Upper limb robot rehabilitation training control method based on electromyographic signals

Abstract

The invention discloses an upper limb robot rehabilitation training control method based on electromyographic signals, and relates to the technical field of rehabilitation medical treatment, and the method comprises the following steps: collecting electromyographic signals of a human body upper limb target muscle group in different action modes, and constructing an electromyographic signal feature library, screening an electromyographic signal matched with the rehabilitation training action from an electromyographic signal feature library and marking the electromyographic signal as a training trigger signal; extracting a signal effective value, a spectrum feature and an action correlation degree of the training trigger signals, judging whether feature enhancement processing needs to be performed on the at least two groups of training trigger signals according to the signal effective value and the spectrum feature, and marking the corresponding training trigger signals as control instruction generation signals; acquiring an upper limb movement track parameter corresponding to the control instruction generation signal; the method has the advantages that quantitative analysis of the electromyographic signals can reflect the neuromuscular function state of a patient in real time, objective data support is provided for training scheme adjustment, and training scientificity and repeatability are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of rehabilitation medicine, more particularly, it relates to a control method for upper limb robot rehabilitation training based on electromyographic signals. BACKGROUND

[0002] In the field of rehabilitation treatment of stroke patients, upper limb motor dysfunction is a common sequelae with an incidence of more than 70%, which not only causes patients to be unable to independently complete basic life actions such as eating and dressing, but also reduces the social participation of patients, and brings long-term care burden to families and society.

[0003] However, most of the existing upper limb rehabilitation robots adopt a preset trajectory traction mode, and do not recognize the movement intention in combination with the patient's own electromyographic signals, so the patient is always in a driven state and cannot stimulate neural plasticity; Even a few systems collect electromyographic signals, but only determine the action according to the signal amplitude, do not analyze the key parameters of the effective value and frequency spectrum characteristics, and do not recognize the multi-muscle linkage characteristics in combination with the muscle group cooperation relationship, and the system cannot distinguish between single muscle action and multi-muscle coordinated action, which is easy to misjudge the coordinated effort as a single action, resulting in the disconnection between the control command and the physiological law; At the same time, the existing system lacks an electromyographic signal feature library and a precise screening mechanism, only collects signals under a small number of action modes, which is difficult to match the diversified action requirements of rehabilitation training, and the signals are not processed for feature enhancement, when the signal has noise or weak amplitude, the system is easy to misjudge, not only reduces the training effect, but also may cause muscle compensation or secondary injury due to action deviation. SUMMARY

[0004] In view of the deficiencies in the prior art, the purpose of the present application is to provide a control method for upper limb robot rehabilitation training based on electromyographic signals.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: A control method for upper limb robot rehabilitation training based on electromyographic signals, the method comprising the following steps: Collecting electromyographic signals of a target muscle group of a human upper limb under different action modes, constructing an electromyographic signal feature library, and screening electromyographic signals matched with rehabilitation training actions from the electromyographic signal feature library and marking them as training trigger signals; Extracting signal effective value, frequency spectrum characteristics and action correlation degree of the training trigger signals, judging whether at least two groups of training trigger signals need to be processed for feature enhancement according to the signal effective value and the frequency spectrum characteristics, and marking the corresponding training trigger signals as control instruction generation signals; Obtaining upper limb motion trajectory parameters corresponding to the control instruction generation signals; Judging whether there is a multi-muscle group linkage feature in the control instruction generation signals according to the action correlation degree and the muscle group cooperation relationship; If the control instruction generation signal corresponds to the action characteristics of a single muscle group, the basic rehabilitation control instruction of the upper limb robot is generated according to the signal effective value and the upper limb motion trajectory parameter; If the control instruction generation signal corresponds to the multi-muscle group linkage characteristics, the cooperative rehabilitation control instruction of the upper limb robot is generated according to the frequency spectrum characteristics and the upper limb motion trajectory parameter; The upper limb robot is driven to complete the rehabilitation training action according to the basic rehabilitation control instruction or the cooperative rehabilitation control instruction.

[0006] Preferably, whether the feature enhancement processing is needed for at least two groups of training trigger signals is judged according to the signal effective value and the frequency spectrum characteristics, and the corresponding training trigger signal is marked as the control instruction generation signal, which comprises: If the signal effective value and the frequency spectrum characteristics corresponding to the same action mode are completely presented in a single group of training trigger signals, the training trigger signal is directly marked as the control instruction generation signal; If the signal effective value and the frequency spectrum characteristics corresponding to the same action mode are scattered in at least two groups of training trigger signals, the corresponding training trigger signal is subjected to the frequency domain decomposition, and the signal component containing the core characteristics is extracted; According to the muscle force exertion law corresponding to the action mode, the noise reduction enhancement processing is performed on each signal component, and the processed signal component is reconstructed to form the control instruction generation signal.

[0007] Preferably, according to the muscle force exertion law corresponding to the action mode, the noise reduction enhancement processing is performed on each signal component, and the processed signal component is reconstructed to form the control instruction generation signal, which comprises: It is judged whether the frequency domain characteristics of at least two signal components exist a characteristic overlapping frequency band; If the characteristic overlapping frequency band exists, the corresponding signal component is marked as the matching signal component, and the signal reconstruction is performed based on the energy characteristics of the overlapping frequency band to form the control instruction generation signal; If the characteristic overlapping frequency band does not exist, the frequency domain characteristics of each signal component are standardized, and if the characteristic parameters of each signal component after the standardization meet the characteristic requirements of the action mode, the corresponding signal component is integrated and reconstructed to form the control instruction generation signal.

[0008] Preferably, whether the multi-muscle group linkage characteristics exist in the control instruction generation signal is judged according to the action correlation degree and the muscle group cooperation relationship, which comprises: If the action correlation degree of the control instruction generation signal corresponds to a single muscle group, it is determined that the multi-muscle group linkage characteristics do not exist in the control instruction generation signal; If the action correlation degree of the control instruction generation signal corresponds to at least two groups of muscle groups, whether the electromyographic signals of each muscle group exist the time sequence cooperation is judged according to the muscle group cooperation relationship; If the muscle signals of each muscle group have no time sequence synergy, it is determined that the control instruction generation signal does not have a multi-muscle group linkage feature; If the muscle signals of each muscle group have time sequence synergy, it is determined that the control instruction generation signal has a multi-muscle group linkage feature.

[0009] Preferably, if the control instruction generation signal only corresponds to the action feature of a single muscle group, the basic rehabilitation control instruction of the upper limb robot is generated according to the signal effective value and the upper limb motion trajectory parameter, including: The feature parameters of the control instruction generation signal are matched with the electromyographic feature motion trajectory mapping table to obtain the joint motion amplitude and motion acceleration of the robot corresponding to the muscle group; The joint motion amplitude and motion acceleration are divided into basic training parameters and advanced training parameters; The joint motion amplitude and motion acceleration are iteratively optimized to generate the basic rehabilitation control instruction according to the adaptability of the basic training parameters and the advanced training parameters.

[0010] Preferably, the basic rehabilitation control instruction is generated by iteratively optimizing the joint motion amplitude and motion acceleration according to the adaptability of the basic training parameters and the advanced training parameters, including: It is determined whether the basic training parameters and the advanced training parameters meet the adaptability requirements of the rehabilitation training stage; If the adaptability requirements are not met, the joint motion amplitude and motion acceleration are adjusted by limiting value based on the basic training parameters to generate the basic rehabilitation control instruction; If the adaptability requirements are met, the adaptation coefficient of the basic training parameters and the advanced training parameters is calculated, the optimal training parameters are obtained by iteratively correcting the joint motion amplitude and motion acceleration with the adaptation coefficient as the optimization weight, and the basic rehabilitation control instruction is generated according to the optimal training parameters.

[0011] Preferably, the cooperative rehabilitation control instruction of the upper limb robot is generated according to the frequency spectrum feature and the upper limb motion trajectory parameter, specifically including the following steps: It is determined whether the frequency spectrum feature corresponding to the multi-muscle group linkage feature has a frequency band superposition effect; If there is a frequency band superposition effect, the muscle synergy level is determined according to the energy distribution of the superimposed frequency band to obtain the first cooperative control parameter of the joint linkage motion; If there is no frequency band superposition effect, the muscle force priority is determined according to the frequency spectrum feature proportion of each muscle group to obtain the second cooperative control parameter of the joint linkage motion; The first cooperative control parameter and the second cooperative control parameter are fused and matched with the upper limb motion trajectory parameter to generate the cooperative rehabilitation control instruction.

[0012] Preferably, the level of muscle synergy is determined based on the energy distribution of the superimposed frequency band to obtain the first synergistic control parameter for joint linkage motion, including: Determine whether the energy distribution of the superimposed frequency band exceeds the preset collaborative force threshold; If the coordinated force threshold is not exceeded, the current muscle coordinated force level is determined to be the basic linkage level; The preset basic angle coordination value and basic velocity coordination value are used as the angle coordination value and velocity coordination value of the joint linkage motion; If the coordinated force threshold is exceeded, the energy peak value and duration information of the superimposed frequency band are obtained. The energy peak value and duration information are compared with the electromyographic energy coordination level mapping table to obtain the angle coordination value and velocity coordination value of the joint linkage movement. The angle coordination value and the velocity coordination value are marked as the first coordination control parameters.

[0013] Preferably, the muscle exertion priority is determined based on the proportion of spectral characteristics of each muscle group to obtain the second coordinated control parameters for joint linkage motion, including: By comparing the proportion of spectral characteristics of each muscle group with the electromyographic spectrum force priority mapping table, the force priority coefficient corresponding to each muscle group can be obtained. Based on the force priority coefficient, the synthetic coordination angle and synthetic coordination velocity of the joint linkage motion are calculated, and the synthetic coordination angle and synthetic coordination velocity are marked as the second coordination control parameters.

[0014] Compared with existing technologies, this invention has the following advantages: By constructing an electromyographic signal feature library covering different movement modes and combining the parameters of signal effective value and spectral characteristics to screen training trigger signals, standardized matching of rehabilitation training movements is achieved; the quantitative analysis of electromyographic signals can reflect the neuromuscular functional state of patients in real time, providing objective data support for adjusting training programs and improving the scientific nature and repeatability of training; by analyzing the correlation of movements to identify active movement intentions, the robot training movements are deeply bound to the patient's force exertion intentions, changing from passive traction to an active triggering and assisted coordination mode, effectively stimulating neural plasticity and improving training efficiency; by extracting signal effective values ​​and spectral characteristics and judging the linkage characteristics of multiple muscle groups, single muscle movements and multi-muscle coordinated movements can be accurately distinguished, generating basic rehabilitation control commands for single muscle group movements and coordinated rehabilitation control commands for multi-muscle coordinated movements, avoiding muscle compensation or secondary damage caused by misjudgment, and improving the safety and accuracy of training. Attached Figure Description

[0015] Fig. 1 This is a schematic diagram illustrating the steps of an upper limb robotic rehabilitation training control method based on electromyographic signals, as provided in an embodiment of the present invention. Fig. 2A step schematic diagram for forming a control instruction generation signal in an upper limb robot rehabilitation training control method based on an electromyographic signal is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0016] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0017] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from the description, and those skilled in the art can make similar generalizations without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0018] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is independent of or selectively excludes other embodiments.

[0019] Reference Figs. 1-2 is shown.

[0020] The present embodiment further illustrates an upper limb robot rehabilitation training control method based on an electromyographic signal proposed by the present application.

[0021] An upper limb robot rehabilitation training control method based on an electromyographic signal, the method comprising the following steps: Collecting electromyographic signals of a target muscle group of a human upper limb under different motion modes, constructing an electromyographic signal feature library, screening electromyographic signals matched with rehabilitation training motions from the electromyographic signal feature library and marking the electromyographic signals as training trigger signals; Extracting signal effective values, spectral features and motion correlation degrees of the training trigger signals, judging whether at least two groups of training trigger signals need to be subjected to feature enhancement processing according to the signal effective values and the spectral features, and marking the corresponding training trigger signals as control instruction generation signals.

[0022] The system first collects the electromyographic signals of the target muscle groups of the upper limbs of the human body comprehensively. For patients who need rehabilitation after stroke, the system selects the biceps brachii, triceps brachii, wrist extensor, and muscles directly related to the flexion and extension of the upper limbs and the activity of the wrist as the collection objects. The collection process covers the electromyographic signals of the patients in different action modes such as actively trying to clench the fist, extend the elbow, and flex the wrist. These signals are sorted and constructed into an electromyographic signal feature library by the system. Then, the system screens the electromyographic signals matching the training action from the feature library according to the preset rehabilitation training target, and marks them as training trigger signals. For example, when the wrist joint extension training is set as the rehabilitation training plan, the system screens the electromyographic signals generated by the wrist extensor in the active extension action from the feature library as the trigger basis of the training action.

[0023] Then, the system deeply analyzes and processes the training trigger signals screened. First, the key features in the signals are extracted, including the signal effective value reflecting the muscle contraction strength, the frequency spectrum feature reflecting the signal frequency distribution, and the action correlation degree reflecting the close degree of the signal and the target action. The system determines whether feature enhancement processing is needed based on the signal effective value and the frequency spectrum feature. For example, if two groups of training trigger signals are collected, one from the elbow flexion action of the biceps brachii and the other from the elbow extension action of the triceps brachii, their signal effective values are low and the frequency spectrum features overlap, which can easily lead to misjudgment of the robot. At this time, the system will perform feature enhancement processing on the two groups of signals, such as amplifying the feature difference of the effective signal and reducing noise interference. The training trigger signal after processing will be marked as a control instruction generation signal, providing a reliable signal basis for generating accurate control instructions of the robot subsequently.

[0024] Obtaining the upper limb motion trajectory parameters corresponding to the control instruction generation signal; Determining whether there is a multi-muscle group linkage feature in the control instruction generation signal according to the action correlation degree and the muscle group cooperation relationship; If the control instruction generation signal only corresponds to a single muscle group linkage feature, generating a basic rehabilitation control instruction of the upper limb robot according to the signal effective value and the upper limb motion trajectory parameters; If the control instruction generation signal corresponds to a multi-muscle group linkage feature, generating a cooperative rehabilitation control instruction of the upper limb robot according to the frequency spectrum feature and the upper limb motion trajectory parameters; Driving the upper limb robot to complete the rehabilitation training action according to the basic rehabilitation control instruction or the cooperative rehabilitation control instruction.

[0025] The basic rehabilitation control instruction corresponds to the action characteristics of a single muscle group, for example, the patient only needs to complete a simple wrist extension action, at this time the system only collects the electromyogram of the wrist extensor muscle. The system generates a control instruction based on the signal effective value and the preset upper limb motion trajectory parameters, wherein the signal effective value is a core indicator reflecting the muscle contraction intensity, and the calculation formula is: , wherein RMS is the signal effective value, x i is the sample point value of the electromyogram, and N is the sample point quantity. This value directly reflects the contraction intensity of the biceps brachii muscle, and the larger the RMS, the stronger the muscle contraction. The system combines this value with the preset upper limb motion trajectory parameters, such as the target angle range of the elbow flexion action being 0° to 120° and the motion speed being 5° / s. When the RMS is detected to be 0.8 mV, the system generates a basic rehabilitation control instruction to drive the elbow joint actuator of the robot to drive the patient's arm to complete the elbow flexion action at the set speed, while monitoring the deviation of the actual motion angle from the target trajectory in real time, adjusting the motor torque through closed-loop control to ensure that the action is accurate and meets the training requirements.

[0026] The collaborative rehabilitation control instruction corresponds to the action characteristics of multiple muscle groups, for example, the patient needs to complete a composite action of lifting the hand to touch the shoulder, which requires multiple muscle groups such as the deltoid muscle, biceps brachii muscle, and levator scapulae muscle to contract synchronously. The system generates a control instruction based on the frequency spectrum characteristics and the upper limb motion trajectory parameters, and the frequency spectrum characteristics are usually represented by the average power frequency MPF, and the calculation formula is: , wherein f is the frequency component of the electromyogram, and P(f) is the power spectral density corresponding to the frequency, which reflects the frequency distribution characteristics of muscle contraction. Different muscles have obvious differences in MPF, and the system can distinguish the activation state of different muscles through this feature. Combined with the preset upper limb motion trajectory parameters of the taking action, such as the shoulder joint flexion angle of 30°, the elbow joint flexion angle of 90°, and the wrist joint neutral position, the system generates a collaborative rehabilitation control instruction to drive the shoulder joint, elbow joint, and wrist joint actuators of the robot to move collaboratively. For example, when the MPF of the biceps brachii muscle is detected to be 50 Hz and the MPF of the triceps brachii muscle is 45 Hz, the system controls the shoulder joint actuator to flex at a speed of 3° / s, the elbow joint actuator to flex at a speed of 4° / s, and the wrist joint actuator to remain in the neutral position, driving the patient's arm to complete the taking action. During the action execution process, the system continuously receives feedback of the electromyogram, and adjusts the motion speed and torque of each joint in real time to ensure that the multi-muscle group action is smooth and meets the rehabilitation training standards.

[0027] Finally, whether it is a basic rehabilitation control instruction or a collaborative rehabilitation control instruction, it will be converted into mechanical motion through the drive module of the robot to drive the patient's upper limb to complete the corresponding rehabilitation training action, realizing the whole-link closed-loop control from electromyogram collection, feature analysis to robot precise execution, effectively improving the pertinence and effectiveness of rehabilitation training.

[0028] According to the signal effective value and the spectrum feature, it is judged whether the feature enhancement processing is needed for at least two groups of training trigger signals, and the corresponding training trigger signal is marked as a control instruction generation signal, comprising: If the signal effective value and the spectrum feature corresponding to the same action mode are completely presented in a single group of training trigger signals, the training trigger signal is directly marked as a control instruction generation signal; If the signal effective value and the spectrum feature corresponding to the same action mode are scattered in at least two groups of training trigger signals, the corresponding training trigger signal is subjected to frequency domain decomposition, and the signal component containing the core feature is extracted; According to the muscle force exertion law corresponding to the action mode, the noise reduction enhancement processing is performed on each signal component, and the processed signal component is reconstructed to form a control instruction generation signal.

[0029] Firstly, the system will judge whether the signal feature corresponding to the same action mode is complete based on the signal effective value and the spectrum feature. For example, when the patient performs a simple elbow flexion action, if the signal effective value of the biceps brachii is 0.7 mV and the average power frequency is 52 Hz in a single group of training trigger signals, and the two features completely correspond to the muscle force exertion law of elbow flexion action, the system will directly mark the training trigger signal as a control instruction generation signal without additional processing.

[0030] If the signal effective value and the spectrum feature corresponding to the same action mode are scattered in at least two groups of training trigger signals, the system will start the feature enhancement process. For example, when the patient performs a fist clenching action, the signal effective value of the flexor muscle is 0.5 mV and the average power frequency is 48 Hz, which are scattered in two groups of training trigger signals, one group contains part of the effective value feature, and the other group contains part of the spectrum feature. At this time, the system will first perform frequency domain decomposition on the two groups of training trigger signals, extract the signal component containing the core feature through Fourier transform, and separate the effective electromyographic feature mixed in the noise.

[0031] Next, the system will combine the muscle force exertion law corresponding to the fist clenching action to perform noise reduction enhancement processing on each signal component, amplify the amplitude of the effective feature by suppressing power frequency interference and motion artifacts. After processing, the system will perform time domain reconstruction on the enhanced signal component, synthesize a group of signals containing complete effective value and spectrum feature, and finally mark the reconstructed signal as a control instruction generation signal, so as to ensure that the subsequent generated robot control instruction can accurately match the action intention of the patient.

[0032] According to the muscle force exertion law corresponding to the action mode, the noise reduction enhancement processing is performed on each signal component, and the processed signal component is reconstructed to form a control instruction generation signal, comprising: It is judged whether the frequency domain features of at least two signal components exist in the feature overlapping frequency band; If there is a characteristic overlapping frequency band, the corresponding signal component is marked as a matching signal component, and the signal reconstruction is performed based on the energy characteristics of the overlapping frequency band to form a control instruction generation signal; If there is no characteristic overlapping frequency band, the frequency domain characteristics of each signal component are standardized, and if the characteristic parameters of each signal component after standardization meet the characteristic requirements of the action mode, the corresponding signal component is integrated and reconstructed to form a control instruction generation signal.

[0033] First, the system compares the frequency domain characteristics of at least two signal components to determine whether there is a characteristic overlapping frequency band. The frequency domain characteristics usually take average power frequency and power spectral density as core indicators. Taking the patient performing elbow extension as an example, the system decomposes two groups of signal components, one from the triceps brachii, whose frequency domain characteristics cover the 20Hz to 60Hz frequency band, and the other from the elbow muscle, whose frequency domain characteristics cover the 40Hz to 70Hz frequency band. At this time, the system will identify that the two groups of components have a characteristic overlapping frequency band in the 40Hz to 60Hz interval, and will mark the two groups of signal components as matching signal components. Next, the system will extract the energy characteristics of the overlapping frequency band, calculate the total energy E overlap , and the formula is: , where P1(f) and P2(f) are the power spectral densities of the two groups of components at frequency f. The system will weight and reconstruct the signal components based on the energy characteristics to strengthen the effective characteristics of the overlapping frequency band, and finally generate a control instruction generation signal containing the complete elbow extension action characteristics.

[0034] If there is no characteristic overlapping frequency band after comparison, the system will standardize the frequency domain characteristics of each signal component. The standardization formula is: , where MPF i is the average power frequency of the i-th signal component, and μMPF and σMPF are the mean and standard deviation of the average power frequency of all signal components in this action mode.

[0035] Taking the patient performing wrist rotation as an example, the system decomposes two groups of signal components without overlapping frequency bands, one from the pronator teres muscle and the other from the supinator muscle. After standardization, if the characteristic parameters of the two groups of components meet the preset characteristic requirements of the wrist rotation, such as the standardized average power frequency being within the reasonable interval of -1 to 1, the system will integrate and reconstruct the two groups of signal components to synthesize a complete control instruction generation signal through time domain superposition, ensuring that the signal can accurately reflect the action intention of multi-muscle coordination.

[0036] The whole process realizes the effective enhancement and reconstruction of scattered signal components through the comparison and standardization of frequency domain characteristics, laying a solid foundation for subsequent generation of accurate robot control instructions.

[0037] According to the action correlation and muscle group cooperation relationship, it is judged whether there is a multi-muscle group linkage feature in the control instruction generation signal, comprising: If the action correlation of the control instruction generation signal only corresponds to a single muscle group, it is determined that there is no multi-muscle group linkage feature in the control instruction generation signal; If the action correlation of the control instruction generation signal corresponds to at least two groups of muscle groups, it is judged according to the muscle group cooperation relationship whether the electromyographic signals of each muscle group have time sequence cooperation; If the electromyographic signals of each muscle group have no time sequence cooperation, it is determined that there is no multi-muscle group linkage feature in the control instruction generation signal; If the electromyographic signals of each muscle group have time sequence cooperation, it is determined that there is a multi-muscle group linkage feature in the control instruction generation signal.

[0038] Firstly, the system will analyze the action correlation of the control instruction generation signal. The action correlation is used to measure the close degree of the signal and the action of a specific muscle group, and its calculation formula is: Wherein xi is the sampling point of the current control instruction generation signal, yi is the standard signal sampling point of the preset single muscle group action, and N is the number of sampling points. The closer the value is to 1, the closer the signal is to the target muscle group action. For example, when the patient performs a simple wrist flexion action, the system detects that the action correlation of the control instruction generation signal is highly related to the wrist flexor muscle, and at this time it is determined that the signal has no multi-muscle group linkage feature, and the subsequent basic rehabilitation control instruction will be generated.

[0039] If the action correlation corresponds to at least two groups of muscle groups, the system will further judge according to the muscle group cooperation relationship whether the electromyographic signals of each muscle group have time sequence cooperation. The time sequence cooperation refers to whether the activation time and peak value occurrence time of the electromyographic signals of multiple muscle groups in the action process conform to the physiological cooperation rule. Taking the patient's action of completing the water cup from the desktop as an example, the action correlation corresponds to three muscle groups of deltoid, biceps and wrist flexor. The system extracts the peak time points of the electromyographic signals of the three muscle groups, the peak time of the deltoid is 100ms after the action starts, the peak time of the biceps is 200ms, and the peak time of the wrist flexor is 300ms. The interval of each time point is within 100ms, which conforms to the time sequence logic of physiological action. The deltoid is started first to lift the arm, then the biceps is contracted to bend the elbow, and finally the wrist flexor is contracted to adjust the cup holding posture. At this time, the system determines that the signal has a multi-muscle group linkage feature, and the subsequent cooperative rehabilitation control instruction will be generated.

[0040] If the muscle electrical signals of each muscle group have no obvious time sequence cooperation, for example, the time difference of the peak values of the muscle electrical signals of the three muscle groups exceeds 300 ms, it indicates that the muscle activation has no clear cooperation rule, and the system determines that the signal does not have the multi-muscle group linkage feature. Through the preliminary screening of the action correlation degree and the verification of the time sequence cooperation, the system can accurately distinguish between single muscle action and multi-muscle linkage action, and ensure that the control command generated subsequently is highly matched with the real action intention of the patient If the control command generation signal only corresponds to the action characteristics of a single muscle group, the basic rehabilitation control command of the upper limb robot is generated according to the signal effective value and the upper limb motion trajectory parameters, including: The characteristic parameters of the control command generation signal are compared with the muscle electrical characteristic motion trajectory mapping table to obtain the joint motion amplitude and motion acceleration of the robot corresponding to the muscle group; The joint motion amplitude and motion acceleration are divided into basic training parameters and advanced training parameters; The joint motion amplitude and motion acceleration are iteratively optimized to generate the basic rehabilitation control command according to the adaptability of the basic training parameters and the advanced training parameters.

[0041] Firstly, the system will compare the characteristic parameters of the control command generation signal with the preset muscle electrical characteristic motion trajectory mapping table. The characteristic parameters mainly include the signal effective value and the upper limb motion trajectory parameters. Taking the case of the patient performing a simple elbow extension action, the system detects that the signal effective value of the triceps brachii muscle is 0.6 mV, and after comparing the mapping table, the corresponding robot elbow joint motion amplitude is 0° to 100°, and the motion acceleration is 3° / s².

[0042] Next, the system will divide the obtained joint motion amplitude and motion acceleration into basic training parameters and advanced training parameters. The basic training parameters usually correspond to low-intensity and small-range action modes, for example, the elbow joint motion amplitude is 0° to 50° and the acceleration is 1° / s², which is suitable for patients in the early stage of rehabilitation; the advanced training parameters correspond to high-intensity and large-range action modes, for example, the motion amplitude is 50° to 100° and the acceleration is 3° / s², which is suitable for patients in the later stage of rehabilitation.

[0043] After that, the system will iteratively optimize the joint motion amplitude and motion acceleration according to the adaptability of the basic training parameters and the advanced training parameters. The iterative optimization process can be represented by the following formula: , where P kT is the target adaptation parameter, and a is the step coefficient. For example, for a patient in the middle stage of rehabilitation, the system will first use the basic training parameters as the initial value and gradually transition to the advanced training parameters through iteration. In the first iteration, the movement amplitude is adjusted from 0° to 50° to 0° to 60°, and the acceleration is adjusted from 1° / s² to 1.5° / s². In each subsequent iteration, the step coefficient is adjusted according to the patient's action completion degree until the parameters adapt to the patient's current rehabilitation ability, and finally a stable basic rehabilitation control instruction is generated.

[0044] Through mapping, division, and iterative optimization, the generated control instruction not only meets the physiological characteristics of muscle contraction but also accurately matches the patient's rehabilitation stage, thereby improving the safety and effectiveness of basic rehabilitation training.

[0045] According to the adaptability of the basic training parameters and the advanced training parameters, the joint movement amplitude and the movement acceleration are iteratively optimized to generate the basic rehabilitation control instruction, including: determining whether the basic training parameters and the advanced training parameters meet the adaptation requirements of the rehabilitation training stage; If the adaptation requirements are not met, the basic training parameters are used as the reference to limit the value adjustment of the joint movement amplitude and the movement acceleration to generate the basic rehabilitation control instruction. If the adaptation requirements are met, the adaptation coefficient of the basic training parameters and the advanced training parameters is calculated, the adaptation coefficient is used as the optimization weight to iteratively correct the joint movement amplitude and the movement acceleration to obtain the optimal training parameters, and the basic rehabilitation control instruction is generated according to the optimal training parameters.

[0046] First, the system determines whether the basic training parameters and the advanced training parameters meet the adaptation requirements of the current rehabilitation training stage. The adaptation requirements are usually set based on the patient's muscle strength, joint range of motion, and action completion degree. For example, for a patient in the early stage of rehabilitation, if the basic training parameters are wrist joint movement amplitude 0° to 30° and acceleration 1° / s², and the advanced training parameters are amplitude 30° to 60° and acceleration 2° / s², the system will evaluate whether the patient can complete the action corresponding to the advanced parameters. If the patient's muscle strength is weak and cannot complete the action of the advanced amplitude, the system determines that the adaptation requirements are not met, and the basic rehabilitation control instruction is generated directly based on the limit value adjustment of the joint movement amplitude and the acceleration, ensuring that the training action is within the range that the patient can safely withstand.

[0047] If the system determines that the parameters meet the adaptation requirements, the adaptation coefficient of the basic training parameters and the advanced training parameters is further calculated, and the calculation formula is: , which reflects the matching degree of the current ability of the patient and the requirement of the advanced training, and the value range is between 0 and 1. For example, the patient's action completion score is 80, and the advanced parameter action standard score is 100, so the adaptation coefficient is 0.8. Then, the system will use the adaptation coefficient as the optimization weight to iteratively correct the joint motion amplitude and acceleration, and the iterative correction formula is: , where P k is the parameter value of the kth iteration, and P adv is the advanced training parameter value. Taking the wrist joint motion amplitude as an example, the initial basic parameter is 0° to 30°, and the advanced parameter is 0° to 60°. In the first iteration, the new motion amplitude is 30+0.8×(60 30)=54, i.e. 0° to 54°; and the acceleration is initially 1° / s², the advanced is 2° / s², and the corrected is 1+0.8×(2 1)=1.8° / s2. The system will continue to iteratively correct until the parameters converge to the optimal training parameters, and finally generate the basic rehabilitation control instruction.

[0048] Through the adaptability judgment and weighted iterative optimization, it is ensured that the generated control instruction can adapt to the current rehabilitation ability of the patient and gradually guide the patient to transition to higher intensity training, thereby improving the safety and effectiveness of the basic rehabilitation training.

[0049] The upper limb robot cooperative rehabilitation control instruction is generated according to the spectrum characteristics and upper limb motion trajectory parameters, specifically including the following steps: determine whether the spectrum characteristics corresponding to the multi-muscle group linkage feature exist a frequency band superposition effect; if the frequency band superposition effect exists, determine the muscle cooperative force level according to the energy distribution of the superimposed frequency band to obtain the first cooperative control parameter of the joint linkage motion; if the frequency band superposition effect does not exist, determine the muscle force priority according to the spectrum characteristic proportion of each muscle group to obtain the second cooperative control parameter of the joint linkage motion; fuse and match the first cooperative control parameter and the second cooperative control parameter with the upper limb motion trajectory parameter to generate the cooperative rehabilitation control instruction.

[0050] Firstly, the system will determine whether the spectrum characteristics corresponding to the multi-muscle group linkage feature exist a frequency band superposition effect. The frequency band superposition effect refers to the phenomenon that the electromyographic signals of different muscle groups appear energy superposition in the same frequency interval, which usually indicates that multiple muscles are in synchronous force. Taking the patient's compound action of lifting the hand to touch the shoulder as an example, the electromyographic signals of the deltoid muscle and the biceps brachii muscle appear energy superposition in the frequency band of 30Hz to 50Hz, so the system determines that there is a frequency band superposition effect. At this time, the system will extract the energy distribution of the superimposed frequency band, calculate the energy proportion of each muscle in the superimposed frequency band, and the calculation formula is: wherein, P i (f) is the power spectral density of the ith muscle at frequency f, M is the number of muscles involved in the linkage, f1 to f2 are the frequency ranges of the superimposed frequency bands. In this way, the muscle synergistic force level is determined, for example, the deltoid muscle energy accounts for 60%, and the biceps brachii muscle accounts for 40%, it is determined that the deltoid muscle is the main force muscle, and the biceps brachii muscle is the auxiliary force muscle, and then the first synergistic control parameter of the joint linkage movement is obtained, including the shoulder joint abduction speed and the elbow joint flexion speed.

[0051] If the frequency band superimposition effect does not exist in the spectrum feature, the system determines the muscle force priority according to the spectrum feature proportion of each muscle group. The spectrum feature proportion refers to the proportion of the spectrum energy of a single muscle group in the overall signal energy, and the calculation formula is: wherein F is the total frequency point number of the spectrum, and the higher the proportion, the higher the muscle force priority. For example, when the patient completes the fist flexion action, the spectrum feature proportion of the biceps brachii muscle is 70%, and the flexor muscle is 30%, the system determines that the biceps brachii muscle is a high-priority force muscle, and the flexor muscle is a low-priority force muscle, and then the second synergistic control parameter is obtained, including the elbow joint flexion amplitude and the wrist joint grip strength.

[0052] Finally, the system will fuse and match the first synergistic control parameter or the second synergistic control parameter with the preset upper limb movement trajectory parameter. Taking the upper limb movement trajectory parameter as the benchmark, the movement amplitude, speed and torque of each joint are adjusted in combination with the synergistic control parameter to generate the final synergistic rehabilitation control instruction. For example, in the action of lifting the hand to touch the shoulder, the system will fuse the shoulder joint abduction speed and the elbow joint flexion speed in the first synergistic control parameter with the preset trajectory parameters such as the shoulder joint abduction angle 120° and the elbow joint flexion angle 90°, adjust the movement timing and intensity of each joint, and drive the robot to complete the precise synergistic action, ensuring that the linkage of multiple muscle groups not only conforms to the physiological law, but also adapts to the rehabilitation training needs of the patient.

[0053] According to the energy distribution of the superimposed frequency band, the muscle synergistic force level is determined, and the first synergistic control parameter of the joint linkage movement is obtained, including: determining whether the energy distribution of the superimposed frequency band exceeds the preset synergistic force threshold value; if the synergistic force threshold value is not exceeded, it is determined that the current muscle synergistic force level is the basic linkage level; the preset basic angle synergistic value and the basic speed synergistic value are taken as the angle synergistic value and the speed synergistic value of the joint linkage movement; if the synergistic force threshold value is exceeded, the energy peak value and the duration information of the superimposed frequency band are obtained, and the energy peak value and the duration information are compared with the electromyographic energy synergy level mapping table to obtain the angle synergistic value and the speed synergistic value of the joint linkage movement; The angle synergy value and the speed synergy value are marked as first synergy control parameters.

[0054] Firstly, the system calculates the total energy of the superimposed frequency band, and the calculation formula is: wherein P i (f) is the power spectral density of the ith muscle at frequency f, M is the number of muscles participating in linkage, f1 to f2 are the frequency range of the superimposed frequency band. Subsequently, the system compares the total energy with the preset synergy threshold. Taking the patient's action of lifting the hand to touch the shoulder as an example, if the superimposed frequency band is 30Hz to 50Hz, the calculated total energy is 0.4mV², and the preset synergy threshold is 0.5mV², at this time the system determines that the energy does not exceed the threshold, and the current muscle synergy level is the basic linkage level. The system directly calls the preset basic angle synergy value and basic speed synergy value, for example, the angle synergy value of the shoulder joint and the elbow joint is 1.2:1, and the speed synergy value is 0.8:1. These values are marked as first synergy control parameters, which are used to guide the basic linkage movement of the joint.

[0055] If the total energy of the superimposed frequency band exceeds the synergy threshold, for example, the calculated total energy is 0.6mV², the system will further extract the energy peak value and duration information of the superimposed frequency band. The energy peak value refers to the maximum energy value in the superimposed frequency band, and the duration refers to the time when the energy maintains above 80% of the peak value. The system will compare these two information with the preset myoelectric energy synergy level mapping table, for example, when the energy peak value is 0.3mV² and the duration is 1.2 seconds, the mapping table will return the corresponding angle synergy value as 1.5:1 and the speed synergy value as 1.1:1. These values will be marked as first synergy control parameters, which are used to guide the high-intensity linkage movement of the joint.

[0056] Through energy threshold judgment and mapping table matching, the muscle synergy level is accurately determined, and the generated first synergy control parameters can directly guide the angle and speed synergy of the robot joint, ensuring the accuracy and safety of the multi-muscle group linkage action.

[0057] According to the frequency spectrum characteristic proportion of each muscle group, the muscle force priority is determined, and the second synergy control parameters of the joint linkage movement are obtained, including: The frequency spectrum characteristic proportion of each muscle group is compared with the myoelectric spectrum force priority mapping table to obtain the force priority coefficient corresponding to each muscle group. Based on the force priority coefficient, the synthesized synergy angle and the synthesized synergy speed of the joint linkage movement are calculated, and the synthesized synergy angle and the synthesized synergy speed are marked as second synergy control parameters.

[0058] Firstly, the system calculates the spectral feature proportion of each muscle group, which reflects the contribution of the spectral energy of a single muscle to the overall signal. The higher the proportion, the greater the muscle force weight in the action. For example, the patient completes the compound action of clenching the fist and bending the elbow. The system calculates that the spectral feature proportion of biceps brachii is 70%, and that of flexor digitorum is 30%. Then, the system maps this proportion to the preset electromyogram spectrum force priority mapping table to obtain the corresponding force priority coefficient. For example, the force priority coefficient of biceps brachii is 0.7, and that of flexor digitorum is 0.3.

[0059] Next, the system calculates the synthetic synergy angle and synthetic synergy speed of the joint linkage motion based on the force priority coefficient. The calculation formula of the synthetic synergy angle is: wherein θ i is the joint angle corresponding to the ith muscle, and w i is the force priority coefficient of the muscle. The calculation formula of the synthetic synergy speed is: wherein v i is the joint motion speed corresponding to the ith muscle. In the clenching and elbow bending action, the elbow joint angle corresponding to biceps brachii is 90°, and the speed is 5° / s. The wrist joint angle corresponding to flexor digitorum is 45°, and the speed is 3° / s. Substituting the formula can obtain: The synthetic synergy angle is θ syn = 90 x 0.7 + 45 x 0.3 = 76.5°. The synthetic synergy speed is v syn = 5 x 0.7 + 3 x 0.3 = 4.4° / s. The system marks the synthetic synergy angle and synthetic synergy speed as the second synergy control parameter, which is directly used to guide the linkage motion of the robot joint, ensuring that the force of multiple muscle groups meets the priority order and accurately adapts to the rehabilitation training needs of the patient.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some technical features. These modifications or substitutions do not change the essence of the corresponding technical solutions beyond the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for controlling upper limb robotic rehabilitation training based on electromyographic signals, characterized in that, The method includes the following steps: Electromyographic (EMG) signals of target muscle groups in the upper limbs of the human body under different movement modes were collected, an EMG signal feature library was constructed, and EMG signals that matched rehabilitation training movements were screened from the EMG signal feature library and marked as training trigger signals. Extract the signal effective value, spectral characteristics and action correlation of the training trigger signal, determine whether feature enhancement processing is needed for at least two sets of training trigger signals based on the signal effective value and spectral characteristics, and mark the corresponding training trigger signal as the control command generation signal; Acquire the upper limb motion trajectory parameters corresponding to the control command generation signal; Determine whether there are multi-muscle group linkage features in the control command generation signal based on the degree of motion correlation and muscle group cooperation; If the control command generation signal corresponds only to the motion characteristics of a single muscle group, then the basic rehabilitation control command for the upper limb robot is generated based on the effective value of the signal and the upper limb motion trajectory parameters. If the control command generation signal corresponds to the multi-muscle group linkage characteristics, then the collaborative rehabilitation control command for the upper limb robot is generated based on the spectral characteristics and upper limb motion trajectory parameters. Drive the upper limb robot to complete rehabilitation training movements according to basic rehabilitation control instructions or collaborative rehabilitation control instructions.

2. The upper limb robotic rehabilitation training control method based on electromyography signals according to claim 1, characterized in that, Based on the signal's effective value and spectral characteristics, determine whether feature enhancement processing is needed for at least two sets of training trigger signals, and mark the corresponding training trigger signals as control command generation signals, including: If the effective value and spectral characteristics of the signal corresponding to the same action mode are fully presented in a single set of training trigger signals, then the training trigger signal is directly marked as the control command generation signal; If the effective value and spectral characteristics of the signal corresponding to the same action mode are scattered in at least two sets of training trigger signals, then the corresponding training trigger signals are decomposed in the frequency domain to extract the signal components containing the core features. Based on the muscle exertion patterns corresponding to the action patterns, noise reduction and enhancement processing is performed on each signal component, and the processed signal components are reconstructed to form control command generation signals.

3. The upper limb robotic rehabilitation training control method based on electromyography signals according to claim 2, characterized in that, Based on the muscle exertion patterns corresponding to the action patterns, noise reduction and enhancement processing are performed on each signal component. The processed signal components are then reconstructed to form the control command generation signal, including: Determine whether there are overlapping frequency bands in the frequency domain characteristics of at least two signal components; If there are overlapping frequency bands, the corresponding signal components are marked as matching signal components, and the signal is reconstructed based on the energy characteristics of the overlapping frequency bands to form the control command generation signal. If there are no overlapping frequency bands, the frequency domain characteristics of each signal component are standardized. If the characteristic parameters of each signal component after standardization meet the characteristic requirements of the action mode, the corresponding signal components are integrated and reconstructed to form the control command generation signal.

4. The upper limb robotic rehabilitation training control method based on electromyography signals according to claim 3, characterized in that, The presence of multi-muscle group linkage features in the control command generation signal is determined based on the correlation between movements and the coordination of muscle groups, including: If the motion correlation of the control command generated signal corresponds only to a single muscle group, then it is determined that there is no multi-muscle group linkage feature in the control command generated signal. If the action correlation of the control command generated signal corresponds to at least two muscle groups, then the temporal synergy of the electromyographic signals of each muscle group is determined based on the muscle group cooperation relationship. If the electromyographic signals of different muscle groups do not exhibit temporal coordination, it is determined that the control command generation signal does not contain multi-muscle group linkage characteristics. If the electromyographic signals of different muscle groups exhibit temporal coordination, then the control command generation signal is determined to contain multi-muscle group linkage characteristics.

5. The upper limb robotic rehabilitation training control method based on electromyography signals according to claim 4, characterized in that, If the control command generation signal corresponds only to the linkage characteristics of a single muscle group, then basic rehabilitation control commands for the upper limb robot are generated based on the effective value of the signal and the upper limb motion trajectory parameters, including: By comparing the characteristic parameters of the control command generated signal with the electromyographic feature motion trajectory mapping table, the amplitude of robot joint motion and motion acceleration corresponding to the muscle group can be obtained. Joint range of motion and acceleration are divided into basic training parameters and advanced training parameters. Based on the compatibility of basic training parameters and advanced training parameters, the range of motion and acceleration of joint movement are iteratively optimized to generate basic rehabilitation control commands.

6. The upper limb robotic rehabilitation training control method based on electromyography signals according to claim 5, characterized in that, Based on the compatibility of basic and advanced training parameters, the range of motion and acceleration of joint movement are iteratively optimized to generate basic rehabilitation control commands, including: Determine whether the basic and advanced training parameters meet the adaptation requirements of the rehabilitation training stage; If the adaptation requirements are not met, basic rehabilitation control instructions will be generated by adjusting the limits of joint range of motion and acceleration based on the basic training parameters. If the adaptation requirements are met, the adaptation coefficient between the basic training parameters and the advanced training parameters is calculated. Using the adaptation coefficient as the optimization weight, the joint range of motion and acceleration are iteratively corrected to obtain the optimal training parameters. Basic rehabilitation control instructions are then generated based on the optimal training parameters.

7. The upper limb robotic rehabilitation training control method based on electromyographic signals according to claim 6, characterized in that, Based on the aforementioned spectral characteristics and upper limb motion trajectory parameters, collaborative rehabilitation control commands for the upper limb robot are generated, specifically including the following steps: Determine whether there is a frequency band superposition effect in the spectral features corresponding to the multi-muscle group linkage characteristics; If there is a frequency band superposition effect, the level of muscle synergistic force exertion is determined based on the energy distribution of the superimposed frequency band, and the first synergistic control parameter of joint linkage movement is obtained. If there is no frequency band superposition effect, the priority of muscle exertion is determined based on the proportion of the spectral characteristics of each muscle group, and the second collaborative control parameter of joint linkage movement is obtained. The first and second collaborative control parameters are fused and matched with the upper limb movement trajectory parameters to generate collaborative rehabilitation control instructions.

8. The upper limb robotic rehabilitation training control method based on electromyography signals according to claim 7, characterized in that, Based on the energy distribution of the superimposed frequency band, the level of muscle synergy is determined, and the first synergistic control parameters of joint linkage motion are obtained, including: Determine whether the energy distribution of the superimposed frequency band exceeds the preset collaborative force threshold; If the coordinated force threshold is not exceeded, the current muscle coordinated force level is determined to be the basic linkage level; The preset basic angle coordination value and basic velocity coordination value are used as the angle coordination value and velocity coordination value of the joint linkage motion; If the coordinated force threshold is exceeded, the energy peak value and duration information of the superimposed frequency band are obtained. The energy peak value and duration information are compared with the electromyographic energy coordination level mapping table to obtain the angle coordination value and velocity coordination value of the joint linkage movement. The angle coordination value and the velocity coordination value are marked as the first coordination control parameters.

9. The upper limb robotic rehabilitation training control method based on electromyographic signals according to claim 8, characterized in that, Based on the proportion of spectral characteristics of each muscle group, the muscle force priority is determined, and the second coordinated control parameters for joint linkage motion are obtained, including: By comparing the proportion of spectral characteristics of each muscle group with the electromyographic spectrum force priority mapping table, the force priority coefficient corresponding to each muscle group can be obtained. Based on the force priority coefficient, the synthetic coordination angle and synthetic coordination velocity of the joint linkage motion are calculated, and the synthetic coordination angle and synthetic coordination velocity are marked as the second coordination control parameters.

Images