Rehabilitation balance system and method based on virtual reality and unstable platform

Abstract

The present application relates to the technical field of medical rehabilitation, in particular to a rehabilitation balance system and method based on virtual reality and an unstable platform, comprising: synchronously collecting real-time posture, surface electromyography and motion trajectory data of a user on the unstable platform, and fusing to generate a current balance ability evaluation parameter. The evaluation parameter is compared with a preset rehabilitation target in real time, and a set of linked virtual reality scenes, platform disturbance modes and target action specification parameters are dynamically generated. The system drives the virtual reality engine to generate an interactive environment and controls the platform to produce multidimensional physical disturbances according to these parameters, and provides real-time specification action guidance through AI voice and visual channels. Through multidimensional data fusion and closed-loop control, the method realizes the collaborative adaptive adjustment of the training environment, physical challenges and real-time guidance, effectively improving the individualization level of balance rehabilitation training and the accuracy of neuromuscular reeducation.

Description

Technical Field

[0001] This invention relates to the field of medical rehabilitation technology, and in particular to a rehabilitation balance system and method based on virtual reality and an unstable platform. Background Technology

[0002] Existing rehabilitation balance training systems often employ a separate combination of technologies, with virtual reality, unstable platforms, and biosignal acquisition modules frequently operating independently or existing only in simple signal triggering relationships. These systems typically rely on a single type of data for assessment, failing to comprehensively reflect the integrated state of neuromuscular control. Training programs are usually based on initial assessments or fixed protocols, lacking effective dynamic adjustment mechanisms during training.

[0003] The main drawback of existing technologies lies in the lack of coordination and real-time linkage between various intervention methods. The challenges of virtual scenarios, the disturbance patterns of physical platforms, and the standards for movement guidance are usually pre-set or can only be adjusted in isolation. In this mode, the training system cannot adaptively adjust the above-mentioned multiple intervention parameters synchronously and in a coordinated manner based on the changes in the user's comprehensive abilities in real time during training. As a result, the training content is difficult to match with the user's current actual functional level in real time, the degree of personalization is insufficient, and the accuracy and efficiency of rehabilitation training are affected.

[0004] A technical solution is needed that can deeply integrate multidimensional biomechanical data for real-time comprehensive assessment, and automatically and collaboratively control the virtual environment, physical disturbances, and real-time guidance system based on the assessment results. This system should be able to achieve closed-loop dynamic adaptation from assessment to multimodal intervention, in order to solve the problems of fragmented intervention methods, lagging adjustments, and lack of overall integration in existing technologies. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a rehabilitation balance system and method based on virtual reality and an unstable platform.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a rehabilitation balance method based on virtual reality and an unstable platform, comprising: Collect real-time posture data of the user on the unstable platform, surface electromyography signal data of the user, and motion trajectory data of the user; The user's current balance ability assessment parameters are generated based on the real-time posture data, the surface electromyography signal data, and the motion trajectory data. Based on the comparison results between the current balance ability assessment parameters and the preset rehabilitation stage goals, a personalized rehabilitation training plan is generated. The personalized rehabilitation training plan includes virtual reality scene dynamic parameters, unstable platform disturbance mode parameters, and target movement specification parameters. The virtual reality scene rendering engine is invoked to generate and present a virtual reality interactive environment adapted to the personalized rehabilitation training program based on the dynamic parameters of the virtual reality scene. Based on the unstable platform disturbance mode parameters, the unstable platform is controlled to generate multi-dimensional physical disturbances; Through AI voice prompts and AI visual prompts, real-time action guidance information is provided to the user based on the target action specification parameters to correct the user's posture when performing actions in the virtual reality interactive environment.

[0007] Preferably, generating the user's current balance ability assessment parameters based on the real-time posture data, the surface electromyography signal data, and the motion trajectory data includes: The real-time posture data includes the body's center of gravity shift and changes in joint angles; The current balance ability assessment parameters include stability index, muscle activation mode coordination coefficient, and movement trajectory deviation. Extract the time series of the body center of gravity offset and the spatial distribution matrix of the joint angle changes from the real-time posture data; A stability analysis is performed on the time series of the body's center of gravity offset to calculate the stability index, which is a composite function of the envelope area of ​​the center of gravity swinging in a predetermined plane and the average velocity. The activation timing and intensity of the primary and antagonistic muscle groups are analyzed from the surface electromyography signal data, and the coordination coefficient of the muscle activation pattern is calculated based on the matching degree of the activation timing and intensity. The motion trajectory data is spatially registered with a preset standard motion trajectory template, the positional deviation of the key points of the trajectory is calculated, and the motion trajectory offset is determined based on the cumulative value of the positional deviation.

[0008] Preferably, the stability analysis of the time series of the body center of gravity offset to calculate the stability index includes: The time series of the body's center of gravity offset is projected and decomposed in the horizontal plane to obtain the center of gravity displacement components in the front-back and left-right directions. Calculate the standard deviation, average moving speed, and oscillation frequency of the center of gravity displacement components in the front-back and left-right directions, respectively. A multidimensional stability assessment model is constructed based on the standard deviation, average moving speed and oscillation frequency. The multidimensional stability assessment model outputs a comprehensive stability index by weighted integration of the variation characteristics of each component. Based on the decay trend of the comprehensive stability index during the rehabilitation training cycle, the comprehensive stability index is normalized to generate a stability index that characterizes the user's immediate balance state.

[0009] Preferably, the step of generating a personalized rehabilitation training plan based on the comparison results between the current balance ability assessment parameters and the preset rehabilitation stage goals includes: The preset rehabilitation stage goals are obtained, which include target stability thresholds, target muscle coordination thresholds, and target trajectory accuracy thresholds required to be achieved at different rehabilitation stages. The stability index is compared with the target stability threshold to generate a stability training requirement level. The muscle activation mode coordination coefficient is compared with the target muscle coordination threshold to generate a muscle coordination training requirement level. The motion trajectory offset is compared with the target trajectory accuracy threshold to generate a trajectory accuracy training requirement level. Based on the combination of stability training requirement level, muscle coordination training requirement level and trajectory accuracy training requirement level, the preset training strategy mapping database is queried to match the corresponding virtual reality scene dynamic parameters, unstable platform disturbance mode parameters and target action specification parameters. The matched parameters are integrated into a personalized rehabilitation training program with clear execution steps, difficulty gradients, and feedback mechanisms.

[0010] Preferably, the step of querying a preset training strategy mapping database and matching the corresponding virtual reality scene dynamic parameters based on the combination of the stability training requirement level, muscle coordination training requirement level, and trajectory accuracy training requirement level includes: The dynamic parameters of the virtual reality scene include at least the complexity of the virtual scene, the frequency and speed of the appearance of interference objects in the scene, and the task objectives that the user's virtual avatar needs to complete. When the stability training requirement level is high, the matched virtual reality scene dynamic parameters point to a virtual environment containing a mobile platform or a narrow passage. When the muscle coordination training demand level is high, the matching virtual reality scene dynamic parameters are directed to the task that requires the user's virtual avatar to perform a compound action task with a specific rhythm or resistance. When the trajectory accuracy training requirement level is high, the matched virtual reality scene dynamic parameters point to a virtual environment that includes precise path guidance or requires avoiding dynamic obstacles.

[0011] Preferably, controlling the unstable platform to generate multi-dimensional physical disturbances based on the unstable platform disturbance mode parameters includes: The disturbance mode parameters of the unstable platform are analyzed to obtain the direction sequence, amplitude sequence, frequency sequence and duration of the disturbance; The unstable platform is controlled to rotate around one or more axes of the roll axis, pitch axis and yaw axis according to the direction sequence. The rotation angle or translation displacement of the unstable platform in each direction is controlled according to the amplitude sequence. The rate of change of the rotational motion or translational displacement of the unstable platform is controlled according to the frequency sequence. The direction sequence, amplitude sequence, and frequency sequence are combined and arranged according to the duration to generate a composite perturbation stimulus that changes dynamically over time, which is then applied to the user's support plane.

[0012] Preferably, the step of parsing the unstable platform disturbance mode parameters to obtain the direction sequence, amplitude sequence, frequency sequence, and duration of the disturbance includes: Read the perturbation mode configuration file encoded in the perturbation mode parameters of the unstable platform. The configuration file uses a hierarchical data structure to store perturbation sequence information. Extract the total duration of the disturbance and the division information of the disturbance phases from the top-level structure of the configuration file; Based on the information on the division of the disturbance stages, the set of disturbance events contained in each disturbance stage is analyzed layer by layer; For each disturbance event, parameter decoding is performed to separate the direction sequence describing the spatial properties of the disturbance, the amplitude sequence describing the intensity properties of the disturbance, the frequency sequence describing the rhythm properties of the disturbance, and the independent duration of each disturbance event. The obtained direction sequence, amplitude sequence, frequency sequence and duration are integrated according to the timeline to generate a timing control command stream that can be directly read by the actuator of the unstable platform.

[0013] Preferably, the step of providing real-time action guidance information to the user based on the target action specification parameters through the AI ​​voice prompt channel and the AI ​​visual prompt channel includes: Continuously acquire the user's real-time posture data and motion trajectory data during the training process; The real-time posture data is compared with the target joint angle range in the target action specification parameters in real time. When a deviation is detected, a voice command containing the specific joint name and adjustment direction is broadcast through the AI ​​voice prompt channel. The motion trajectory data is overlaid and displayed in real time with the target motion trajectory path in the target motion specification parameters. The trajectory guidance and deviation feedback are presented in the form of highlighted paths, arrows, or deviation heatmaps in the virtual reality interactive environment through the AI ​​visual prompt channel. Based on the difference between the muscle activation mode coordination coefficient and the target mode, the names of the muscle groups that need to be activated or relaxed are simultaneously prompted through the AI ​​voice prompt channel and the AI ​​visual prompt channel.

[0014] Preferably, the real-time comparison of the real-time posture data with the target joint angle range in the target motion specification parameters is performed in real time. When a deviation is detected, a voice command containing the specific joint name and adjustment direction is broadcast through the AI ​​voice prompt channel, including: The three-dimensional angle data of each major joint is extracted in real time from the real-time attitude data stream; The extracted joint 3D angle data is compared frame by frame with the predefined target angle range of each joint in the target motion specification parameters; When the angle data of a specific joint continuously exceeds the target angle range for a preset threshold duration, it is determined that the specific joint has deviated from its movement. The direction of joint movement to be adjusted is determined based on the difference between the angle data of the deviation from the joint and the median value of the target angle range; The speech synthesis engine is invoked to combine the name of the deviated joint, the detected direction of deviation, and the target direction to be adjusted into a natural language command; The generated voice commands are broadcast in real time through the AI ​​voice prompt channel, and subsequent posture data is monitored after the broadcast to evaluate the execution effect of the commands.

[0015] Preferably, the present invention also includes a rehabilitation balance system based on virtual reality and an unstable platform. The system includes a processor and a memory, the memory and the processor being connected. The memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to realize the rehabilitation balance method based on virtual reality and an unstable platform as described above.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: By simultaneously collecting and fusing real-time posture, surface electromyography (EMG) signals, and motion trajectory data, balance ability assessment parameters are generated. These parameters are then compared with stage targets in real time, directly driving the generation of a set of linked virtual reality scene dynamic parameters, unstable platform disturbance mode parameters, and target movement standard parameters. This technical solution enables the coordinated adjustment of training difficulty, physical interference, and movement standards based on the user's real-time, multi-dimensional, integrated physiological and motor state. An adaptive closed-loop control core is constructed, achieving dynamic and precise matching between training challenge and the user's current functional level, avoiding the problem of unsuitable training intensity caused by using fixed schemes or single-dimensional adjustments.

[0017] Based on the target movement specifications parameters in the personalized training plan, specific movement guidance information is provided through AI voice and visual prompts when the user is in a dynamic virtual environment or experiencing physical disturbances. Personalized movement standards are embedded into the execution of immersive training. In complex multi-task training environments, users are provided with immediate, clear, and directly relevant operational guidance, forming a real-time biofeedback and correction loop. This effectively guides users towards more standardized and efficient movement patterns, improving the efficiency of movement learning and the accuracy of neuromuscular re-education during training. Attached Figure Description

[0018] Figure 1 This is a flowchart of the rehabilitation balance method based on virtual reality and an unstable platform as described in this invention; Figure 2 A flowchart for calculating the stability index; Figure 3 A flowchart for generating a personalized rehabilitation training plan; Figure 4 A multi-dimensional bar chart for assessing balance ability during the rehabilitation phase; Figure 5 A bar chart comparing the effectiveness indicators of different rehabilitation training programs. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are used only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have an orientation, or be constructed and operated in an orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1The overall implementation scheme of the rehabilitation balance method based on virtual reality and an unstable platform is as follows: A sensor system deployed on the unstable platform and the user's body synchronously collects the user's real-time posture data, surface electromyography (EMG) signal data, and motion trajectory data on the unstable platform. The real-time posture data is captured by an inertial measurement unit (IMU), the EMG signal data is collected by electrodes attached to relevant muscle groups in the user's lower limbs and trunk, and the motion trajectory data is captured by an optical or depth camera. A data processing unit receives the above multi-source data and, based on the real-time posture data, EMG signal data, and motion trajectory data, generates the user's current balance ability assessment parameters using a built-in evaluation algorithm. After generating the assessment parameters, the scheme generation module compares the current balance ability assessment parameters with preset rehabilitation stage goals. Based on the comparison results, it retrieves and combines corresponding parameter sets from the training strategy library to generate a personalized rehabilitation training scheme. This scheme specifically includes virtual reality scene dynamic parameters, unstable platform perturbation mode parameters, and target movement specification parameters. During the scheme execution phase, the virtual reality rendering subsystem calls the virtual reality scene rendering engine to generate and present a virtual reality interactive environment adapted to the personalized rehabilitation training scheme based on the virtual reality scene dynamic parameters. The platform control subsystem drives the actuators of the unstable platform according to the unstable platform disturbance mode parameters, causing them to generate multi-dimensional physical disturbances. Furthermore, the guidance feedback subsystem analyzes the collected real-time user data based on the target action specification parameters through AI voice and visual prompt channels, and provides the user with real-time action specification guidance information to correct the user's posture when performing actions in the virtual reality interactive environment.

[0022] In one embodiment of the present invention, see [reference] Figure 2 The collected real-time posture data includes body center of gravity offset and joint angle changes. The current balance ability assessment parameters to be generated include a stability index, a muscle activation pattern coordination coefficient, and a movement trajectory offset. The time series of body center of gravity offset and the spatial distribution matrix of joint angle changes are extracted from the real-time posture data. Stability analysis is performed on the time series of body center of gravity offset to calculate the stability index, which is a composite function of the envelope area and average velocity of the center of gravity swinging within a predetermined plane. The activation sequence and intensity of the primary and antagonistic muscle groups are analyzed from the surface electromyography (EMG) signal data, and the muscle activation pattern coordination coefficient is calculated based on the matching degree of the activation sequence and intensity. The movement trajectory data is spatially registered with a preset standard movement trajectory template, the positional deviation of key points in the trajectory is calculated, and the movement trajectory offset is determined based on the cumulative value of the positional deviation.

[0023] In practice, the current balance ability assessment parameters of the user are generated based on real-time multi-source data. The data acquisition system simultaneously acquires the user's real-time posture data, surface electromyography (EMG) signal data, and motion trajectory data on the unstable platform. The real-time posture data is provided by an inertial measurement unit group installed on key parts of the user's torso and limbs, including the body's center of gravity shift and joint angle changes. The surface EMG signal data is acquired by an electrode array attached to the surface of the user's lower limbs and core muscle groups. The motion trajectory data is recorded by an optical motion capture system or depth sensor deployed in the training area. The data processing unit receives the above data streams. The current balance ability assessment parameters include a stability index calculated based on posture data, a muscle activation mode coordination coefficient calculated based on EMG data, and a motion trajectory offset calculated based on trajectory data.

[0024] In specific implementation, the time series of the body center of gravity offset and the spatial distribution matrix of the joint angle change are extracted from the real-time posture data. The time series of the body center of gravity offset is obtained by fusing the data of the inertial measurement unit group and obtaining the coordinate change sequence of the user's center of gravity in three-dimensional space at a fixed sampling frequency within the training time. The spatial distribution matrix of the joint angle change describes the matrix formed by the angle values ​​of the user's major joints such as hip, knee, and ankle in the sagittal, coronal, and horizontal planes at a specific training moment.

[0025] In specific implementation, stability analysis is performed on the time series of the body center of gravity offset to calculate the stability index. The time series of the projected displacement of the body center of gravity in the horizontal plane is decomposed into forward and backward displacement components and left and right displacement components. The standard deviation, average speed and dominant frequency component of the forward and backward displacement components within the analysis time window are calculated respectively. The standard deviation, average speed and dominant frequency component of the left and right displacement components are also calculated. The constructed multidimensional stability evaluation model fuses the characteristic values ​​of these components. In specific implementation, the activation sequence and intensity of the primary and antagonistic muscle groups are analyzed from the surface electromyography (SEMG) signal data. The acquired raw SEMG signals are subjected to bandpass filtering, full-wave rectification, and smoothing to obtain the linear envelope of each target muscle. The start point, peak point, and end point of muscle activation are identified from the linear envelope to determine the activation sequence, including the activation order and activation interval. The activation intensity is obtained by calculating the integral or average value of the linear envelope during the activation period. The muscle activation pattern coordination coefficient is calculated based on the matching degree between the activation sequence and activation intensity. The calculation of the matching degree involves comparing the co-activation sequence relationship of the agonist and antagonist muscles, as well as the degree of closeness of their intensity ratio to the preset ideal ratio. The muscle activation pattern coordination coefficient is a weighted combination of these timing and intensity matching degree scores.

[0026] In specific implementation, the motion trajectory data is spatially registered with a preset standard motion trajectory template. The standard motion trajectory template defines the ideal motion path of key body landmarks (such as hands, feet, and hips) in three-dimensional space when completing a specific training action. Through rigid or non-rigid registration algorithms, the real-time collected motion trajectory data is spatially aligned with the standard motion trajectory template, and the positional deviation of the key points of the trajectory is calculated. On the aligned trajectory in the registered space, the Euclidean distance between the real-time trajectory points and the corresponding points on the standard trajectory template is calculated. The motion trajectory offset is determined based on the cumulative value of the positional deviation. The calculation of the cumulative value can be the summation of the positional deviations of all sampling points in the entire action cycle or the root mean square value. The motion trajectory offset is a direct mapping or normalized representation of this cumulative value.

[0027] In one embodiment of the present invention, see [reference] Figure 3 The time series of the body's center of gravity offset is projected and decomposed in a horizontal plane to obtain the center of gravity displacement components in the forward-backward and left-right directions. The standard deviation, average movement speed, and oscillation frequency of each of the forward-backward and left-right displacement components are calculated. A multidimensional stability assessment model is constructed based on the standard deviation, average movement speed, and oscillation frequency. This model outputs a comprehensive stability index by weighted fusion of the variation characteristics of each component. Based on the decay trend of the comprehensive stability index during the rehabilitation training period, the comprehensive stability index is normalized to generate a stability index characterizing the user's immediate balance state.

[0028] The process of calculating the stability index begins with processing the time series of body center of gravity offset, which is a continuous record of the user's center of gravity in three-dimensional spatial coordinates during training. The time series of body center of gravity offset is projected and decomposed in the horizontal plane. This decomposition operation converts the three-dimensional spatial coordinates into two-dimensional planar coordinates by ignoring the vertical height change, thereby obtaining the center of gravity displacement components in the front-back direction and the center of gravity displacement components in the left-right direction. The front-back and left-right directions are usually defined to correspond to the user's sagittal and coronal axes.

[0029] In specific implementation, the standard deviation, average moving speed, and swing frequency of the center of gravity displacement components in the front-back direction are calculated respectively. At the same time, the standard deviation, average moving speed, and swing frequency of the center of gravity displacement components in the left-right direction are also calculated. For the center of gravity displacement components in the front-back direction, the standard deviation is used to quantify the dispersion of the user's body swing in the front-back direction. The average moving speed is obtained by calculating the mean of the derivative of the displacement amount in that direction with respect to time. The swing frequency is extracted from the dominant frequency component in the displacement signal through spectrum analysis. The corresponding characteristic values ​​of the center of gravity displacement components in the left-right direction are obtained in parallel using the same calculation method.

[0030] In some embodiments, a multidimensional stability assessment model is constructed based on the standard deviation, average moving speed, and oscillation frequency. This multidimensional stability assessment model is a mathematical function or algorithmic framework whose input consists of six sets of feature values ​​in two directions: forward / backward and left / right. The multidimensional stability assessment model outputs a comprehensive stability index by weighted fusion of the variation characteristics of each component. The multidimensional stability assessment model can be designed as follows: in: The subscript represents the comprehensive stability index output by the multidimensional stability assessment model. and They represent the front-back direction and the left-right direction, respectively. This represents the standard deviation of the displacement components in the corresponding direction. The average velocity representing the displacement component in the corresponding direction. The oscillation frequency represents the displacement component in the corresponding direction. , , These are the weighting coefficients assigned to the eigenvalues ​​in each direction, used to reflect the degree of contribution of different features to the overall balance stability.

[0031] It is understood that, based on the decay trend of the comprehensive stability index during the rehabilitation training cycle, the comprehensive stability index is normalized. The decay trend during the rehabilitation training cycle refers to the long-term pattern of the comprehensive stability index of the same user changing towards a better value in multiple consecutive training phases. The normalization process is based on the maximum, minimum, or distribution range of the user's historical comprehensive stability index, and the currently calculated comprehensive stability index is normalized accordingly. Mapped to a standardized numerical range, such as 0 to 1, a stability index is generated to characterize the user's immediate balance state. This stability index is a scalar value whose magnitude directly reflects the user's level of balance control at the current moment. Optionally, weighting coefficients... , , The determination can be based on statistical analysis of a large amount of rehabilitation training data, or it can be pre-set by rehabilitation therapists based on clinical experience.

[0032] In some embodiments, the oscillation frequency can be extracted using a fast Fourier transform or autoregressive model spectral estimation method to identify the main periodic oscillation components from the discrete center-of-gravity displacement time series, and their frequency values ​​are used as the oscillation frequency. Used. Optional, average movement speed. The calculation needs to consider the directionality of the displacement, that is, to use the average absolute velocity to avoid the influence of positive and negative displacements canceling each other out on the velocity assessment. This can be understood as a comprehensive stability index. The smaller the value, the smaller the swing amplitude of the user's center of gravity on the platform, the slower the speed and the more controllable the rhythm. This corresponds to better balance stability. After normalization, the closer the stability index is to 1, the better the balance state.

[0033] In one embodiment of the present invention, the preset rehabilitation stage target is obtained, which includes a target stability threshold, a target muscle coordination threshold, and a target trajectory accuracy threshold required to be achieved at different rehabilitation stages. The stability index is compared with the target stability threshold to generate a stability training requirement level. The muscle activation mode coordination coefficient is compared with the target muscle coordination threshold to generate a muscle coordination training requirement level. The movement trajectory offset is compared with the target trajectory accuracy threshold to generate a trajectory accuracy training requirement level. Based on the combination of the stability training requirement level, muscle coordination training requirement level, and trajectory accuracy training requirement level, a preset training strategy mapping database is queried to match the corresponding virtual reality scene dynamic parameters, unstable platform disturbance mode parameters, and target movement specification parameters. The virtual reality scene dynamic parameters include at least the complexity of the virtual scene, the frequency and speed of interference within the scene, and the task objectives that the user's virtual avatar needs to complete. When the stability training requirement level is high, the matched virtual reality scene dynamic parameters point to a virtual environment containing a moving platform or a narrow passage. When the muscle coordination training requirement level is high, the matched virtual reality scene dynamic parameters point to a composite movement task that requires the user's virtual avatar to perform a specific rhythm or resist resistance. When the required level of trajectory accuracy training is high, the matched virtual reality scene dynamic parameters are directed to a virtual environment that includes precise path guidance or requires avoidance of dynamic obstacles. Finally, the matched parameters are integrated into a personalized rehabilitation training program with clear execution steps, difficulty gradients, and feedback mechanisms.

[0034] In practice, the process of generating a personalized rehabilitation training plan begins with obtaining preset rehabilitation stage goals. These preset goals are a set of digital goals pre-set by rehabilitation therapists based on the specific rehabilitation period of the user. The preset goals include target stability thresholds, target muscle coordination thresholds, and target trajectory accuracy thresholds required to be achieved in different rehabilitation periods. The target stability threshold defines the minimum allowable value of the stability index that the user should achieve in the current rehabilitation period. The target muscle coordination threshold defines the minimum allowable value of the muscle activation mode coordination coefficient. The target trajectory accuracy threshold defines the maximum allowable value of the movement trajectory deviation.

[0035] In practice, the calculated stability index is compared with the target stability threshold in the preset rehabilitation stage goals. This comparison operation quantifies the difference between the stability index and the target stability threshold through subtraction or division. Based on the size of the difference, discrete stability training requirement levels are divided. The levels can be set to three levels: low, medium, and high, or a numerical grading system can be used. The comparison between the muscle activation mode coordination coefficient and the target muscle coordination threshold follows the same logic to generate a muscle coordination training requirement level. The comparison between the movement trajectory deviation and the target trajectory accuracy threshold generates a trajectory accuracy training requirement level. The larger the value of the movement trajectory deviation, the worse the accuracy usually is. Therefore, attention should be paid to the directionality of the threshold when comparing.

[0036] In some embodiments, based on the combination of stability training requirement level, muscle coordination training requirement level, and trajectory accuracy training requirement level, a preset training strategy mapping database is queried. The training strategy mapping database is a structured data storage, where the key value is an ordered combination of the three requirement levels, and the associated value is the specific configuration data of the corresponding recommended virtual reality scene dynamic parameters, unstable platform disturbance mode parameters, and target action specification parameters. The matching process uses the currently generated three-element requirement level combination as the search key to find and extract the corresponding parameter set from the training strategy mapping database.

[0037] It is understandable that the dynamic parameters of a virtual reality scene include at least the complexity of the virtual scene, the frequency and speed of the appearance of interference objects within the scene, and the task objectives that the user's virtual avatar needs to complete. When the stability training requirement level is high, the matched virtual reality scene dynamic parameters are directed to a virtual environment containing moving platforms or narrow passages. Moving platforms refer to surfaces in which the user stands or walks in the virtual environment and will move back and forth, left and right, or rotate. Narrow passages require the user to control the virtual avatar to accurately pass through a space of limited width. When the muscle coordination training requirement level is high, the matched virtual reality scene dynamic parameters are directed to a composite action task that requires the user's virtual avatar to perform a specific rhythm or resist resistance. A specific rhythm task may require the user to step according to the flashing sequence of cue lights, while a resist resistance task simulates pushing and pulling heavy objects or walking against the wind in a virtual environment. When the trajectory accuracy training requirement level is high, the matched virtual reality scene dynamic parameters are directed to a virtual environment containing precise path guidance or dynamic obstacles that need to be avoided. Precise path guidance will display a light strip in the scene that the user must follow, while dynamic obstacles refer to virtual objects that move along a set path and that the user needs to avoid in time.

[0038] In practical implementation, the matched virtual reality scene dynamic parameters, unstable platform disturbance mode parameters, and target action specification parameters are integrated into a personalized rehabilitation training program with clear execution steps, difficulty gradients, and feedback mechanisms. The execution steps are arranged in a timeline sequence of training stages, including warm-up, main training task, relaxation, and specific action instructions for each stage. The difficulty gradient is reflected in the numerical settings of the virtual reality scene dynamic parameters and unstable platform disturbance mode parameters, such as a gradual change in the speed of the interfering object from slow to fast, and a gradual change in the amplitude of platform disturbance from small to large. The feedback mechanism clearly defines the specific triggering conditions and manifestations for the system to perform real-time comparisons based on the target action specification parameters and provide guidance information during the user's action execution. Optionally, the correspondence between the demand level combinations and parameter configurations in the training strategy mapping database can be achieved through a function: in: Represents the set of matched parameters. This represents the mapping relationship used for query matching from the training policy mapping database. This represents the level of stability training requirements. This represents the level of muscle coordination training needs. This represents the level of trajectory accuracy training requirements. Optionally, personalized rehabilitation training programs are generated in the form of structured data files or scripts for subsequent reading and execution by the virtual reality rendering engine, unstable platform controller, and AI guidance module.

[0039] In one embodiment of the present invention, the disturbance mode parameters of the unstable platform are parsed to obtain the direction sequence, amplitude sequence, frequency sequence, and duration of the disturbance. A disturbance mode configuration file encoded in the disturbance mode parameters of the unstable platform is read; the configuration file stores disturbance sequence information using a hierarchical data structure. The total duration of the disturbance and the division information of the disturbance stages are extracted from the top-level structure of the configuration file. Based on the division information of the disturbance stages, the set of disturbance events contained in each disturbance stage is parsed layer by layer. Parameter decoding is performed on each disturbance event to separate the direction sequence describing the spatial attributes of the disturbance, the amplitude sequence describing the intensity attributes of the disturbance, the frequency sequence describing the rhythm attributes of the disturbance, and the independent duration of each disturbance event. The parsed direction sequence, amplitude sequence, frequency sequence, and duration are integrated according to a timeline to generate a timing control command stream that can be directly read by the unstable platform actuator. The unstable platform is controlled to rotate around one or more axes of the roll, pitch, and yaw axes according to the direction sequence. The rotation angle or translational displacement of the unstable platform in each direction is controlled according to the amplitude sequence. The rate of change of the rotational motion or translational displacement of the unstable platform is controlled according to the frequency sequence. The direction sequence, amplitude sequence, and frequency sequence are combined and arranged according to the duration to generate a composite perturbation stimulus that changes dynamically over time, which is then applied to the user's support plane.

[0040] In specific implementation, the unstable platform is controlled to generate multi-dimensional physical disturbances based on the unstable platform disturbance mode parameters. The starting point of this process is to parse the received unstable platform disturbance mode parameters and obtain the direction sequence, amplitude sequence, frequency sequence and duration of the disturbance encoded in the parameters. The direction sequence defines the spatial axial order of the platform disturbance, the amplitude sequence defines the intensity change of the disturbance along the corresponding axis, the frequency sequence defines the speed of the disturbance action, and the duration specifies the time length of the entire disturbance sequence or its sub-stage.

[0041] In specific implementation, the perturbation mode configuration file encoded in the unstable platform perturbation mode parameters is read. The configuration file uses a hierarchical data structure to store perturbation sequence information. A typical hierarchical structure includes a top-level overview layer, a stage division layer, and an event description layer. The top-level overview layer defines the global attributes of the scheme. The stage division layer divides the total duration into several stages with different training intentions. The event description layer defines in detail the parameters of one or more specific perturbation events contained in each stage. The total duration of the perturbation and the division information of the perturbation stages are extracted from the top-level structure of the configuration file. Based on the division information of the perturbation stages, the set of perturbation events contained in each perturbation stage is parsed layer by layer.

[0042] In some embodiments, each disturbance event is parameter-decoded to separate a direction sequence describing the spatial properties of the disturbance, an amplitude sequence describing the intensity properties of the disturbance, a frequency sequence describing the rhythm properties of the disturbance, and the independent duration of each disturbance event. The direction sequence may be represented as an array of instruction codes for rotation or translation along the roll, pitch, and yaw axes, while the amplitude and frequency sequences are numerical arrays that correspond to the elements of the direction sequence. Referring to Table 1, the parsed parameters of a stage containing three disturbance events can be organized internally in the form of Table 1.

[0043] Table 1: Disturbance Event Parameter Table Understandably, integrating the parsed direction, amplitude, frequency, and duration sequences along a timeline generates a timing control command stream that can be directly read by the actuators of the unstable platform. This integration process requires converting the parameters of each disturbance event into high-frequency real-time control signals based on their independent duration. The amplitude sequence values ​​may be directly used as the target angle or displacement, or an interpolation function may be needed to smoothly generate a continuous amplitude variation curve over the event's duration. For example, for an amplitude sequence value... The event, at the moment ( , The actual target amplitude (for the duration of the event) It can be done through the formula: in: The perturbation amplitude of the platform target at time t. This represents the baseline amplitude value of the event extracted from the amplitude sequence. This represents the perturbation frequency of the event as resolved from the frequency sequence. This represents the current time since the start of the disturbance event, and this formula makes the platform disturbance exhibit a sinusoidal fluctuation pattern.

[0044] In specific implementation, the unstable platform is controlled to rotate around one or more axes of roll, pitch, and yaw according to the direction sequence. The rotation angle or translational displacement of the unstable platform in each direction is controlled according to the amplitude sequence. The rate of change of the rotational or translational displacement is controlled according to the frequency sequence. The platform's control system receives a timing control command stream and drives the corresponding electric cylinder, servo motor, or torque motor to generate physical motion precisely according to the axial, amplitude, and time relationships specified in the command stream. Optionally, the direction sequence, amplitude sequence, and frequency sequence are combined and arranged according to the duration to generate a composite perturbation stimulus that dynamically changes over time, acting on the user's support plane. The composite perturbation stimulus means that the platform may simultaneously perform movements of different amplitudes and frequencies around multiple axes, such as superimposing a high-frequency, low-amplitude roll tremor while pitching, thereby simulating complex real-world environmental disturbances. In some embodiments, the values ​​in the amplitude sequence can be absolute values ​​or relative changes relative to a certain reference position, while the frequency sequence determines the number of cycles the platform completes for a specific amplitude movement per unit time. Optionally, the timing control instruction stream will be preloaded into the cache of the unstable platform controller to ensure the real-time and continuous execution of instructions and avoid disturbances and inconsistencies caused by data processing delays.

[0045] See Figure 4 This is a multi-dimensional bar chart assessing balance ability during the rehabilitation phase. As rehabilitation progresses, "ability-related indicators (stability, coordination)" improve, while "deviation-related indicators (trajectory deviation)" decrease, reflecting the positive effects of rehabilitation training. It visually demonstrates the rehabilitation program's improvement in balance ability. In the later stages, the trajectory deviation has decreased to 12, suggesting a possible reduction in training intensity for this dimension, focusing on more refined movement optimization. The differences in the rate of improvement of different indicators (e.g., faster improvement in stability) can help adjust the focus of subsequent training. By transforming the abstract concept of "improved balance ability" into the dynamic changes of three types of indicators, this chart clearly quantifies the effects from the initial to the later stages of rehabilitation, providing intuitive data support for the effectiveness of the rehabilitation program.

[0046] In one embodiment of the present invention, real-time posture data and motion trajectory data of the user during training are continuously acquired. The real-time posture data is compared in real-time with the target joint angle range in the target motion specification parameters. When a deviation is detected, a voice command containing the specific joint name and adjustment direction is broadcast through the AI ​​voice prompt channel. Three-dimensional angle data of each major joint is extracted in real-time from the real-time posture data stream. The extracted joint three-dimensional angle data is compared frame-by-frame with the predefined target angle range of each joint in the target motion specification parameters. When the angle data of a specific joint continuously exceeds the target angle range for a preset threshold duration, it is determined that the specific joint has deviated from its motion. Based on the difference between the angle data of the deviating joint and the median value of the target angle range, the motion direction that the joint needs to adjust is determined. A speech synthesis engine is invoked to combine the name of the deviating joint, the detected deviation direction, and the target direction to be adjusted into a natural language command. The generated voice command is broadcast in real-time through the AI ​​voice prompt channel, and subsequent posture data is monitored after broadcasting to evaluate the command execution effect. Simultaneously, the motion trajectory data is overlaid with the target motion trajectory path in the target action specification parameters in real time. The AI ​​visual prompting channel provides trajectory guidance and deviation feedback in the virtual reality interactive environment through highlighted paths, arrow indicators, or deviation heatmaps. Based on the difference between the muscle activation mode coordination coefficient and the target mode, the AI ​​voice prompting channel and the AI ​​visual prompting channel simultaneously indicate the names of the muscle groups that need to be activated or relaxed.

[0047] In practice, AI voice prompts and AI visual prompts provide users with real-time guidance on proper movement based on the target movement specifications. This process begins with the continuous acquisition of real-time posture and motion trajectory data during training. Real-time posture data is streamed at high frequency by the inertial measurement unit, while motion trajectory data is provided synchronously by the motion capture system. The real-time posture data is then compared in real-time with the target joint angle range in the target movement specifications. The target joint angle range is a preset allowable range of motion for major joints such as the hip, knee, ankle, and trunk for each standard movement in the training program. When a deviation is detected, a voice command containing the specific joint name and adjustment direction is broadcast through the AI ​​voice prompt channel.

[0048] In practice, the three-dimensional angle data of each major joint is extracted in real time from the real-time attitude data stream. This extraction process involves parsing the quaternion or Euler angle data of the inertial measurement unit and converting it to a unified biomechanical coordinate system to obtain the current angles of each joint in the sagittal, coronal, and horizontal planes. The extracted joint three-dimensional angle data is then compared frame-by-frame with the predefined target angle ranges for each joint in the target motion specification parameters. The comparison operation calculates the difference between the current angle value and the upper and lower boundaries of the target angle range. When the angle data of a specific joint continuously exceeds the target angle range for a preset threshold duration, it is determined that the specific joint has deviated from its motion. The preset threshold duration is used to filter out brief, unintentional attitude fluctuations; for example, it is set to trigger a judgment only after 0.5 consecutive seconds of exceeding the range.

[0049] In some embodiments, the direction of motion to be adjusted for the joint is determined based on the difference between the angle data of the deviation from the joint and the median of the target angle range. The median of the target angle range is the arithmetic mean of the upper and lower limits of the range. The sign and magnitude of the difference between the current angle and the median are calculated. The sign of the difference indicates whether the deviation direction is greater than or less than the median. This difference is then combined with the anatomical direction of joint motion to determine the adjustment direction. A speech synthesis engine is invoked to combine the name of the deviating joint, the detected deviation direction, and the target direction to be adjusted into a natural language command, such as "Right knee hyperextension detected, please flex slightly." The generated voice command is broadcast in real time through an AI voice prompt channel. After broadcasting, subsequent posture data is monitored to evaluate the command execution effect. If subsequent data indicates that the deviation has been corrected, the relevant prompts are stopped; if not corrected, the prompt content may be repeated or reinforced.

[0050] In practical implementation, the motion trajectory data is overlaid and displayed in real time with the target motion trajectory path in the target action specification parameters. The target motion trajectory path is one or more standard motion curves pre-drawn in virtual space. Through an AI visual cues channel, trajectory guidance and deviation feedback are presented in the form of highlighted paths, arrow indicators, or deviation heatmaps within the virtual reality interactive environment. Highlighted paths indicate the ideal path the user should follow; arrow indicators dynamically show the deviation between the real-time motion direction and the target direction; and deviation heatmaps visually map the degree of positional deviation on the user's actual trajectory using color intensity. Optionally, the color mapping function of the deviation heatmap can be expressed as... , in the formula, Represents the pixel color value rendered in the virtual environment. It represents the shortest spatial distance (i.e., deviation) between the real-time trajectory point and the target trajectory path. This represents a mapping function from deviation to color value; for example, the greater the deviation, the more reddish the color. Based on the difference between the muscle activation pattern coordination coefficient and the target pattern, the names of the muscle groups that need to be activated or relaxed are simultaneously prompted through the AI ​​voice prompt channel and the AI ​​visual prompt channel. The target pattern is preset by the rehabilitation plan. When the real-time calculated muscle activation pattern coordination coefficient shows that a specific muscle group is under-activated or over-activated, the AI ​​voice prompt channel broadcasts an instruction such as "Please strengthen the quadriceps muscles," while the AI ​​visual prompt channel highlights the corresponding muscle group position on the user's virtual avatar.

[0051] See Figure 5 This is a bar chart comparing the effectiveness indicators of different rehabilitation training programs. Multimodal combined training (VR + unstable platform) is significantly more effective than a single training program, demonstrating the synergistic rehabilitation value of "virtual reality + physical perturbation." It clearly shows that combined training is the most effective program and can be promoted as a core rehabilitation method; single programs (such as traditional training) have limited effectiveness, and their proportion in training can be reduced while increasing the duration of combined training. It demonstrates that the combination of VR and an unstable platform can simultaneously improve ability indicators and training efficiency, providing data support for the integration and innovation of rehabilitation technologies. The clear comparison shows that "VR + unstable platform combined training" is optimal in terms of stability, coordination improvement rate, and training efficiency, providing data support for rehabilitation institutions to select core training programs and avoiding the waste of resources from inefficient programs.

[0052] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A rehabilitation balance method based on virtual reality and an unstable platform, characterized in that, The method includes: Collect real-time posture data of the user on the unstable platform, surface electromyography signal data of the user, and motion trajectory data of the user; The user's current balance ability assessment parameters are generated based on the real-time posture data, the surface electromyography signal data, and the motion trajectory data. Based on the comparison results between the current balance ability assessment parameters and the preset rehabilitation stage goals, a personalized rehabilitation training plan is generated. The personalized rehabilitation training plan includes virtual reality scene dynamic parameters, unstable platform disturbance mode parameters, and target movement specification parameters. The virtual reality scene rendering engine is invoked to generate and present a virtual reality interactive environment adapted to the personalized rehabilitation training program based on the dynamic parameters of the virtual reality scene. Based on the unstable platform disturbance mode parameters, the unstable platform is controlled to generate multi-dimensional physical disturbances; Through AI voice prompts and AI visual prompts, real-time action guidance information is provided to the user based on the target action specification parameters to correct the user's posture when performing actions in the virtual reality interactive environment.

2. The rehabilitation balance method based on virtual reality and an unstable platform according to claim 1, characterized in that, The step of generating the user's current balance ability assessment parameters based on the real-time posture data, the surface electromyography signal data, and the motion trajectory data includes: The real-time posture data includes the body's center of gravity shift and changes in joint angles; The current balance ability assessment parameters include stability index, muscle activation mode coordination coefficient, and movement trajectory deviation. Extract the time series of the body center of gravity offset and the spatial distribution matrix of the joint angle changes from the real-time posture data; A stability analysis is performed on the time series of the body's center of gravity offset to calculate the stability index, which is a composite function of the envelope area of ​​the center of gravity swinging in a predetermined plane and the average velocity. The activation timing and intensity of the primary and antagonistic muscle groups are analyzed from the surface electromyography signal data, and the coordination coefficient of the muscle activation pattern is calculated based on the matching degree of the activation timing and intensity. The motion trajectory data is spatially registered with a preset standard motion trajectory template, the positional deviation of the key points of the trajectory is calculated, and the motion trajectory offset is determined based on the cumulative value of the positional deviation.

3. The rehabilitation balance method based on virtual reality and an unstable platform according to claim 2, characterized in that, The stability analysis of the time series of the body's center of gravity offset, and the calculation of the stability index, includes: The time series of the body's center of gravity offset is projected and decomposed in the horizontal plane to obtain the center of gravity displacement components in the front-back and left-right directions. Calculate the standard deviation, average moving speed, and oscillation frequency of the center of gravity displacement components in the front-back and left-right directions, respectively. A multidimensional stability assessment model is constructed based on the standard deviation, average moving speed and oscillation frequency. The multidimensional stability assessment model outputs a comprehensive stability index by weighted integration of the variation characteristics of each component. Based on the decay trend of the comprehensive stability index during the rehabilitation training cycle, the comprehensive stability index is normalized to generate a stability index that characterizes the user's immediate balance state.

4. The rehabilitation balance method based on virtual reality and an unstable platform according to claim 1, characterized in that, The process of generating a personalized rehabilitation training plan based on the comparison results between the current balance ability assessment parameters and the preset rehabilitation stage goals includes: The preset rehabilitation stage goals are obtained, which include target stability thresholds, target muscle coordination thresholds, and target trajectory accuracy thresholds required to be achieved at different rehabilitation stages. The stability index is compared with the target stability threshold to generate a stability training requirement level. The muscle activation mode coordination coefficient is compared with the target muscle coordination threshold to generate a muscle coordination training requirement level. The motion trajectory offset is compared with the target trajectory accuracy threshold to generate a trajectory accuracy training requirement level. Based on the combination of stability training requirement level, muscle coordination training requirement level and trajectory accuracy training requirement level, the preset training strategy mapping database is queried to match the corresponding virtual reality scene dynamic parameters, unstable platform disturbance mode parameters and target action specification parameters. The matched parameters are integrated into a personalized rehabilitation training program with clear execution steps, difficulty gradients, and feedback mechanisms.

5. The rehabilitation balance method based on virtual reality and an unstable platform according to claim 4, characterized in that, The step involves querying a preset training strategy mapping database based on a combination of the stability training requirement level, muscle coordination training requirement level, and trajectory accuracy training requirement level to match the corresponding dynamic parameters of the virtual reality scene, including: The dynamic parameters of the virtual reality scene include at least the complexity of the virtual scene, the frequency and speed of the appearance of interference objects in the scene, and the task objectives that the user's virtual avatar needs to complete. When the stability training requirement level is high, the matched virtual reality scene dynamic parameters point to a virtual environment containing a mobile platform or a narrow passage. When the muscle coordination training demand level is high, the matching virtual reality scene dynamic parameters are directed to the task that requires the user's virtual avatar to perform a compound action task with a specific rhythm or resistance. When the trajectory accuracy training requirement level is high, the matched virtual reality scene dynamic parameters point to a virtual environment that includes precise path guidance or requires avoiding dynamic obstacles.

6. The rehabilitation balance method based on virtual reality and an unstable platform according to claim 1, characterized in that, The method of controlling the unstable platform to generate multi-dimensional physical disturbances based on the unstable platform disturbance mode parameters includes: The disturbance mode parameters of the unstable platform are analyzed to obtain the direction sequence, amplitude sequence, frequency sequence and duration of the disturbance; The unstable platform is controlled to rotate around one or more axes of the roll axis, pitch axis and yaw axis according to the direction sequence. The rotation angle or translation displacement of the unstable platform in each direction is controlled according to the amplitude sequence. The rate of change of the rotational motion or translational displacement of the unstable platform is controlled according to the frequency sequence. The direction sequence, amplitude sequence, and frequency sequence are combined and arranged according to the duration to generate a composite perturbation stimulus that changes dynamically over time, which is then applied to the user's support plane.

7. The rehabilitation balance method based on virtual reality and an unstable platform according to claim 6, characterized in that, The step of parsing the unstable platform disturbance mode parameters to obtain the disturbance direction sequence, amplitude sequence, frequency sequence, and duration includes: Read the perturbation mode configuration file encoded in the perturbation mode parameters of the unstable platform. The configuration file uses a hierarchical data structure to store perturbation sequence information. Extract the total duration of the disturbance and the division information of the disturbance phases from the top-level structure of the configuration file; Based on the information on the division of the disturbance stages, the set of disturbance events contained in each disturbance stage is analyzed layer by layer; For each disturbance event, parameter decoding is performed to separate the direction sequence describing the spatial properties of the disturbance, the amplitude sequence describing the intensity properties of the disturbance, the frequency sequence describing the rhythm properties of the disturbance, and the independent duration of each disturbance event. The obtained direction sequence, amplitude sequence, frequency sequence and duration are integrated according to the timeline to generate a timing control command stream that can be directly read by the actuator of the unstable platform.

8. The rehabilitation balance method based on virtual reality and an unstable platform according to claim 1, characterized in that, The provision of real-time action guidance information to the user based on the target action specification parameters through AI voice prompt channels and AI visual prompt channels includes: Continuously acquire the user's real-time posture data and motion trajectory data during the training process; The real-time posture data is compared with the target joint angle range in the target action specification parameters in real time. When a deviation is detected, a voice command containing the specific joint name and adjustment direction is broadcast through the AI ​​voice prompt channel. The motion trajectory data is overlaid and displayed in real time with the target motion trajectory path in the target motion specification parameters. The trajectory guidance and deviation feedback are presented in the form of highlighted paths, arrows, or deviation heatmaps in the virtual reality interactive environment through the AI ​​visual prompt channel. Based on the difference between the muscle activation mode coordination coefficient and the target mode, the names of the muscle groups that need to be activated or relaxed are simultaneously prompted through the AI ​​voice prompt channel and the AI ​​visual prompt channel.

9. The rehabilitation balance method based on virtual reality and an unstable platform according to claim 8, characterized in that, The real-time posture data is compared with the target joint angle range in the target motion specification parameters in real time. When a deviation is detected, a voice command containing the specific joint name and adjustment direction is broadcast through the AI ​​voice prompt channel, including: The three-dimensional angle data of each major joint is extracted in real time from the real-time attitude data stream; The extracted joint 3D angle data is compared frame by frame with the predefined target angle range of each joint in the target motion specification parameters; When the angle data of a specific joint continuously exceeds the target angle range for a preset threshold duration, it is determined that the specific joint has deviated from its movement. The direction of joint movement to be adjusted is determined based on the difference between the angle data of the deviation from the joint and the median value of the target angle range; The speech synthesis engine is invoked to combine the name of the deviated joint, the detected direction of deviation, and the target direction to be adjusted into a natural language command; The generated voice commands are broadcast in real time through the AI ​​voice prompt channel, and subsequent posture data is monitored after the broadcast to evaluate the execution effect of the commands.

10. A rehabilitation balance system based on virtual reality and an unstable platform, characterized in that: The device includes a processor and a memory, the memory and the processor being connected. The memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to implement the rehabilitation balance method based on virtual reality and an unstable platform as described in any one of claims 1-9.

Images