Children rehabilitation training system based on scene interaction

Abstract

The invention discloses a children rehabilitation training system based on scene interaction, relates to the technical field of rehabilitation training, and is technically characterized in that the system comprises a scene simulation module, a motion capture module, a data processing module, an interaction feedback module and a training planning module, and a complete technical closed loop is formed through cooperation of multiple modules. The situation simulation module is used for creating a dynamic situation by utilizing a virtual reality or augmented reality technology and generating personalized training parameters; the motion capture module obtains child motion data in real time through a multi-source sensor fusion technology; the data processing module evaluates the matching degree of the child action and the target action through a quantification algorithm, and generates an action matching score; the interactive feedback module is combined with multi-mode real-time feedback to improve the training interestingness and participation degree; the training plan module dynamically adjusts the training target and difficulty according to the real-time performance and the rehabilitation trend. According to the system, the interestingness, scientificity and individuation level of rehabilitation training of children are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of rehabilitation training, and in particular to a children's rehabilitation training system based on situational interaction. Background Art

[0002] With the development of medical rehabilitation technology, child rehabilitation training has gradually become an important intervention method for children with cerebral palsy, autism, developmental delay, etc. The goal of child rehabilitation training is to help children improve their motor skills, cognitive functions and social skills through scientific training methods. However, the current child rehabilitation training technology and methods have many shortcomings in practical application, which limits the improvement of rehabilitation effects, mainly reflected in the following aspects:

[0003] Traditional rehabilitation training for children is usually completed through single repetitive movement exercises, such as limb stretching, grasping movements, etc. This training method is often boring and difficult to attract children's interest and attention, resulting in low compliance with the training. Especially for young children or children with attention deficit, monotonous training content can easily lead to interruptions in training, thus affecting the long-term rehabilitation effect.

[0004] Most of the current rehabilitation training for children relies on the therapist's observation and subjective evaluation to determine whether the child's movements meet the standard, which makes it difficult to achieve accurate and real-time data collection and movement evaluation. A small number of systems that use motion capture devices can only provide limited limb movement data and fail to combine multi-source data (such as position, speed, trajectory, etc.) for comprehensive analysis. In addition, the sensors in existing systems are often affected by noise, and data reliability is not fully considered, resulting in insufficient accuracy of the evaluation results.

[0005] Since children's disease types, rehabilitation progress and individual characteristics vary significantly, rehabilitation training needs to dynamically adjust the difficulty of training according to the children's real-time performance. However, most existing systems use a preset fixed training process and lack dynamic adaptation functions for individuals. This method cannot optimize the training plan according to the children's actual ability and recovery status, resulting in the inability to fully exert the training effect.

[0006] Real-time feedback in rehabilitation training is crucial to enhancing children's sense of participation and improving movement accuracy. However, in existing rehabilitation training systems, the interactive feedback mechanism is relatively simple, usually limited to visual or voice prompts, and cannot achieve multi-modal in-depth feedback. For example, after children complete a training task, the system may not be able to quickly generate targeted motivational feedback, making it difficult to enhance the fun and sense of accomplishment of the training.

[0007] Although some rehabilitation training systems use advanced technologies such as virtual reality (VR) and augmented reality (AR), they are usually separated from data processing, training plan adjustment and other links, making it difficult to form a complete closed loop from training scenarios, motion capture, data analysis to feedback optimization. For example, some systems provide high-precision motion capture functions, but fail to combine real-time feedback and personalized optimization, resulting in limited practical application effects.

[0008] To this end, the present application designs and provides a children's rehabilitation training system based on situational interaction to solve the above problems. Summary of the invention

[0009] The purpose of the present invention is to solve the technical problems raised in the above background technology and to provide a children's rehabilitation training system based on situational interaction.

[0010] The above-mentioned of the present invention is currently achieved by the following technical solutions:

[0011] The solution of the present invention provides a child rehabilitation training system based on scenario interaction, the system comprising:

[0012] The scenario simulation module is used to create multi-dimensional interactive scenarios through virtual reality and augmented reality technologies and generate scenario parameter sets:

[0013] S m ={s m,1 ,s m,2 ,…,s m,Np}

[0014] Where: S m is the parameter set of the mth scenario; s m,n The nth characteristic parameter (n=1,2,…,N) of the mth scenario p ), where N p is the total number of feature parameters, including scene complexity C and number of dynamic targets N d and the difficulty of the task target D; C is the complexity of the scenario, N d is the number of dynamic targets, and D is the difficulty of the task target.

[0015] Motion capture module, used to capture children's real-time motion data through sensor arrays and cameras, generating:

[0016]

[0017] Where: X t is the global motion capture data at time t; For the mth (m=1,2,…,N s )Data captured by the sensor; N s is the total number of sensors.

[0018] Data processing module, used to process the captured motion data X t Perform preprocessing, feature extraction and motion pattern analysis, and generate an action matching score M;

[0019] The interactive feedback module is used to dynamically generate a feedback signal R according to the action matching score M. The feedback content includes voice, image and tactile information.

[0020] The training plan module is used to generate a personalized training plan P based on the child's rehabilitation data H and adjust the training parameters according to the real-time performance.

[0021] Furthermore, the motion capture module fuses multi-source data, and the fusion algorithm is a weighted average model:

[0022]

[0023] Where: X t is the global motion capture result at time t; For the mth (m=1,2,…,N s )Data captured by the sensor; m is the weight coefficient, satisfying:

[0024]

[0025] The weight formula is:

[0026]

[0027] in: represents the noise variance of the mth sensor; N s Indicates the total number of sensors.

[0028] Furthermore, the scenario simulation module dynamically adjusts the scenario parameter S according to the child's real-time action matching score M. m , the adjustment rules are:

[0029] s m,n,t+1 =s m,n,t +Δs m,n

[0030] Where: s m,n,t is the value of the nth parameter of the mth scenario at time t; Δs m,n =k·(1-M), where k is the adjustment gain coefficient and M is the action matching score, which is defined as:

[0031]

[0032] Where: X i Actual action data for children; Xref,i is the reference standard motion data; σ is the tolerance parameter for motion matching; n is the number of sampling points.

[0033] Furthermore, the feedback signal R of the interactive feedback module is generated based on the child's performance, and the feedback intensity calculation formula is:

[0034]

[0035] Where: R is the feedback signal strength; M is the current action matching score; M min and M max are the preset minimum and maximum matching scores respectively; β is the feedback strength adjustment factor.

[0036] Furthermore, the training plan module dynamically adjusts the training difficulty D according to the child's performance score M and rehabilitation history data H. The adjustment rule is:

[0037]

[0038] Where: D t and D t+1 are the training difficulty of the current and next stage respectively; M ref Score the target action match.

[0039] Furthermore, the data processing module predicts the child's rehabilitation trend through a deep learning model based on movement characteristics, and the prediction model is defined as:

[0040] Y t =g(W·F t +b)

[0041] Where: Y t is the predicted value of the recovery status at time t; F t is the action feature vector, including action speed, acceleration and displacement; W is the weight matrix; b is the bias; g is the activation function, such as ReLU or sigmoid.

[0042] Furthermore, the deep learning model is trained with historical data, and the target optimization function is:

[0043]

[0044] Where: L is the loss function; Y i is the predicted value of the i-th sample; Y target,i is the target value of the i-th sample; N is the number of training samples.

[0045] Furthermore, the interactive feedback module optimizes the training effect by adjusting the feedback intensity adjustment factor β. The adjustment rule is:

[0046]

[0047] Where: ΔL is the difference in loss function between the last two trainings; L old is the loss function value of the previous stage.

[0048] Furthermore, the training plan module is based on the predicted value Y of the rehabilitation status t The training content is adjusted jointly with the real-time performance M, and the optimization function is:

[0049] Q=α·Y t +(1-α)·M

[0050] Where: Q is the comprehensive optimization goal; α is the weight of the predicted value and performance score, satisfying 0≤α≤1.

[0051] Furthermore, the system uses the scenario parameter S m , action data X t , feedback signal R, training plan P and predicted value Y t The cycle optimization forms a complete technical closed loop, realizing dynamic optimization and intelligent interaction of children's rehabilitation training.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. The present invention constructs a situational interactive training environment by introducing virtual reality (VR) and augmented reality (AR) technologies, integrates rehabilitation training tasks into interesting simulation scenes, and effectively improves children's interest and participation in training. In addition, the system uses multi-sensor data fusion technology to capture and analyze children's motion data in real time, and combines the matching score algorithm to quantitatively evaluate the completion of training, which greatly improves the accuracy and scientificity of the training process. Through the combination of fun and accuracy, the present invention effectively improves the boring and monotonous shortcomings of traditional rehabilitation training and improves the compliance of children's rehabilitation.

[0054] 2. The present invention can dynamically adjust the training goals and difficulty according to the child's real-time performance. For example, by analyzing the matching score M, the system can increase the difficulty of training when the child performs well, or reduce the complexity of the task when the child performs poorly, so that the training process matches the child's actual ability. Combined with the child's historical training data, the system can also generate personalized training plans to ensure that the rehabilitation needs of different types of child patients are met in a targeted manner. This personalized and dynamic optimization mechanism significantly improves the efficiency and effectiveness of rehabilitation training.

[0055] 3. The present invention organically combines scenario simulation, motion capture, data processing, interactive feedback and training plan modules to form a complete closed-loop system from data collection to training optimization. The system predicts the rehabilitation trend of children through deep learning algorithms, helps therapists to grasp the recovery of children in real time, and formulate scientific rehabilitation plans based on the predicted results. The implementation of closed-loop design not only improves the intelligence of the system, but also reduces the workload of therapists, making the rehabilitation process more efficient and scientific, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 This is a system block diagram of a children's rehabilitation training system based on situational interaction;

[0058] Figure 2 This is a logical block diagram of a children's rehabilitation training system based on situational interaction. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] The following combines multiple embodiments, Figure 1 and Figure 2 , the specific implementation methods of the present invention are described in detail.

[0061] The child rehabilitation training system of the present invention operates in coordination through the situation simulation module, motion capture module, data processing module, interactive feedback module and training plan module. The system adopts multi-sensor fusion, dynamic scene adjustment, real-time feedback optimization and other technologies to provide personalized solutions for child rehabilitation. The details are as follows:

[0062] Situation simulation module: Functional description: The system builds interactive scenarios through virtual reality (VR) or augmented reality (AR) technology. Situation parameter set S m Including scene complexity C, number of dynamic targets N d , task objective difficulty D.

[0063] Adjustment logic: The system adjusts the scenario parameters based on the child’s real-time performance:

[0064] s m,n,t+1 =s m,n,t +k·(1-M)

[0065] Where: k is the adjustment gain coefficient; M is the action matching score.

[0066] Motion capture module: Use inertial sensors and depth cameras to capture children’s body movements in real time and generate data sets

[0067] Data fusion: The multi-source data captured by the sensor is fused through a weighted average model:

[0068] The weight w m Dynamic calculation:

[0069] The noise variance of the mth sensor; N s : The total number of sensors.

[0070] Data processing module: Processing captured data X t , and standard action data X ref Compare and generate the action matching score M.

[0071] Match score calculation:

[0072] Where: X i :Children's actual action data; X ref,i : standard action data; σ: matching tolerance;

[0073] n: number of sampling points.

[0074] Interactive feedback module: provides real-time feedback (voice, image, touch) based on M.

[0075] Feedback Strength:

[0076] Where: M min ,M max : minimum and maximum matching scores; β: feedback strength adjustment factor.

[0077] Training plan module: dynamically adjust the training difficulty D according to the child’s rehabilitation data H and real-time performance M.

[0078] Adjustment rules:

[0079] Where: Mref : Target matching score.

[0080] Example 1: Single-target rehabilitation training;

[0081] Goal: To train children to complete a single action (such as extending their arms to a target point).

[0082] Implementation process: Situation initialization: Setting the target point P target =(x,y,z), scenario parameter S m ={C=1,N d =1,D=1};

[0083] Motion capture: Use 3 inertial sensors to capture the child's arm position in real time t ;

[0084] Data processing: Calculate the matching score M between the current position of the arm and the target point;

[0085] Feedback and adjustment: Provide voice feedback based on M, such as "The action is done very well";

[0086] Dynamically adjust the position of the target point and update S m .

[0087] Calculation process:

[0088] Data Fusion:

[0089] Match score:

[0090] Through simple single target point tasks, children are trained to complete basic movements, such as stretching their arms or moving a specific limb to a target position. This embodiment is suitable for children with milder conditions who need basic functional training, especially children who are receiving rehabilitation training for the first time or have weak motor skills.

[0091] Example 2: Multi-objective dynamic training;

[0092] Objective: To train children to touch multiple moving target points continuously in a dynamic situation.

[0093] Implementation process: Situation initialization: Set 3 dynamic target points P target,1 ,P target,2 ,P target,3 , scenario parameter S m ={C=3,N d =3,D=2};

[0094] Motion capture: Use 5 sensors to capture children's movements in real time t ;Data processing: Calculate the matching score between the child trajectory and the target trajectory;

[0095] Feedback and adjustment: Display the number of successful touches in real time and play encouraging voice; dynamically adjust the speed or number of target points according to M.

[0096] Calculation process:

[0097] Data Fusion:

[0098] Match score:

[0099] For children with certain motor abilities, multi-target scenarios (such as dynamic target points or continuous tasks) are simulated to improve their motor coordination and concentration. It is suitable for children who need to improve their motor continuity and accuracy.

[0100] Example 3: Rehabilitation trend prediction and personalized optimization;

[0101] Goal: To predict recovery trends based on historical training data and dynamically optimize training plans.

[0102] Implementation process: Scenario initialization: The system reads historical data and sets the initial scenario S m ;

[0103] Motion capture: real-time collection of children's motion dataX t ;

[0104] Data processing: Using deep learning models to predict recovery trends Y t ;

[0105] Feedback and optimization: Dynamically adjust the difficulty D according to the prediction results and generate a personalized training plan.

[0106] Calculation process:

[0107] Recovery Prognosis:

[0108] Y t =g(W·F t +b)

[0109] Optimization difficulty:

[0110]

[0111] In view of the need for long-term rehabilitation training, a deep learning model is used to analyze the rehabilitation trend of children and dynamically adjust the training plan. This embodiment is suitable for children who need long-term rehabilitation training, such as children with cerebral palsy or severely limited limb function.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A children's rehabilitation training system based on situational interaction, characterized in that: The system includes: The scenario simulation module is used to create multi-dimensional interactive scenarios through virtual reality and augmented reality technologies and generate scenario parameter sets: Where: S m is the parameter set of the mth scenario; s m,n The nth characteristic parameter (n=1,2,…,N) of the mth scenario p ), where N p is the total number of feature parameters, including scene complexity C and number of dynamic targets N d and the difficulty of the task target D; C is the complexity of the scenario, N d is the number of dynamic targets, and D is the difficulty of the task target. Motion capture module, used to capture children's real-time motion data through sensor arrays and cameras, generating: Where: X t is the global motion capture data at time t; For the mth (m=1,2,…,N s )Data captured by the sensor; N s is the total number of sensors. Data processing module, used to process the captured motion data X t Perform preprocessing, feature extraction and motion pattern analysis, and generate an action matching score M; The interactive feedback module is used to dynamically generate a feedback signal R according to the action matching score M. The feedback content includes voice, image and tactile information. The training plan module is used to generate a personalized training plan P based on the child's rehabilitation data H and adjust the training parameters according to the real-time performance.

2. The child rehabilitation training system based on situational interaction according to claim 1 is characterized in that: The motion capture module fuses multi-source data, and the fusion algorithm is a weighted average model: Where: X t is the global motion capture result at time t; For the mth (m=1,2,…,N s )Data captured by the sensor; m is the weight coefficient, satisfying: The weight formula is: in: represents the noise variance of the mth sensor; N s Indicates the total number of sensors.

3. The child rehabilitation training system based on situational interaction according to claim 1 is characterized in that: The situation simulation module dynamically adjusts the situation parameter S according to the child's real-time action matching score M m , the adjustment rules are: s m,n,t+1 =s m,n,t +Δs m,n Where: s m,n,t is the value of the nth parameter of the mth scenario at time t; Δs m,n =k·1-M), where k is the adjustment gain coefficient and M is the action matching score, which is defined as: Where: X i Actual action data for children; X ref,i is the reference standard motion data; σ is the tolerance parameter for motion matching; n is the number of sampling points.

4. The child rehabilitation training system based on situational interaction according to claim 1 is characterized in that: The feedback signal R of the interactive feedback module is generated based on the child's performance, and the feedback intensity calculation formula is: Where: R is the feedback signal strength; M is the current action matching score; M min and M max are the preset minimum and maximum matching scores respectively; β is the feedback strength adjustment factor.

5. The child rehabilitation training system based on situational interaction according to claim 1 is characterized in that: The training plan module dynamically adjusts the training difficulty D according to the child's performance score M and rehabilitation history data H. The adjustment rules are: Where: D t and D t+1 are the training difficulty of the current and next stage respectively; M ref Score the target action match.

6. The child rehabilitation training system based on scenario interaction according to claim 1 is characterized in that the data processing module predicts the child rehabilitation trend through a deep learning model based on motion characteristics, and the prediction model is defined as: Y t =g(W·F t +b) in: Y t is the predicted value of the recovery status at time t; F t is the action feature vector, including action speed, acceleration and displacement; W is the weight matrix; b is the bias; g is the activation function, such as ReLU or Sigmoid.

7. The child rehabilitation training system based on situational interaction according to claim 6 is characterized in that: The deep learning model is trained with historical data, and the target optimization function is: Where: L is the loss function; Y i is the predicted value of the i-th sample; Y target,i is the target value of the i-th sample; N is the number of training samples.

8. The child rehabilitation training system based on situational interaction according to claim 1 is characterized in that: The interactive feedback module optimizes the training effect by adjusting the feedback intensity adjustment factor β. The adjustment rule is: Where: ΔL is the difference in loss function between the last two trainings; L old is the loss function value of the previous stage.

9. The child rehabilitation training system based on situational interaction according to claim 1, characterized in that: The training plan module is based on the predicted value Y of the rehabilitation status t The training content is adjusted jointly with the real-time performance M, and the optimization function is: Q=α·Y t +(1-a)·M Where: Q is the comprehensive optimization goal; α is the weight of the predicted value and performance score, satisfying 0≤α≤1.

10. The child rehabilitation training system based on scenario interaction according to any one of claims 1 to 9, characterized in that: The system uses the scenario parameter S m , action data X t , feedback signal R, training plan P and predicted value Y t The cycle optimization forms a complete technical closed loop, realizing dynamic optimization and intelligent interaction of children's rehabilitation training.

Images