Hand function rehabilitation robot anti-resistance training method, device and system based on motion intention recognition and storage medium

Abstract

The invention discloses a hand function rehabilitation robot anti-resistance training method, device and system based on motion intention recognition and a storage medium, relates to the technical field of rehabilitation machine learning, and aims to solve the problem that a hand function rehabilitation robot cannot adapt to different patients in a personalized manner in the prior art. The method comprises the steps of collecting forearm electromyographic signals and actual motion angles of a patient; calculating an anti-resistance training evaluation index according to the finger driving force error and the position angle error; outputting and obtaining a resistance coefficient corresponding to the maximum resistance training evaluation index based on the fitting model, namely the optimal resistance coefficient; the optimal resistance is obtained according to the optimal resistance coefficient, and the hand function rehabilitation robot is set to use the optimal resistance to train the patient. According to the method, real-time electromyographic signal detection is combined, the resistance coefficient is dynamically adjusted to obtain the optimal resistance, and high individuation and self-adaptability of rehabilitation are achieved.

Description

Technical Field

[0001] The present invention relates to a human-in-the-loop hand function rehabilitation robot resistance training method, device, system and storage medium based on movement intention recognition, belonging to the technical field of rehabilitation machine learning. Background Art

[0003] In recent years, with the rapid development of robotics technology, including advances in robotic arms, sensors, and control systems, hand function rehabilitation robots have emerged. Resistance training is a key method to enhance muscle strength and endurance, and can help patients with mild to moderate hand dysfunction improve their motor control ability. By applying external resistance during exercise, patients can perform targeted strength training in a safe and controlled environment. Rehabilitation robots achieve this training through sophisticated mechanical design and control algorithms, making it more efficient. The technical system consists of sensor technology, control algorithms, feedback systems, and user interfaces. It monitors the patient's movement status and neural signals in real time, and dynamically adjusts training parameters through intelligent control algorithms. The visual, auditory, or tactile feedback provided by the system helps patients understand their motor performance, thereby enhancing the training effect.

[0004] However, since each patient's recovery condition and needs are different, completely personalized hand function rehabilitation cannot be achieved, which has a certain negative impact on the training effect.

[0005] By interpreting and utilizing the electromyographic signals generated by the nervous system to interact with rehabilitation robots, new possibilities are provided for personalized rehabilitation. Because each patient's recovery situation and needs are different, the human-in-the-loop system improves the adaptability and personalization of the system response by introducing human feedback, which is particularly important in hand function rehabilitation. Rehabilitation training based on movement intention recognition promotes the reorganization and recovery of neural pathways by repeating and reinforcing movements related to intentions. Summary of the invention

[0006] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide a human-in-the-loop hand function rehabilitation robot resistance training method, device, system and storage medium based on motion intention recognition, combined with real-time electromyographic signal detection, dynamically adjust the resistance coefficient to obtain the optimal resistance, and achieve high personalization and adaptability of rehabilitation.

[0007] To achieve the above object, the present invention is implemented by adopting the following technical solutions: In a first aspect, the present invention provides a hand function rehabilitation robot resistance training method based on movement intention recognition, comprising: S1. Randomly select a resistance coefficient within a preset resistance coefficient range, and determine the resistance force according to the resistance coefficient; S2. Train the patient through the hand function rehabilitation robot. During the training process, collect the patient's forearm electromyographic signal and actual movement angle; S3, extracting the time domain features and frequency domain features from the electromyographic signal, and using the time domain features and frequency domain features as input, and obtaining the finger driving force based on the output of the convolutional neural network model; S4, comparing the finger driving force and the actual movement angle with the preset finger driving force and the preset movement angle to calculate the finger driving force error and the position angle error; S5, calculating a resistance training evaluation index according to the finger driving force error and the position angle error, and forming a data group with the resistance coefficient and the resistance training evaluation index; S6, repeat S1 to S6 to obtain multiple data groups, wherein the multiple data groups constitute a data sample set; S7, taking the data sample set as input, obtaining the resistance coefficient corresponding to the maximum resistance training evaluation index based on the fitting model output, and adding the data group consisting of the maximum resistance training evaluation index and its corresponding resistance coefficient to the data sample set; S8, taking the resistance coefficient corresponding to the maximum resistance training evaluation index as the model hyperparameter, repeating S7 to S8 until a preset number of cycles is reached, and obtaining the resistance coefficient corresponding to the maximum resistance training evaluation index, which is the optimal resistance coefficient; S9. Obtain an optimal resistance according to the optimal resistance coefficient, and set a hand function rehabilitation robot to train the patient using the optimal resistance.

[0008] Furthermore, the expression for the resistance calculated according to the resistance coefficient is: ; in, F Indicates resistance, k represents the resistance coefficient, p Indicates the airbag gas pressure of the hand function rehabilitation robot.

[0009] Furthermore, the collection of the patient's forearm electromyographic signal and the actual movement angle is achieved through a MYO electromyographic acquisition armband, which includes a nine-degree-of-freedom motion sensor and eight dry electrodes. Specifically: Eight dry electrodes are evenly arranged on the patient's forearm, and the electromyographic signals are collected by the dry electrodes at a sampling frequency of 200 Hz and transmitted to a nine-degree-of-freedom motion sensor, which obtains the actual movement angle according to the electromyographic signals.

[0010] Furthermore, before extracting the time domain features and frequency domain features in the electromyographic signal, the method further includes processing the electromyographic signal to obtain a processed electromyographic signal, which includes: The electromyographic signal is filtered by using a bandpass filter to remove low-frequency motion artifacts and high-frequency electrical noise to obtain a filtered electromyographic signal; The filtered electromyographic signal is normalized by the maximum voluntary contraction method, and the signal intensity of the electromyographic signal is expressed as the maximum contraction percentage to obtain the processed electromyographic signal.

[0011] Furthermore, the time domain features include the root mean square value of the electromyographic signal voltage, the average absolute value of the electromyographic signal voltage, and the zero crossing rate of the electromyographic signal; The frequency domain features include the power spectrum density of the electromyographic signal voltage.

[0012] Furthermore, the expression for the resistance training evaluation index calculated based on the finger driving force error and the position angle error is: ; in, represents the resistance training evaluation index, represents the finger driving force error, Indicates the training time, Indicates the position angle error.

[0013] Furthermore, the fitting model is a Gaussian mixture model or a random forest classification model.

[0014] In a second aspect, the present invention further provides a hand function rehabilitation robot resistance training device based on movement intention recognition, comprising: The initial training module is configured to randomly select a resistance coefficient within a preset resistance coefficient range and determine the resistance according to the resistance coefficient; train the patient through the hand function rehabilitation robot, and collect the patient's forearm electromyographic signal and actual movement angle during the training process; A feature extraction module is configured to extract time domain features and frequency domain features from the electromyographic signal, and use the time domain features and the frequency domain features as input to obtain the finger driving force based on the output of the convolutional neural network model; An error acquisition module is configured to compare the finger driving force and the actual movement angle with the preset finger driving force and the preset movement angle to calculate the finger driving force error and the position angle error; A data group generating module is configured to calculate a resistance training evaluation index according to the finger driving force error and the position angle error, and form a data group with the resistance coefficient and the resistance training evaluation index; repeat the operation to obtain multiple data groups, and the multiple data groups constitute a data sample set; The optimal resistance coefficient acquisition module is configured to take the data sample set as input, obtain the resistance coefficient corresponding to the maximum resistance training evaluation index based on the fitting model output, add the data group consisting of the maximum resistance training evaluation index and its corresponding resistance coefficient to the data sample set; take the resistance coefficient corresponding to the maximum resistance training evaluation index as the model hyperparameter, repeat the operation until a preset number of cycles is reached, and obtain the resistance coefficient corresponding to the maximum resistance training evaluation index, which is the optimal resistance coefficient; The training implementation module is configured to obtain an optimal resistance according to an optimal resistance coefficient, and to set a hand function rehabilitation robot to train a patient using the optimal resistance.

[0015] In a third aspect, the present invention further provides a hand function rehabilitation robot resistance training system based on movement intention recognition, comprising: A memory for storing instructions; The processor is used to execute the instructions so that the system implements the operation of the hand function rehabilitation robot resistance training method based on movement intention recognition as described in any one of the above.

[0016] In a fourth aspect, the present embodiment further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the hand function rehabilitation robot resistance training method based on motion intention recognition as described in any one of the above.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The present invention monitors electromyographic signals in real time, dynamically adjusts the resistance coefficient through Bayesian optimization, and calculates the optimal resistance using the resistance coefficient corresponding to the maximum resistance training, so that the hand function rehabilitation machine can use the optimal resistance to train patients, so that the training method can be adaptively adjusted according to the specific needs and ability level of each patient, with high personalization and adaptability, and greatly improves the rehabilitation training effect; By monitoring electromyographic signals, the present invention can accurately map the required driving force, ensure that the training intensity matches the patient's ability, and enable the patient's subjective effort level to be directly reflected in the training process, which helps to enhance the patient's sense of participation and initiative and motivate the patient to actively participate in rehabilitation training. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of a hand function rehabilitation robot resistance training method based on movement intention recognition in one embodiment of the present invention; Figure 2 This is a schematic diagram of hand function rehabilitation robot training in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention. Example 1

[0020] like Figure 1 As shown, an embodiment of the present invention provides a hand function rehabilitation robot resistance training method based on movement intention recognition, comprising the following steps: S1. Preset the range of resistance coefficient values. The reasonable determination of the range of resistance coefficient values ​​needs to be obtained according to the actual situation. In this embodiment, it is achieved by setting the baseline resistance. Specifically: Professional therapists or technicians explained the requirements of the trajectory following task to the subjects in detail, helping them to become familiar with the trajectory following task and the operation of the pneumatic rehabilitation robot. The subjects experienced the trajectory following task without strict evaluation, and through 3-5 cycles, became familiar with the robot operation and the dynamic feedback mechanism of the trajectory following task.

[0021] Then, combine Figure 2 , ask the patient to make a full grip gesture, with the palm based on the horizontal plane and the fingers able to rotate 90° as an indicator. The resistance at this time is obtained according to the MYO electromyography acquisition armband, which is the baseline resistance. The resistance coefficient corresponding to the baseline resistance force should be within the preset resistance coefficient value range.

[0022] The MYO EMG acquisition armband includes a nine-degree-of-freedom motion sensor and eight dry electrodes. During acquisition, the eight dry electrodes are evenly distributed on the patient's forearm. The dry electrodes are used to collect EMG signals at a sampling frequency of 200 Hz and transmitted to the nine-degree-of-freedom motion sensor. The nine-degree-of-freedom motion sensor can obtain the actual movement angle based on the EMG signals.

[0023] S2. Randomly select a resistance coefficient in the preset resistance coefficient value range, and calculate the resistance according to the resistance coefficient. The calculation expression is: ; in, F Indicates resistance, k represents the resistance coefficient, p Indicates the airbag gas pressure of the hand function rehabilitation robot.

[0024] It should be noted that the resistance coefficient that has been selected cannot be selected again.

[0025] S3. Set up a hand function rehabilitation robot to train patients using resistance. During the training process, use the MYO electromyography acquisition armband to collect the patient's forearm electromyography signals and actual movement angles.

[0026] S4. Processing the patient's forearm electromyographic signal. The processing operation includes: filtering the electromyographic signal with a bandpass filter to remove low-frequency motion artifacts and high-frequency electrical noise to obtain a filtered electromyographic signal.

[0027] The filtered EMG signal was normalized by the maximum voluntary contraction method (MVC), and the signal intensity of the EMG signal was expressed as the maximum contraction percentage to obtain the processed EMG signal.

[0028] Then, the time domain features and frequency domain features of the patient's forearm EMG signal are extracted. The time domain features include the root mean square value (RMS) of the EMG signal voltage, the mean absolute value (MAV) of the EMG signal voltage, and the zero crossing rate (ZCR) of the EMG signal. The zero crossing rate is the number of times the EMG signal crosses the zero point from positive to negative in unit time. The frequency domain features include the power spectral density (PSD) of the EMG signal voltage. The power spectral density is used to reflect the energy distribution of the voltage signal over frequency.

[0029] The time domain features and frequency domain features are used as input, and the finger driving force is obtained based on the output of the convolutional neural network model. The internal processing of the convolutional neural network model includes: first, the feature map of the input data is extracted through the convolution layer and the pooling layer, and these feature maps are converted into a vector through the fully connected layer, and finally the vector is mapped to the finger driving force through the Softmax layer.

[0030] S5. Compare the finger driving force and the actual movement angle with the preset finger driving force and the preset movement angle to obtain the finger driving force error and the position angle error. The finger driving force error is the difference between the finger driving force and the preset finger driving force, and the position angle error is the difference between the actual movement angle and the preset movement angle.

[0031] S6. When the resistance is small, the muscle activation is small and the finger driving force is small, while when the resistance is large, the muscle activation is large and the finger driving force is large. Therefore, if the resistance is too small, the training intensity of the subjects is not enough, the participation is low, and it is not conducive to neural remodeling; if the resistance is too large, the subjects are not able to complete the task and lose the training effect; therefore, we need to find a set of optimal training parameters. Therefore, according to the goal of adaptive training, a resistance training evaluation index is defined, which takes into account both the finger driving force and the angle of the motion trajectory.

[0032] The resistance training evaluation index is calculated based on the finger driving force error and position angle error, and its calculation expression is: ; in, represents the resistance training evaluation index, represents the finger driving force error, Indicates the training time, Indicates the position angle error.

[0033] The obtained resistance coefficient and the resistance training evaluation index are combined into a data group, and the data group is expressed as The obtained resistance coefficient and the resistance training evaluation index are combined into a data group, and the data group is expressed as , represents the i-th selected impedance system, represents the i-th resistance training evaluation index .

[0034] S7. Repeat steps S2 to S7 continuously to obtain multiple data groups, which constitute a data sample set.

[0035] S8. Take the data sample set as input, obtain the resistance coefficient corresponding to the maximum resistance training evaluation index based on the fitting model output, add the data group consisting of the maximum resistance training evaluation index and its corresponding resistance coefficient to the data sample set, and form a new data sample set.

[0036] In some embodiments, the fitting model may select a Gaussian mixture model or a random forest classification model.

[0037] S9, taking the resistance coefficient corresponding to the maximum resistance training evaluation index as the model hyperparameter, repeating S8-S9 until the preset number of cycles is reached, and obtaining the resistance coefficient corresponding to the maximum resistance training evaluation index, which is the optimal resistance coefficient. Repeating the process of S8-S9 is Bayesian optimization.

[0038] S10. Obtain the optimal resistance according to the optimal resistance coefficient. The calculation method is the same as above and will not be elaborated here. Finally, the hand function rehabilitation robot is set to train the patient using the optimal resistance. Example 2

[0039] On the basis of Example 1, this embodiment further provides a hand function rehabilitation robot resistance training device based on movement intention recognition, comprising: The initial training module is configured to randomly select a resistance coefficient within a preset resistance coefficient range and determine the resistance according to the resistance coefficient; train the patient through the hand function rehabilitation robot, and collect the patient's forearm electromyographic signal and actual movement angle during the training process; A feature extraction module is configured to extract time domain features and frequency domain features from the electromyographic signal, and use the time domain features and the frequency domain features as input to obtain the finger driving force based on the output of the convolutional neural network model; An error acquisition module is configured to compare the finger driving force and the actual movement angle with the preset finger driving force and the preset movement angle to calculate the finger driving force error and the position angle error; A data group generating module is configured to calculate a resistance training evaluation index according to the finger driving force error and the position angle error, and form a data group with the resistance coefficient and the resistance training evaluation index; repeat the operation to obtain multiple data groups, and the multiple data groups constitute a data sample set; The optimal resistance coefficient acquisition module is configured to take the data sample set as input, obtain the resistance coefficient corresponding to the maximum resistance training evaluation index based on the fitting model output, add the data group consisting of the maximum resistance training evaluation index and its corresponding resistance coefficient to the data sample set; take the resistance coefficient corresponding to the maximum resistance training evaluation index as the model hyperparameter, repeat the operation until a preset number of cycles is reached, and obtain the resistance coefficient corresponding to the maximum resistance training evaluation index, which is the optimal resistance coefficient; The training implementation module is configured to obtain an optimal resistance according to an optimal resistance coefficient, and to set a hand function rehabilitation robot to train a patient using the optimal resistance. Example 3

[0040] On the basis of Example 1, this embodiment further provides a hand function rehabilitation robot resistance training system based on movement intention recognition, comprising: A memory for storing instructions; The processor is used to execute the instructions so that the system implements the operation of the hand function rehabilitation robot resistance training method based on movement intention recognition as described in Example 1. Example 4

[0041] On the basis of Example 1, this embodiment further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the hand function rehabilitation robot resistance training method based on motion intention recognition as described in Example 1 is implemented.

[0042] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A hand function rehabilitation robot resistance training method based on movement intention recognition, characterized in that: include: S1. Randomly select a resistance coefficient within a preset resistance coefficient range, and determine the resistance force according to the resistance coefficient; S2. Train the patient through the hand function rehabilitation robot. During the training process, collect the patient's forearm electromyographic signal and actual movement angle; S3, extracting the time domain features and frequency domain features from the electromyographic signal, and using the time domain features and frequency domain features as input, and obtaining the finger driving force based on the output of the convolutional neural network model; S4, comparing the finger driving force and the actual movement angle with the preset finger driving force and the preset movement angle to calculate the finger driving force error and the position angle error; S5, calculating a resistance training evaluation index according to the finger driving force error and the position angle error, and forming a data group with the resistance coefficient and the resistance training evaluation index; S6, repeat S1 to S6 to obtain multiple data groups, wherein the multiple data groups constitute a data sample set; S7, taking the data sample set as input, obtaining the resistance coefficient corresponding to the maximum resistance training evaluation index based on the fitting model output, and adding the data group consisting of the maximum resistance training evaluation index and its corresponding resistance coefficient to the data sample set; S8, taking the resistance coefficient corresponding to the maximum resistance training evaluation index as the model hyperparameter, repeating S7 to S8 until a preset number of cycles is reached, and obtaining the resistance coefficient corresponding to the maximum resistance training evaluation index, which is the optimal resistance coefficient; S9. Obtain an optimal resistance according to the optimal resistance coefficient, and set a hand function rehabilitation robot to train the patient using the optimal resistance.

2. The hand function rehabilitation robot resistance training method based on movement intention recognition according to claim 1 is characterized in that: The expression for determining the resistance according to the resistance coefficient is: ； in, F Indicates resistance, k represents the resistance coefficient, p Indicates the airbag gas pressure of the hand function rehabilitation robot.

3. The hand function rehabilitation robot resistance training method based on movement intention recognition according to claim 1 is characterized in that: The collection of the patient's forearm electromyographic signal and the actual movement angle is achieved through the MYO electromyographic acquisition armband, which includes a nine-degree-of-freedom motion sensor and 8 dry electrodes. Specifically: Eight dry electrodes are evenly arranged on the patient's forearm, and the electromyographic signals are collected by the dry electrodes at a sampling frequency of 200 Hz and transmitted to a nine-degree-of-freedom motion sensor, which obtains the actual movement angle according to the electromyographic signals.

4. The hand function rehabilitation robot resistance training method based on movement intention recognition according to claim 1 is characterized in that: The step of extracting the time domain features and the frequency domain features from the electromyographic signal also includes processing the electromyographic signal to obtain a processed electromyographic signal, which includes: The electromyographic signal is filtered by using a bandpass filter to remove low-frequency motion artifacts and high-frequency electrical noise to obtain a filtered electromyographic signal; The filtered electromyographic signal is normalized by the maximum voluntary contraction method, and the signal intensity of the electromyographic signal is expressed as the maximum contraction percentage to obtain the processed electromyographic signal.

5. The hand function rehabilitation robot resistance training method based on movement intention recognition according to claim 1 is characterized in that: The time domain features include the root mean square value of the electromyographic signal voltage, the average absolute value of the electromyographic signal voltage, and the zero crossing rate of the electromyographic signal; The frequency domain features include the power spectrum density of the electromyographic signal voltage.

6. The hand function rehabilitation robot resistance training method based on movement intention recognition according to claim 1 is characterized in that: The expression for the resistance training evaluation index calculated based on the finger driving force error and position angle error is: ； in, represents the resistance training evaluation index, represents the finger driving force error, Indicates the training time, Indicates the position angle error.

7. The hand function rehabilitation robot resistance training method based on movement intention recognition according to claim 1 is characterized in that: The fitting model is a Gaussian mixture model or a random forest classification model.

8. A hand function rehabilitation robot resistance training device based on movement intention recognition, characterized in that: include: The initial training module is configured to randomly select a resistance coefficient within a preset resistance coefficient range and determine the resistance according to the resistance coefficient; train the patient through the hand function rehabilitation robot, and collect the patient's forearm electromyographic signal and actual movement angle during the training process; A feature extraction module is configured to extract time domain features and frequency domain features from the electromyographic signal, and use the time domain features and the frequency domain features as input to obtain the finger driving force based on the output of the convolutional neural network model; An error acquisition module is configured to compare the finger driving force and the actual movement angle with the preset finger driving force and the preset movement angle to calculate the finger driving force error and the position angle error; A data group generating module is configured to calculate a resistance training evaluation index according to the finger driving force error and the position angle error, and form a data group with the resistance coefficient and the resistance training evaluation index; repeat the operation to obtain multiple data groups, and the multiple data groups constitute a data sample set; The optimal resistance coefficient acquisition module is configured to take the data sample set as input, obtain the resistance coefficient corresponding to the maximum resistance training evaluation index based on the fitting model output, add the data group consisting of the maximum resistance training evaluation index and its corresponding resistance coefficient to the data sample set; take the resistance coefficient corresponding to the maximum resistance training evaluation index as the model hyperparameter, repeat the operation until a preset number of cycles is reached, and obtain the resistance coefficient corresponding to the maximum resistance training evaluation index, which is the optimal resistance coefficient; The training implementation module is configured to obtain an optimal resistance according to an optimal resistance coefficient, and to set a hand function rehabilitation robot to train a patient using the optimal resistance.

9. A hand function rehabilitation robot resistance training system based on movement intention recognition, characterized in that: include: A memory for storing instructions; The processor is used to execute the instructions so that the system implements the operation of the hand function rehabilitation robot resistance training method based on movement intention recognition as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the hand function rehabilitation robot resistance training method based on movement intention recognition as described in any one of claims 1 to 8 is implemented.

Images