Induced muscle strength rehabilitation training method and device

Abstract

The present application discloses an induced muscle rehabilitation training method and device, which includes obtaining the patient's EEG signals and electromyographic signals, obtaining a first movement intention based on the EEG signals, obtaining a second movement intention based on the electromyographic signals, and determining the movement intention based on the first movement intention and the second movement intention; determining the patient's fatigue level based on the electromyographic signals; and determining the corresponding training mode based on the movement intention and fatigue level, the training mode includes an induced training mode and an on-demand auxiliary training mode. The patient's movement intention is determined by the fusion of EEG signals and electromyographic signals, which improves the accuracy and reliability of movement intention recognition; based on the patient's movement intention and fatigue state, the corresponding training mode is determined for training, which can not only induce the patient to appropriately increase the intensity of exercise, speed up the rehabilitation treatment process, and improve the treatment effect, but also make adaptive adjustments in time when the patient is in poor condition, avoid patient injuries, and improve the safety of training.

Description

Technical Field

[0001] The present technical solution mainly relates to the field of rehabilitation training, and in particular to an induced muscle strength rehabilitation training method and device. Background Art

[0002] Stroke has become the number one killer of Chinese residents. On average, one person suffers a first-time or recurrent stroke every 10 seconds. After the onset of the disease, it often causes motor dysfunction in patients, requiring a large amount of limb movement rehabilitation training to rebuild the patient's motor function and improve his or her living ability and quality of life.

[0003] With the development of rehabilitation medicine, rehabilitation robot training systems that integrate mechanical design, computer technology, sensor technology, signal processing and control technology have begun to be increasingly used in the field of rehabilitation training, making patients' rehabilitation training more scientific and effective, and greatly reducing the pressure on rehabilitation training physicians. The control strategy of rehabilitation robots has an important impact on the rehabilitation training effect of patients. In order to adapt to the individual differences between different patients, adaptive on-demand auxiliary control strategy has become an important research direction, that is, rehabilitation robots adjust the auxiliary force or movement trajectory in real time according to the patient's movement intention and movement performance, helping patients to make up the required auxiliary force in real time when the movement force is insufficient. However, it is difficult for patients to maintain a strong willingness to train. When the patient's willingness to train decreases or for patients with weak willingness to train, this control strategy can easily lead to patients' slackness and decreased movement strength. They mainly rely on the auxiliary force provided by the rehabilitation robot to exercise, which cannot achieve the expected training effect, and the patient's ability potential cannot be stimulated. The increase in training difficulty is delayed, which makes the rehabilitation training effect progress slowly.

[0004] Therefore, there is an urgent need for an inductive muscle strength rehabilitation training method to solve the problems existing in the above-mentioned prior art. Summary of the invention

[0005] The embodiments of the present application provide an induced muscle strength rehabilitation training method and device to solve the technical problems in the prior art of delayed improvement of patient training difficulty and slow progress of rehabilitation training effect.

[0006] In a first aspect, an embodiment of the present application provides an induced muscle rehabilitation training method, comprising respectively acquiring an electroencephalogram (EEG) signal and an electromyographic (EMG) signal of a patient; identifying movement intentions based on the EEG signals to obtain a first movement intention; identifying movement intentions based on the electromyographic (EMG) signals to obtain a second movement intention; determining movement intentions based on the first movement intention and the second movement intention, wherein the movement intentions include movement intentions and no movement intentions; evaluating fatigue based on the electromyographic (EMG) signals to determine the patient's fatigue level, wherein the fatigue levels include relaxed, general, and fatigued; determining a corresponding training mode based on the movement intentions and fatigue levels for training, wherein the training modes include an induced training mode and an on-demand auxiliary training mode.

[0007] In combination with the first aspect, in a possible implementation method, before obtaining the first movement intention by identifying the movement intention based on the EEG signal and obtaining the second movement intention by identifying the movement intention based on the EMG signal, it also includes: before the training task starts, allowing the patient to complete multiple cycles of trial movements in an on-demand assisted training mode; using the first machine learning model to identify the movement intention of the EEG signal to obtain the first initial movement intention of each cycle; using the second machine learning model to identify the movement intention of the EMG signal to obtain the second initial movement intention of each cycle; after each cycle of trial movement is completed, obtaining the patient's true movement intention for each cycle; training the first machine learning model based on the true movement intention of each cycle and the first initial movement intention of the corresponding cycle; training the second machine learning model based on the true movement intention of each cycle and the second initial movement intention of the corresponding cycle.

[0008] In combination with the first aspect, in a possible implementation method, the identification of movement intention based on the EEG signal to obtain the first movement intention includes: performing signal preprocessing and feature extraction on the collected original EEG signal in sequence, and then inputting the signal into the first machine learning model to obtain the first movement intention; the identification of movement intention based on the electromyography signal to obtain the second movement intention includes: performing signal preprocessing and feature extraction on the collected original electromyography signal in sequence, and then inputting the signal into the second machine learning model to obtain the second movement intention.

[0009] In combination with the first aspect, in a possible implementation method, the acquisition of the electromyographic signal includes: acquiring electromyographic signals of multiple channels; acquiring preset training actions; determining the weights of the electromyographic signals of each channel based on the preset training actions; and obtaining the electromyographic signal according to the electromyographic signals of each channel and their weights.

[0010] In combination with the first aspect, in a possible implementation method, determining the movement intention based on the first movement intention and the second movement intention includes: assigning weights to the first movement intention and the second movement intention respectively according to the classification accuracy of the electromyographic signal and the electroencephalographic signal in the trial movement; and performing weighted summation of the first movement intention and the second movement intention according to the weights to determine the movement intention.

[0011] In combination with the first aspect, in a possible implementation method, the corresponding training mode is determined based on the exercise intention and fatigue level, including: the induced training mode includes a regional induced training mode and a periodic induced training mode; when the fatigue level is in a relaxed state, if the patient has no exercise intention, the regional induced training mode is adopted, and if the patient has exercise intention, the periodic induced training mode is adopted; when the fatigue level is in a general state, the on-demand auxiliary training mode is adopted; when the fatigue level is in a fatigue state, the training is stopped.

[0012] In combination with the first aspect, in a possible implementation, the regional induction training mode includes: within a movement cycle, when the patient's limb moves to the first area, the driving force of the muscle rehabilitation robot is changed from on-demand assistance to quantitative compensation force, and the quantitative compensation force F is less than the patient's average on-demand assistance compensation force F0; if the pause time during the quantitative compensation movement does not exceed the preset time, the quantitative compensation force F is maintained and the first area is gradually expanded to the entire limb movement within the subsequent first preset cycles, that is, entering the periodic induction training mode, if the pause time during the quantitative compensation movement exceeds the preset time, the on-demand assistance is re-triggered; the periodic induction training mode includes: within a complete movement cycle, the driving force of the muscle rehabilitation robot is quantitative compensation force, and the quantitative compensation force F is less than the patient's average on-demand assistance compensation force F0; if the pause time during the quantitative compensation movement does not exceed the preset time, the quantitative compensation force F is maintained for the second preset cycle and then reduced to the preset threshold value ∆F, if the pause time during the quantitative compensation movement exceeds the preset time, the on-demand assistance is re-triggered.

[0013] In combination with the first aspect, in a possible implementation manner, the first area is an area where the limbs exert force better.

[0014] In combination with the first aspect, in a possible implementation, if the pause time during the quantitative force compensation movement exceeds the preset time, the on-demand assistance is re-triggered, including: the single pause time exceeds the first preset time; or the cumulative pause time within the third preset cycle exceeds the second preset time; the second preset time is greater than the first preset time.

[0015] In the second aspect, an embodiment of the present application provides an inductive muscle rehabilitation training device, including: an EEG acquisition module for acquiring a patient's EEG signal; an EMG acquisition module for acquiring a patient's EMG signal; a movement intention recognition module for identifying a movement intention based on the EEG signal to obtain a first movement intention, identifying a movement intention based on the EMG signal to obtain a second movement intention, and then determining a movement intention based on the first movement intention and the second movement intention, wherein the movement intention includes movement intention and no movement intention; a fatigue assessment module for assessing fatigue based on the EMG signal to determine the patient's fatigue level, wherein the fatigue level includes relaxed, general, and fatigue; a training module for determining a corresponding training mode for training based on the movement intention and fatigue level, wherein the training mode includes an induced training mode and an on-demand auxiliary training mode.

[0016] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects:

[0017] The embodiment of the present application provides an inductive muscle rehabilitation training method, which determines the patient's movement intention by the decision fusion of EEG signals and electromyographic signals, thereby improving the accuracy and reliability of movement intention recognition. Based on the patient's movement intention and fatigue state, the corresponding training mode is determined, which can not only induce the patient to appropriately increase the intensity of exercise when the patient is in good condition, maintain the patient's participation and excitement, speed up the rehabilitation treatment process, and improve the treatment effect; but also can make adaptive adjustments in time when the patient is in poor condition, avoid excessive training and cause injury to the patient, and improve the safety of training. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following is a brief introduction to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 A flowchart of an inductive muscle strength rehabilitation training method provided in an embodiment of the present application;

[0020] Figure 2 A schematic diagram of the structure of an inductive muscle strength rehabilitation training device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0022] The following describes some of the techniques involved in the embodiments of the present application to facilitate understanding, and they should be considered as merely exemplary. Therefore, it should be appreciated by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for the sake of clarity and conciseness, some descriptions of well-known functions and structures are omitted in the following description.

[0023] The present application embodiment provides an inductive muscle strength rehabilitation training method, such as Figure 1 As shown, the method includes steps S101 to S106. Among them, Figure 1 This is only an execution order shown in the embodiment of the present application, and does not represent the only execution order of the induction muscle strength rehabilitation training method. If the final result can be achieved, Figure 1 The steps shown may be performed in parallel or reversed.

[0024] S101: Obtain the patient's electroencephalogram (EEG) signal and electromyography (EMG) signal respectively.

[0025] It should be noted that in order to improve the anti-interference ability of the signal and accurately analyze the signal, the acquisition of EEG signals and electromyographic signals can adopt a multi-channel mode. Specifically, when acquiring electromyographic signals for analysis, the weights of the electromyographic signals of each channel need to be adjusted based on the action to be trained and then weighted summed, and the summed result is used as the electromyographic signal for the subsequent analysis process. The specific process of acquiring the electromyographic signal includes: respectively acquiring the electromyographic signals of multiple channels and the preset training action, and then determining the weights of the electromyographic signals of each channel based on the preset training action, and finally obtaining the electromyographic signal according to the electromyographic signals of each channel and their weights. The electromyographic signal is a very weak signal with an amplitude of only between uV and mV levels, which is easily disturbed by the surrounding environment. The training action is different, and the muscle blocks that exert force are also different. The muscle fibers contract quickly or forcefully, and the amplitude of the electromyographic signal is higher, and the anti-interference ability is stronger. Improving its weight is conducive to improving the accuracy of movement intention recognition.

[0026] Exemplarily, in a specific embodiment of the present application, when conducting rehabilitation training for the upper limbs, multi-channel electromyographic signals are respectively collected from the biceps long head, deltoid-clavicle, triceps long head and deltoid-scapula, wherein during the pulling exercise training, the weights of the biceps long head and deltoid-clavicle channels are higher than the weights of the triceps long head and deltoid-scapula channels; during the pushing exercise training, the weights of the triceps long head and deltoid-scapula channels are higher than the weights of the biceps long head and deltoid-clavicle channels.

[0027] S102: Identify movement intention based on the EEG signal to obtain a first movement intention.

[0028] Specifically, for the collected electroencephalogram (EEG) signal, wavelet transform is used to filter the electroencephalogram (EEG) signal to filter out noises such as electromyographic signals, electrocardiographic signals, and power frequency noise, and the electroencephalogram (EEG) signal is extracted therefrom for baseline correction and bad track processing. The processed electroencephalogram (EEG) signal is segmented according to the time window segmentation scheme, and the characteristic values ​​of each electroencephalogram (EEG) signal unit after segmentation are calculated. After forming a feature group, the feature group is input into the first machine learning model trained in advance using the electroencephalogram (EEG) signal sample to obtain the result of motion intention recognition. The present application does not limit the first machine learning model, which can be a random forest model or a support vector machine model. The classification result 0 represents no motion intention, and 1 represents motion intention.

[0029] S103: Identify the movement intention based on the electromyographic signal to obtain a second movement intention.

[0030] Studies have shown that myoelectric signals are generated tens of microseconds before the start of an action, so myoelectric signals have a certain predictive power for human body movements. A surface electromyographic sensor (such as a sEMG sensor) is used to collect myoelectric signals, and the collected raw myoelectric signals are sequentially preprocessed and feature extracted before being input into a second machine learning model for identification of movement intention. This application also does not limit the second machine learning model, which may be a random forest model or a support vector machine model. The classification result 0 represents no movement intention, and 1 represents movement intention. It should be noted that the above-mentioned first machine learning model and the second machine learning model may be the same or different.

[0031] The original electromyographic signal often contains noise and artifacts, so the signal preprocessing process includes filtering, amplification, rectification and smoothing to obtain clean electromyographic signals and improve their signal-to-noise ratio and stability. Filtering can remove high-frequency noise and low-frequency interference, amplification can enhance signal strength, and rectification and smoothing can make the signal smoother and easier to analyze. The preprocessed electromyographic signal needs to be feature extracted. Feature extraction is the process of extracting effective information that can reflect the state of muscle activity from the signal. Features include time domain features and frequency domain features, among which time domain features use mean, variance, and root mean square, and frequency domain features use power spectrum density, spectrum peak frequency, etc. The extracted feature groups are input into the pre-trained electromyographic signal support vector machine (SVM) model or random forest model for motion intention recognition.

[0032] S104: Determine a movement intention based on the first movement intention and the second movement intention, where the movement intention includes movement intention and no movement intention.

[0033] Specifically, according to the classification accuracy of the electromyographic signal and the electroencephalographic signal in the test movement, weights are respectively assigned to the first movement intention and the second movement intention, and then the first movement intention and the second movement intention are weightedly summed according to the assigned weights to determine the movement intention.

[0034] Due to the differences between individual patients, while improving the universality of the model, the applicability to individuals should also be taken into account. Therefore, in a specific embodiment of the present application, before the training task begins, the patient is asked to complete multiple cycles of test exercises in the on-demand auxiliary training mode; the first machine learning model is used to identify the movement intention of the EEG signal to obtain the first initial movement intention of each cycle; the second machine learning model is used to identify the movement intention of the electromyographic signal to obtain the second initial movement intention of each cycle; after each cycle of test exercise is completed, the patient's true movement intention for each cycle is obtained; based on the true movement intention of each cycle and the first initial movement intention of the corresponding cycle, the first machine learning model is trained, and the classification accuracy of the EEG signal in the test exercise is obtained; based on the true movement intention of each cycle and the second initial movement intention of the corresponding cycle, the second machine learning model is trained, and the classification accuracy of the electromyographic signal in the test exercise is obtained.

[0035] It should be noted that the above-mentioned training of the first machine learning model based on the real movement intention of each cycle and the first initial movement intention of the corresponding cycle is mainly to optimize the reward function in the first machine learning model, map the optimized reward function with the patient and store it in the data pool. When the patient undergoes subsequent rehabilitation training, the corresponding reward function can be called, thereby improving the applicability of the first machine learning model to different individuals. The optimization process of the second machine learning model is the same, so I will not go into details.

[0036] In addition, during the patient's training, there are active feedback and passive feedback. Active feedback refers to randomly or periodically selecting a training cycle during the training process to guide the patient to feedback the true movement intention of the cycle; passive feedback means that the patient can trigger feedback at any time during the training process to feedback the true movement intention of the current training cycle. Among them, the triggering method can include any one of voice, emergency button or specific actions (such as blinking, shaking head, etc.). All active feedback and passive feedback data are used to optimize the reward function for the patient, so that the corresponding machine learning model has good growth potential to adapt to the different recovery stages of patients.

[0037] When receiving active feedback results, the patient's real movement intention can be mapped with the collected data and recognition results of the corresponding period and stored for optimization of the reward function for the patient in the future; when receiving passive feedback results, the patient's real movement intention is compared with the recognized movement intention to see if they are consistent. If they are inconsistent, the corresponding machine learning model is optimized in combination with the pre-set active feedback results, and after the model optimization is completed, it is updated to the optimized model for subsequent training. In this way, the model is corrected in a timely manner.

[0038] S105: Evaluate the fatigue degree based on the electromyographic signal to determine the patient's fatigue level, which includes relaxed, general and fatigue. De-noise the collected electromyographic signal and extract the characteristic value, which includes the time domain characteristic value and the frequency domain characteristic value, and then input the extracted characteristic value into the convolutional neural network model, namely the CNN (Convolutional Neural Network) network model to classify the fatigue degree.

[0039] S106: Based on the exercise intention and fatigue level, determine a corresponding training mode for training.

[0040] It should be noted that the above-mentioned training modes include an induced training mode and an on-demand assisted training mode. When the fatigue level is in a relaxed state, the induced training mode is adopted; when the fatigue level is in a general state, the on-demand assisted training mode is adopted; when the fatigue level is in a fatigue state, the training is stopped. Specifically, the induced training mode includes a regional induced training mode and a periodic induced training mode; when the fatigue level is in a relaxed state, if the patient has no intention of exercising, the regional induced training mode is adopted, and if the patient has the intention of exercising, the periodic induced training mode is adopted. When the patient adopts the regional induced training mode, the transition to the periodic induced training mode is made by gradually enlarging the induced area.

[0041] The regional induction training mode includes: within a movement cycle, when the patient's limb moves to the first area, the driving force of the muscle rehabilitation robot is changed from on-demand assistance to quantitative supplementary force, and the quantitative supplementary force F is less than the patient's average on-demand assistance supplementary force F0; if the pause time during the quantitative supplementary force movement does not exceed the preset time, then the quantitative supplementary force F is maintained and the first area is gradually expanded to the entire limb movement within the first preset cycle, that is, entering the periodic induction training mode, if the pause time during the quantitative supplementary force movement exceeds the preset time, the on-demand assistance is re-triggered. Specifically, the quantitative supplementary force in this application refers to the method of providing supplementary force, and the quantitative supplementary force refers to the size of the supplementary force required to be provided when the quantitative supplementary force method is adopted.

[0042] It should be noted that the above-mentioned on-demand auxiliary average supplementary force F0 may be an average value of the on-demand auxiliary supplementary force in the previous adjacent preset cycles, and the value range of the adjacent preset cycles is 1-10. For example, the previous 5 adjacent cycles may be selected.

[0043] The periodic induction training mode includes: in a complete exercise cycle, the driving force of the muscle rehabilitation robot is a quantitative supplementary force, and the quantitative supplementary force F is less than the average supplementary force F0 of the patient's on-demand assistance; if the pause time during the quantitative supplementary force exercise does not exceed the preset time, the quantitative supplementary force F is maintained for the second preset cycle and then reduced by the preset threshold ∆F; if the pause time during the quantitative supplementary force exercise exceeds the preset time, the on-demand assistance is re-triggered.

[0044] It should be noted that in the present application, a force sensor is arranged at the handheld end of the muscle rehabilitation robot to collect the human-computer interaction force signal, and a PID and impedance control algorithm is adopted to control the force required for compensation during the movement of the rehabilitation machine according to the human-computer interaction force, and the actual output supplementary force is determined based on the current on-demand assistance or quantitative force compensation mode.

[0045] Specifically, the first area is the area where the limbs exert force better. For example, when the arm is pushed outward, the angle between the upper arm and the forearm is about 90°, which is the best force area. If the pause time during the quantitative force supplementation exercise exceeds the preset time, the on-demand assistance will be re-triggered, including: the single pause time exceeds the first preset time; or the cumulative pause time within the third preset cycle exceeds the second preset time; the second preset time is greater than the first preset time.

[0046] The present application also provides an inductive muscle strength rehabilitation training device 200, the structural diagram of which is shown in FIG. Figure 2 As shown, including:

[0047] The EEG acquisition module 201 is used to acquire the patient's EEG signal;

[0048] The electromyography acquisition module 202 is used to obtain the electromyography signal of the patient;

[0049] A movement intention recognition module 203 is used to obtain a first movement intention by recognizing the movement intention based on the EEG signal, obtain a second movement intention by recognizing the movement intention based on the EMG signal, and then determine the movement intention based on the first movement intention and the second movement intention, where the movement intention includes movement intention and no movement intention;

[0050] A fatigue assessment module 204 is used to assess fatigue based on electromyographic signals and determine the patient's fatigue level, where the fatigue level includes relaxed, normal and fatigue;

[0051] The training module 205 is used to determine the corresponding training mode for training based on the exercise intention and fatigue level. The training mode includes an induction training mode and an on-demand auxiliary training mode.

[0052] Some modules in the apparatus of the present application may be described in the general context of computer executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media including storage devices.

[0053] The devices or modules described in the above application embodiments can be implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described in various modules according to their functions. When implementing the embodiments of the present application, the functions of each module can be implemented in the same or multiple software and / or hardware. Of course, the module that implements a certain function can also be implemented by combining multiple sub-modules or sub-units.

[0054] The methods, devices or modules described in this application can be implemented in the form of computer-readable program codes. The controller can be implemented in any appropriate manner. For example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program codes (such as software or firmware) that can be executed by the (micro)processor, logic gates, switches, application-specific integrated circuits (English: Application Specific Integrated Circuit, referred to as: ASIC), programmable logic controllers and embedded microcontrollers. Examples of controllers include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in the form of pure computer-readable program codes, the controller can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers and embedded microcontrollers by logically programming the method steps. Therefore, this controller can be considered as a hardware component, and the devices included in it for implementing various functions can also be regarded as structures within the hardware component. Or even, the means for realizing various functions may be regarded as both a software module for realizing the method and a structure within a hardware component.

Claims

1. An inductive muscle strength rehabilitation training method, characterized in that: include: Acquire the patient's electroencephalogram (EEG) signal and electromyography (EMG) signal respectively; Recognizing the movement intention based on the EEG signal to obtain a first movement intention; Recognize the movement intention based on the electromyographic signal to obtain a second movement intention; Determine a movement intention based on the first movement intention and the second movement intention, wherein the movement intention includes a movement intention and a non-movement intention; Evaluate fatigue based on the electromyographic signal to determine the patient's fatigue level, where the fatigue level includes relaxed, general, and fatigue; Based on the exercise intention and fatigue level, determine a corresponding training mode for training, wherein the training mode includes an induction training mode and an on-demand auxiliary training mode; The induction training mode includes a regional induction training mode and a periodic induction training mode; The regional induction training mode includes: in one movement cycle, when the patient's limb moves to the first area, the driving force of the muscle rehabilitation robot is changed from on-demand assistance to quantitative supplementary force, and the quantitative supplementary force F is less than the average on-demand assistance supplementary force F0 of the patient; The periodic induction training mode includes: in a complete movement cycle, the driving force of the muscle rehabilitation robot is a quantitative supplementary force, and the quantitative supplementary force F is less than the average supplementary force F0 of the patient's on-demand assistance.

2. The method according to claim 1, characterized in that: Before obtaining the first movement intention by identifying the movement intention based on the EEG signal and obtaining the second movement intention by identifying the movement intention based on the EMG signal, the method further includes: Before the training task begins, the patient is asked to complete multiple cycles of trial exercise in the on-demand assistance training mode; Using a first machine learning model to perform movement intention recognition on the EEG signal to obtain a first initial movement intention in each cycle; Using a second machine learning model to identify the movement intention of the electromyographic signal to obtain a second initial movement intention of each cycle; After each cycle of trial exercise, the patient's actual exercise intention for each cycle was obtained; Training the first machine learning model based on the actual motion intention of each cycle and the first initial motion intention of the corresponding cycle; The second machine learning model is trained based on the actual motion intention of each cycle and the second initial motion intention of the corresponding cycle.

3. The method according to claim 2, characterized in that The identifying of the movement intention based on the EEG signal to obtain the first movement intention includes: preprocessing and extracting the features of the collected original EEG signal in sequence, and then inputting the preprocessed signal into a first machine learning model to obtain the first movement intention; The identification of movement intention based on the electromyographic signal to obtain the second movement intention includes: preprocessing and feature extracting the collected original electromyographic signal in sequence, and then inputting the signal into the second machine learning model to obtain the second movement intention.

4. The method according to claim 1, characterized in that: The acquisition of the electromyographic signal includes: Acquire multiple channels of myoelectric signals; Get preset training actions; Determining the weight of the muscle electronic signal of each channel based on the preset training action; The myoelectric signal is obtained according to the myoelectric signals of each channel and their weights.

5. The method according to claim 2, characterized in that: The determining of the movement intention based on the first movement intention and the second movement intention includes: According to the classification accuracy of the electromyographic signal and the electroencephalographic signal in the test movement, weights are respectively assigned to the first movement intention and the second movement intention; The first movement intention and the second movement intention are weightedly summed according to the weight to determine the movement intention.

6. The method according to claim 1, characterized in that The determining of a corresponding training mode based on the exercise intention and fatigue level includes: When the fatigue level is in a relaxed state, if the patient has no intention to exercise, the regional induction training mode is used; if the patient has the intention to exercise, the periodic induction training mode is used; When the fatigue level is normal, use the on-demand auxiliary training mode; When the fatigue level reaches the fatigue state, stop training.

7. The method according to claim 6, characterized in that The regional induction training mode also includes: if the pause time during the quantitative force supplementation movement does not exceed the preset time, the quantitative supplementation force F is maintained and the first area is gradually expanded to the entire range of limb movement within the first preset period, that is, the periodic induction training mode is entered, and if the pause time during the quantitative force supplementation movement exceeds the preset time, the on-demand assistance is re-triggered; The periodic induction training mode also includes: if the pause time during the quantitative force supplementation exercise does not exceed the preset time, then the quantitative supplementation force F is maintained for a second preset period and then reduced by a preset threshold value ΔF; if the pause time during the quantitative force supplementation exercise exceeds the preset time, the on-demand assistance is re-triggered.

8. The method according to claim 7, characterized in that The first area is the area where the limbs exert better force.

9. The method according to claim 7, characterized in that If the pause time during the quantitative force supplementation movement exceeds the preset time, the on-demand assistance is re-triggered, including: The single pause time exceeds the first preset time; or The accumulated pause time in the third preset period exceeds the second preset time; The second preset time is greater than the first preset time.

10. An inductive muscle strength rehabilitation training device, characterized in that: include: EEG acquisition module, used to obtain the patient's EEG signal; Myoelectric acquisition module, used to obtain the patient's electromyographic signal; A movement intention recognition module, used for obtaining a first movement intention by recognizing the movement intention based on the EEG signal, obtaining a second movement intention by recognizing the movement intention based on the EMG signal, and then determining the movement intention based on the first movement intention and the second movement intention, wherein the movement intention includes having a movement intention and having no movement intention; A fatigue assessment module, used to assess fatigue based on the electromyographic signal and determine the patient's fatigue level, wherein the fatigue level includes relaxed, general and fatigue; A training module, used to determine a corresponding training mode for training based on the exercise intention and fatigue level, wherein the training mode includes an induction training mode and an on-demand auxiliary training mode; The induction training mode includes a regional induction training mode and a periodic induction training mode; The regional induction training mode includes: in one movement cycle, when the patient's limb moves to the first area, the driving force of the muscle rehabilitation robot is changed from on-demand assistance to quantitative supplementary force, and the quantitative supplementary force F is less than the average on-demand assistance supplementary force F0 of the patient; The periodic induction training mode includes: in a complete movement cycle, the driving force of the muscle rehabilitation robot is a quantitative supplementary force, and the quantitative supplementary force F is less than the average supplementary force F0 of the patient's on-demand assistance.

Images