Hand rehabilitation equipment and system based on multi-source data analysis

Abstract

The invention relates to the technical field of limb training equipment, in particular to hand rehabilitation equipment and system based on multi-source data analysis. The hand rehabilitation equipment comprises a wearable glove, a positive and negative pressure air pump module, an electromagnetic valve module, a temperature control module, a human-computer interaction interface and a memory storage module. According to multi-source hand sensing data, an actual intention motion polar coordinate layout of a current hand of a target user, standardization dislocation analysis and multi-dimensional lag compensation actual intention motion polar coordinate layout are constructed, and then independent angle change of virtual execution posture rotation when the hand rehabilitation equipment applies stroke driving compensation amount is solved through reverse kinematics. Joint flexion and extension control parameters required by the hand rehabilitation equipment are obtained, and air pressure control accumulation calculation is executed according to the joint flexion and extension control parameters, so that the hand rehabilitation equipment is intelligently trained and controlled. Accurate control feedback is provided for the hand rehabilitation equipment by analyzing the multi-source data, the training efficiency and accuracy are effectively improved, and the control error risk is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of limb training equipment, and in particular to a hand rehabilitation device and system based on multi-source data analysis. Background Art

[0002] With the development of neurorehabilitation medicine and intelligent rehabilitation devices, hand rehabilitation devices have gained widespread application in areas such as post-stroke hand dysfunction and post-operative rehabilitation of hand injuries. Traditional hand rehabilitation devices often use a single sensor (such as an electromyographic sensor, force sensor, or angle sensor) for signal acquisition, and a pre-programmed mechanical actuator to perform rehabilitation training. However, patients' intended movements are complex and diverse, with significant individual differences. A single data source often fails to accurately reflect a patient's true motor intentions and recovery status.

[0003] With advances in wearable technology, sensor fusion, and artificial intelligence, multi-source data fusion has become a key path to improving the intelligence of rehabilitation training. By integrating multiple sources of information, such as electromyography (EMG), electroencephalography, hand posture, joint angles, or physiological parameters, a patient's motor ability can be more comprehensively assessed, rehabilitation strategies can be dynamically adjusted, and personalized, refined control can be achieved. However, existing hand rehabilitation devices still rely solely on a single type of data, making it difficult to cope with ambiguity and interference in movement intention recognition. This can easily cause training movements to deviate from the patient's intentions, making it impossible for hand rehabilitation devices to make reasonable and accurate control decisions based on the current hand condition of the target user. Secondly, most existing devices lack real-time perception and analysis of user movement feedback (such as force output, muscle response, or coordinated movements), making it difficult to achieve accurate human-machine collaboration and closed-loop control. This can lead to errors such as training incoordination, significantly reducing the effectiveness and efficiency of hand training. Summary of the Invention

[0004] The present invention overcomes the deficiencies of the prior art and provides a hand rehabilitation device and system based on multi-source data analysis.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: A first aspect of the present invention provides a hand rehabilitation device based on multi-source data analysis, the hand rehabilitation device comprising: Wearing gloves: The wearing gloves are used to completely cover the finger joints and palm joints of the target user; Positive and negative pressure air pump module: The positive and negative pressure air pump module is used to control the generation of positive and negative pressure gas, provide air pressure power for auxiliary flexion and extension training activities of finger joints, and adjust the training intensity and strength; Solenoid valve module: The solenoid valve module is connected to the positive and negative pressure air pump modules and is responsible for controlling the on / off and flow of compressed air to simulate the natural grasping action of the hand; Temperature control module: The temperature control module is used to adjust the hand temperature according to the training status of the target user, providing a heating effect for hand rehabilitation training; Human-computer interaction interface: The human-computer interaction interface is used to provide the target user with the ability to switch working modes and personalize the time, strength, and frequency of hand training; Memory storage module: The memory storage module is responsible for memorizing and saving the last hand training program, so that the effect can be directly applied next time without repeated debugging.

[0006] A second aspect of the present invention provides a control method for a hand rehabilitation device based on multi-source data analysis, which is applied to the hand rehabilitation device based on multi-source data analysis and includes the following steps: S102: Acquire multi-source hand sensing data from a hand rehabilitation device and a human body intended motion system, perform polar coordinate radial fusion of the multi-source hand sensing data according to a logical and standardized trend in accordance with the human body intended motion system, and obtain an actual intended motion polar coordinate layout of the multi-source hand sensing data; S104: Analyzing the polar coordinate layout of the actual intended motion based on the standardized offset of the ideal working condition parameter set of the human body intended motion when the hand rehabilitation device executes the selected working mode, thereby performing multi-dimensional hysteresis compensation to obtain the stroke drive compensation amount required for the current intended motion of the hand rehabilitation device to reach the selected point working mode; S106: Constructing a chain joint motion structure for the target user, and simultaneously injecting a virtual end effector into the hand rehabilitation device model. Using an inverse kinematics solution algorithm, the independent angle changes of each virtual end effector's posture rotation when the hand rehabilitation device applies the stroke drive compensation in accordance with the chain joint motion structure are solved to obtain joint flexion and extension control parameters for the hand rehabilitation device to execute the stroke drive compensation. S108: Based on the rotational posture scalar field derived from the simulation process and the preset training control strategy of the hand rehabilitation device, a quaternion mapping field is constructed, the joint flexion and extension control parameters are mapped to the quaternion mapping field and the air pressure control accumulation calculation is performed, so as to intelligently train and control the hand rehabilitation device.

[0007] More specifically, the step S102 includes the following steps: Acquire a multi-source sensor array of the hand rehabilitation device. When the target user wears the hand rehabilitation device, the multi-source sensor array collects data of the target user within a preset time period to obtain multi-source hand sensing data; Centrally preprocess each group of multi-source hand sensing data, introduce a covariance algorithm to calculate the autocorrelation dependency of each group of multi-source hand sensing data after centralized preprocessing, and output the coupling variance between each data variable in each group of multi-source hand sensing data; Obtain anthropometric knowledge graphs through big data, use the anthropometric knowledge graphs to identify multi-source hand sensing data, output a human body intention movement system that conforms to the target user's stage characteristics, and extract different human body intention movement indicators and logical normative trend indexes for each human body intention movement indicator through the human body intention movement system; According to different human intention motion indices, the principal component source coordinate angles are allocated in the form of Cartesian coordinates to form a polar coordinate framework for human intention motion. Maintaining the coupling variance as the autonomous projection criterion, a radial joint projection vector constraint is established according to the logical norm trend index. Based on the radial joint projection vector constraint, each group of multi-source hand sensing data is radially decomposed one by one according to the autonomous projection criterion in the covariance algorithm, and a cross calculation of the collaborative correlation is performed. At this time, the radial covariance scale and cross-projection covariance matrix of each group of multi-source hand sensing data are output; If the weighted sum of the cross-projection covariance matrices exceeds the maximum weighted sum threshold, each set of multi-source hand sensing data is projected onto the fitting coordinate radius on the radial covariance scale of the subordinate response within the human body intended motion polar coordinate architecture to generate the actual intended motion polar coordinate layout of the multi-source hand sensing data.

[0008] More specifically, the step S104 includes the following steps: Obtain the target user's selected working mode for the hand rehabilitation device and the conceptual design drawings of the hand rehabilitation device, and obtain the ideal working condition parameter set of the human body's intended movement when the hand rehabilitation device executes the selected working mode through the conceptual design drawings; Based on the ideal working condition parameter set, a multi-source hand sensing data is constructed to tend towards the ideal intended motion polar coordinate layout under the selected working mode. A hash misalignment algorithm is introduced to calculate the hash misalignment area function of the non-overlapping area between the actual intended motion polar coordinate layout and the ideal intended motion polar coordinate layout. Based on the hash misalignment area function, the steady-state misalignment degree of the current hand rehabilitation device that enables the target user to operate in the selected working mode is determined. If the steady-state misalignment is greater than the preset steady-state misalignment, an intended motion state space model of the current hand rehabilitation device is constructed based on the coupling correlation of multi-source hand sensing data on different human intended motion indicators, and a decoupling analysis between different input channels and output channels of the intended motion state space model is performed by calculating zero points and poles to determine the open-loop operation transfer function of the current hand rehabilitation device; Obtain the desired diagonal control system for the selected operating mode, determine the compensation cutoff frequency on the Bode plot based on the phase margin of the desired diagonal control system, and establish a gain compensation matrix for different multi-source hand sensing data based on the steady-state misalignment. The open-loop gain of the intended motion state space model is globally adjusted in a steady-state manner through the gain compensation matrix, so that the center frequency of the hysteresis network of the open-loop operation transfer function continuously approaches the compensation cutoff frequency, generating a multi-dimensional hysteresis compensation structure. The multi-dimensional hysteresis compensation structure is deployed in the actual diagonal control system of the current hand rehabilitation device to obtain the stroke drive compensation amount required for the target user to achieve the point selection working mode when using the current hand rehabilitation device.

[0009] More specifically, the step S106 includes the following steps: The target user's hand features are copied using a multi-sensor array to obtain the target user's actual joint linear contours. The joint linear contours with the greatest similarity to the actual joint linear contours are extracted from the hand feature database preset by the hand rehabilitation device and defined as the joint reference blueprint. Obtaining standard product drawings of a hand rehabilitation device, reconstructing a simulation model of the hand rehabilitation device according to the standard product drawings using SolidWorks model simulation software, and injecting a virtual end effector into each hand joint in the simulation model; Based on the ideal working condition parameter set, the virtual end effector presets the final posture rotation matrix of the selected working mode. According to the joint reference blueprint, the joint degrees of freedom, bone connection relationship and starting motion angle that match the target user's hand line are defined from the root to the virtual end effector, and the chain joint motion structure is output; Obtain personalized setting parameters for the target user using the selected working mode, synchronously obtain the movable points of the current hand rehabilitation device, and the real-time joint angles of the target user maintaining the previous default posture state when capturing multi-source hand sensing data; An inverse kinematics solution algorithm is introduced. Starting from the real-time joint angle as the established origin, the stroke drive compensation is applied in the inverse kinematics solution algorithm in accordance with the motion premise constraints of the chain joint motion structure and personalized setting parameters. The actual position of each virtual end effector is deduced from this, and the end posture rotation matrix is ​​obtained. The deviation between the end posture rotation matrix and the end posture rotation matrix is ​​calculated to obtain the matrix error of each movable point. Matrix error is analyzed to locate and assign bone binding weights and perform weighted transformation rendering to generate an active weighted vertex grid. The active weighted vertex grid is used to solve the independent angle changes of the virtual end effector, and the joint flexion and extension control parameters of the hand rehabilitation device's execution stroke drive compensation are obtained.

[0010] More specifically, the analysis matrix error is used to locate and assign bone binding weights and perform weighted transformation rendering to generate an active weighted vertex grid. The active weighted vertex grid is used to solve the independent angle changes of the virtual end effector to obtain the joint flexion and extension control parameters of the hand rehabilitation device to execute the stroke drive compensation. Specifically, the steps include: If the matrix error is less than the preset matrix error, the movable point is calibrated as a low-amplitude response active point; If the matrix error is greater than the preset matrix error, the active point is calibrated as a high-amplitude response active point; Based on the matrix error, the skeleton binding weights of different low-amplitude response active points and high-amplitude response active points are assigned. The active points are rendered with weighted transformation using the final posture rotation matrix according to the skeleton binding weights to generate an active weighted vertex grid. During the weighted transformation process, a synchronous linear approximation is used to establish the Jacobian matrix between the joint angle changes and the virtual end-effector posture changes. Based on the Jacobian matrix, a damped least squares solution is performed on the joint angles of the active trade-off vertex grid to obtain an independent angle change chain for each virtual end-effector. The joint angle corresponding to each finger joint on the current hand rehabilitation device is continuously debugged and transformed according to the independent angle change chain until the matrix errors of the low-amplitude response active point and the high-amplitude response active point are eliminated, and the joint flexion and extension control parameters of the hand rehabilitation device for executing the stroke drive compensation are obtained.

[0011] More specifically, the step S108 includes the following steps: Obtaining the preset training control strategy for the hand rehabilitation device, and extracting the various rotation change vectors planned by the hand rehabilitation device for different joint flexion and extension processes, as well as the air pressure control levels output by the positive and negative air pressure pumps to achieve the decision-making output of each rotation change vector through the preset training control strategy; The simulation model of the hand rehabilitation device is used to import the preset training control strategy for simulation. At this time, the rotational posture scalar field of the hand rehabilitation device during the process of executing the stroke drive compensation from the real-time joint angle simulation to achieve the selected working mode is exported through SolidWorks model simulation software. The quaternion mapping field is constructed based on the structural layout of the rotational posture scalar field. Each rotation change vector is defined as a prefix label in the quaternion mapping domain, and the corresponding air pressure control magnitude of each rotation change vector is set as a suffix label to form a prefix-suffix mapping bundled control query item, and each rotation change vector is preset to be assigned a quaternion string; Performing rotational differential conversion according to the joint flexion and extension angle, angular velocity, angle switching step, and joint flexion and extension increment recorded in the joint flexion and extension control parameters to obtain the necessary rotation change vector for the hand rehabilitation device to execute the stroke drive compensation amount; An exponential mapping formula is introduced to convert the necessary rotation change vector into an incremental quaternion to obtain a necessary rotation incremental quaternion, and the necessary rotation incremental quaternion is projected into a quaternion mapping field, so that the necessary rotation incremental quaternion is traversed one by one through the quaternion string corresponding to each control query item for approximate comparison; If the approximate agreement between the necessary rotation increment quaternion and the quaternion string is greater than a preset approximate agreement, a cumulative superposition cap mark is added to the quaternion string; otherwise, the quaternion string is ignored, and a series of candidate superposition quaternions are obtained; Quaternion multiplication is used to rotate and accumulate the air pressure control magnitudes of a series of candidate superposition quaternions corresponding to the control query items to update the posture quaternion of the hand rehabilitation device. Finally, the air pressure control accumulation parameters are output, and intelligent training control of the hand rehabilitation device is performed based on the air pressure control accumulation parameters.

[0012] The third aspect of the present invention provides a hand rehabilitation system based on multi-source data analysis, the hand rehabilitation system includes a memory and a processor, the memory stores a hand rehabilitation method program based on multi-source data analysis, when the hand rehabilitation method program is executed by the processor, any one of the hand rehabilitation method steps is implemented.

[0013] The present invention solves the technical defects existing in the background technology, and the beneficial technical effects of the present invention are: This invention integrates multi-source sensor information, including electromyographic signals, movement posture, joint angles, and force changes, to perceive in real time the patient's hand movement intentions, status, and functional changes. By intelligently adjusting training intensity, frequency, and duration through positive and negative pressure, it enables personalized, phased hand rehabilitation training. Compared to traditional devices, this invention boasts greater adaptability and feedback control capabilities, effectively improving training efficiency and accuracy, reducing the risk of manual error, and enabling quantitative assessment and visual tracking of rehabilitation outcomes. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in 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 only some embodiments of the present invention. For ordinary technicians in this field, without paying any creative work, they can also obtain drawings of other embodiments based on these drawings.

[0015] Figure 1 A device architecture diagram of a hand rehabilitation device based on multi-source data analysis is shown; Figure 2 A first method flow chart of the control method of the present device is shown; Figure 3 A second method flow chart showing the control method of the device is shown; Figure 4 A system framework diagram of a hand rehabilitation system based on multi-source data analysis is shown. DETAILED DESCRIPTION

[0016] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0017] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0018] The first aspect of the present invention provides a hand rehabilitation device based on multi-source data analysis, such as Figure 1 As shown, the hand rehabilitation device includes: Wearing gloves 11: The wearing gloves are used to completely cover the finger joints and palm joints of the target user; Positive and negative pressure air pump module 12: The positive and negative pressure air pump module is used to control the generation of positive pressure gas and negative pressure gas, provide air pressure power for auxiliary flexion and extension training activities of finger joints, and adjust the training intensity and strength; Solenoid valve module 13: The solenoid valve module is connected to the positive and negative pressure air pump modules and is responsible for controlling the on / off and flow of compressed air to simulate the natural grasping action of the hand; Temperature control module 14: The temperature control module is used to adjust the hand temperature according to the training status of the target user, providing a heating effect for hand rehabilitation training; Human-computer interaction interface 15: The human-computer interaction interface is used to provide the target user with the ability to switch working modes and personalize the time, strength, and frequency of hand training; Memory storage module 16: The memory storage module is responsible for memorizing and saving the last hand training program, so that the effect can be directly applied next time without repeated debugging.

[0019] The second aspect of the present invention provides a control method for a hand rehabilitation device based on multi-source data analysis, which is applied to the hand rehabilitation device based on multi-source data analysis, such as Figure 2 As shown, the following steps are included: S102: Acquire multi-source hand sensing data from a hand rehabilitation device and a human body intended motion system, perform polar coordinate radial fusion of the multi-source hand sensing data according to a logical and standardized trend in accordance with the human body intended motion system, and obtain an actual intended motion polar coordinate layout of the multi-source hand sensing data; S104: Analyzing the polar coordinate layout of the actual intended motion based on the standardized offset of the ideal working condition parameter set of the human body intended motion when the hand rehabilitation device executes the selected working mode, thereby performing multi-dimensional hysteresis compensation to obtain the stroke drive compensation amount required for the current intended motion of the hand rehabilitation device to reach the selected point working mode; S106: Constructing a chain joint motion structure for the target user, and simultaneously injecting a virtual end effector into the hand rehabilitation device model. Using an inverse kinematics solution algorithm, the independent angle changes of each virtual end effector's posture rotation when the hand rehabilitation device applies the stroke drive compensation in accordance with the chain joint motion structure are solved to obtain joint flexion and extension control parameters for the hand rehabilitation device to execute the stroke drive compensation. S108: Based on the rotational posture scalar field derived from the simulation process and the preset training control strategy of the hand rehabilitation device, a quaternion mapping field is constructed, the joint flexion and extension control parameters are mapped to the quaternion mapping field and the air pressure control accumulation calculation is performed, so as to intelligently train and control the hand rehabilitation device.

[0020] More specifically, the S102, as Figure 3 As shown, the specific steps include: S202: Acquire a multi-source sensor array of the hand rehabilitation device. When the target user wears the hand rehabilitation device, collect data of the target user within a preset time period through the multi-source sensor array to obtain multi-source hand sensing data; S204: Centrally preprocessing each group of multi-source hand sensing data, introducing a covariance algorithm to perform autocorrelation dependency calculation on each group of multi-source hand sensing data after the centralized preprocessing, and outputting the coupling variance between each data variable in each group of multi-source hand sensing data; S206: Obtaining anthropometric knowledge graphs through big data, using the anthropometric knowledge graphs to identify multi-source hand sensing data, outputting a human body intended motion system that conforms to the target user's stage characteristics, and extracting different human body intended motion indicators and logical normative trend indices for each human body intended motion indicator through the human body intended motion system; S208: allocating principal component source coordinate angles in Cartesian coordinate form according to different human intention motion indices to form a human intention motion polar coordinate framework, maintaining coupling variance as an autonomous projection criterion, and establishing a radial joint projection vector constraint according to a logical norm trend index; S210: Based on the radial joint projection vector constraint, radially decompose each group of multi-source hand sensing data one by one according to the autonomous projection criterion in the covariance algorithm, and perform cross calculation of collaborative correlation, thereby outputting the radial covariance scale and cross projection covariance matrix of each group of multi-source hand sensing data; S212: If the weighted sum value of the cross-projection covariance matrix exceeds the maximum weighted sum threshold, each set of multi-source hand sensing data is projected onto the fitting coordinate radius on the radial covariance scale of the subordinate response within the human body intended motion polar coordinate architecture to generate the actual intended motion polar coordinate layout of the multi-source hand sensing data.

[0021] It should be noted that this device will control the positive and negative air pressure pumps to drive hand training according to the intention and movement of the target user, but the multi-source data types of the hand are numerous, and it is difficult to uniformly feedback the current hand status in real time from the two levels of intention and movement, resulting in errors in the subsequent air pressure propulsion control. To this end, this method first obtains multi-source hand sensing data through the multi-source sensor array of the hand rehabilitation device, wherein the multi-source hand sensing data includes electromyographic signal data, electroencephalographic signal data, skin electrical response data, joint angle data, finger and palm posture data, grip data, finger force data and basic information of the target user. Subsequently, the multi-source hand sensing data is centrally preprocessed, which can eliminate the data offset and avoid the correlation calculation of mean interference, so that the covariance matrix correctly reflects the true relationship between the data variables, significantly improving the effectiveness and consistency of the subsequent polar coordinate fusion of multi-source data. Covariance calculations calculate the autocorrelation of each set of pre-processed, centralized multi-source hand sensor data, outputting the coupling variance of each data point. This coupling variance quantifies the typical dependencies between each data variable within each set of multi-source hand sensor data. Based on the magnitude of the coupling variance, a basic matrix is ​​provided for solving the projection direction of the multi-source data, characterizing the expression structure between different data sources. Because multi-source data contain diverse intentional movement trends, such as feedback from EEG signals, joint angle data, and finger force data indicating a user's intention to extend rather than flex their fingers, feedback on current hand status requires a diverse and standardized set of intended movement references. Specifically, different human intentional movement indicators (diversity) are established. These include grasping intention, pinching intention, opening intention, fisting intention, finger erection intention, wrist flexion, single finger movement, pronation gesture movement, and supination gesture movement. Each indicator is assigned an ergonomically defined logical normative index (standardization), namely the logical normative trend index.

[0022] It should be noted that this method constructs a polar coordinate architecture for human intentional motion, which uses Cartesian coordinates to define and divide multiple equal angles starting from the center position. The polar coordinate area corresponding to each equal angle is used to represent a human intentional motion indicator. This can fully display the real-time intentional motion status of multi-source data in a polar coordinate manner, providing a multi-dimensional visualization fusion basis for multi-source data feedback. Among them, the logical norm trend index constrains the projection benchmark of each polar coordinate angle to meet the logical norm of the human intention movement indicator, preventing the projection vector of multi-source data from being arbitrarily scaled and providing standardized projection results. The polar coordinate projection of multi-source data must always maintain the typical correlation between them to ensure the real-time and reproducibility of multi-source data. Therefore, the coupled variance is set as a projection criterion that is highly autonomous, autonomous, and spontaneous. Then, based on the radial joint projection vector constraint and in accordance with the autonomous projection criterion, the multi-source hand sensing data are radially decomposed one by one to obtain the corresponding radial covariance scale and cross-projection covariance matrix. The radial covariance scale measures the projection vector that jointly shares the current hand condition between each group of multi-source hand sensing data, maximizing the total correlation between the projection results of each multi-source data and reflecting the common structure between the multi-source data. The cross-projection covariance matrix represents the total correlation between different data variables in the multi-source hand sensing data. If the weighted sum value of the cross-projection covariance matrix exceeds the maximum weighted sum threshold, it means that the total correlation between all multi-source hand sensing data at a certain human intention motion index level reaches the maximum effect on the radial covariance scale, that is, the current hand status of the target user can be collaboratively fed back from the index level. Therefore, each group of multi-source hand sensing data is projected on the radial covariance scale, so that the multi-source hand sensing data is fitted into a polar coordinate radius amplitude representation, forming a polar coordinate summary view that expresses the real-time status of the current hand at the intention and motion levels in terms of angle and radius. Through this method, multi-source data can be analyzed in a multi-dimensional, diversified and standardized manner, and the real-time intention and motion status of the target user's hand can be fed back in real time, providing an analysis basis for subsequent precise control of the device.

[0023] More specifically, the step S104 includes the following steps: Obtain the target user's selected working mode for the hand rehabilitation device and the conceptual design drawings of the hand rehabilitation device, and obtain the ideal working condition parameter set of the human body's intended movement when the hand rehabilitation device executes the selected working mode through the conceptual design drawings; Based on the ideal working condition parameter set, a multi-source hand sensing data is constructed to tend towards the ideal intended motion polar coordinate layout under the selected working mode. A hash misalignment algorithm is introduced to calculate the hash misalignment area function of the non-overlapping area between the actual intended motion polar coordinate layout and the ideal intended motion polar coordinate layout. Based on the hash misalignment area function, the steady-state misalignment degree of the current hand rehabilitation device that enables the target user to operate in the selected working mode is determined. If the steady-state misalignment is greater than the preset steady-state misalignment, an intended motion state space model of the current hand rehabilitation device is constructed based on the coupling correlation of multi-source hand sensing data on different human intended motion indicators, and a decoupling analysis between different input channels and output channels of the intended motion state space model is performed by calculating zero points and poles to determine the open-loop operation transfer function of the current hand rehabilitation device; Obtain the desired diagonal control system for the selected operating mode, determine the compensation cutoff frequency on the Bode plot based on the phase margin of the desired diagonal control system, and establish a gain compensation matrix for different multi-source hand sensing data based on the steady-state misalignment. The open-loop gain of the intended motion state space model is globally adjusted in a steady-state manner through the gain compensation matrix, so that the center frequency of the hysteresis network of the open-loop operation transfer function continuously approaches the compensation cutoff frequency, generating a multi-dimensional hysteresis compensation structure. The multi-dimensional hysteresis compensation structure is deployed in the actual diagonal control system of the current hand rehabilitation device to obtain the stroke drive compensation amount required for the target user to achieve the point selection working mode when using the current hand rehabilitation device.

[0024] It should be noted that since the target user can pre-select the working mode, each working mode has a preset corresponding training template. The movement from the current real-time hand state to the intended movement in the training template necessarily requires a control stroke. However, existing control methods typically roughly estimate the movement angle deviation between the real-time hand state and the training template to obtain the required drive stroke for the finger joints. This makes the subsequent output control compensation of the positive and negative pressure air pumps less accurate, making it difficult for hand training to meet the standard control requirements of the selected working mode. To address this, the present method first obtains the target user's selected working mode and the ideal working condition parameter set (training template) under that mode. The working modes of this device mainly include automatic mode, single-finger cycle mode, anti-interference mode, and mirror mode. According to the ideal working condition parameter set, the multi-source hand sensing data tends to the ideal intention motion polar coordinate layout under the selected working mode condition. The ideal intention motion polar coordinate layout can reflect the current real-time hand condition to achieve the standard degree of the selected working mode, that is, it is globally reflected from the hash dislocation area function of the non-overlapping area between the two. The hash dislocation area function represents a steady-state dislocation degree of the current hand rehabilitation device running the selected working mode; if the steady-state dislocation degree is greater than the preset steady-state dislocation degree, it means that the phase margin of the current hand rehabilitation device is insufficient, making it difficult for the current hand condition to complete the selected working mode, and it is necessary to use positive and negative pressure air pumps and electromagnetic The valve applies a certain amount of air pressure to propel compensation control. At this time, this method constructs the intended motion state space model of the current hand rehabilitation device based on the coupling correlation of multi-source hand sensing data on different human intention motion indicators, and then clarifies the dynamic relationship between the input and output of the controlled object, including coupling structure, stability or response, etc., and then decouples the intended motion state space model by calculating the zero points and poles, analyzes the directionality and coupling strength of multi-source data from different input channels and output channels of multi-dimensional feedback of the current hand rehabilitation device, and further locks the control plane, link and amplitude to be compensated, improves steady-state performance, and effectively improves the compensation accuracy of stroke drive.

[0025] It should be noted that the desired diagonal control system provides an approximate diagonalized architecture for achieving the desired decoupling effect (optimal decoupling of the directionality and coupling strength of multi-source hand sensor data) for the current hand rehabilitation device after compensation in the selected operating mode. The compensation cutoff frequency determined by the desired diagonal control system serves as the center point of the compensation structure, which can set a target for the compensation range. The steady-state misalignment is then used to establish a gain compensation matrix for different multi-source hand sensor data, globally calibrating the open-loop gain of the intended motion state-space model in a steady-state manner. This ensures that the center frequency of the lag network in the open-loop transfer function consistently approaches the compensation cutoff frequency, thereby precisely controlling the compensation structure's influence frequency band and improving the specificity and directionality of the pneumatic propulsion compensation to meet the requirements of the selected-point operating mode. This further enhances the steady-state compensation accuracy and smoothness of the current hand rehabilitation device, ensuring the accuracy of the assisted operation control of the current hand rehabilitation device for the target user's intended motion in the selected-point operating mode.

[0026] More specifically, the step S106 includes the following steps: The target user's hand features are copied using a multi-sensor array to obtain the target user's actual joint linear contours. The joint linear contours with the greatest similarity to the actual joint linear contours are extracted from the hand feature database preset by the hand rehabilitation device and defined as the joint reference blueprint. Obtaining standard product drawings of a hand rehabilitation device, reconstructing a simulation model of the hand rehabilitation device according to the standard product drawings using SolidWorks model simulation software, and injecting a virtual end effector into each hand joint in the simulation model; Based on the ideal working condition parameter set, the virtual end effector presets the final posture rotation matrix of the selected working mode. According to the joint reference blueprint, the joint degrees of freedom, bone connection relationship and starting motion angle that match the target user's hand line are defined from the root to the virtual end effector, and the chain joint motion structure is output; Obtain personalized setting parameters for the target user using the selected working mode, synchronously obtain the movable points of the current hand rehabilitation device, and the real-time joint angles of the target user maintaining the previous default posture state when capturing multi-source hand sensing data; An inverse kinematics solution algorithm is introduced. Starting from the real-time joint angle as the established origin, the stroke drive compensation is applied in the inverse kinematics solution algorithm in accordance with the motion premise constraints of the chain joint motion structure and personalized setting parameters. The actual position of each virtual end effector is deduced from this, and the end posture rotation matrix is ​​obtained. The deviation between the end posture rotation matrix and the end posture rotation matrix is ​​calculated to obtain the matrix error of each movable point. Matrix error is analyzed to locate and assign bone binding weights and perform weighted transformation rendering to generate an active weighted vertex grid. The active weighted vertex grid is used to solve the independent angle changes of the virtual end effector, and the joint flexion and extension control parameters of the hand rehabilitation device's execution stroke drive compensation are obtained.

[0027] It should be noted that the training templates currently used by hand rehabilitation devices to achieve the selected working mode using stroke drive compensation are actually the motion process of the sequential joint rotation of different finger joints. However, existing control methods have difficulty accurately inferring the degree of flexion and extension rotation of the finger joints based on the known stroke drive compensation. Furthermore, the finger joints of the glove of this device each have corresponding node-controlled positive and negative pressure air pumps. This makes it difficult for the corresponding positive and negative pressure air pumps to output accurate and reliable propulsion control air pressure, which can lead to incoordination and excessive finger movement during training. To address this issue, this method constructs a standardized hand rehabilitation device simulation model, primarily for simulating stroke drive compensation. Notably, this method virtualizes the finger joint end with an actuator (virtual end effector) in this simulation model. This virtual end effector can be used to capture and infer the position and posture of the current hand rehabilitation device during the stroke drive compensation simulation process, providing a basis for the reverse calculation of parameters such as the angle or displacement of subsequent joint flexion and extension. Compared with traditional manual experience-based intervention, this method can significantly improve the accuracy of joint rotation flexion and extension deduction and ensure control reliability. Synchronously obtain the joint linear profile (joint reference blueprint) that matches the target user's hand linear profile (maximum similarity) as a copy template for the joint chain topology and degree of freedom restrictions, thereby constructing a chain joint motion structure that defines the joint degrees of freedom, skeletal connection relationships, and starting motion angles that match the target user's hand linear profile from the root to the virtual end effector, thereby clarifying the adjustable joint position nodes and their physical limitations, forming a structural clue basis for reverse deduction. Since the point selection working mode allows the target user to customize different personalized setting parameters such as force, frequency, and time, the hand control is extremely flexible, which may cause drift in the simulation. Therefore, the personalized setting parameters already chosen by the target user need to be added as directional constraints for the motion premise during the simulation process, effectively improving the traceability reliability of the hand rehabilitation equipment for joint-driven errors.

[0028] It should be noted that this method, through the introduction of an inverse kinematics algorithm, applies stroke drive compensation based on the chained joint motion structure and the motion constraints of the personalized parameters, starting from the real-time joint angles. This method reversely infers the stroke rotation landing points of different active joints (movable points) when the hand rehabilitation device executes the stroke drive compensation. The matrix error between the end-pose rotation matrix and the final pose rotation matrix for each movable point is calculated. This matrix error quantifies the deviation between the current rotational posture and the target rotational posture when achieving the selected working mode. This matrix error drives the coordinated inverse kinematics process of different joints and serves as critical feedback information for coordinated scheduling. This method uses simulation to inversely infer the joint flexion and extension landing points, angles, and displacements of the current hand rehabilitation device during the time-series execution of the stroke drive compensation. This provides a complete flexion and extension clue chain for achieving the selected working mode when the positive and negative pressure air pumps operate the stroke drive compensation at different joint nodes of the hand rehabilitation device, ensuring the coordination and rationality of hand training and improving the accuracy of the air pressure propulsion control of the positive and negative pressure air pumps.

[0029] More specifically, the analysis matrix error is used to locate and assign bone binding weights and perform weighted transformation rendering to generate an active weighted vertex grid. The active weighted vertex grid is used to solve the independent angle changes of the virtual end effector to obtain the joint flexion and extension control parameters of the hand rehabilitation device to execute the stroke drive compensation. Specifically, the steps include: If the matrix error is less than the preset matrix error, the movable point is calibrated as a low-amplitude response active point; if the matrix error is greater than the preset matrix error, the movable point is calibrated as a high-amplitude response active point; Based on the matrix error, the skeleton binding weights of different low-amplitude response active points and high-amplitude response active points are assigned. The active points are rendered with weighted transformation using the final posture rotation matrix according to the skeleton binding weights to generate an active weighted vertex grid. During the weighted transformation process, a synchronous linear approximation is used to establish the Jacobian matrix between the joint angle changes and the virtual end-effector posture changes. Based on the Jacobian matrix, a damped least squares solution is performed on the joint angles of the active trade-off vertex grid to obtain an independent angle change chain for each virtual end-effector. The joint angle corresponding to each finger joint on the current hand rehabilitation device is continuously debugged and transformed according to the independent angle change chain until the matrix errors of the low-amplitude response active point and the high-amplitude response active point are eliminated, and the joint flexion and extension control parameters of the hand rehabilitation device for executing the stroke drive compensation are obtained.

[0030] It should be noted that this method uses matrix error to analyze the posture rotation scheduling of different joint nodes of the current hand rehabilitation device. If the matrix error is less than the preset matrix error, it means that the necessary scheduling ratio of the joint participating in the posture rotation at this node is low, and it is a joint point that does not require frequent rotation activities. The movable point is calibrated as a low-amplitude response active point; otherwise, it means that the necessary scheduling ratio of the joint participating in the posture rotation is high. If its rotation definition and support are missing, it may lead to the inability to complete the posture requirements of the selected working mode, so it is calibrated as a high-amplitude response active point; and the matrix error responds to the low-amplitude response active point and the high-amplitude response active point. The positive and negative pressure air pumps at the active point are affected by the coordinated scheduling of the joint skeleton posture rotation corresponding to the virtual end effector. Therefore, this method assigns a bone binding weight to the active point on the hand rehabilitation device and the joint skeleton corresponding to the virtual end effector according to the matrix error. The weight allows the hand rehabilitation device to respond to the bone transformation, and finally renders a vertex grid that controls the stroke drive compensation of the current hand rehabilitation device, namely the active weight vertex grid. This realizes the visualization of inverse kinematic deduction, improves the expressiveness, restoration and realism of joint flexion and extension rotation, and makes the calculation of joint flexion and extension control parameters more accurate. In addition, this method establishes the Jacobian matrix between the joint angle change and the virtual end effector posture change through linear approximation. This Jacobian matrix provides detailed information on the trend of the movable joint's change toward the target direction, thereby accurately capturing the flexibility of bone movement during simulation and improving the accuracy of the linear change of the independent angle opening of different virtual end effectors as time changes.

[0031] More specifically, the step S108 includes the following steps: Obtaining the preset training control strategy for the hand rehabilitation device, and extracting the various rotation change vectors planned by the hand rehabilitation device for different joint flexion and extension processes, as well as the air pressure control levels output by the positive and negative air pressure pumps to achieve the decision-making output of each rotation change vector through the preset training control strategy; The simulation model of the hand rehabilitation device is used to import the preset training control strategy for simulation. At this time, the rotational posture scalar field of the hand rehabilitation device during the process of executing the stroke drive compensation from the real-time joint angle simulation to achieve the selected working mode is exported through SolidWorks model simulation software. The quaternion mapping field is constructed based on the structural layout of the rotational posture scalar field. Each rotation change vector is defined as a prefix label in the quaternion mapping domain, and the corresponding air pressure control magnitude of each rotation change vector is set as a suffix label to form a prefix-suffix mapping bundled control query item, and each rotation change vector is preset to be assigned a quaternion string; Performing rotational differential conversion according to the joint flexion and extension angle, angular velocity, angle switching step, and joint flexion and extension increment recorded in the joint flexion and extension control parameters to obtain the necessary rotation change vector for the hand rehabilitation device to execute the stroke drive compensation amount; An exponential mapping formula is introduced to convert the necessary rotation change vector into an incremental quaternion to obtain a necessary rotation incremental quaternion, and the necessary rotation incremental quaternion is projected into a quaternion mapping field, so that the necessary rotation incremental quaternion is traversed one by one through the quaternion string corresponding to each control query item for approximate comparison; If the approximate agreement between the necessary rotation increment quaternion and the quaternion string is greater than a preset approximate agreement, a cumulative superposition cap mark is added to the quaternion string; otherwise, the quaternion string is ignored, and a series of candidate superposition quaternions are obtained; Quaternion multiplication is used to rotate and accumulate the air pressure control magnitudes of a series of candidate superposition quaternions corresponding to the control query items to update the posture quaternion of the hand rehabilitation device. Finally, the air pressure control accumulation parameters are output, and intelligent training control of the hand rehabilitation device is performed based on the air pressure control accumulation parameters.

[0032] It should be noted that the device's preset control strategy typically determines an air pressure control level based on the glove joint's rotation vector to provide appropriate driving force to control the selected training mode. However, traditional control methods rely solely on fixed matching control variables based on manual control experience, making it difficult to determine the appropriate air pressure control level for more refined rotation vectors. This can easily lead to large air pressure control errors in the positive and negative pressure pumps, resulting in unreasonable training ranges and failure to meet the standards of the selected working mode, greatly increasing the risk of training accidents. To address this, this method constructs a quaternion mapping domain based on the structural layout of the rotational posture scalar field derived during hand rehabilitation device simulation. This quaternion mapping domain provides a reliable index query platform for subsequent joint flexion and extension control parameters. This quaternion mapping domain records control query items that correspond to each other in the preset training control strategy. An example control query item is: prefix x, y, z (rotation change vector) - suffix 1.5 MPa (air pressure control level) [0.133, 0.87, 1.62, 0.5225 (quaternion string)]. Then, the necessary rotation change vector after the conversion of the joint flexion and extension control parameters is converted into an incremental quaternion form to obtain the necessary rotation incremental quaternion, and it is mapped to the constructed quaternion mapping field to perform an ergodic approximate query comparison on each control query item. If the approximate consistency between the necessary rotation incremental quaternion and the quaternion string is greater than the preset approximate consistency, it means that the rotation change vector corresponding to the prefix of the quaternion string recorded in the control query item and the necessary rotation incremental quaternion match with a high degree of precision. Therefore, the air pressure control magnitude of the corresponding suffix can be used for the air pressure propulsion control of the joint flexion and extension control parameters. Therefore, this method uses a cumulative superposition cap mark to mark the control query item, which can improve the index accuracy of the quaternion rotation accumulation, avoid the phenomenon of cumulative omissions, errors or missing, and ensure the reliability and stability of the air pressure control of the positive and negative pressure air pumps. Finally, the marked candidate superposition quaternions are accumulated to form the control quantity that meets the joint flexion and extension control parameters, which conforms to the changing characteristics of joint flexion and extension as time progresses. It can significantly reduce the risk and probability of training accidents caused by unreasonable air pressure control and improve the coordination and stability of transition control of hand rehabilitation equipment.

[0033] The third aspect of the present invention provides a hand rehabilitation system based on multi-source data analysis, such as Figure 4 As shown, the hand rehabilitation system includes a memory 41 and a processor 42. The memory 41 stores a hand rehabilitation method program based on multi-source data analysis. When the hand rehabilitation method program is executed by the processor 42, any one of the hand rehabilitation method steps is implemented.

[0034] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A hand rehabilitation device based on multi-source data analysis, characterized in that: The hand rehabilitation device comprises: Wearing gloves: The wearing gloves are used to completely cover the finger joints and palm joints of the target user; Positive and negative pressure air pump module: The positive and negative pressure air pump module is used to control the generation of positive and negative pressure gas, provide air pressure power for auxiliary flexion and extension training activities of finger joints, and adjust the training intensity and strength; Solenoid valve module: The solenoid valve module is connected to the positive and negative pressure air pump modules and is responsible for controlling the on / off and flow of compressed air to simulate the natural grasping action of the hand; Temperature control module: The temperature control module is used to adjust the hand temperature according to the training status of the target user, providing a heating effect for hand rehabilitation training; Human-computer interaction interface: The human-computer interaction interface is used to provide the target user with the ability to switch working modes and personalize the time, strength, and frequency of hand training; Memory storage module: The memory storage module is responsible for memorizing and saving the last hand training program, so that the effect can be directly applied next time without repeated debugging.

2. A control method for a hand rehabilitation device based on multi-source data analysis, applied to the hand rehabilitation device based on multi-source data analysis according to claim 1, characterized in that: The steps include: S102: Acquire multi-source hand sensing data from a hand rehabilitation device and a human body intended motion system, perform polar coordinate radial fusion of the multi-source hand sensing data according to a logical and standardized trend in accordance with the human body intended motion system, and obtain an actual intended motion polar coordinate layout of the multi-source hand sensing data; S104: Analyzing the polar coordinate layout of the actual intended motion based on the standardized offset of the ideal working condition parameter set of the human body intended motion when the hand rehabilitation device executes the selected working mode, thereby performing multi-dimensional hysteresis compensation to obtain the stroke drive compensation amount required for the current intended motion of the hand rehabilitation device to reach the selected point working mode; S106: Constructing a chain joint motion structure for the target user, and simultaneously injecting a virtual end effector into the hand rehabilitation device model. Using an inverse kinematics solution algorithm, the independent angle changes of each virtual end effector's posture rotation when the hand rehabilitation device applies the stroke drive compensation in accordance with the chain joint motion structure are solved to obtain joint flexion and extension control parameters for the hand rehabilitation device to execute the stroke drive compensation. S108: Based on the rotational posture scalar field derived from the simulation process and the preset training control strategy of the hand rehabilitation device, a quaternion mapping field is constructed, the joint flexion and extension control parameters are mapped to the quaternion mapping field and the air pressure control accumulation calculation is performed, so as to intelligently train and control the hand rehabilitation device.

3. The control method of hand rehabilitation equipment based on multi-source data analysis according to claim 1, characterized in that: The S102 specifically includes the following steps: Acquire a multi-source sensor array of the hand rehabilitation device. When the target user wears the hand rehabilitation device, the multi-source sensor array collects data of the target user within a preset time period to obtain multi-source hand sensing data; Centrally preprocess each group of multi-source hand sensing data, introduce a covariance algorithm to calculate the autocorrelation dependency of each group of multi-source hand sensing data after centralized preprocessing, and output the coupling variance between each data variable in each group of multi-source hand sensing data; Obtain anthropometric knowledge graphs through big data, use the anthropometric knowledge graphs to identify multi-source hand sensing data, output a human body intention movement system that conforms to the target user's stage characteristics, and extract different human body intention movement indicators and logical normative trend indexes for each human body intention movement indicator through the human body intention movement system; According to different human intention motion indices, the principal component source coordinate angles are allocated in the form of Cartesian coordinates to form a polar coordinate framework for human intention motion. Maintaining the coupling variance as the autonomous projection criterion, a radial joint projection vector constraint is established according to the logical norm trend index. Based on the radial joint projection vector constraint, each group of multi-source hand sensing data is radially decomposed one by one according to the autonomous projection criterion in the covariance algorithm, and a cross calculation of the collaborative correlation is performed. At this time, the radial covariance scale and cross-projection covariance matrix of each group of multi-source hand sensing data are output; If the weighted sum of the cross-projection covariance matrices exceeds the maximum weighted sum threshold, each set of multi-source hand sensing data is projected onto the fitting coordinate radius on the radial covariance scale of the subordinate response within the human body intended motion polar coordinate architecture to generate the actual intended motion polar coordinate layout of the multi-source hand sensing data.

4. The control method of hand rehabilitation equipment based on multi-source data analysis according to claim 1, characterized in that: The S104 specifically includes the following steps: Obtain the target user's selected working mode for the hand rehabilitation device and the conceptual design drawings of the hand rehabilitation device, and obtain the ideal working condition parameter set of the human body's intended movement when the hand rehabilitation device executes the selected working mode through the conceptual design drawings; Based on the ideal working condition parameter set, a multi-source hand sensing data is constructed to tend towards the ideal intended motion polar coordinate layout under the selected working mode. A hash misalignment algorithm is introduced to calculate the hash misalignment area function of the non-overlapping area between the actual intended motion polar coordinate layout and the ideal intended motion polar coordinate layout. Based on the hash misalignment area function, the steady-state misalignment degree of the current hand rehabilitation device that enables the target user to operate in the selected working mode is determined. If the steady-state misalignment is greater than the preset steady-state misalignment, an intended motion state space model of the current hand rehabilitation device is constructed based on the coupling correlation of multi-source hand sensing data on different human intended motion indicators, and a decoupling analysis between different input channels and output channels of the intended motion state space model is performed by calculating zero points and poles to determine the open-loop operation transfer function of the current hand rehabilitation device; Obtain the desired diagonal control system for the selected operating mode, determine the compensation cutoff frequency on the Bode plot based on the phase margin of the desired diagonal control system, and establish a gain compensation matrix for different multi-source hand sensing data based on the steady-state misalignment. The open-loop gain of the intended motion state space model is globally adjusted in a steady-state manner through the gain compensation matrix, so that the center frequency of the hysteresis network of the open-loop operation transfer function continuously approaches the compensation cutoff frequency, generating a multi-dimensional hysteresis compensation structure. The multi-dimensional hysteresis compensation structure is deployed in the actual diagonal control system of the current hand rehabilitation device to obtain the stroke drive compensation amount required for the target user to achieve the point selection working mode when using the current hand rehabilitation device.

5. The control method of hand rehabilitation equipment based on multi-source data analysis according to claim 1, characterized in that: The S106 specifically includes the following steps: The target user's hand features are copied using a multi-sensor array to obtain the target user's actual joint linear contours. The joint linear contours with the greatest similarity to the actual joint linear contours are extracted from the hand feature database preset by the hand rehabilitation device and defined as the joint reference blueprint. Obtaining standard product drawings of a hand rehabilitation device, reconstructing a simulation model of the hand rehabilitation device according to the standard product drawings using SolidWorks model simulation software, and injecting a virtual end effector into each hand joint in the simulation model; Based on the ideal working condition parameter set, the virtual end effector presets the final posture rotation matrix of the selected working mode. According to the joint reference blueprint, the joint degrees of freedom, bone connection relationship and starting motion angle that match the target user's hand line are defined from the root to the virtual end effector, and the chain joint motion structure is output; Obtain personalized setting parameters for the target user using the selected working mode, synchronously obtain the movable points of the current hand rehabilitation device, and the real-time joint angles of the target user maintaining the previous default posture state when capturing multi-source hand sensing data; An inverse kinematics solution algorithm is introduced. Starting from the real-time joint angle as the established origin, the stroke drive compensation is applied in the inverse kinematics solution algorithm in accordance with the motion premise constraints of the chain joint motion structure and personalized setting parameters. The actual position of each virtual end effector is deduced from this, and the end posture rotation matrix is ​​obtained. The deviation between the end posture rotation matrix and the end posture rotation matrix is ​​calculated to obtain the matrix error of each movable point. Matrix error is analyzed to locate and assign bone binding weights and perform weighted transformation rendering to generate an active weighted vertex grid. The active weighted vertex grid is used to solve the independent angle changes of the virtual end effector, and the joint flexion and extension control parameters of the hand rehabilitation device's execution stroke drive compensation are obtained.

6. The control method of hand rehabilitation equipment based on multi-source data analysis according to claim 5, characterized in that: The analysis matrix error positioning allocation of bone binding weights and weighted transformation rendering are performed to generate an active weighted vertex grid, and the active weighted vertex grid is used to solve the independent angle change of the virtual end effector to obtain the joint flexion and extension control parameters of the hand rehabilitation device execution stroke drive compensation amount, which specifically includes the following steps: If the matrix error is less than the preset matrix error, the movable point is calibrated as a low-amplitude response active point; if the matrix error is greater than the preset matrix error, the movable point is calibrated as a high-amplitude response active point; Based on the matrix error, the skeleton binding weights of different low-amplitude response active points and high-amplitude response active points are assigned. The active points are rendered with weighted transformation using the final posture rotation matrix according to the skeleton binding weights to generate an active weighted vertex grid. During the weighted transformation process, a synchronous linear approximation is used to establish the Jacobian matrix between the joint angle changes and the virtual end-effector posture changes. Based on the Jacobian matrix, a damped least squares solution is performed on the joint angles of the active trade-off vertex grid to obtain an independent angle change chain for each virtual end-effector. The joint angle corresponding to each finger joint on the current hand rehabilitation device is continuously debugged and transformed according to the independent angle change chain until the matrix errors of the low-amplitude response active point and the high-amplitude response active point are eliminated, and the joint flexion and extension control parameters of the hand rehabilitation device for executing the stroke drive compensation are obtained.

7. The control method of hand rehabilitation equipment based on multi-source data analysis according to claim 1, characterized in that: The S108 specifically includes the following steps: Obtaining the preset training control strategy for the hand rehabilitation device, and extracting the various rotation change vectors planned by the hand rehabilitation device for different joint flexion and extension processes, as well as the air pressure control levels output by the positive and negative air pressure pumps to achieve the decision-making output of each rotation change vector through the preset training control strategy; The simulation model of the hand rehabilitation device is used to import the preset training control strategy for simulation. At this time, the rotational posture scalar field of the hand rehabilitation device during the process of executing the stroke drive compensation from the real-time joint angle simulation to achieve the selected working mode is exported through SolidWorks model simulation software. The quaternion mapping field is constructed based on the structural layout of the rotational posture scalar field. Each rotation change vector is defined as a prefix label in the quaternion mapping domain, and the corresponding air pressure control magnitude of each rotation change vector is set as a suffix label to form a prefix-suffix mapping bundled control query item, and each rotation change vector is preset to be assigned a quaternion string; Performing rotational differential conversion according to the joint flexion and extension angle, angular velocity, angle switching step, and joint flexion and extension increment recorded in the joint flexion and extension control parameters to obtain the necessary rotation change vector for the hand rehabilitation device to execute the stroke drive compensation amount; An exponential mapping formula is introduced to convert the necessary rotation change vector into an incremental quaternion to obtain a necessary rotation incremental quaternion, and the necessary rotation incremental quaternion is projected into a quaternion mapping field, so that the necessary rotation incremental quaternion is traversed one by one through the quaternion string corresponding to each control query item for approximate comparison; If the approximate agreement between the necessary rotation increment quaternion and the quaternion string is greater than a preset approximate agreement, a cumulative superposition cap mark is added to the quaternion string; otherwise, the quaternion string is ignored, and a series of candidate superposition quaternions are obtained; Quaternion multiplication is used to rotate and accumulate the air pressure control magnitudes of a series of candidate superposition quaternions corresponding to the control query items to update the posture quaternion of the hand rehabilitation device. Finally, the air pressure control accumulation parameters are output, and intelligent training control of the hand rehabilitation device is performed based on the air pressure control accumulation parameters.

8. A hand rehabilitation system based on multi-source data analysis, characterized in that: The hand rehabilitation system includes a memory and a processor. The memory stores a hand rehabilitation method program based on multi-source data analysis. When the hand rehabilitation method program is executed by the processor, the hand rehabilitation method steps described in any one of claims 1 to 7 are implemented.

Images