Wrist joint rehabilitation method based on multi-stage on-demand assistance grading and related device

Abstract

The embodiment of the application relates to the technical fields of data processing and rehabilitation medical instrument, and provides a wrist joint rehabilitation method based on multi-stage on-demand assistance grading and related devices, the method comprising: acquiring compensation interaction torque, real-time angle and surface electromyogram signal; based on the compensation interaction torque and the surface electromyogram signal, extracting characteristic values through a sliding window method to obtain an average absolute value set and a root mean square value set; performing normalization processing on the average absolute value set and the root mean square value set to obtain a normalized characteristic value set; performing weight dynamic distribution based on the real-time angle to obtain muscle weight coefficients; performing scoring calculation according to the normalized characteristic value set and the muscle weight coefficients to obtain a comprehensive score set; and performing assistance interval division based on the comprehensive score set to obtain multi-stage assistance intervals, which can accurately divide the multi-stage assistance intervals, realize personalized and dynamic assistance grading of wrist joint rehabilitation training, and are beneficial to improving the pertinence and effectiveness of wrist joint rehabilitation training.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical fields of data processing and rehabilitation medical devices, in particular to a wrist joint rehabilitation method based on multi-level on-demand assistance grading and related devices. BACKGROUND

[0002] The wrist joint is an important joint of the human body and plays a crucial role in daily life and work. However, due to various diseases, trauma or aging factors, wrist joint dysfunction is relatively common in clinical practice, which seriously affects the quality of life of patients. Stroke is one of the main causes of disability in adults worldwide, resulting in a large number of patients with upper limb motor dysfunction, especially wrist joint injury. The wrist joint is a key part connecting the hand and forearm, bearing the functions of support, stability and movement. Once damaged, the movement of the hand and forearm will be significantly limited, therefore, wrist joint dysfunction rehabilitation is crucial. Traditional wrist joint rehabilitation training relies on manual assessment, which cannot accurately quantify the real-time movement state and muscle activation level of patients, making it difficult to dynamically adjust according to individual differences and training stages, seriously affecting the effectiveness and pertinence of rehabilitation training. Therefore, how to improve the effectiveness of wrist joint rehabilitation training has become a problem to be solved. SUMMARY

[0003] The embodiments of the present application provide a wrist joint rehabilitation method based on multi-level on-demand assistance grading and related devices, which can accurately divide multi-level assistance intervals based on objective quantitative indicators, realize personalized and dynamic assistance grading for wrist joint rehabilitation training, and is beneficial to improving the effectiveness and pertinence of wrist joint rehabilitation training.

[0004] The first aspect of the embodiments of the present application provides a wrist joint rehabilitation method based on multi-level on-demand assistance grading, the method comprising:

[0005] obtaining a compensation interaction torque, a real-time angle and a surface electromyography signal;

[0006] based on the compensation interaction torque and the surface electromyography signal, extracting feature values through a sliding window method to obtain a mean absolute value set and a root mean square value set;

[0007] normalizing the mean absolute value set and the root mean square value set to obtain a normalized feature value set;

[0008] dynamically distributing weights based on the real-time angle to obtain a muscle weight coefficient;

[0009] performing scoring calculation based on the normalized feature value set and the muscle weight coefficient to obtain a comprehensive score set;

[0010] dividing an assistance interval based on the comprehensive score set to obtain a multi-level assistance interval.

[0011] In this example, by acquiring the compensation interaction torque, the real-time angle and the surface electromyogram signal, the average absolute value set and the root mean square value set can be obtained based on the compensation interaction torque and the surface electromyogram signal by the sliding window method, and the normalized feature value set can be obtained by normalizing the average absolute value set and the root mean square value set, so that the muscle weight coefficient can be obtained based on the real-time angle by dynamic weight distribution, and the comprehensive score set can be obtained by further performing score calculation based on the normalized feature value set and the muscle weight coefficient, and then the multi-level assistance interval can be obtained by dividing the assistance interval based on the comprehensive score set, so that the wrist joint rehabilitation based on the multi-level on-demand assistance grading can be provided, and the effectiveness of the rehabilitation training and the initiative participation of the patient can be improved.

[0012] The second aspect of the embodiment of the present application provides a wrist joint rehabilitation device based on multi-level on-demand assistance grading, which comprises:

[0013] The acquisition module is configured to acquire the compensation interaction torque, the real-time angle and the surface electromyogram signal.

[0014] The first processing module is configured to obtain the average absolute value set and the root mean square value set by extracting the feature value based on the compensation interaction torque and the surface electromyogram signal by the sliding window method.

[0015] The second processing module is configured to obtain the normalized feature value set by normalizing the average absolute value set and the root mean square value set.

[0016] The third processing module is configured to obtain the muscle weight coefficient by dynamically distributing the weight based on the real-time angle.

[0017] The fourth processing module is configured to obtain the comprehensive score set by performing score calculation based on the normalized feature value set and the muscle weight coefficient.

[0018] The fifth processing module is configured to obtain the multi-level assistance interval by dividing the assistance interval based on the comprehensive score set.

[0019] The third aspect of the embodiment of the present application provides a terminal, which comprises a processor, an input device, an output device and a memory, and the processor, the input device, the output device and the memory are connected with each other, wherein the memory is configured to store a computer program, the computer program comprises program instructions, and the processor is configured to invoke the program instructions to execute the steps as in the first aspect of the embodiment of the present application.

[0020] A fourth aspect of the embodiments of the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program for electronic data exchange, and the computer program causes a computer to perform some or all of the steps described in the first aspect of the embodiments of the present application.

[0021] A fifth aspect of the embodiments of the present application provides a computer program product, which includes a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform some or all of the steps described in the first aspect of the embodiments of the present application. The computer program product can be a software installation package. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of these drawings.

[0023] Figure 1A A structural design schematic diagram of a wrist joint rehabilitation device provided by the embodiments of the present application;

[0024] Figure 1B A schematic diagram of support adjustment provided by the embodiments of the present application;

[0025] Figure 2 A flowchart of a wrist joint rehabilitation method based on multi-stage on-demand assistance grading provided by the embodiments of the present application;

[0026] Figure 3A A schematic diagram of wrist joint dorsiflexion movement provided by the embodiments of the present application;

[0027] Figure 3B A schematic diagram of a compensation processing process for wrist joint rehabilitation training provided by the embodiments of the present application;

[0028] Figure 3C A wrist joint dorsiflexion movement total resistance torque trend chart provided by the embodiments of the present application;

[0029] Figure 4 A dorsiflexion movement schematic diagram of a wrist joint rehabilitation device provided by the embodiments of the present application;

[0030] Figure 5 A dynamic muscle weight change curve chart provided by the embodiments of the present application;

[0031] Figure 6is a trend function diagram of a feature value intensity changing with an angle provided by an embodiment of the present application;

[0032] Figure 7 is a myoelectric signal energy distribution and intention recognition threshold diagram provided by an embodiment of the present application;

[0033] Figure 8 is a on-demand assistance interval division diagram provided by an embodiment of the present application;

[0034] Figure 9 is a multi-level on-demand assistance grading flowchart provided by an embodiment of the present application;

[0035] Figure 10 is a structure diagram of a terminal provided by an embodiment of the present application;

[0036] Figure 11 is a structure diagram of a wrist joint rehabilitation device based on multi-level on-demand assistance grading provided by an embodiment of the present application. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0038] The terms "first", "second", and the like in the specification of the present application, claims and above-described drawings are used to distinguish different objects, rather than to describe a particular order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed or can optionally include other steps or units inherent to the process, method, product or device.

[0039] In the present application, "embodiment" means that the specific features, structures or characteristics described in conjunction with the embodiment can be included in at least one embodiment of the present application. The phrase appears at various places in the specification does not necessarily refer to the same embodiment, nor is it independent or alternative to other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described in the present application can be combined with other embodiments.

[0040] In order to better understand the wrist rehabilitation method based on multi-stage on-demand assistive grading provided by the embodiments of the present application, the prior art will be briefly introduced first. At present, the traditional wrist rehabilitation treatment mainly relies on physical therapists to perform manual assisted training. This way not only has high labor cost, but also the treatment effect is difficult to quantify and evaluate. With the development of robot-assisted rehabilitation technology, various wrist rehabilitation devices have been gradually applied in clinical practice. However, the existing rehabilitation devices mostly use preset programs for training, and cannot accurately assist control according to the real-time state of the patient, which has the following problems:

[0041] 1) The active movement intention of the patient cannot be accurately identified, resulting in insufficient active participation of the patient during the training process;

[0042] 2) The assistance strength is difficult to accurately adjust according to the real-time muscle activity state of the patient, resulting in excessive or insufficient assistance force;

[0043] 3) There is a lack of comprehensive evaluation of the muscle synergy of the patient, and it is difficult to provide differentiated assistance according to the functional characteristics of different muscles;

[0044] 4) The existing devices mostly use a single physical parameter (such as position, speed, torque, etc.) as the control basis, ignoring the fusion advantages of bioelectric signals and mechanical signals.

[0045] Based on the rehabilitation training strategy of the patient, the wrist rehabilitation robot can be divided into three training modes: passive training mode, assistive training mode and resistance training mode. The passive training mode assists the generation of wrist movement through an actuator, which is usually used for patients with limited wrist movement, and most devices have this mode; the assistive / resistance training mode determines the assistive or resistive force provided by the actuator by detecting the interaction force information between the patient and the rehabilitation robot to indirectly reflect the movement performance of the patient. Medical research has long proven that the active participation of patients can promote neural plasticity and motor recovery during treatment, but most patients require training strategies between passive and resistance modes, i.e. in a situation where they want to move but cannot move, therefore, the assistive training mode is more important than the resistance training mode.

[0046] Multi-source information fusion technology is widely used in rehabilitation robot system for intention recognition and control strategy optimization. For example, three-dimensional motion monitoring is achieved through motion capture and inertial sensors, motion intention judgment and speed estimation are achieved through force / laser sensors, and walking state analysis is achieved through force-angle sensors. Surface electromyography (sEMG) is a bioelectric signal generated autonomously by the human body, which can directly reflect the neural-muscular function state and is an important part of multi-source information fusion, which helps rehabilitation robot system to realize intention recognition and control decision. The use of sEMG signals to predict joint angles, torques, and motion recognition has mature applications, which proves that there is considerable potential in adaptive adjustment and classification in on-demand assistance strategies. Therefore, developing a wrist rehabilitation method that can provide on-demand assistance according to the muscle state and motion intention of the patient is of great significance to improve the effectiveness of rehabilitation training and enhance the patient's active participation.

[0047] As shown in Figure 1A The wrist rehabilitation device (WReD) provided by the present application includes a support unit 10, a driving unit 20, a forearm support 30 and a handle support 40. The support unit 10 serves as the basic structure and is composed of a bottom plate 11 and three vertical support rods (12 / 13 / 14), two of which (12 / 13) carry the driving unit and a mechanical limiting device with physical blocking function, and the third support rod 14 maintains the geometric stability of the overall frame through a virtual constraint mechanism. The driving unit 20 uses a reduction gear box 21 to cooperate with a rubber coupling 22 to ensure high torque output and concentricity compensation, and integrates a static torque sensor 23 and an angle sensor 24 to monitor the interaction force and angle change in real time. The forearm support 30 realizes precise alignment of the wrist rotation axis and the driving shaft through a vertical slider 31, and the horizontal slider 32 adaptively compensates for the eccentricity between the flexion / extension and radial / ulnar axes based on the dovetail groove structure 33 pre-tightened by the spring. As shown in Figure 1B , Figure 1B A schematic diagram of the support adjustment is exemplarily shown. Specifically, the forearm can be fixed on the support of the wrist rehabilitation device, and the support can be manually moved up and down and forward and backward to align the wrist joint motion rotation center with the device rotation center (motion axis). The handle support 40 realizes two-degree-of-freedom switching training through a rotatable inner and outer ring structure 41, and cooperates with a magic tape to quickly fix the patient's hand, forming a dynamic cooperative training system.

[0048] The present application is suitable for wrist rehabilitation training of various neuromuscular disorder patients, such as stroke, peripheral nerve injury, postoperative recovery in orthopedics, etc. According to the rehabilitation stage and muscle function state of different patients, the corresponding level of assistance force is provided, which can not only ensure the training effect, but also maximize the active participation of patients and promote the recovery of motor function.

[0049] The wrist joint rehabilitation method based on multi-level on-demand assistance grading provided in the present application is based on multi-source information fusion technology, and a comprehensive scoring system is constructed by collecting surface electromyography (sEMG), interaction torque and angle information to realize accurate evaluation of the motion state of a user and on-demand assistance control.

[0050] Referring to Figure 2 , Figure 2 A flowchart of a wrist joint rehabilitation method based on multi-level on-demand assistance grading is provided for the embodiments of the present application, and the method comprises:

[0051] S10: acquiring a compensation interaction torque, a real-time angle and a surface electromyography signal.

[0052] The compensation interaction torque can refer to the interaction torque value obtained after compensation processing of the acquired initial interaction torque. Specifically, the initial interaction torque can include comprehensive torques such as patient active force, wrist joint self-gravity, device friction, motor rotating torque, etc. By subtracting the total resistance torque through compensation processing, the interference of non-active force can be eliminated, so that the net torque reflecting the patient active force can be obtained.

[0053] In a possible implementation, before acquiring the compensation interaction torque, the acquired initial interaction torque can also be compensated first. That is, in order to ensure that the user's force size and muscle activation degree when expressing the wrist extension movement of the hand during the use of the device are monitored, the remaining forces caused by the wrist joint self-gravity, motor rotating torque and friction need to be compensated. That is, the wrist joint rehabilitation method based on multi-level on-demand assistance grading can also comprise the following steps:

[0054] S11: acquiring an initial interaction torque of a target user's wrist joint extension movement, a wrist joint movement resistance torque, a hand support self-gravity and a distance from a rotating shaft to the hand support;

[0055] S12: determining a hand support gravity torque according to the real-time angle, the hand support self-gravity and the distance from the rotating shaft to the hand support;

[0056] S13: calculating and obtaining a total resistance torque according to the hand support gravity torque and the wrist joint movement resistance torque;

[0057] S14: compensating the initial interaction torque based on the total resistance torque to obtain a compensation interaction torque.

[0058] The target user can refer to an object receiving wrist joint rehabilitation training, and can specifically refer to a patient who needs to be rehabilitated through a wrist joint rehabilitation method based on multi-level on-demand assistance grading, and has wrist joint dysfunction (such as limited wrist joint movement function caused by stroke, peripheral nerve injury, and postoperative recovery in orthopedics). Figure 3A Figure 3A A schematic diagram of wrist joint dorsiflexion movement is exemplarily shown.

[0059] The initial interaction torque can refer to the interaction torque between the patient and the wrist joint rehabilitation device collected by the device for the first time. The initial interaction torque can include various torques, such as the torque generated by the patient's active force, the torque caused by the gravity of the wrist joint, the friction torque of the device during operation, the torque of the motor rotation, and the like. It can be understood that the initial interaction torque can be raw data without processing, and irrelevant interference needs to be removed through subsequent compensation processing to reflect the patient's real active force.

[0060] The wrist joint movement resistance torque can refer to the total sum of the device's own resistance torque that needs to be overcome during the wrist joint movement. The resistance torque corresponding to the resistance type included in the wrist joint movement resistance torque can include the rotation torque of the device motor, the friction torque between the mechanical structures, the resistance torque generated by the component connection, and the like, which are not limited in the present application.

[0061] The hand support gravity can refer to the weight of the support on which the patient's hand is placed (fixed parameter, calibrated before the device is shipped). The hand support gravity can be one of the key parameters for calculating the hand support gravity torque, and its gravity will change with the wrist joint movement angle to affect the interaction torque, which needs to be compensated and eliminated.

[0062] The distance from the rotating shaft to the hand support can refer to the straight-line distance between the rotating shaft (the axis of the wrist joint movement) and the hand support in the device (fixed parameter, determined when the device is designed). Please refer to Figure 3B Figure 3B A wrist joint rehabilitation training force schematic diagram is exemplarily shown. As shown in Figure 3B The distance from the rotating shaft to the hand support can be understood as Figure 3B the distance from the support holding center G to the wrist joint rotation center 0, which can include the distance from the support holding center G to the hand centroid C and the distance from the hand centroid C to the wrist joint rotation center 0. The distance from the rotating shaft to the hand support can be used to calculate the torque generated by the hand support gravity (the farther the distance, the greater the torque under the same weight), which is a core parameter of the hand support gravity torque formula.

[0063] The hand support gravity torque can refer to the torque generated by the hand support gravity around the rotating shaft of the device. The hand support gravity torque can change with the real-time angle ​​Changes, for example, if the real-time angle is 0° (horizontal position), cos( )=1, at this time the hand support gravity torque is maximum; if the real-time angle increases, the hand support gravity torque will gradually decrease.

[0064] The real-time angle can refer to the angle of the wrist joint rotating in real time during the dorsiflexion movement. The real-time angle can be directly measured by a device sensor (such as an angle encoder). Optionally, the real-time angle can be used to dynamically allocate muscle weight coefficients (it can be understood that different angles, the contribution of the wrist joint dorsiflexion related muscles is different), and calculate the hand support gravity torque and the wrist joint movement to overcome the resistance torque, which is not limited in the present application.

[0065] The total resistance torque can refer to the sum of the interference torque that needs to be subtracted from the initial interaction torque, which is the patient's active force. By subtracting the total resistance torque, the device resistance, hand support gravity and other irrelevant factors can be eliminated to obtain the torque reflecting only the patient's active force.

[0066] As shown in Figure 3B , the present application provides a compensation processing method for wrist joint rehabilitation training. When the patient holds the device handle with his hand, drives the hand support 40 to make the wrist joint palmar flexion / dorsiflexion movement (as shown in Figure 3A ), an auxiliary torque is generated by controlling the driving unit 20 to compensate for the total resistance torque M D required to be overcome by the patient. The total resistance torque M D mainly includes two parts, the device gravity resistance torque M s generated by the hand support self-gravity G s , and the hand movement resistance torque M d . The hand movement resistance torque is composed of the patient's own hand gravity G h , and the viscous resistance torque M h generated by physiological factors such as postoperative joint activity friction and increased muscle viscosity. It can be understood that because the device user is mainly the postoperative patient with hand movement disorder, such patients generally have the above physiological resistance characteristics. The total resistance torque M D can comprehensively reflect the physical and physiological resistance that needs to be overcome by the patient when moving the hand.

[0067] Due to the force arm between the action point of the above resistance source and the center of rotation O of the wrist joint, it finally shows the resistance torque that hinders the movement. Therefore, the present application calculates the target compensation torque M D , and controls the driving device to apply an auxiliary torque equal in value and opposite in direction, so as to cancel the above resistance torque in effect, and then realizes the compensation processing, which is beneficial to reduce the force burden of the patient.

[0068] The target compensation torque M D is the device gravity resistance torque Ms The sum of the hand movement resistance moments (M d ) are different (as shown in Figure 3B ), and need to be calculated based on the moment = force x force arm respectively, and the calculation formula is:

[0069] M D =M s +M d

[0070] M s =G s ×(L s +L h )×cosθ

[0071] M d =G h ×L h ×cosθ+M h

[0072] Wherein, M D represents the total resistance moment value, which can realize the effect of gravity moment, movement resistance moment compensation, that is, the auxiliary moment provided by the driving unit; M s is the hand support gravity moment; M d is the wrist joint movement resistance moment, that is, the hand movement resistance moment described above; G s is the hand support gravity; L s is the distance from the center of the support holding to the center of mass of the hand C in the distance from the shaft to the hand support; L h is the distance from the center of mass of the hand C to the center of rotation of the wrist joint O in the distance from the shaft to the hand support; θ is the real-time angle, that is, the rotation angle of the wrist joint movement deviating from the horizontal position; cos is the operation of operating through the cosine function; G h is the gravity of the hand itself; M h is the viscous resistance moment.

[0073] The surface electromyography signal can refer to the bioelectric signal collected from the surface of the wrist joint extension movement related muscle, reflecting the muscle activation degree. Among them, the wrist joint extension movement related muscle can include Extensor Carpi Radialis Longus (ECRL), Extensor Carpi Radialis Brevis (ECRB), and Extensor Carpi Ulnaris (ECU). The surface electromyography signal can be used to extract characteristic values (such as average absolute value, root mean square value) in subsequent steps, and can be combined with the interaction moment to evaluate the patient's movement intention and effort level, which is not limited in the present application.

[0074] It should be noted that in the compensation processing stage, a gravity evaluation in the range of ±60° can be performed before the device is used. For example, the user can place his hand on the wrist joint rehabilitation device and keep it relaxed, and the device can drive the wrist joint to perform a periodic passive motion evaluation in the specified range, so as to obtain the overall motion resistance of the wrist joint at different angles. As shown in Figure 3C the stress of the wrist joint is sharply increased and the motion resistance is overcome, and at the same time, the hand motion resistance is in the same direction as the coulomb friction resistance, and the motion resistance tends to be stable; when the motion is near 60°, the hand changes from dorsiflexion to palmar flexion, the hand motion resistance is opposite to the device friction resistance, the hand support resistance is superimposed on the motion resistance, and the compensation value presents a reverse increasing trend compared with dorsiflexion; with the increase of the palmar flexion angle, the motion resistance compensation value tends to be stable.

[0075] As shown in Figure 4 in the process of training of the user through the wrist joint rehabilitation device, data can be collected in real time, such as the compensated compensation interaction torque of the wrist joint rehabilitation device, the real-time angle θ, and the sEMG of the three main control muscles of the wrist joint in dorsiflexion motion, which are the radial wrist long extensor (ECRL), the radial wrist short extensor (ECRB) and the ulnar wrist extensor (ECU). Optionally, the training process can be three times of training as a group, and once of training process can refer to the dorsiflexion motion trajectory from 0° to 60° and back to 0°, which is not limited in the present application. The compensation interaction torque and the angle data can be collected at a frequency of 200Hz, and the sEMG signal can be collected at a frequency of 1200Hz. Further, the collected data can be processed to provide assistance for subsequent assistance scoring. Specifically, in order to better combine the three groups of data together, the angle data and the compensation interaction torque can be up-sampled to make them consistent with the sampling frequency of the sEMG signal.

[0076] S20: Based on the compensation interaction torque and the surface electromyogram signal, a feature value is extracted by a sliding window method to obtain a mean absolute value set and a root mean square value set.

[0077] The sliding window method is a signal feature extraction method, which can divide the continuous signal into multiple overlapping or non-overlapping "windows" (such as fixed time length segments), and can calculate the feature value of the signal in each window. Through the sliding window method, the mean absolute value (Mean Absolute Value, MAV) and the root mean square value (Root Mean Square, RMS) can be extracted from the surface electromyogram signal and the compensation interaction torque to reflect the intensity and energy of the signal in the local time period. For example, the length of the sliding window can be set to 200ms (corresponding to 240 data points at a sampling frequency of 1200Hz), the sliding step is 50ms, and the window overlap ratio is 75%, to balance the real-time performance and the stability of the feature, which is not limited in the present application.

[0078] The average absolute value set can include one or more average absolute values, which can be one of the time-domain features included in the feature values. The average absolute value set can include an average absolute value subset corresponding to the surface electromyography signal, which can reflect the average level of muscle activity; and an average absolute value subset corresponding to the compensatory interaction torque, which can reflect the average size of the torque in the range of motion.

[0079] The root mean square value set can include one or more root mean square values, which can be one of the time-domain features included in the feature values. The root mean square value set can include a root mean square value subset corresponding to the surface electromyography signal, which can reflect the energy level of muscle activation; and a root mean square value subset corresponding to the compensatory interaction torque, which can reflect the energy level of the torque in the range of motion. Optionally, the root mean square value set can be used together with the average absolute value set as core features for evaluating the strength of the patient's force.

[0080] The present application adopts a MAV and RMS dual-parameter joint analysis strategy to simultaneously extract sEMG signals and compensatory interaction torques. Through the sliding window method, the dynamic changes in muscle activity level and activation degree can be accurately quantified, and the average torque amplitude and energy characteristic parameter changes of the active force in the range of motion can be simultaneously represented.

[0081] Optionally, taking the extraction of the average absolute value set and the root mean square value set based on the surface electromyography signal through the sliding window method as an example, the time-domain features MAV and RMS are extracted from the sEMG signal and the compensatory interaction torque through the sliding window method: ;

[0082] ; wherein MAV represents the average absolute value set, which can reflect the activity level of the muscle; N represents the number of data points; i R(i) represents the i-th data point corresponding to the surface electromyography signal; i represents the index of the data point; RMS represents the root mean square value set, which can reflect the activation degree of the muscle.

[0083] Optionally, the above formula can be used to extract the average absolute value set and the root mean square value set based on the compensatory interaction torque through the sliding window method. That is, the same processing method described above is applied to the compensatory interaction torque, that is, i R(i) represents the i-th data point corresponding to the compensatory interaction torque, to obtain the average absolute value set and the root mean square value set that can reflect the average size and energy level of the torque in the range of motion.

[0084] It should be noted that the sEMG signal and the energy change of the compensation interaction torque during the wrist joint movement both show a single-peak symmetric characteristic. In a possible implementation, Gaussian distribution fitting can be performed based on the average absolute value set and the root mean square value set to obtain a Gaussian distribution mean and a Gaussian distribution standard deviation.

[0085] Optionally, the Gaussian distribution fitting based on the average absolute value set can refer to the following formula: , wherein (x) represents a probability density function of the average absolute value set and the root mean square value set, which is a superposition function of multiple Gaussian distributions in this application; represents a random variable, for example, in sEMG signal analysis, x usually refers to the amplitude of the electromyographic signal (i.e., the specific numerical value of the signal); N represents the number of superimposed Gaussian distributions, which is a positive integer; i represents an index variable for summation, taking a value from 1 to N, for traversing each Gaussian distribution term to distinguish different Gaussian components; σ i represents the standard deviation of the i-th Gaussian distribution, which is used to describe the dispersion degree of the Gaussian component; μ i represents the mean (expectation) of the i-th Gaussian distribution, which is the center position of the Gaussian component; 0 represents the noise level of the sEMG signal, which is approximately a constant.

[0086] It should be noted that in the analysis process of the compensation interaction torque, x in the above formula can represent the specific numerical value of the compensation interaction torque, which is used to analyze the probability density characteristics of the compensation interaction torque through Gaussian distribution fitting, and to provide a basis for subsequent rehabilitation training data analysis (such as assistance interval division, force feature recognition, etc.). Optionally, the compensation interaction torque signal has a noise floor, which can be represented by 0, which is approximately a constant.

[0087] S30: Normalizing the average absolute value set and the root mean square value set to obtain a normalized feature value set.

[0088] The normalized feature value set can include one or more normalized feature values, which can refer to the feature values after standardizing MAV and RMS. It can be understood that normalizing can eliminate the influence of individual differences of different patients (such as natural differences in muscle strength and signal intensity).

[0089] In order to unify the influence of different user personal data, the statistical characteristics of personal data can be taken as a reference benchmark, such as extracting the 95% quantile of the MAV and RMS two feature values as the maximum value of normalization to avoid the interference of possible extreme values, and then extracting the 5% quantile as the minimum value of normalization to simulate the resting level in the non-movement state:

[0090] ;

[0091] ; wherein norm MAV represents a normalized mean absolute value set in the normalized feature value set; norm RMS represents a normalized root mean square value set in the normalized feature value set; MAV i and RMS i represent the current feature value, i.e., MAV i represents the i-th mean absolute value, RMS i represents the i-th root mean square value; MAV max represents the maximum mean absolute value, which is the 95th percentile of the mean absolute values extracted in the sliding window; MAV rest represents the minimum mean absolute value, which is the 5th percentile of the mean absolute values extracted in the sliding window; RMS max represents the maximum root mean square value, which is the 95th percentile of the root mean square values extracted in the sliding window; RMS rest represents the minimum root mean square value, which is the 5th percentile of the root mean square values extracted in the sliding window.

[0092] Based on the above normalized feature value set, the normalized feature matrix is defined as:

[0093] ;

[0094] ; wherein E norm is the sEMG signal normalized feature matrix; is a normalized mean absolute value subset of the extensor carpi radialis longus (ECRL); is a normalized root mean square value subset of the extensor carpi radialis longus (ECRL); is a normalized mean absolute value subset of the extensor carpi radialis brevis (ECRB); is a normalized root mean square value subset of the extensor carpi radialis brevis (ECRB); is a normalized mean absolute value subset of the extensor carpi ulnaris (ECU); is a normalized root mean square value subset of the extensor carpi ulnaris (ECU); M norm is the normalized feature matrix for compensating the interaction torque; is a normalized mean absolute value subset for compensating the interaction torque; is a normalized root mean square value subset for compensating the interaction torque.

[0095] S40: dynamically allocating weights based on the real-time angle to obtain muscle weight coefficients.

[0096] The muscle weight coefficient can refer to a coefficient dynamically allocated based on the real-time angle of the wrist joint. The muscle weight coefficient can reflect the contribution proportion of different muscles at the current angle. It can be understood that the sum of the muscle weight coefficients corresponding to the different muscles can be 1.

[0097] To achieve wrist extension movement, the muscles of extensor carpi radialis brevis (ECRB), extensor carpi radialis longus (ECRL) and extensor carpi ulnaris (ECU) will work together according to their respective attachment points and dynamic functions. During wrist joint extension, the weight ratio of the three muscles will change with the increase of the angle:

[0098] 0°-20°: ECRB > ECRL > ECU;

[0099] 20°-40°: ECRB ≈ ECRL > ECU;

[0100] 40°-60°: ECRB ≈ ECRL ≈ ECU;

[0101] As shown in Figure 5 , in the initial stage, ECRB can provide the maximum power output because its insertion point is closest to the center of the wrist joint, so its weight ratio is the highest; as the extension angle increases, the role of ECU gradually increases, when approaching the set maximum extension range, ECU provides additional support through the vector decomposition of the ulnar moment (the component force in the extension direction), its activation degree increases, and the weight ratio of ECRB gradually decreases; ECRL takes into account both power output and support, so there is little change.

[0102] As shown in Figure 6 , the change trend of the characteristic value at the angle can be divided into two stages: an increasing stage and a decreasing stage. The characteristic value intensity is described by a parameterized function:

[0103] When , ;

[0104] When , ;

[0105] wherein, and are the extension angle range, is the angle corresponding to the maximum activation of the sEMG signal.

[0106] Based on the characteristic value intensity , the weights of the three muscles are dynamically allocated:

[0107] When , ;

[0108] When , ;

[0109] wherein w I is the muscle weight matrix when the wrist joint is in a resting state , w T is the muscle weight matrix at the end of the segment , and K2 = [k4 k5 k6] T is the coefficient matrix for adjusting the weight ratio corresponding to the different stages of the three muscles, and needs to satisfy .

[0110] S50: Perform score calculation according to the normalized feature value set and the muscle weight coefficient to obtain a comprehensive score set.

[0111] The comprehensive score set can include one or more comprehensive scores (also referred to as total scores), which can refer to a comprehensive score index fused after the surface electromyogram score and the compensation interaction torque score, and can be used to divide the assistance interval in the subsequent steps. Specifically, the electromyogram score and the torque score can be obtained based on the muscle weight coefficient and the normalized feature value set respectively, and then integrated and obtained by a weighted exponential fusion function, which is not limited in the present application.

[0112] Based on the normalized feature matrix and the dynamic weight, the process of calculating the comprehensive score set can first calculate the score of the sEMG signal:

[0113] ;

[0114] ;

[0115] ;

[0116] wherein, represents the mean absolute value set (MAV) feature score of the surface electromyogram (sEMG); is the muscle weight matrix dynamically allocated based on the real-time angle of the wrist joint; E norm is the sEMG signal feature value normalization matrix; E norm (:,1) represents the first column of the sEMG signal normalization feature matrix, which contains the normalized MAV values of the three muscles; represents the root mean square value set (RMS) feature score of the surface electromyogram (sEMG); E norm (:,2) represents the second column of the sEMG signal normalization feature matrix, which contains the normalized RMS values of the three muscles; score sEMGa set of comprehensive scores representing the sEMG signals (fusion of MAV and RMS features); w M represents the weight of MAV in the comprehensive scores; w R represents the weight of RMS in the comprehensive scores. It should be noted that in the embodiments of the present application, the weights w M and w R may be equal, and in actual applications, the two weights can be adjusted according to the rehabilitation training effect of the target user, and the present application does not limit this.

[0117] Then, the score of the compensation interaction torque is calculated: wherein score M represents a set of comprehensive scores of the compensation interaction torque; w M represents the weight of MAV in the comprehensive scores; M norm is a normalized matrix of the feature values of the compensation interaction torque; M norm (1) is a set of normalized average absolute values of the compensation interaction torque, which is described by the parameter M norm (1) without limiting the present application; w R represents the weight of RMS in the comprehensive scores; M norm (2) is a set of normalized root mean square values of the compensation interaction torque, which is described by the parameter M norm (2) without limiting the present application. It should be noted that the MAV and the RMS herein are respectively the weights of MAV and RMS in the comprehensive scores, so as to ensure that the weights are equal to those in the sEMG comprehensive scores. and are respectively the weights of MAV and RMS in the comprehensive scores, so as to ensure that the weights are equal to those in the sEMG comprehensive scores.

[0118] Finally, the two scores are fused together by using a weighted exponential fusion function: wherein score total represents a set of comprehensive scores, which is a total index of the fusion of the sEMG signal score and the compensation interaction torque score, and is used for finally dividing the assistance interval; w sEMG is a weight coefficient of the sEMG signal score, which is used for adjusting the proportion of the electromyographic feature in the set of comprehensive scores; w M is a compensation interaction torque feature score weight, which is used for adjusting the proportion of the mechanical feature in the set of comprehensive scores; is a nonlinear adjustment factor, satisfying > 0.

[0119] The weighted exponential function effectively enhances the robustness of the system, fuses two heterogeneous signals, eliminates differences caused by different physical dimensions, and eliminates magnitude differences through power normalization, preventing any one signal from dominating the score. By adjusting the weights of the sEMG signal and the compensated interaction torque characteristic score, scoring strategies can be personalized for users at different stages of rehabilitation. The nonlinear adjustment factor effectively controls the nonlinear intensity, increases the weight of high-scoring signals, reduces the difficulty of initial user motivation, and strengthens the user's ability to exert maximum effort.

[0120] S60: Divide the assistance intervals based on the comprehensive score set to obtain multi-level assistance intervals.

[0121] The multi-level assistance interval can refer to multiple patient exertion level intervals obtained based on a comprehensive score set. Optionally, the device assistance level can be further determined based on the multi-level assistance interval, but this application does not limit this.

[0122] In one possible implementation, the starting threshold and ending threshold of the active motion intention can be determined based on the aforementioned Gaussian distribution mean and standard deviation; the starting threshold of the active motion intention is then converted into a score to obtain the minimum motion threshold score.

[0123] For example, such as Figure 7 As shown, the MAV of the sEMG signal is used as a reference standard, and a normal distribution is fitted to it. Since the sEMG energy parameters of a single wrist flexion / extension movement approximately follow a normal distribution... Based on the statistical properties of the Gaussian distribution function, (μ-2σ, μ+2σ) contains 95.4% of the energy distribution. Therefore, this can be used as a threshold to define the start and end intervals of energy (action). In other words, on the energy curve, the left endpoint (μ-2σ) of the energy interval can correspond to the threshold of the starting point of the active movement intention, and the right endpoint (μ+2σ) of the energy interval can correspond to the threshold of the ending point of the active movement intention.

[0124] It should be noted that for back extension exercises repeated N times, the Gaussian energy thresholds obtained by averaging the thresholds of the N energy intervals can be used to represent the active movement intention start point threshold and active movement intention end point threshold corresponding to the N back extension exercises: ;

[0125] Where A can be the threshold for the start of the active motion intention; B can be the threshold for the end of the active motion intention; N represents the total number of movements; Q() represents the probability density function of sEMG (using sEMG as an example here, without limiting this application); μ i σ represents the mean of the Gaussian distribution corresponding to the i-th motion; i Let represent the standard deviation of the Gaussian distribution corresponding to the i-th motion.

[0126] The active motion intention starting point threshold value and the active motion intention ending point threshold value can effectively serve as the start and end of the minimum force of the user, guarantee the minimum motion threshold value of the user, can serve as the judgment of the motion intention, and ensure that the device does not provide additional assistance when the user does not reach the threshold value. It can be understood that when determining the minimum motion threshold value score, the judgment can be made based on the comprehensive score set of the sEMG signal and the compensation interaction torque. The above example takes the MAV of the sEMG signal as an example, which does not limit the present application. When the comprehensive score of the patient exceeds the minimum motion threshold value score, it is determined that there is an active motion intention, and the device can start to respond.

[0127] For example, taking the comprehensive score range as [0, 10] as an example, if the minimum motion threshold value score is 3. That is, when the patient performs wrist extension motion, if the comprehensive score calculated based on the multi-source signal fusion exceeds 3, the device determines that the patient has an active motion intention, and thus starts to start the training mode and provides corresponding assistance and other support for the patient; if the comprehensive score is less than 3, the device determines that the patient does not have an effective active motion intention, and does not start the assistance and other functions of the training mode.

[0128] It should be understood that the division of the assistance interval based on the comprehensive score set to obtain the multi-level assistance interval refers to the process of how to determine the dynamic threshold value based on the comprehensive score set, and further divide and obtain the multi-level assistance interval according to the dynamic threshold value. In step S60, that is, the division of the assistance interval based on the comprehensive score set to obtain the multi-level assistance interval can include the following steps:

[0129] S61: calculating and obtaining the score mean of the comprehensive score set based on the comprehensive score set;

[0130] S62: calculating and obtaining the score standard deviation of the comprehensive score set based on the comprehensive score set;

[0131] S63: calculating a dynamic threshold value based on the score mean and the score standard deviation;

[0132] S64: dividing the multi-level assistance interval according to the dynamic threshold value, the comprehensive score set and the minimum motion threshold value score.

[0133] Optionally, the total score mean μ global and the standard deviation σ global of all the training data can be calculated, and the multi-level assistance interval is divided according to the total score mean μ global and the standard deviation σ globalThe dynamic threshold is calculated, so as to further divide the multi-level assistance interval based on the dynamic threshold. Wherein, the dynamic threshold is calculated according to the mean μ global and the standard deviation σ global The process of calculating the dynamic threshold, that is, the process of setting the high / low dynamic threshold to divide the assistance interval, can refer to the following formula:

[0134] ;

[0135] ;

[0136] Wherein, score low represents the dynamic low threshold, which is used to divide the boundary between the low force interval and the transition interval; μ global represents the global score statistical average value of the current user higher than the minimum motion threshold score; represents the coverage adjustment coefficient; σ global represents the standard deviation of the comprehensive score set, score high represents the dynamic high threshold. By obtaining the individualized μ global and σ global values, the individual differences caused by using fixed thresholds by different users can be avoided.

[0137] Based on the normal distribution statistical characteristics, the threshold interval is divided in combination with the scene requirements of rehabilitation training, that is, the confidence interval coverage range can be controlled by adjusting the n value, so as to effectively balance the sensitivity of force change and system stability while ensuring the statistical regularity, and the width of the assistance level interval can also be adjusted. The dynamic threshold strategy can adapt to the individual feature differences of patients in different rehabilitation stages, and improve the individualized adaptation ability.

[0138] As shown in Figure 8 , through the comprehensive scoring system, the comprehensive score set of each stage of the back extension movement in the set range can be obtained, which can also be understood as the real force score (score total ) of the target user. By combining the foregoing minimum motion threshold and dynamic threshold, and comparing score total with them, the assistance interval can be divided into four intervals: , wherein, when score total <s0, the assistance interval is Zone0; Zone0 represents the rest interval, and the score in this interval does not reach s0 score; score total represents the comprehensive score, that is, the real force score of the target user; s0 represents the minimum motion threshold score; when s0≤score total <score low , the assistance interval is Zone1; Zone1 is the start interval; score lowDynamic low threshold value, used to divide the boundary between low force interval and transition interval; when score low ≤ score total < score high , the assistance interval is Zone2; Zone2 is the transition interval; score high Dynamic high threshold value; when score total ≥ score high , the assistance interval is Zone3; Zone3 is the comfortable force interval.

[0139] When performing the back extension movement, the force starts from the starting phase, and rapidly increases to the maximum value in the comfortable force interval, and then gradually decreases as the wrist joint reaches the set maximum back extension angle, so as to return to the force interval level similar to the starting phase.

[0140] In a possible implementation, the device assistance score can be determined according to the comprehensive score of the target user and the threshold of the assistance interval currently occupied by the target user; wherein the determination of the device assistance score according to the comprehensive score of the target user and the threshold of the assistance interval currently occupied by the target user comprises:

[0141] Case one: when the comprehensive score of the target user is between the upper and lower thresholds of the reference assistance interval, the difference between the upper threshold of the reference assistance interval and the comprehensive score of the target user is taken as the device assistance score; the reference assistance interval is an interval in the multi-level assistance interval that needs to provide gradient assistance; the upper and lower thresholds of the reference assistance interval are determined by the dynamic threshold and the minimum movement threshold score;

[0142] Case two: when the comprehensive score of the target user is lower than the minimum movement threshold score, or when the comprehensive score of the target user is higher than the upper threshold in the dynamic threshold, the device assistance score is zero, and no assistance is output at this time.

[0143] For example, the wrist joint rehabilitation method based on the multi-level on-demand assistance grading further comprises: based on the divided assistance interval, the step of implementing on-demand assistance control, that is, the calculation step of the device assistance score:

[0144] When and i = 0, 1,

[0145] ,

[0146] When and i = 0 or and i = 1,

[0147] S A = 0,

[0148] S A is the assistance score provided by the device; S R is the real force score of the target user, i.e. the comprehensive score of the target user, here the parameter S R is described does not constitute a limitation on the present application; S Li is the lower threshold of the current assistance interval; S H(i+1) is the upper threshold of the current assistance interval; i is the index of the reference assistance interval, corresponding to two reference assistance intervals that need to be provided by the device. i = 0 can correspond to the low force interval (Zone1), i.e. the patient has weak force but has exceeded the minimum movement threshold; i = 1 can correspond to the transition interval (Zone2), i.e. the patient has moderate force.

[0149] It should be noted that when the real force score of the target user S and i = 0, 1, the device needs to provide assistance according to S H(i+1) -S R the interval that needs to provide gradient assistance, which is the reference assistance interval that needs to provide gradient assistance; when the real force score of the target user S and i = 0 or and i = 1, the device assistance score is 0, at which time no gradient assistance needs to be provided.

[0150] The device assistance score S A is a dimensionless score index, which needs to be converted into the actual output auxiliary torque M assist of the device. The conversion relationship can be seen as follows: M assist =S A ·M max， , wherein M assist is the actual output auxiliary torque of the device; S A is the assistance score provided by the device; M max is the maximum auxiliary torque preset value that the rehabilitation device can provide.

[0151] That is, when the device assistance score S A is zero, the auxiliary torque M assist output by the device is also zero, i.e. the device does not provide any assistance. When score total s0 (Zone0), it is determined that the user has no movement intention, S A = 0, M assist = 0, and the device is in standby state. When s0≤score total < score low (Zone1), the user has weak force, the device provides assistance, S A =S H1 S R , M assist=S A ·M max ; when score low ≤ score total < score high (Zone2), the user exerts moderate force, the device provides partial assistance, S A =S H2 S R , M assist =S A ·M max ; when score total ≥ score high (Zone3), the user exerts full force, no device assistance is needed, S A = 0, M assist = 0, the training is actively completed by the user.

[0152] The assistance score provided by the device will gradually decrease according to the increase of the user's real force score within the set assistance interval, achieving on-demand assistance. When the user's real force score exceeds the upper threshold of the current interval, the user will automatically enter the next assistance interval, otherwise, it will drop back to the previous assistance interval. In the resting interval (Zone0), if the user's real force score does not reach the minimum movement threshold score, the device will remain in the inactive state to avoid the user's dependence on assistance.

[0153] The present application proposes a multi-level on-demand assistance grading method based on multi-source information fusion. This method combines surface electromyography (sEMG), compensation interaction torque and angle information, dynamic weight distribution and normalization processing to build a comprehensive scoring system, effectively avoiding the problems of insufficient accuracy and large individual differences caused by relying on a single signal. Through a dynamic threshold function, a 4-level assistance interval division method is proposed. Within a single assistance interval, the system can adaptively adjust the assistance force according to the real-time state of the patient, thereby avoiding the patient's excessive dependence on the device and the training difficulty caused by insufficient assistance.

[0154] As shown in Figure 9 , the complete process of the multi-level on-demand assistance grading of the present application includes data acquisition, compensation processing, feature extraction and normalization, muscle weight calculation, comprehensive score calculation, assistance interval division and on-demand assistance control. Specifically, it can be as follows:

[0155] a) Collect the interaction torque, real-time angle θ and surface electromyography sEMG during the wrist joint dorsiflexion movement of the user, wherein the surface electromyography includes the electromyography of the radial wrist long extensor (ECRL), the radial wrist short extensor (ECRB) and the ulnar wrist extensor (ECU);

[0156] b) compensation processing is performed to remove the remaining force caused by the wrist joint self-gravity, motor rotation torque and friction force;

[0157] c) based on the interaction torque and surface electromyogram sEMG, feature values are extracted by a sliding window method, the feature values including a mean absolute value set (MAV) and a root mean square value set (RMS), and Gaussian fitting is performed;

[0158] d) the extracted feature values are normalized;

[0159] e) the weight coefficients of three muscles are dynamically assigned based on the wrist joint motion angle;

[0160] f) a comprehensive score set is calculated according to the normalized feature value set and the weight coefficients;

[0161] g) it is judged whether the comprehensive score in the comprehensive score set exceeds a minimum motion threshold, if not, no assistance force is provided, and the process ends, if yes, a dynamic threshold is set based on the comprehensive score, and a multi-level assistance force interval is divided;

[0162] h) the user is provided with corresponding auxiliary force according to the divided assistance force interval, and then the process returns to the data acquisition step for the next cycle.

[0163] Compared with the prior art, the application has the following beneficial effects:

[0164] 1) By collecting and analyzing surface electromyogram signals, interaction torques and angle information, the user's motion intention is accurately recognized, and the initiative of training is improved;

[0165] 2) The compensation processing mechanism is introduced to eliminate the interference factors such as wrist joint self-gravity and device friction, and to ensure the accuracy of the measurement data;

[0166] 3) The active motion intention starting point threshold and the active motion intention ending point threshold are set based on the Gaussian distribution characteristics, which effectively avoids the interference of noise signals on the starting point and the ending point of training;

[0167] 4) The muscle weight coefficients are dynamically assigned, considering the contribution difference of different muscles at different motion angles, so that the evaluation is more accurate;

[0168] 5) By using the weighted exponential fusion function, the feature advantages of electromyogram signals and torque signals are effectively integrated, and the accuracy and robustness of the score are improved;

[0169] 6) The multi-level on-demand assistance force mechanism based on the dynamic threshold can adaptively adjust the size of the auxiliary force according to the real-time state of the user, and realizes precise and personalized rehabilitation training;

[0170] 7) By segmented on-demand assistance, it avoids dependence due to excessive assistance or movement difficulty caused by insufficient assistance, and strengthens the user's active rehabilitation awareness.

[0171] It can be seen that, in the above scheme, by acquiring the compensation interaction torque, the real-time angle and the surface electromyogram signal, the feature values can be extracted by the sliding window method based on the compensation interaction torque and the surface electromyogram signal, the average absolute value set and the root mean square value set are obtained, and the average absolute value set and the root mean square value set can be normalized to obtain the normalized feature value set, so that the muscle weight coefficient can be obtained based on the real-time angle for weight dynamic allocation, and the comprehensive score set can be obtained by further scoring calculation based on the normalized feature value set and the muscle weight coefficient, and then the multi-level assistance interval can be obtained by dividing the assistance interval based on the comprehensive score set, which can provide accurate wrist joint rehabilitation based on multi-level on-demand assistance classification, and is beneficial to improve the effectiveness of rehabilitation training and the active participation of patients.

[0172] Consistent with the above embodiments, please refer to Figure 10 , Figure 10 A structure schematic diagram of a terminal provided by the embodiment of the present application, as shown in Figure 10 , including a processor, an input device, an output device and a memory, the processor, the input device, the output device and the memory are connected with each other, wherein the memory is used for storing a computer program, the computer program includes program instructions, the processor is configured to call the program instructions, the above program includes instructions for executing the following steps;

[0173] Acquire the compensation interaction torque, the real-time angle and the surface electromyogram signal;

[0174] Based on the compensation interaction torque and the surface electromyogram signal, the feature values are extracted by the sliding window method to obtain the average absolute value set and the root mean square value set;

[0175] The average absolute value set and the root mean square value set are normalized to obtain the normalized feature value set;

[0176] Based on the real-time angle, the muscle weight coefficient is obtained by dynamic weight allocation;

[0177] According to the normalized feature value set and the muscle weight coefficient, the comprehensive score set is obtained by scoring calculation;

[0178] Based on the comprehensive score set, the multi-level assistance interval is obtained by dividing the assistance interval.

[0179] In this example, by acquiring the compensation interaction torque, the real-time angle and the surface electromyogram signal, the average absolute value set and the root mean square value set can be obtained based on the compensation interaction torque and the surface electromyogram signal by extracting the characteristic value through the sliding window method, and the normalization characteristic value set can be obtained by normalizing the average absolute value set and the root mean square value set, so that the muscle weight coefficient can be obtained based on the real-time angle by performing dynamic weight distribution, and the comprehensive score set can be obtained by performing score calculation based on the normalization characteristic value set and the muscle weight coefficient, and then the multi-level assistance interval can be obtained by dividing the assistance interval based on the comprehensive score set, so that the wrist joint rehabilitation based on the multi-level on-demand assistance classification can be provided, and the effectiveness of the rehabilitation training and the initiative participation of the patient can be improved.

[0180] The above describes the scheme of the embodiments of the present application mainly from the perspective of the process of executing the method. It can be understood that the terminal includes a hardware structure and / or a software module corresponding to the execution of each function in order to implement the above functions. Those skilled in the art should easily realize that, in combination with the unit and algorithm steps of each example described in the embodiments provided herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0181] The embodiments of the present application can divide the functional units of the terminal according to the above method examples, for example, each functional unit can be divided corresponding to each function, or two or more functions can be integrated in one processing unit. The integrated unit can be implemented in the form of hardware or software functional unit. It should be noted that the division of units in the embodiments of the present application is illustrative, and is only a logical function division. There can be another division method when actually implemented.

[0182] Consistent with the above, please refer to Figure 11 , Figure 11 A structure diagram of a wrist joint rehabilitation device based on multi-level on-demand assistance classification is provided for the embodiments of the present application. As shown in Figure 11 , the device includes:

[0183] The acquisition module is configured to acquire the compensation interaction torque, the real-time angle and the surface electromyogram signal.

[0184] The first processing module is configured to extract the characteristic value through the sliding window method based on the compensation interaction torque and the surface electromyogram signal, and obtain the average absolute value set and the root mean square value set.

[0185] The second processing module is configured to perform normalization processing on the average absolute value set and the root mean square value set to obtain a normalized characteristic value set.

[0186] The third processing module is configured to perform dynamic weight allocation based on the real-time angle to obtain a muscle weight coefficient.

[0187] The fourth processing module is configured to perform score calculation based on the normalized characteristic value set and the muscle weight coefficient to obtain a comprehensive score set.

[0188] The fifth processing module is configured to perform assistance interval division based on the comprehensive score set to obtain a multi-level assistance interval.

[0189] In one possible implementation, the acquisition module is further configured to:

[0190] acquire an initial interaction torque, a wrist movement overcoming resistance torque, a hand support self-gravity, and a distance from a rotating shaft to the hand support when the target user performs wrist joint dorsiflexion movement;

[0191] determine a hand support gravity torque according to the real-time angle, the hand support self-gravity, and the distance from the rotating shaft to the hand support;

[0192] calculate and obtain a total resistance torque according to the hand support gravity torque and the wrist movement overcoming resistance torque;

[0193] perform compensation processing on the initial interaction torque based on the total resistance torque to obtain a compensated interaction torque.

[0194] In one possible implementation, the fifth processing module is configured to perform assistance interval division based on the comprehensive score set to obtain a multi-level assistance interval, and specifically configured to:

[0195] calculate and obtain a score mean value of the comprehensive score set based on the comprehensive score set;

[0196] calculate and obtain a score standard deviation of the comprehensive score set based on the comprehensive score set;

[0197] calculate a dynamic threshold value based on the score mean value and the score standard deviation;

[0198] divide the multi-level assistance interval according to the dynamic threshold value, the comprehensive score set, and a minimum movement threshold score.

[0199] In one possible implementation, the fifth processing module is further configured to:

[0200] determine a device assistance score according to the comprehensive score of the target user and a threshold value of an assistance interval currently occupied by the target user;

[0201] The fifth processing module is further configured to determine a device assistance score according to the comprehensive score of the target user and a threshold of a current assistance interval of the target user, and specifically configured to:

[0202] When the comprehensive score of the target user is between the upper threshold and the lower threshold of a reference assistance interval, a difference between the upper threshold of the reference assistance interval and the comprehensive score of the target user is taken as the device assistance score; the reference assistance interval is an interval in the multi-level assistance interval that needs to provide gradient assistance.

[0203] When the comprehensive score of the target user is lower than the minimum motion threshold score, or when the comprehensive score of the target user is higher than the upper threshold of the dynamic threshold, the device assistance score is zero.

[0204] In one possible implementation, the first processing module is further configured to:

[0205] perform Gaussian distribution fitting based on the set of average absolute values and the set of root mean square values to obtain a Gaussian distribution mean and a Gaussian distribution standard deviation;

[0206] determine a start threshold of active motion intention and an end threshold of active motion intention according to the Gaussian distribution mean and the Gaussian distribution standard deviation;

[0207] perform score conversion on the start threshold of active motion intention to obtain a minimum motion threshold score.

[0208] The embodiment of the application further provides a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, and the computer program causes a computer to perform part or all steps of any one of the wrist joint rehabilitation methods based on multi-level on-demand assistance grading described in the above method embodiments.

[0209] The embodiment of the application further provides a computer program product, which comprises a non-transitory computer-readable storage medium storing a computer program, and the computer program causes a computer to perform part or all steps of any one of the wrist joint rehabilitation methods based on multi-level on-demand assistance grading described in the above method embodiments.

[0210] It should be noted that, for each of the above method embodiments, in order to simply describe, each is described as a combination of a series of actions, but those skilled in the art should know that the application is not limited to the action order described, because according to the application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily necessary for the application.

[0211] In the above-described embodiments, the description of each embodiment focuses on different aspects, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0212] In several embodiments provided in the present application, it should be understood that the disclosed apparatus can be implemented by other means. For example, the apparatus embodiments described above are only illustrative, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units or components shown or discussed can be indirect coupling or communication connection through some interfaces, apparatuses or units, which can be electrical or other forms.

[0213] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or they can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0214] In addition, the functional units in each embodiment of the application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software program module.

[0215] The integrated unit, if realized in the form of a software program module and sold or used as an independent product, can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part of the prior art that contributes to the technical solutions or the whole or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0216] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiments of various methods can be completed by a program instructing relevant hardware, and the program can be stored in a computer-readable memory, which can include a flash disk, a read-only memory, a random access memory, a magnetic disk or an optical disk, etc.

[0217] The above has carried out the detailed introduction to the embodiments of the application, and the principle and implementation mode of the application are described by applying specific examples; the above embodiment explanation is only for helping to understand the method of the application and its core idea; meanwhile, for the general technical personnel in the art, according to the idea of the application, the specific implementation mode and application range will have changes; in conclusion, the content of the specification should not be understood as the limitation of the application.

Claims

1. A wrist rehabilitation device based on multi-stage assistive force grading, characterized in that, The device comprises: An acquisition module is configured to acquire a compensation interaction torque, a real-time angle, and a surface electromyography signal; the compensation interaction torque is a value of an interaction torque obtained by compensating an initial interaction torque; the initial interaction torque is an interaction torque between a patient and a wrist joint rehabilitation device collected for the first time; the real-time angle is an angle of real-time rotation of the wrist joint during a dorsiflexion movement; and the surface electromyography signal is a bioelectric signal collected from a surface of a muscle related to the dorsiflexion movement of the wrist joint; A first processing module is configured to extract feature values based on the compensation interaction torque and the surface electromyography signal by using a sliding window method, to obtain a mean absolute value set and a root mean square value set; A second processing module is configured to perform normalization processing on the mean absolute value set and the root mean square value set, to obtain a normalized feature value set; A third processing module is configured to perform dynamic weight distribution based on the real-time angle, to obtain a muscle weight coefficient; A fourth processing module is configured to perform scoring calculation based on the normalized feature value set and the muscle weight coefficient, to obtain a comprehensive score set; A fifth processing module is configured to perform assistance interval division based on the comprehensive score set, to obtain a multi-level assistance interval; The first processing module is configured to extract feature values based on the compensation interaction torque and the surface electromyography signal by using the sliding window method, to obtain the mean absolute value set and the root mean square value set, and specifically configured to: extract feature values based on the surface electromyography signal by using the sliding window method, to obtain a mean absolute value subset corresponding to the surface electromyography signal and a root mean square value subset corresponding to the surface electromyography signal; extract feature values based on the compensation interaction torque by using the sliding window method, to obtain a mean absolute value subset corresponding to the compensation interaction torque and a root mean square value subset corresponding to the compensation interaction torque.

2. The multi-stage on-demand assistance based hierarchical wrist rehabilitation device according to claim 1, wherein, The acquisition module is further configured to: acquire an initial interaction torque, a wrist joint movement resistance torque, a hand support self-gravity, and a distance from a rotating shaft to the hand support when a target user performs a wrist joint dorsiflexion movement; determine a hand support gravity torque according to the real-time angle, the hand support self-gravity, and the distance from the rotating shaft to the hand support; calculate and obtain a total resistance torque according to the hand support gravity torque and the wrist joint movement resistance torque; perform compensation processing on the initial interaction torque based on the total resistance torque, to obtain the compensation interaction torque.

3. The multi-stage on-demand assistance based hierarchical wrist rehabilitation device according to claim 2, wherein, The fifth processing module is configured to perform assistance interval division based on the comprehensive score set, to obtain the multi-level assistance interval, and specifically configured to: calculate and obtain a score mean value of the comprehensive score set based on the comprehensive score set; calculate and obtain a score standard deviation of the comprehensive score set based on the comprehensive score set; calculate a dynamic threshold value based on the score mean value and the score standard deviation; divide the multi-level assistance interval according to the dynamic threshold value, the comprehensive score set, and a minimum movement threshold score.

4. The multi-stage on-demand assistance based hierarchical wrist rehabilitation device according to claim 3, wherein, The fifth processing module is further configured to: determine a device assistance score according to a comprehensive score of the target user and a threshold value of an assistance interval currently occupied by the target user. The device assistance score is determined according to the comprehensive score of the target user and a threshold value of the assistance interval currently occupied by the target user. When the comprehensive score of the target user is between the upper threshold value and the lower threshold value of the reference assistance interval, the difference between the upper threshold value of the reference assistance interval and the comprehensive score of the target user is taken as the device assistance score; and the reference assistance interval is an interval in the multi-level assistance interval that needs to provide gradient assistance. When the comprehensive score of the target user is lower than the minimum motion threshold value, or when the comprehensive score of the target user is higher than the upper threshold value in the dynamic threshold value, the device assistance score is zero.

5. The multi-stage on-demand assist level based wrist rehabilitation device according to any one of claims 1-4, characterized in that, The first processing module is further configured to: perform Gaussian distribution fitting based on the average absolute value set and the root mean square value set to obtain a Gaussian distribution mean value and a Gaussian distribution standard deviation; determine a starting point threshold value of the active motion intention and an ending point threshold value of the active motion intention according to the Gaussian distribution mean value and the Gaussian distribution standard deviation; and perform score conversion on the starting point threshold value of the active motion intention to obtain a minimum motion threshold value.

6. A terminal, characterized by comprising: The device comprises a processor, an input device, an output device and a memory, which are connected to each other, wherein the memory is configured to store a computer program, the computer program comprises program instructions, and the processor is configured to invoke the program instructions to execute the following method: obtain a compensation interaction torque, a real-time angle and a surface electromyography signal; the compensation interaction torque is a value of an interaction torque obtained by compensating an initial interaction torque; the initial interaction torque is an interaction torque between a patient and a wrist joint rehabilitation device collected for the first time; the real-time angle is an angle of real-time rotation of the wrist joint during a dorsiflexion movement; and the surface electromyography signal is a bioelectric signal collected from a muscle surface related to the dorsiflexion movement of the wrist joint; extract feature values based on the compensation interaction torque and the surface electromyography signal by using a sliding window method to obtain an average absolute value set and a root mean square value set; perform normalization processing on the average absolute value set and the root mean square value set to obtain a normalized feature value set; dynamically distribute weights based on the real-time angle to obtain a muscle weight coefficient; perform score calculation according to the normalized feature value set and the muscle weight coefficient to obtain a comprehensive score set; divide an assistance interval based on the comprehensive score set to obtain a multi-level assistance interval. The method of extracting feature values based on the compensation interaction torque and the surface electromyography signal by using the sliding window method to obtain the average absolute value set and the root mean square value set comprises: extract feature values based on the surface electromyography signal by using the sliding window method to obtain an average absolute value subset corresponding to the surface electromyography signal and a root mean square value subset corresponding to the surface electromyography signal; extract feature values based on the compensation interaction torque by using the sliding window method to obtain an average absolute value subset corresponding to the compensation interaction torque and a root mean square value subset corresponding to the compensation interaction torque.

7. A computer readable storage medium characterized by The computer readable storage medium stores a computer program, the computer program comprising program instructions which, when executed by a processor, cause the processor to perform the following method: obtaining a compensation interaction torque, a real-time angle and a surface electromyography signal; the compensation interaction torque is a value of an interaction torque obtained by compensation processing on an initial interaction torque collected, and the initial interaction torque is an interaction torque between a patient and a wrist joint rehabilitation device collected for the first time; the real-time angle is an angle of real-time rotation of the wrist joint during the dorsiflexion movement; the surface electromyography signal is a bioelectric signal collected from the surface of the dorsiflexion movement related muscle of the wrist joint; based on the compensation interaction torque and the surface electromyography signal, extracting feature values through a sliding window method to obtain an average absolute value set and a root mean square value set; normalizing the average absolute value set and the root mean square value set to obtain a normalized feature value set; dynamically distributing weights based on the real-time angle to obtain a muscle weight coefficient; performing scoring calculation according to the normalized feature value set and the muscle weight coefficient to obtain a comprehensive score set; dividing the assistance interval based on the comprehensive score set to obtain a multi-level assistance interval; the compensation interaction torque and the surface electromyography signal, extracting feature values through a sliding window method to obtain an average absolute value set and a root mean square value set, comprising: based on the surface electromyography signal, extracting feature values through a sliding window method to obtain an average absolute value subset corresponding to the surface electromyography signal and a root mean square value subset corresponding to the surface electromyography signal; based on the compensation interaction torque, extracting feature values through a sliding window method to obtain an average absolute value subset corresponding to the compensation interaction torque and a root mean square value subset corresponding to the compensation interaction torque.

Images