 4A flowchart of the steps for obtaining the multi-dimensional prediction results of the present invention; Figure 5 A flowchart of the steps of constructing a relationship diagram of a patient's whole body movement according to the present invention; Figure 6 A flow chart of steps for obtaining the motion pattern analysis results of the present invention; Figure 7 A flowchart of the steps of adjusting the training content and sequence of multiple parts of the present invention; Figure 8 The present invention is a flowchart of the steps for obtaining an optimized rehabilitation plan. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0018] Example 1: Please refer to Figure 1 , the computer-assisted rehabilitation movement planning system includes: The data decomposition and feature extraction module collects rehabilitation data, sorts them in chronological order, decomposes the data points to obtain a single set of intrinsic mode functions and residual trends, merges the results of all intrinsic mode functions, extracts key features for multiple parts of the body, and generates rehabilitation data decomposition results; The neural network time series prediction module reconstructs the time series of the intrinsic mode function and residual trend based on the decomposition results of the rehabilitation data, calculates the predicted value of each component, merges the prediction results, predicts each part feature and movement pattern, and generates multi-dimensional prediction results; The motion pattern and correlation analysis module extracts the prediction data of each part from the multi-dimensional prediction results, constructs a relationship diagram of the patient's whole body movement, uses connecting edges to represent the motion correlation between parts, aggregates neighbor node information, updates node features, extracts key motion patterns, and generates motion pattern analysis results; The rehabilitation plan and training optimization module is based on the results of movement pattern analysis, identifies key movement patterns, adjusts the training content and sequence of multiple parts, automatically weights the information of neighbor nodes, and adjusts the local movement training intensity to generate an optimized rehabilitation plan.

[0019] The rehabilitation data decomposition results include the intrinsic mode function of time series decomposition, the residual trend of time series decomposition, and the data point location feature extraction results. The multi-dimensional prediction results include the intrinsic mode function prediction data, the residual trend prediction data, and the merged time series prediction results. The motion pattern analysis results include the motion relationship graph node update results, the motion correlation edge connection information between parts, and the extraction of key motion patterns. The optimization of the rehabilitation plan includes the adjustment of the part training content, the training sequence optimization results, and the local training intensity adjustment results.

[0020] See also Figure 2 , the specific steps for obtaining the intrinsic mode function and residual trend are: Sort the rehabilitation data in chronological order to generate a sorted data sequence D; After sorting the data sequence , decomposed by applying Hilbert-Huang transform, using the formula: ; ; Introducing dynamic adjustment coefficient and , calculate the natural mode components and residual trend , get the intrinsic modal components and residual trends before processing; in, and is a coefficient adjusted based on the characteristics of the data, is the first Points, is the total number of data points, is the number of components, Indicates frequency; For each natural mode component Perform energy normalization and retain the residual trend As part of the analysis, the formula was used: ; Get the normalized intrinsic mode function and residual trend; in, is the normalized The intrinsic mode function changes with time changes, is the integral, representing the sum of the energy over the entire data set.

[0021] Intrinsic modal components The calculation process Parameter Description: : Coefficient used to adjust the amplitude of the modal component, usually depends on the amplitude of the data. Assume that the amplitude of the data ranges from , you can take As an example; : Coefficient, used to adjust the impact of frequency components, generally related to the frequency of data changes. Assume that the frequency adjustment coefficient is Within the scope, we set ; : The first points, for example, in a specific set of data, ; : The total number of data points, for example 5; : The number of natural mode components, here set to 1; : The basic frequency of the data, set to .

[0022] Formula derivation: Assuming time The range is Seconds are divided into 100 time points.

[0023] ; Bring in specific parameters: ; Specific calculation (in time): ; ; Calculated .

[0024] Residual trend The calculation process: Formula derivation: ; ; Assumptions and Related, such as : ; Normalized Intrinsic Mode Function : Formula derivation: ; To simplify the calculation, assume The value of is const. (based on by is a simplified integral estimate of the variable): ; .

[0025] See also Figure 3 ,The specific steps for obtaining the rehabilitation data decomposition results are: Integrated Normalized Intrinsic Mode Function and residual trend To a data series, use the formula: ; Calculate the synthetic modal data sequence; in, is the comprehensive modal data sequence, are weights adjusted according to the variance of each mode function, is the standardized weight of the residual trend; Key feature extraction is performed on the synthetic modal data sequence using the formula: ; Calculate the slope and curvature of the data to identify variation points and trend characteristics, and obtain key feature quantitative indicators; in, Quantitative indicators for key features, represents the first-order derivative of the comprehensive modal data sequence M, i.e., the slope, represents the second-order derivative of the integrated modal data sequence M, i.e., the curvature; Quantify indicators based on key characteristics Reconstruct the data points using the formula: ; Using functions Adjust the contribution of each data point to generate the rehabilitation data decomposition result; in, Decomposition results for rehabilitation data, is a threshold function, according to Adjust the value of contribution.

[0026] Formula 1: Comprehensive modal data sequence: ; Parameter explanation: : Each intrinsic mode function The weight of is usually proportional to its contribution to signal reconstruction. In practical applications, Can be It is determined by the proportion of the total energy of the signal; : Residual trend The weights of , ensure that the residual trend is appropriately represented in the overall model. depending on The energy proportion in the total signal; : No. Intrinsic mode functions; : residual trend; :They are and The maximum value of is used for data normalization.

[0027] Derivation and examples: Assume there are two intrinsic mode functions and , and a residual trend , the given values ​​are as follows: In time The value of ; In time The value of ; In time The value of ; ; Calculate max .

[0028] ;

[0029] ; ; Formula 2: Key feature quantitative indicators: ; Parameter explanation: :They are The first and second time derivatives of The speed and acceleration of change over time.

[0030] Derivation and examples: Based on the previously calculated , , ; Calculate the K value: .

[0031] See also Figure 4 , the specific steps for obtaining multi-dimensional prediction results are: Applying Hilbert-Huang transform, the rehabilitation data decomposition results are decomposed into differentiated intrinsic mode functions to obtain the intrinsic mode time series; The future state of each intrinsic mode function time series is predicted using the autoregressive moving average model, and the formula is: ; Obtain the enhanced prediction results for each modal component; in, is the predicted value, is the autoregressive coefficient, It is before The observed value of the period, is the moving average coefficient, It is before The error term of the period, is the adjustment factor, represents the minimum weight of the time series, It is the maximum value of the error; The enhanced prediction results of each modal component are combined and a weighted linear superposition method is used to generate a time series of prediction results; For differentiated rehabilitation parts and movement patterns, multivariate linear regression analysis is performed using the prediction result time series to determine the prediction results corresponding to each rehabilitation part and movement pattern, and to obtain multi-dimensional prediction results for each part feature and movement pattern.

[0032] formula: ; Parameter description and calculation process : At time The predicted value of : Autoregression coefficient, indicating the previous The strength of the impact of the previous observation on the current value; :time The observed value of Before Data at a certain moment; : Moving average coefficient, indicating the previous The impact of past forecast errors on current forecasts; : At time The prediction error of : Adjustment coefficient, used to balance the impact of the moving average term; : Weighting coefficient of the minimum value of the time series; : Time series to time The minimum value of : To time The maximum prediction error.

[0033] Specific calculation examples Assume the following time series data Observed values ​​and prediction errors of : Observations ; Prediction Error ; Assume the coefficients are set as follows: ; ; ; ; ; .

[0034] Calculation process: Autoregressive term calculation: Autoregressive term ; Moving average calculation: Moving average term ; Minimum weighted term and maximum error term calculation: Minimum weighted

[0035] Maximum error term

[0036] Total predicted value: .

[0037] See also Figure 5 ,The steps of constructing the relationship diagram of the patient's whole body movement are as follows: From the multi-dimensional prediction results, the data of the target part is filtered through the index filtering algorithm to generate a part data sequence; Apply smoothing to the position data series to reduce the impact of noise and outliers, using the formula: ; Generate smoothed data series; in, is the smoothed predicted value, is the original data sequence at time point The data value of is the neighborhood size, is the attenuation coefficient; Using the smoothed data series, vectorize the data using vector representation and calculate the fit between the vectors using the formula: ; Generate goodness-of-fit results for the data vector; in, Indicates the part and parts The goodness of fit of the data vector between , Respectively represent the parts and The data vector, Representation vector and The elements, is the vector length; According to the fitting results of the data vector, a relationship graph of whole-body movement is constructed, and the connecting edges are used to represent the movement correlation between parts. Each node represents a part, and the weight of each edge is calculated according to the fitting results of the data vector. The dynamic pattern of whole-body movement is represented by the node and edge relationship of the graph, and a relationship graph of the patient's whole-body movement is obtained.

[0038] Detailed explanation of the formula and derivation process: Smoothing formula: ; Parameter Description: : Time point The smoothing value of : Time point in the original data sequence The value of : Neighborhood size, which determines how many time points before and after the data are considered for smoothing; : Attenuation coefficient, used to reduce the weight of neighborhood edge data.

[0039] Assumptions and examples: Assume that there is a data series ,choose .for (i.e. the value 40 in the original data), the smoothing calculation process is as follows: ; In this example, the data points The smoothing value of is 26.67, which reduces the influence of edge points by considering their own and neighboring point values ​​and applying a falloff coefficient.

[0040] The calculation formula of goodness of fit is: ; Parameter Description: :vector and The degree of fit between :vector and No. elements; : Dimension of the vector.

[0041] Assumptions and examples: Assume vector and , the fitting calculation process is as follows: ; This calculation shows and High degree of fit between .

[0042] See also Figure 6 , the specific steps for obtaining the motion pattern analysis results are: Using the whole body motion relationship graph, the neighbor node information is aggregated by calculating the weighted average of the adjacent node features, using the formula: ; Generate aggregated neighbor information; in, Is a node The aggregate eigenvalues ​​of yes The set of neighbor nodes of Neighbor node The characteristic value of Is a node and The weight of the edge between them; Use the aggregated neighbor node information to update the features of each node, combining the original features with the neighbor aggregate features, using the formula: ; Determine the influence of original features and aggregated features, and generate updated node features; in, is the updated node feature, is the balance coefficient; According to the updated node features, key motion patterns are identified and extracted using the formula: ; generating motion pattern analysis results; in, Is a node The critical rating of It is The weight of the feature, yes The update feature of Features.

[0043] Formula explanation and calculation process: formula: ; Parameter explanation: :node The aggregate feature value of :node and The weight of the edge between them is assumed to be in the range of 0.1 to 1.0, indicating the strength of the connection; : Neighboring nodes The characteristic value of can be any relevant measure, such as speed or power; :node The set of neighbor nodes.

[0044] Calculation process and examples: Assume Node There are three neighbors . Given the eigenvalue , The edge weights are .

[0045] calculate : ; formula: ; Parameter explanation: : Updated node features; : Balance coefficient, assuming a value of 0.5, to balance the influence of original features and aggregated features; :node The original characteristics of : Aggregate features calculated in the previous step.

[0046] Calculation process and examples: Given , calculated previously ; .

[0047] formula: ; Parameter explanation: :node Criticality ratings; : No. The weight of the feature is assumed to be ; : Update the first features, obtained from the above calculations.

[0048] Calculation process and examples: Assumptions It includes three features: .

[0049] .

[0050] See also Figure 7 , the specific steps to adjust the training content and sequence of multiple parts are: To identify key patterns from the motion pattern analysis results, the formula is used: ; Quantitatively analyze the influence of each mode and generate key movement mode scores; in, For Mode The critical rating of Is factor The influence weight of Yes Mode In the factor The performance indicators below, Is factor The attenuation coefficient of Impact analysis of key motion modes using the formula: ; Determine its impact on the training area and obtain the impact assessment of each key mode; in, Yes Mode The impact index, It is a feature The importance weight of Yes Mode In Features performance, It is a feature The enhancement factor of According to the impact assessment results of key modes, the training content and sequence are optimized using the formula: ; Get adjustments to the training content and sequence of multiple parts; in, Indicates the training order, through the influence index Sort in descending order.

[0051] Criticality scoring formula: ; Impact index formula: ; Detailed explanation of formula and examples: Criticality Scoring Formula : Parameter description and assumed values: : The total number of factors, assumed to be 3 (e.g., frequency, statistical significance, and complexity); :factor The influence weights of ; :model In the factor The performance indicators under the assumption are ; :factor The attenuation coefficients are assumed to be .

[0052] Calculation process: Calculate the sum of the factor influence weights: ; Calculate a weighted performance index for each factor, taking into account decay: ; Total weighted performance: ; Critical Rating: ; Impact Index Formula : Parameter description and assumed values: : Total number of characteristics, assumed to be 3 (e.g., strength, flexibility, endurance); :feature The importance weights are assumed to be ; :model In Features The performance of ; :feature The enhancement coefficients are assumed to be .

[0053] Calculation process: Calculate the sum of feature importance weights: ; Calculate the weighted performance for each feature, taking into account the enhancement factor: ; Total Weighted Performance: ; Impact Index: .

[0054] See also Figure 8 , the steps to obtain the optimized rehabilitation plan are as follows: Using the whole body motion relationship graph, automatically calculate the neighbor node information of each node. Use the formula: ; Aggregate the features of neighbor nodes to obtain weighted neighbor node information; in, Representation Node The aggregate eigenvalues ​​of yes The set of neighbor nodes of Neighbor node The characteristic value of Is a node and The connection weights between , is the nonlinear adjustment parameter, is a small constant to avoid division by zero errors; According to the weighted neighbor node information, adjust the training intensity of each training part using the formula: ; generating adjusted local exercise training intensity; in, is the adjusted training intensity, Is a node Basic training intensity, is the adjustment factor; Integrate the weighted neighbor node information and the adjusted local motion training intensity, using the formula: ; Generate optimized rehabilitation plans; in, Indicates optimization of rehabilitation program.

[0055] formula: ; Parameter Description: :node and The connection weight between . Assume ; : Neighboring nodes Assume that ; , : Nonlinear adjustment parameter, assumed to be 2; : A small constant to avoid division by zero error, assuming .

[0056] Calculation process: Assume Node There are two neighbors and , then the calculation is as follows: ; ; ; ; ; ; formula: ; Parameter Description: : Basic training intensity, assumed to be 50; : Adjustment coefficient, assumed to be 0.01.

[0057] Calculation process: ; formula: ; Parameter description and calculation process: Assume this is a simple integration with only one value, so: .

[0058] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A computer-assisted rehabilitation exercise planning system, characterized in that: The system comprises: The data decomposition and feature extraction module collects rehabilitation data, sorts them in chronological order, decomposes the data points to obtain a single set of intrinsic mode functions and residual trends, merges the results of all intrinsic mode functions, extracts key features for multiple parts of the body, and generates rehabilitation data decomposition results; The neural network time series prediction module reconstructs the time series of the intrinsic mode function and the residual trend based on the decomposition result of the rehabilitation data, calculates the predicted value of each component, combines the prediction results, predicts each part feature and movement pattern, and generates a multi-dimensional prediction result; The motion pattern and correlation analysis module extracts the prediction data of each part from the multi-dimensional prediction results, constructs a relationship diagram of the patient's whole body movement, uses connecting edges to represent the motion correlation between parts, aggregates neighbor node information, updates node features, extracts key motion patterns, and generates motion pattern analysis results; The rehabilitation plan and training optimization module identifies key movement patterns based on the movement pattern analysis results, adjusts the training content and sequence of multiple parts, automatically weights the information of neighbor nodes, and adjusts the local movement training intensity to generate an optimized rehabilitation plan.

2. The computer-aided rehabilitation exercise planning system according to claim 1, characterized in that: The steps for obtaining the intrinsic mode function and the residual trend are specifically as follows: Sort the rehabilitation data in chronological order to generate a sorted data sequence D; The sorted data sequence , decomposed by applying Hilbert-Huang transform, using the formula: ； ； Introducing dynamic adjustment coefficient and , calculate the natural mode components and residual trend , get the intrinsic modal components and residual trends before processing; in, and is a coefficient adjusted based on the characteristics of the data, is the first Points, is the total number of data points, is the number of components, Indicates frequency; For each of the natural mode components Perform energy normalization and retain the residual trend As part of the analysis, the formula was used: ； Get the normalized intrinsic mode function and residual trend; in, is the normalized The intrinsic mode function changes with time changes, is the integral, representing the sum of the energy over the entire data set.

3. The computer-aided rehabilitation movement planning system according to claim 2, characterized in that: The steps for obtaining the rehabilitation data decomposition result are specifically as follows: The normalized intrinsic mode function is summarized as follows and residual trend To a data series, use the formula: ； Calculate the synthetic modal data sequence; in, is the comprehensive modal data sequence, are weights adjusted according to the variance of each mode function, is the standardized weight of the residual trend; The key feature extraction is performed on the comprehensive modal data sequence using the formula: ； Calculate the slope and curvature of the data to identify variation points and trend characteristics, and obtain key feature quantitative indicators; in, Quantitative indicators for key features, represents the first-order derivative of the comprehensive modal data sequence M, i.e., the slope, represents the second-order derivative of the integrated modal data sequence M, i.e., the curvature; Quantify the indicators based on the key characteristics Reconstruct the data points using the formula: ； Using functions Adjust the contribution of each data point to generate the rehabilitation data decomposition result; in, Decomposition results for rehabilitation data, is a threshold function, according to Adjust the value of contribution.

4. The computer-aided rehabilitation exercise planning system according to claim 1, characterized in that: The steps for obtaining the multi-dimensional prediction results are specifically as follows: Applying Hilbert-Huang transform, decomposing the rehabilitation data decomposition result into differentiated intrinsic mode functions to obtain an intrinsic mode time series; The future state of each intrinsic mode function time series is predicted using an autoregressive moving average model, and the formula is: ； Obtain the enhanced prediction results for each modal component; in, is the predicted value, is the autoregressive coefficient, It is before The observed value of the period, is the moving average coefficient, It is before The error term of the period, is the adjustment factor, represents the minimum weight of the time series, It is the maximum value of the error; The enhanced prediction results of each modal component are combined and a weighted linear superposition method is used to generate a prediction result time series; For differentiated rehabilitation parts and movement patterns, multivariate linear regression analysis is performed using the prediction result time series to determine the prediction results corresponding to each rehabilitation part and movement pattern, and obtain multi-dimensional prediction results for each part feature and movement pattern.

5. The computer-aided rehabilitation exercise planning system according to claim 1, characterized in that: The steps of constructing the relationship diagram of the patient's whole body movement are specifically as follows: From the multi-dimensional prediction results, filter the data of the target part by an index filtering algorithm to generate a part data sequence; Smoothing is applied to the site data series to reduce the impact of noise and outliers, using the formula: ； Generate smoothed data series; in, is the smoothed predicted value, is the original data sequence at time point The data value of is the neighborhood size, is the attenuation coefficient; Using the smoothed data sequence, the data is vectorized by vector representation, and the fit between the vectors is calculated using the formula: ； Generate goodness-of-fit results for the data vector; in, Indicates the part and parts The goodness of fit of the data vector between , Respectively represent the parts and The data vector, Representation vector and The elements, is the vector length; A relationship graph of whole-body movement is constructed based on the fitting results of the data vector, and the movement correlation between parts is represented by connecting edges. Each node represents a part, and the weight of each edge is calculated based on the fitting results of the data vector. The dynamic pattern of whole-body movement is represented by the node and edge relationship of the graph, and a relationship graph of the patient's whole-body movement is obtained.

6. The computer-aided rehabilitation exercise planning system according to claim 1, characterized in that: The steps for obtaining the motion pattern analysis result are specifically as follows: Using the whole body motion relationship graph, the neighbor node information is aggregated by calculating the weighted average of the adjacent node features, using the formula: ； Generate aggregated neighbor information; in, Is a node The aggregate eigenvalues ​​of yes The set of neighbor nodes of Neighbor node The characteristic value of Is a node and The weight of the edge between them; Use the aggregated neighbor node information to update the features of each node, combine the original features with the neighbor aggregate features, and use the formula: ； Determine the influence of original features and aggregated features, and generate updated node features; in, is the updated node feature, is the balance coefficient; According to the updated node features, the key motion patterns are identified and extracted using the formula: ； generating motion pattern analysis results; in, Is a node The critical rating of It is The weight of the feature, yes The update feature of Features.

7. The computer-aided rehabilitation exercise planning system according to claim 1, characterized in that: The steps of adjusting the training content and sequence of multiple parts are specifically as follows: The key pattern is identified from the motion pattern analysis results using the formula: ； Quantitatively analyze the influence of each mode and generate key movement mode scores; in, For Mode The critical rating of Is factor The influence weight of Yes Mode In the factor The performance indicators below, Is factor The attenuation coefficient of The impact analysis of the key motion modes is performed using the formula: ； Determine its impact on the training area and obtain the impact assessment of each key mode; in, Yes Mode The impact index, It is a feature The importance weight of Yes Mode In Features performance, It is a feature The enhancement factor of According to the impact assessment results of the key modes, the training content and sequence are optimized using the formula: ； Get adjustments to the training content and sequence of multiple parts; in, Indicates the training order, through the influence index Sort in descending order.

8. The computer-aided rehabilitation exercise planning system according to claim 1, characterized in that: The steps for obtaining the optimized rehabilitation plan are specifically as follows: The whole body motion relationship graph is used to automatically calculate the neighbor node information of each node using the formula: ； Aggregate the features of neighbor nodes to obtain weighted neighbor node information; in, Representation Node The aggregate eigenvalues ​​of yes The set of neighbor nodes of Neighbor node The characteristic value of Is a node and The connection weights between , is the nonlinear adjustment parameter, is a small constant to avoid division by zero errors; According to the weighted neighbor node information, the training intensity of each training part is adjusted using the formula: ； generating adjusted local exercise training intensity; in, is the adjusted training intensity, Is a node Basic training intensity, is the adjustment factor; The weighted neighbor node information and the adjusted local motion training intensity are integrated using the formula: ； Generate optimized rehabilitation plans; in, Indicates optimization of rehabilitation program.

Images