Upper limb joint rehabilitation training method and system based on exoskeleton mechanical arm

Abstract

The invention relates to the field of upper limb joint rehabilitation training, in particular to an upper limb joint rehabilitation training method and system based on an exoskeleton mechanical arm, and the method comprises the steps: collecting sEMG signals of a target muscle group of a user in real time in a training process; calculating the root mean square of the amplitude of the sEMG signal in the current time window; if the root mean square is smaller than a muscle force insufficient threshold value, a training mode is switched to a full-assistance mode; if the root mean square is not smaller than the muscle force insufficient threshold value and smaller than the normal muscle force threshold value, the training mode is switched to a semi-assistance mode; if the root mean square is larger than or equal to the normal muscle force threshold value, the training mode is switched to an anti-resistance training mode. The rehabilitation training effect of the upper limb joints can be effectively improved.

Description

Technical Field

[0001] This invention relates to the field of upper limb joint rehabilitation training. More specifically, this invention relates to an upper limb joint rehabilitation training method and system based on an exoskeleton robotic arm. Background Technology

[0002] Exoskeleton rehabilitation robots are important assistive devices in modern physical therapy. They assist patients with neurological disorders such as stroke and brain injury by providing precise and repeatable motion assistance or resistance, aiming to restore damaged upper limb motor function and muscle strength. Exoskeleton rehabilitation robots are commonly used for rehabilitation training of patients with upper limb joint problems.

[0003] However, traditional rehabilitation training methods often employ preset, fixed training patterns or require therapists to manually adjust parameters. This "one-size-fits-all" approach fails to respond in real-time to changes in the patient's actual muscle strength and fatigue levels during training. This results in training intensity that is either too low for effective results or too high, posing a risk of secondary injury, making it difficult to achieve personalized and optimized rehabilitation outcomes. To address this issue, some existing technologies are beginning to incorporate surface electromyography (sEMG) signals to assess muscle activation levels and use this information for human-computer interaction control. These methods typically employ fixed threshold control strategies; for example, when the sEMG signal exceeds a preset single threshold, the system is triggered to switch from assisted mode to resistance mode.

[0004] However, this strategy has significant drawbacks: the threshold it uses is fixed and non-personalized. It neither considers the vast individual differences in muscle strength among patients nor adapts to changes in muscle strength due to fatigue during a single training session or the long-term rehabilitation process. This simple control logic based on a fixed threshold leads to a continuous deviation between the robotic arm's output and the patient's real-time needs. This not only fails to ensure the effectiveness of training but also poses safety risks when the patient's muscle strength fluctuates. This is a key technical problem that urgently needs to be addressed by existing technologies.

[0005] In summary, existing upper limb joint rehabilitation training methods suffer from poor training effectiveness and safety issues. Summary of the Invention

[0006] To address the technical problems of poor training effect and poor safety in existing upper limb joint rehabilitation training methods, this invention provides solutions in the following aspects.

[0007] In a first aspect, the present invention provides an upper limb joint rehabilitation training method based on an exoskeleton robotic arm, comprising:

[0008] During training, sEMG signals of the user's target muscle groups are collected in real time.

[0009] Calculate the root mean square of the amplitude of the sEMG signal within the current time window;

[0010] If the root mean square (RMS) is less than the muscle strength deficiency threshold, the training mode is switched to full-assisted mode; if the RMS is not less than the muscle strength deficiency threshold but less than the normal muscle strength threshold, the training mode is switched to half-assisted mode; if the RMS is greater than or equal to the normal muscle strength threshold, the training mode is switched to resistance training mode. In full-assisted mode, the exoskeleton arm assists the user's upper limb joint movement; in half-assisted mode, the exoskeleton arm assists and the user's own upper limb strength jointly assist the user's upper limb joint movement; in resistance training mode, the user's own upper limb strength overcomes the resistance of the exoskeleton arm to assist the user's own upper limb joint movement. The assistance of the exoskeleton arm in full-assisted mode is twice that in half-assisted mode.

[0011] Muscle weakness threshold T weak The calculation expression is:

[0012] T weak =R rest +0.2×(R MVC -R rest );

[0013] Normal muscle strength threshold T normal The calculation expression is:

[0014] T normal =R rest +0.5×R MVC ;

[0015] In the above two equations, R rest R represents the root mean square (RMS) amplitude of the sEMG signal in a relaxed state. The RMS amplitude of the sEMG signal in a relaxed state refers to the root mean square (RMS) amplitude of the sEMG signal of the target muscle group acquired when the user is in a relaxed state. MVC The root mean square (RMS) of the sEMG signal amplitude during isometric maximal contraction refers to the root mean square of the sEMG signal amplitude of the target muscle group collected during the user's full contraction of the target muscle group.

[0016] Preferably, it further includes: filtering the acquired sEMG signal of the target muscle group, specifically including:

[0017] Bandpass filtering and notch filtering were applied to the sEMG signals of the target muscle group;

[0018] The sEMG signals of the target muscle group are subjected to full-wave rectification to convert the bidirectional signals into unipolar signals.

[0019] The sEMG signals of the target muscle group were smoothed.

[0020] Preferably, the method for obtaining the root mean square (RMS) of the sEMG signal amplitude during isometric maximal contraction includes: continuously testing the RMS of the sEMG signal amplitude corresponding to multiple instances of a user's full-force contraction of the target muscle group, and averaging the RMS of the multiple obtained sEMG signal amplitudes to obtain the RMS of the sEMG signal amplitude during isometric maximal contraction; for a given test, obtaining the corresponding RMS of the sEMG signal amplitude includes:

[0021] The first sEMG signal of the target muscle group is collected in real time during the user's full contraction of the target muscle group, and the full contraction of the target muscle group lasts for 3-5 seconds.

[0022] The second sEMG signal is obtained by removing the sEMG signals within 500ms of the start time of contraction and the sEMG signals within 500ms of the end time of contraction from the first sEMG signal.

[0023] Calculate the root mean square (R) of the amplitude of the sEMG signal of the target muscle group acquired during the user's full contraction of the target muscle group. MVC The calculation expression is:

[0024]

[0025] In the formula, N represents the number of peaks corresponding to the second sEMG signal, and x i This represents the amplitude corresponding to the i-th peak.

[0026] Preferably, it further includes:

[0027] In resistance training mode, the actual value of the user's current joint rotation angular velocity is collected in real time. Combined with the actual value of the maximum joint rotation angle from the previous training session, the resistance of the exoskeleton robotic arm is adjusted. The formula for calculating the adjusted resistance is:

[0028]

[0029] In the formula, τ represents the adjusted resistance, and K p K represents the stiffness coefficient. d Represents the damping coefficient, q and q represents the actual value and the expected value of the maximum joint rotation angle from the previous training session, respectively; d and G(q) represents the actual value of the user's current joint rotational angular velocity and the expected value of the joint rotational angular velocity, respectively; G(q) represents the sum of the gravitational torque of the exoskeleton robotic arm and the gravitational torque of the user's limb.

[0030] Preferably, the gravitational torque G of the exoskeleton robotic armrobat The expression for calculating (q) is:

[0031]

[0032] In the formula, m i Let g represent the mass of the i-th link, and g represent the acceleration due to gravity. Let Jacobian matrix represent the linear velocity of the i-th link of the exoskeleton robotic arm. represents the direction vector of gravity in polar coordinates, and n represents the total number of links in the exoskeleton robotic arm.

[0033] Preferably, the gravitational torque G of the user's limb human (q) The calculation expression is:

[0034]

[0035] In the formula, m j Let l represent the mass of the j-th limb segment. j The distance from the user's joint to the center of mass is represented by g, and the acceleration due to gravity is q. j The value represents the joint angle, and k represents the total number of limb segments.

[0036] Preferably, it further includes: dynamically adjusting the stiffness coefficient and damping coefficient based on the acquired sEMG signal, including:

[0037] Calculate the normalized value of the amplitude of the acquired sEMG signal;

[0038] The adjusted stiffness coefficient and adjusted damping coefficient are calculated based on the normalized value of the sEMG signal amplitude. The expression for calculating the adjusted stiffness coefficient is as follows:

[0039] K p,1 =K p,0 ×(1+α×S N );

[0040] The formula for calculating the adjusted damping coefficient is as follows:

[0041] K d,1 =K d,0 ×(1+β×S N );

[0042] In the above two equations, K p,1 and K p,0 These represent the adjusted stiffness coefficient and the original stiffness coefficient, respectively; K d,1 and K d,0 These represent the adjusted damping coefficient and the original damping coefficient, respectively; S N This represents the normalized value of the sEMG signal amplitude, where α and β are both adjustment coefficients.

[0043] Preferably, it also includes a training protection mechanism, which is as follows: during the training process, the instantaneous torque output by the robotic arm to the joint is detected in real time. If the amplitude of the torque signal is greater than the first torque threshold, the drive motor of the exoskeleton robotic arm is controlled to stop rotating and an alarm is sent to the medical staff.

[0044] Preferably, it also includes a training protection mechanism, which is to detect the instantaneous torque output by the robotic arm to the joint in real time during the training process. If the instantaneous torque is greater than the second torque threshold, the instantaneous torque output by the robotic arm to the joint is reduced to 5 N·m.

[0045] In a second aspect, the present invention provides an upper limb joint rehabilitation training system based on an exoskeleton robotic arm, comprising: an exoskeleton robotic arm, an electromyography (EMG) device, and a control unit. The control unit is connected to the EMG device for data acquisition and to the exoskeleton robotic arm for control, so as to realize the upper limb joint rehabilitation training method based on the exoskeleton robotic arm of the present invention.

[0046] The beneficial effects of this invention are as follows: Compared with existing technologies that use fixed thresholds or require manual adjustment of training modes by therapists, the method disclosed in this independent claim achieves adaptive adjustment. It abandons the "one-size-fits-all" fixed threshold and instead uses real-time calibration of the user's own resting muscle strength (R). rest ) and maximum muscle strength (R MVC This system dynamically generates personalized training mode switching thresholds for each user. This ensures that training mode switching perfectly matches the user's current actual muscle strength level, resolving the issue of mismatched training intensity due to individual differences or fatigue. This makes rehabilitation training truly personalized, intelligent, and precise, thereby greatly improving the effectiveness of rehabilitation training and patient engagement while ensuring safety.

[0047] Furthermore, the method in this embodiment scientifically divides the user's muscle strength status into three stages—"severely insufficient muscle strength," "muscle strength developing," and "relatively good muscle strength"—using 20% ​​and 50% of the user's range of motion as dividing points. These stages correspond to the three most suitable training modes: full-assisted, half-assisted, and resistance training, respectively. This calculation method based on a specific ratio provides a reproducible and standardized technical basis for generating personalized thresholds, making the system's decision-making logic clearer and more reliable, and avoiding the mode switching disorder that may result from fuzzy calculations. Attached Figure Description

[0048] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein:

[0049] Figure 1 This is a schematic flowchart illustrating an upper limb joint rehabilitation training method based on an exoskeleton robotic arm according to an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram illustrating the structure of an upper limb joint rehabilitation training system based on an exoskeleton robotic arm according to an embodiment of the present invention;

[0051] Figure 3 This is a schematic diagram illustrating the structure of a control unit according to an embodiment of the present invention. Detailed Implementation

[0052] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0053] Example of an upper limb joint rehabilitation training method based on exoskeleton robotic arms:

[0054] like Figure 1 As shown, the upper limb joint rehabilitation training method based on an exoskeleton robotic arm of the present invention includes:

[0055] S1. Collect sEMG signals of the user's target muscle groups in real time during training;

[0056] The sEMG sampling rate is ≥1000Hz. In this embodiment, the biceps brachii can be used as the target muscle group. When acquiring sEMG signals, the motor attachment position follows the SENIAM standard, for example, it can be attached to the belly of the biceps brachii.

[0057] Electrode attachment locations follow SENIAM standards (e.g., the belly of the biceps brachii muscle).

[0058] S2. Calculate the root mean square of the amplitude of the sEMG signal within the current time window;

[0059] S3. The training mode is switched based on the root mean square (RMS) of the sEMG signal amplitude. Specifically: if the RMS is less than the muscle weakness threshold, the training mode is switched to full-assisted mode; if the RMS is not less than the muscle weakness threshold but less than the normal muscle strength threshold, the training mode is switched to half-assisted mode; if the RMS is greater than or equal to the normal muscle strength threshold, the training mode is switched to resistance training mode. In full-assisted mode, the exoskeleton arm assists the user's upper limb joint movement; in half-assisted mode, the exoskeleton arm assists and the user's own upper limb strength jointly assist the user's upper limb joint movement; in resistance training mode, the user's own upper limb strength overcomes the resistance of the exoskeleton arm to assist the user's own upper limb joint movement. The assistance of the exoskeleton arm in full-assisted mode is twice that in half-assisted mode.

[0060] Before training begins, the user wears an exoskeleton robotic arm, with sEMG sensors attached to the outer surface of the target muscle group. Before training, the user's muscle strength can be assessed, and an initial training mode can be set based on the assessment results. If the user's muscle strength level is 0-2, the full-assisted mode is selected; if the user's muscle strength level is 3, the half-assisted mode is selected; and if the user's muscle strength level is 4-5, the resistance training mode is selected. In full-assisted mode, the robotic arm guides the user to complete a preset trajectory (such as a "cup to mouth" motion simulating eating). In half-assisted mode, the system provides assistance when the patient actively exerts force. In resistance training mode, the robotic arm applies adjustable resistance (e.g., 5-20 N·m) to enhance muscle endurance.

[0061] Doctors can monitor joint torque and muscle activation curves in real time through a control console, and can pause training in an emergency if necessary.

[0062] Training data is automatically uploaded to the hospital information system (HIS) to generate rehabilitation progress reports.

[0063] In this embodiment, the exoskeleton robotic arm can be the force feedback exoskeleton robotic arm based on active and passive joints as described in Chinese invention patent application No. CN116619336A, or the upper limb rehabilitation exoskeleton with shoulder-humer linkage mechanism as described in Chinese invention patent application No. CN119950255A.

[0064] The resistance of the exoskeleton robotic arm is provided by the drive motor of the exoskeleton robotic arm. By adjusting the torque of the drive motor, the resistance of the exoskeleton robotic arm can be adjusted. By adjusting the rotation direction of the drive motor, the switching between assist mode and resistance mode can be realized.

[0065] In this embodiment, the muscle weakness threshold T weak The calculation expression is:

[0066] T weak =R rest +0.2×(R MVC -R rest );

[0067] Normal muscle strength threshold T normal The calculation expression is:

[0068] T normal =R rest +0.5×R MVC ;

[0069] In the above two equations, R restR represents the root mean square (RMS) amplitude of the sEMG signal in a relaxed state. The RMS amplitude of the sEMG signal in a relaxed state refers to the root mean square (RMS) amplitude of the sEMG signal of the target muscle group acquired when the user is in a relaxed state. MVC The root mean square (RMS) of the sEMG signal amplitude during isometric maximal contraction refers to the root mean square of the sEMG signal amplitude of the target muscle group collected during the user's full contraction of the target muscle group.

[0070] Compared to existing technologies that use fixed thresholds or require therapists to manually adjust training modes, the method disclosed in this independent claim achieves adaptive adjustment. It abandons the "one-size-fits-all" fixed threshold and dynamically generates personalized training mode switching thresholds for each user by real-time calibration of their resting muscle strength (Rrest) and maximum muscle strength (Rmvc). This ensures that training mode switching perfectly matches the user's current actual muscle strength level, resolving the problem of mismatched training intensity due to individual differences or fatigue. This enables truly personalized, intelligent, and precise rehabilitation training, thereby significantly improving the effectiveness of rehabilitation training and patient engagement while ensuring safety.

[0071] Furthermore, the method in this embodiment scientifically divides the user's muscle strength status into three stages—"severely insufficient muscle strength," "muscle strength developing," and "relatively good muscle strength"—using 20% ​​and 50% of the user's range of motion as dividing points. These stages correspond to the three most suitable training modes: full-assisted, half-assisted, and resistance training, respectively. This calculation method based on a specific ratio provides a reproducible and standardized technical basis for generating personalized thresholds, making the system's decision-making logic clearer and more reliable, and avoiding the mode switching disorder that may result from fuzzy calculations.

[0072] In one embodiment, when the user is in a relaxed state, the acquisition duration of the sEMG signal of the target muscle group is 3-5 seconds.

[0073] In one embodiment, the method for obtaining the root mean square (RMS) amplitude of the sEMG signal during isometric maximal contraction includes: continuously testing the RMS amplitude of the sEMG signal corresponding to multiple instances of a user's full-force contraction of the target muscle group, and averaging the RMS amplitudes of the multiple obtained sEMG signal amplitudes to obtain the RMS amplitude of the sEMG signal during isometric maximal contraction; for a given test, obtaining the RMS amplitude of the corresponding sEMG signal includes:

[0074] S101. The first sEMG signal of the target muscle group is collected in real time during the process of the user fully contracting the target muscle group, and the process of fully contracting the target muscle group lasts for 3-5 seconds.

[0075] S102. Remove the sEMG signals within 500ms from the start time of contraction in the first sEMG signal and the sEMG signals within 500ms from the end time of contraction in the first sEMG signal to obtain the second sEMG signal.

[0076] S103. Calculate the root mean square R of the amplitude of the sEMG signal of the target muscle group acquired during the user's full contraction of the target muscle group. MVC The calculation expression is:

[0077]

[0078] In the formula, N represents the number of peaks corresponding to the second sEMG signal, and x i This represents the amplitude corresponding to the i-th peak.

[0079] Unlike existing technologies that simply average the electromyographic signals throughout the entire contraction process, this embodiment proposes R... MVC The acquisition method is more precise. By eliminating unstable signals at the beginning and end of the contraction phase (which often do not represent true maximum muscle strength), this method effectively eliminates jitter and noise interference during muscle exertion / dissipation, focusing on the smooth contraction phase that best reflects the user's maximum exertion ability. This ensures that R, as a personalized benchmark, is accurate. MVC This ensures the accuracy and stability of the values, thereby further guaranteeing the training effect for users.

[0080] In one embodiment, the method further includes: filtering the acquired sEMG signal of the target muscle group, specifically including:

[0081] S201. Perform bandpass filtering and notch filtering on the sEMG signal of the target muscle group;

[0082] The upper cutoff frequency of the bandpass filter is set to 500 Hz, and the lower cutoff frequency is set to 20 Hz. The notch region of the notch filter is 50-60 Hz.

[0083] S202. Perform full-wave rectification on the sEMG signal of the target muscle group to convert the bidirectional signal into a unipolar signal.

[0084] S203. Smooth the sEMG signal of the target muscle group.

[0085] The smoothing process involves using a moving average or low-pass filter (cutoff frequency 2-5Hz) to obtain a linear envelope.

[0086] In one embodiment, the method further includes: reacquiring the root mean square amplitude of the sEMG signal of the isometric maximum contraction every first preset time interval, and recalculating the corresponding muscle weakness threshold and normal muscle strength threshold.

[0087] Patient fatigue or prolonged training can lead to a decline in muscle strength. By recalculating R_MVC and updating the insufficiency and normal muscle strength thresholds, users can continue to complete training.

[0088] Compared to methods that only perform calibration once before training begins, this embodiment introduces a periodic update mechanism. This allows the system to adapt not only to individual differences among users, but also to changes in the same user's abilities during long-term rehabilitation (such as increased muscle strength) or changes in their state during a single training session (such as decreased muscle strength due to muscle fatigue). By periodically "refreshing" the user's maximum muscle strength baseline, the training program ensures that it dynamically follows the user's rehabilitation curve, always remaining within the most effective training intensity range, achieving full adaptive tracking throughout the entire rehabilitation cycle.

[0089] In one embodiment, it also includes:

[0090] In resistance training mode, the actual value of the user's current joint rotation angular velocity is collected in real time. Combined with the actual value of the maximum joint rotation angle from the previous training session, the resistance of the exoskeleton robotic arm is adjusted. The formula for calculating the adjusted resistance is:

[0091]

[0092] In the formula, τ represents the adjusted resistance, and K p K represents the stiffness coefficient. d Represents the damping coefficient, q and q represents the actual value and the expected value of the maximum joint rotation angle from the previous training session, respectively; d and G(q) represents the actual value of the user's current joint rotational angular velocity and the expected value of the joint rotational angular velocity, respectively; G(q) represents the sum of the gravitational torque of the exoskeleton robotic arm and the gravitational torque of the user's limb.

[0093] Once in resistance mode, this method does not apply constant resistance, but dynamically adjusts the resistance based on the user's real-time electromyographic signal intensity, thus achieving refined and real-time interaction within a single mode.

[0094] In one embodiment, the gravitational torque G of the exoskeleton robotic arm robot The expression for calculating (q) is:

[0095]

[0096] In the formula, m i Let g represent the mass of the i-th link, and g represent the acceleration due to gravity. Let Jacobian matrix represent the linear velocity of the i-th link of the exoskeleton robotic arm. represents the direction vector of gravity in polar coordinates, and n represents the total number of links in the exoskeleton robotic arm.

[0097] The mass and center of mass of the robotic arm can be measured using CAD software (such as SolidWorks) or by actual weighing. The Jacobian matrix is ​​calculated in real time based on the robotic arm's DH parameters.

[0098] The linear velocity of the i-th link is used to characterize the joint rotational speed.

[0099] In one embodiment, the gravitational torque G of the user's limb human (q) The calculation expression is:

[0100]

[0101] In the formula, m j Let l represent the mass of the j-th limb segment. j The distance from the user's joint to the center of mass is represented by g, and the acceleration due to gravity is q. j The value represents the joint angle, and k represents the total number of limb segments.

[0102] When obtaining the mass of the j-th limb segment, an anthropometry database can be consulted. Specifically, the upper arm accounts for 3.3% of the total weight, and its center of mass is located at 43% of its total length; the forearm accounts for 2.2% of the total weight, and its center of mass is located at 43% of its total length; the hand accounts for 0.6% of the total weight, and its center of mass is located at 50% of its total length. The user's total weight is collected using a weighing scale, and the mass of each segment can then be calculated.

[0103] In one embodiment, the method further includes: dynamically adjusting the stiffness coefficient and damping coefficient based on the acquired sEMG signal, including:

[0104] S301. Calculate the normalized value of the amplitude of the acquired sEMG signal;

[0105] In this embodiment, the normalized value of the sEMG signal amplitude is equal to the percentage of the sEMG signal amplitude relative to the root mean square of the sEMG signal amplitude of the maximally contracted sEMG signal.

[0106] S302. Calculate the adjusted stiffness coefficient and adjusted damping coefficient based on the normalized value of the sEMG signal amplitude. The expression for calculating the adjusted stiffness coefficient is as follows:

[0107] K p,1 =K p,0 ×(1+α×S N );

[0108] The formula for calculating the adjusted damping coefficient is as follows:

[0109] Kd,1 =K d,0 ×(1+β×S N );

[0110] In the above two equations, K p,1 and K p,0 These represent the adjusted stiffness coefficient and the original stiffness coefficient, respectively; K d,1 and K d,0 These represent the adjusted damping coefficient and the original damping coefficient, respectively; S N This represents the normalized value of the sEMG signal amplitude, where α and β are both adjustment coefficients.

[0111] In this embodiment, both α and β are 0.5. In other embodiments, other suitable values ​​may also be used.

[0112] Unlike traditional resistance modes that only provide fixed resistance or simple proportional resistance, this invention achieves dynamic shaping of the "texture" of resistance by simultaneously adjusting two key parameters in the control system: stiffness (Kp) and damping (Kd). When the user applies greater force, the system not only provides a larger positional deviation penalty (increased stiffness) but also a greater velocity resistance (increased damping), simulating a more realistic and challenging resistance experience. This method of online adjustment of PID controller parameters based on real-time muscle force makes human-computer interaction smoother and more intelligent, effectively motivating users to actively apply force and providing immediate and nuanced mechanical feedback on the magnitude of force applied.

[0113] In one embodiment, the system also includes developing VR scenes using Unity3D. The tasks within these VR scenes may include space grabbing and kitchen simulation. In the space grabbing task, the patient abducts their shoulder joint 90° to touch a virtual planet. In the kitchen simulation task, the patient "picks up" virtual tableware by flexing their elbow joint 90°.

[0114] The data mapping method is as follows: the joint rotation angle is mapped to the virtual hand position, and the torque value is mapped to the gripping force (such as indicating excessive force when crushing a cup).

[0115] By setting up VR scene training tasks, the fun of training can be greatly enhanced.

[0116] In one embodiment, the method further includes: determining that the user has insufficient muscle strength in response to the sEMG signal amplitude being lower than a second muscle weakness threshold within a series of time windows.

[0117] The second muscle weakness threshold is equal to the root mean square of the sEMG signal amplitude in the relaxed state plus 20 percent of the root mean square of the sEMG signal amplitude during isometric maximal contraction.

[0118] When the sEMG signal amplitude falls below the second muscle weakness threshold for multiple consecutive time windows, it indicates that the user has experienced severe muscle fatigue due to prolonged training. Training can be stopped or the training mode can be switched to full-assisted mode.

[0119] In one embodiment, the method further includes: providing a red emergency stop switch at the waist of the exoskeleton robotic arm, the red emergency stop switch being used to cut off the power supply to the drive motor of the exoskeleton robotic arm.

[0120] In one embodiment, a training protection mechanism is also included, which is as follows: during the training process, the instantaneous torque output by the robotic arm to the joint is detected in real time. If the amplitude of the torque signal is greater than the first torque threshold, the drive motor of the exoskeleton robotic arm is controlled to stop rotating and an alarm is sent to the medical staff.

[0121] In this embodiment, the first torque threshold is set to 30 N·m.

[0122] Torque sensors can be used to collect the instantaneous torque output by the robotic arm to the joint. When installing the torque sensor, it is essential to ensure that its measuring axis is coaxial with the rotation axis of the human joint. This ensures that the sensor measures the torque around the joint, and not additional bending or shearing forces. For example, in elbow joint training: the sensor can be installed between the elbow joint rotation axis and the connecting rod of the exoskeleton robotic arm to measure the torque exerted by the robotic arm on the patient's elbow. This provides real-time feedback on the resistance output by the robotic arm to the patient's joint.

[0123] In another embodiment, the training protection mechanism is as follows: during the training process, the instantaneous torque output by the robotic arm to the joint is detected in real time. If the instantaneous torque is greater than the second torque threshold, the instantaneous torque output by the robotic arm to the joint is reduced to 5 N·m.

[0124] The second torque threshold is 15 N·m.

[0125] In this embodiment, the instantaneous torque output by the robotic arm to the joint is equal to the instantaneous value of the resistance torque applied by the robotic arm to the patient.

[0126] Once a dangerous torque exceeding the threshold is detected, the system will immediately and proactively unload the torque to a safe level, rather than simply shutting down. This not only provides users with "hard protection" to avoid joint and muscle damage, but also maintains the continuity of training by preserving a baseline torque, achieving rapid, intelligent, and non-disruptive safety intervention in extreme situations.

[0127] When the force sensor of the rehabilitation training robotic arm detects an instantaneous torque > 15 Nm, the system automatically reduces the resistance torque to 5 N·m. This design is mainly based on the following risk considerations:

[0128] I. Physiological risks: joint and muscle damage to patients.

[0129] (1) Joint overload.

[0130] Safe range of torque for upper limb joints:

[0131] Shoulder joint: The torque during daily activities is usually <10Nm, while a momentary torque >15Nm may cause labral tears or rotator cuff injuries.

[0132] Elbow joint: Flexion / extension torque >15Nm is prone to causing ligament sprains (such as ulnar collateral ligament).

[0133] Wrist joint: >8Nm may cause carpal tunnel syndrome or triangular fibrocartilage complex (TFCC) injury.

[0134] For patients with spasticity, a sudden increase in muscle tone can cause a rapid increase in torque, which can easily lead to joint dislocation.

[0135] (2) Muscle strain.

[0136] Muscle limit load: The maximum isometric contraction torque of the biceps brachii is about 20-30 Nm, but the muscle strength of patients in the recovery period may only be 10%-30% of that of healthy people. Instantaneous resistance >15 Nm may cause muscle fiber tearing (especially during eccentric contraction, the risk is higher).

[0137] II. Mechanical risks: Equipment damage.

[0138] Motor overload:

[0139] The rated torque of the servo motors commonly used in rehabilitation robotic arms is usually 10-20 Nm. Continuous overload will cause overheating and shorten the lifespan.

[0140] Harmonic reducers may experience increased backlash or tooth surface wear when the torque is greater than 15 Nm.

[0141] Sensor failure:

[0142] The range of a six-axis force sensor is typically ±20Nm. Frequent impacts >15Nm will reduce calibration accuracy.

[0143] III. Risks to training effectiveness.

[0144] Compensatory actions:

[0145] In an attempt to overcome excessive resistance, patients may attempt to complete the training through incorrect movements such as shrugging their shoulders or swinging their trunks, which may actually worsen the motor dysfunction.

[0146] Pain suppression:

[0147] When excessive training triggers pain, patients' active participation decreases, and the rehabilitation effect is compromised.

[0148] The reason for reducing the drag torque to 5 N·m in this embodiment is as follows:

[0149] (1) Clinical evidence:

[0150] Daily activity requirements:

[0151] Lifting a 500ml water cup (approximately 5N force, lever arm 0.1m) requires approximately 0.5Nm, and 5Nm can cover most ADL (Activities of Daily Living) training needs.

[0152] (2) Safety Redundancy

[0153] Below the injury threshold: 5 Nm is approximately 20%-30% of the limit load of the joint ligament (e.g., the rupture moment of the ulnar collateral ligament of the elbow joint is approximately 25 Nm).

[0154] Preserve training stimulus: It can still provide basic resistance to promote muscle recovery (reducing it completely to 0 Nm will render the training ineffective).

[0155] The method used in this embodiment can effectively improve the security of the user training process and the effectiveness of the training.

[0156] Example of an upper limb joint rehabilitation training system based on exoskeleton robotic arms:

[0157] like Figure 2 As shown, the present invention also provides an upper limb joint rehabilitation training system based on an exoskeleton robotic arm, including: an exoskeleton robotic arm, an electromyography (EMG) device, and a control unit. The control unit is connected to the EMG device for data acquisition and to the exoskeleton robotic arm for control, so as to realize the upper limb joint rehabilitation training method based on the exoskeleton robotic arm in the above embodiments.

[0158] like Figure 3 As shown, the control unit system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement an upper limb joint rehabilitation training method based on an exoskeleton robotic arm according to the first aspect of the present invention.

[0159] The control unit also includes other components well known to those skilled in the art, such as communication buses and communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0160] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A method for upper limb joint rehabilitation training based on an exoskeleton robotic arm, characterized in that, include: During training, sEMG signals of the user's target muscle groups are collected in real time. Calculate the root mean square of the amplitude of the sEMG signal within the current time window; If the root mean square is less than the muscle weakness threshold, the training mode will be switched to full-assist mode. If the root mean square is not less than the muscle weakness threshold and less than the normal muscle strength threshold, then the training mode will be switched to the semi-assisted mode. If the root mean square is greater than or equal to the normal muscle strength threshold, the training mode will be switched to resistance training mode. In full-assisted mode, the exoskeleton robotic arm assists in the movement of the user's upper limb joints. In half-assisted mode, the exoskeleton robotic arm assists in the movement of the user's upper limb joints together with the user's own upper limb strength. In resistance training mode, the user's own upper limb strength overcomes the resistance of the exoskeleton robotic arm to drive the movement of the user's own upper limb joints. The assistance provided by the exoskeleton robotic arm in full-assisted mode is twice that in half-assisted mode. Muscle weakness threshold T weak The calculation expression is: T weak =R rest +0.2×(R MVC -R rest ); Normal muscle strength threshold T normal The calculation expression is: T normal =R rest +0.5×R MVC ; In the above two equations, R rest R represents the root mean square (RMS) amplitude of the sEMG signal in a relaxed state. The RMS amplitude of the sEMG signal in a relaxed state refers to the root mean square (RMS) amplitude of the sEMG signal of the target muscle group acquired when the user is in a relaxed state. MVC The root mean square (RMS) of the sEMG signal amplitude during isometric maximal contraction refers to the root mean square of the sEMG signal amplitude of the target muscle group collected during the user's full contraction of the target muscle group.

2. The upper limb joint rehabilitation training method based on an exoskeleton robotic arm as described in claim 1, characterized in that, It also includes: filtering the sEMG signals of the acquired target muscle groups, specifically including: Bandpass filtering and notch filtering were applied to the sEMG signals of the target muscle group; The sEMG signals of the target muscle group are subjected to full-wave rectification to convert the bidirectional signals into unipolar signals. The sEMG signals of the target muscle group were smoothed.

3. The upper limb joint rehabilitation training method based on an exoskeleton robotic arm as described in claim 1, characterized in that, The method for obtaining the root mean square (RMS) amplitude of the sEMG signal during isometric maximal contraction includes: continuously testing the RMS amplitude of the sEMG signal corresponding to multiple instances of the user's full-force contraction of the target muscle group, and averaging the RMS amplitudes of the multiple obtained sEMG signal values ​​to obtain the RMS amplitude of the sEMG signal during isometric maximal contraction; for a given test, obtaining the corresponding RMS amplitude of the sEMG signal includes: The first sEMG signal of the target muscle group is collected in real time during the user's full contraction of the target muscle group, and the full contraction of the target muscle group lasts for 3-5 seconds. The second sEMG signal is obtained by removing the sEMG signals within 500ms of the start time of contraction and the sEMG signals within 500ms of the end time of contraction from the first sEMG signal. Calculate the root mean square (R) of the amplitude of the sEMG signal of the target muscle group acquired during the user's full contraction of the target muscle group. MVC The calculation expression is: In the formula, N represents the number of peaks corresponding to the second sEMG signal, and x i This represents the amplitude corresponding to the i-th peak.

4. The upper limb joint rehabilitation training method based on an exoskeleton robotic arm as described in claim 1, characterized in that, Also includes: In resistance training mode, the actual value of the user's current joint rotation angular velocity is collected in real time. Combined with the actual value of the maximum joint rotation angle from the previous training session, the resistance of the exoskeleton robotic arm is adjusted. The formula for calculating the adjusted resistance is: In the formula, τ represents the adjusted resistance, and K p K represents the stiffness coefficient. d Represents the damping coefficient, q and q represents the actual value and the expected value of the maximum joint rotation angle from the previous training session, respectively; d and G(q) represents the actual value of the user's current joint rotational angular velocity and the expected value of the joint rotational angular velocity, respectively; G(q) represents the sum of the gravitational torque of the exoskeleton robotic arm and the gravitational torque of the user's limb.

5. The upper limb joint rehabilitation training method based on an exoskeleton robotic arm as described in claim 4, characterized in that, The gravitational torque G of the exoskeleton robotic arm robot The expression for calculating (q) is: In the formula, m i Let g represent the mass of the i-th link, and g represent the acceleration due to gravity. Let Jacobian matrix represent the linear velocity of the i-th link of the exoskeleton robotic arm. represents the direction vector of gravity in polar coordinates, and n represents the total number of links in the exoskeleton robotic arm.

6. The upper limb joint rehabilitation training method based on an exoskeleton robotic arm as described in claim 4, characterized in that, The gravitational torque G of the user's limbs human (q) The calculation expression is: In the formula, m j Let l represent the mass of the j-th limb segment. j The distance from the user's joint to the center of mass is represented by g, and the acceleration due to gravity is q. j The angle represents the joint angle, and k represents the total number of limb segments.

7. The upper limb joint rehabilitation training method based on an exoskeleton robotic arm as described in claim 1, characterized in that, Also includes: The stiffness and damping coefficients are dynamically adjusted based on the acquired sEMG signals, including: Calculate the normalized value of the amplitude of the acquired sEMG signal; The adjusted stiffness coefficient and adjusted damping coefficient are calculated based on the normalized value of the sEMG signal amplitude. The expression for calculating the adjusted stiffness coefficient is as follows: K p,1 =K p,0 ×(1+α×S N ); The formula for calculating the adjusted damping coefficient is as follows: K d,1 =K d,0 ×(1+β×S N ); In the above two equations, K p,1 and K p,0 These represent the adjusted stiffness coefficient and the original stiffness coefficient, respectively; K d,1 and K d,0 These represent the adjusted damping coefficient and the original damping coefficient, respectively; S N This represents the normalized value of the sEMG signal amplitude, where α and β are both adjustment coefficients.

8. The upper limb joint rehabilitation training method based on an exoskeleton robotic arm as described in any one of claims 1 to 7, characterized in that, It also includes a training protection mechanism, which involves real-time detection of the instantaneous torque output from the robotic arm to the joints during training. If the amplitude of the torque signal is greater than the first torque threshold, the drive motor of the exoskeleton robotic arm is controlled to stop rotating, and an alarm is sent to the medical staff.

9. The upper limb joint rehabilitation training method based on an exoskeleton robotic arm as described in any one of claims 1 to 7, characterized in that, It also includes a training protection mechanism, which is to detect the instantaneous torque output by the robotic arm to the joint in real time during the training process. If the instantaneous torque is greater than the second torque threshold, the instantaneous torque output by the robotic arm to the joint will be reduced to 5 N·m.

10. An upper limb joint rehabilitation training system based on an exoskeleton robotic arm, characterized in that, include: The exoskeleton robotic arm, electromyography (EMG) device, and control unit are provided. The control unit collects data and connects to the EMG device, and controls the exoskeleton robotic arm to implement the upper limb joint rehabilitation training method based on the exoskeleton robotic arm as described in any one of claims 1 to 9.

Images