Virtual interaction hand mirror image rehabilitation training method and system based on myoelectric control

Abstract

The invention discloses a virtual interaction hand mirror image rehabilitation training method and system based on myoelectricity control, relates to the technical field of rehabilitation machine learning, and aims to solve the problems of low recognition precision and poor tactile feedback in virtual interaction rehabilitation training in the prior art. Performing feature extraction on the real-time electromyographic signal data of the healthy side hand of the patient to obtain time domain features and frequency domain features; taking the time domain feature and the frequency domain feature as input, and outputting based on a convolutional neural network model to obtain a gesture classification result; according to the gesture classification result, the hand rehabilitation robot is used for carrying out mirror image training on the hand on the affected side, and interaction between a virtual game and real actions of the patient is achieved through game software. The motion intention of the patient is predicted in advance by collecting the electromyographic signals, the patient is better guided to conduct mirror image training, interaction between the actual motion of the patient and the virtual game is achieved, and the pleasure and enthusiasm of the patient in rehabilitation training can be increased.

Description

Technical Field

[0001] The invention relates to a virtual interactive hand mirror image rehabilitation training method and system based on myoelectric control, belonging to the technical field of rehabilitation machine learning. Background Art

[0002] The aging problem has led to a significant increase in the prevalence of various diseases, among which hand dysfunction caused by brain diseases such as stroke is particularly significant, bringing significant psychological, family, and life pressure to patients. Due to the differences in patients' physical constitutions and family economic conditions, the recovery rate of patients has not increased significantly, so it is imperative to promote rehabilitation training for patients with hand motor dysfunction.

[0003] During rehabilitation training, patients usually need to complete a series of specific movements along specific trajectories under the guidance of professionals. Among them, the use of rehabilitation robots for rehabilitation training is undoubtedly a powerful tool in the treatment process. Intelligent training methods combined with effective training evaluation can significantly improve the patient's recovery rate.

[0004] Current virtual interactive training systems usually use eye tracking, hand tracking, voice interaction and other methods to achieve interaction in virtual scenes, but they generally have shortcomings such as low recognition accuracy and poor tactile feedback.

[0005] The biggest feature of myoelectric sensing is that the generation of myoelectric signals is ahead of the actual movement, and it contains a lot of joint movement information. Mirror training is widely used in various medical occasions. It was first proposed as a treatment method to relieve phantom limb pain in amputees. Subsequently, mirror therapy has been increasingly used to treat a series of other chronic pain diseases, such as complex regional pain syndrome (CRPS) and persistent pain after wrist fractures and hand surgery. Summary of the invention

[0006] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide a virtual interactive hand mirror rehabilitation training method and system based on electromyographic control. By collecting electromyographic signals, the patient's movement intention can be predicted in advance, and the patient can be better guided to perform mirror training. The interaction between the patient's actual movements and virtual games is realized, which can increase the patient's interest and enthusiasm in rehabilitation training.

[0007] To achieve the above object, the present invention is implemented by adopting the following technical solutions: On the one hand, the present invention provides a virtual interactive hand mirror rehabilitation training method based on myoelectric control, comprising: Collect real-time electromyographic signal data of the patient's healthy hand; The real-time electromyographic signal data of the patient's healthy hand is extracted to obtain time domain features and frequency domain features; Taking the time domain features and frequency domain features as input, the gesture classification results are obtained based on the output of the convolutional neural network model; According to the gesture classification results, a hand rehabilitation robot is used to perform mirror training on the patient's affected hand, and the interaction between the virtual game and the patient's real movements is achieved through game software.

[0008] Furthermore, the real-time electromyographic signal data of the healthy hand of the patient is collected by a MYO electromyographic collection armband, which includes a nine-degree-of-freedom motion sensor and 8 dry electrodes, specifically including: Eight dry electrodes were evenly placed on the patient's healthy forearm; The electromyographic signal of the patient's healthy hand forearm is collected by dry electrodes at a sampling frequency of 200 Hz and a baud rate of 1 Mbps, and transmitted to a nine-degree-of-freedom motion sensor. The real-time electromyographic signal data of the patient's healthy hand is obtained through the output of the nine-degree-of-freedom motion sensor.

[0009] Furthermore, the feature extraction of the real-time electromyographic signal of the healthy hand of the patient to obtain time domain features and frequency domain features includes: Performing signal waveform processing on the real-time electromyographic signal of the patient's healthy hand, wherein the signal waveform processing includes using a sliding window to divide the real-time electromyographic signal of the patient's healthy hand into windows to reduce signal spectrum leakage; Set the bandpass filter to 10~500Hz to remove interference and noise and obtain the processed EMG signal data; The processed electromyographic signal data is subjected to feature extraction to obtain time domain features and frequency domain features.

[0010] Furthermore, the time domain features include the average value of the electromyographic signal voltage, the standard deviation of the electromyographic signal voltage, and the waveform length of the electromyographic signal, and the frequency domain features include the spectrum peak frequency of the electromyographic signal voltage and the frequency bandwidth change of the electromyographic signal voltage; Among them, the average value of the electromyographic signal voltage is used to reflect the overall changes in muscle tension and activation state; the standard deviation of the electromyographic signal voltage is used to reflect the magnitude of muscle contraction force; and the waveform length of the electromyographic signal is used to reflect the complexity of muscle movement.

[0011] Furthermore, the convolutional neural network model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer connected in sequence; The input layer includes 6 neurons, and its activation function adopts the ReLU activation function. The convolution kernel size of the convolution layer is 3×3, the step size is 1, and the convolution kernel includes 50 neurons. The pooling layer adopts a 2×2 pooling window, which is used to reduce the dimension of the data. The fully connected layer adopts 6 neurons, and its activation function adopts the Softmax activation function. The output layer is used to output the gesture probability distribution, and the gesture category corresponding to the maximum value of the gesture probability distribution is the gesture classification result.

[0012] Furthermore, it also includes pre-training the convolutional neural network model, and the pre-training method includes: S1, collecting the electromyographic signals generated when the patient's healthy hand makes different gestures, and obtaining electromyographic signal data corresponding to different gestures; S2, extracting time domain features and frequency domain features of the electromyographic signal data corresponding to different gestures to obtain feature data sets corresponding to different gestures; S3, dividing the feature data sets corresponding to different gestures into training set data and test set data; S4. Use the training set data as input to pre-train the convolutional neural network model. During the training process, calculate the cross entropy loss function and use the Adam optimizer to adjust the model weights according to the loss function. S5. Use the test set data as input to test the convolutional neural network model and calculate the evaluation index. If the evaluation index is higher than the preset value, the pre-trained convolutional neural network model is obtained. If the accuracy of the gesture classification result output by the convolutional neural network model is lower than the preset value, repeat S4 to S5 until the pre-trained convolutional neural network model is obtained.

[0013] Furthermore, the expression of the cross entropy loss function is: ; in, represents the cross entropy loss function value, is the number of data samples, Indicates The gesture probability distribution of data samples, Indicates The true gesture category of the data samples, The value of is 0 or 1; The evaluation index is the harmonic mean, which is calculated by taking the harmonic mean of the precision and the recall.

[0014] Furthermore, the number of the gestures is 6, which are single thumb bending, pointed shape, tripod shape, cylindrical shape, half grip shape and full grip shape; The myoelectric signals generated when the patient's healthy hand makes different gestures are collected to obtain the myoelectric signal data corresponding to different gestures through the MYO myoelectric acquisition armband. The acquisition parameters include collecting each gesture 10 times, each time with an interval of 10 seconds, and stopping for 3 minutes before changing to the next gesture.

[0015] Furthermore, according to the gesture classification result, the hand rehabilitation robot is used to perform mirror training on the patient's affected hand, and the interaction between the virtual game and the patient's real action is realized through the game software, including: The gesture classification results are transmitted to the hand rehabilitation robot and the game software respectively through the host computer; While the hand rehabilitation robot performs mirror training on the affected hand, the game software makes corresponding actions based on the gesture classification results, realizing the interaction between the virtual game and the patient's real actions.

[0016] On the other hand, the present invention also provides a virtual interactive hand mirror rehabilitation training system based on myoelectric control, comprising: The electromyographic signal acquisition module is configured to collect real-time electromyographic signal data of the patient's healthy hand; A feature extraction module is configured to extract features from the real-time electromyographic signal data of the healthy hand of the patient to obtain time domain features and frequency domain features; A gesture classification module is configured to take the time domain features and the frequency domain features as inputs and obtain a gesture classification result based on the output of the convolutional neural network model; The training interaction module is configured to perform mirror training on the affected hand using a hand rehabilitation robot according to the gesture classification results, and to achieve interaction between the virtual game and the patient's real movements through the game software.

[0017] Compared with the prior art, the present invention has the following beneficial effects: Compared with voice signals and visual signals, the electromyographic signal collected by the present invention is more direct and less susceptible to external influences, which improves the refinement and accuracy of recognition. Since the generation of electromyographic signals is ahead of the patient's actual movements, the patient's movement intention can be predicted in advance, and the patient can be better guided to perform mirror training, thereby increasing the user's neural perception of force and movement control. The present invention is also provided with game software to realize the interaction between the patient's actual actions and the virtual game, thereby increasing the patient's interest and enthusiasm in rehabilitation training, and can effectively promote the regeneration and repair of the patient's motor neurons, thereby accelerating the patient's recovery. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A schematic diagram of a flow chart of a virtual interactive hand mirror rehabilitation training method based on myoelectric control in an embodiment of the present invention; Figure 2The game software interface diagram of a virtual interactive hand mirror rehabilitation training method based on myoelectric control in one 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 virtual interactive hand mirror rehabilitation training method based on myoelectric control, including a training phase and an implementation phase.

[0021] Training phase: Construct a convolutional neural network model. In this embodiment, the structure of the convolutional neural network model includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer connected in sequence. The input layer includes 6 neurons, the number of the convolution layer is 1, the convolution kernel size is 3×3, the step size is 1, each convolution kernel corresponds to 50 neurons, and the "same" filling method is used to ensure that the input data dimension is the same as the output data dimension. The activation function of the convolution layer uses the ReLU activation function. The pooling layer uses a 2×2 pooling window and uses the maximum pooling method to reduce the dimension of the data. The fully connected layer is the same as the input layer, using 6 neurons, and the activation function of the fully connected layer uses the Softmax activation function. The output layer is used to output the gesture probability distribution, and the gesture category corresponding to the maximum value of the gesture probability distribution is the gesture classification result.

[0022] The constructed convolutional neural network model is pre-trained. The pre-training process includes: S1. Collect the electromyographic signals generated when the patient's healthy hand performs different gestures to obtain electromyographic signal data corresponding to different gestures.

[0023] In this embodiment, the electromyographic signal is obtained through the MYO electromyographic acquisition armband, which includes a nine-degree-of-freedom motion sensor and eight dry electrodes. The specific operations of acquisition include: Eight dry electrodes were evenly distributed on the patient's healthy forearm.

[0024] The electromyographic signal of the patient's healthy hand forearm is collected by dry electrodes at a sampling frequency of 200 Hz and a baud rate of 1 Mbps, and transmitted to a nine-degree-of-freedom motion sensor. The real-time electromyographic signal data of the patient's healthy hand is obtained through the output of the nine-degree-of-freedom motion sensor.

[0025] There are 6 types of gestures, including single thumb bending, pointed shape, tripod shape, cylindrical shape, half grip and full grip. Each gesture was collected 10 times, with an interval of 10 seconds each time, and stopped for 3 minutes before changing to the next gesture.

[0026] S2. Extract the time domain features and frequency domain features of the electromyographic signal data corresponding to different gestures to obtain the feature data sets corresponding to different gestures. Specifically: The electromyographic signal data corresponding to different gestures are subjected to signal waveform processing. The signal waveform processing includes using a sliding window to split the electromyographic signal to reduce signal spectrum leakage, and setting the bandpass filter to 10-500 Hz to remove interference and noise to obtain the processed electromyographic signal data.

[0027] The processed EMG signal data is extracted for time domain features and frequency domain features. In this embodiment, the time domain features include the average value of the EMG signal voltage, the standard deviation of the EMG signal voltage, and the waveform length of the EMG signal. The average value of the EMG signal voltage is used to reflect the overall changes in muscle tension and activation state, the standard deviation of the EMG signal voltage is used to reflect the size of muscle contraction force, and the waveform length of the EMG signal is used to reflect the complexity of muscle movement. The frequency domain features include the peak frequency of the spectrum of the EMG signal voltage and the frequency bandwidth change of the EMG signal voltage.

[0028] S3. Divide the feature data sets corresponding to different gestures into training set data and test set data, where the training set data accounts for 80% and the test set data accounts for 20%.

[0029] S4. Use the training set data as input to pre-train the convolutional neural network model. During the training process, calculate the cross entropy loss function, and use the Adam optimizer to adjust the model weights according to the loss function.

[0030] The expression of the cross entropy loss function is: ; in, represents the cross entropy loss function value, is the number of data samples, Indicates The gesture probability distribution of data samples, Indicates The true gesture category of the data samples, The value of is 0 or 1.

[0031] , that is, when When the maximum value in the probability distribution of the gesture of a data sample is the same as the true gesture category of the data sample, The value of is 1, otherwise The value of is 0.

[0032] S5. Use the test set data as input to test the convolutional neural network model and calculate the evaluation index. In this embodiment, the evaluation index is the harmonic mean, which is calculated by harmonic averaging the precision and recall.

[0033] If the evaluation index is higher than the preset value, a pre-trained convolutional neural network model is obtained. If the accuracy of the gesture classification result output by the convolutional neural network model is lower than the preset value, S4 to S5 are repeated until a pre-trained convolutional neural network model is obtained.

[0034] Testing phase: The real-time electromyographic signal data of the patient's healthy hand is collected. The electromyographic signal collection method is as shown above and will not be elaborated here.

[0035] The real-time electromyographic signal data of the patient's healthy hand is extracted to obtain time domain features and frequency domain features.

[0036] The time domain features and frequency domain features are used as input, and the gesture classification results are obtained based on the output of the pre-trained convolutional neural network model. It should be noted that the convolutional neural network model outputs a probability distribution, and the gesture category corresponding to the maximum probability value is the gesture classification result.

[0037] According to the gesture classification results, the hand rehabilitation robot is used to conduct mirror training on the affected hand, and the game software is used to achieve interaction between the virtual game and the patient's real movements, including: The gesture classification results are transmitted to the hand rehabilitation robot and game software respectively through the host computer.

[0038] While the hand rehabilitation robot performs mirror training on the affected hand, the game software makes corresponding actions based on the gesture classification results, realizing the interaction between the virtual game and the patient's real actions.

[0039] In this embodiment, if Figure 2 As shown, the game software is designed as follows: Game interface design: Draw a colorful sky background with white cloud elements to create an open game scene, and the ground is green grass to increase the visual layering.

[0040] A cute bird image is designed with bright colors, yellow body and red beak. The size of the bird is moderate, which is convenient for players to observe its movements clearly on the screen. The position of the bird on the screen is in the middle of the first picture. There is a button to start the game.

[0041] Hand motion capture and mapping: The MYO electromyography acquisition armband is used to obtain the electromyography signals generated by the real-time movements of the player's healthy hand. The signal processing is uploaded to the host computer, which processes the data and then sends it to the hand rehabilitation robot through serial communication. The hand rehabilitation robot receives the instructions and controls the affected hand to make the same movements as the healthy hand to achieve mirror training.

[0042] For example, when the player's (patient's) healthy-side fingers bend inward (flex), they will be mapped to the affected-side hand, causing the fingers of the affected-side hand to bend inward. At the same time, the game software will interpret it as an upward flight command, causing the bird in the game to move upward a certain distance. When the healthy-side fingers bend outward (extend), they will be mapped to the affected-side hand. At the same time, the game software will interpret it as an upward flight command, causing the bird to move downward a certain distance. When the healthy-side hand moves left or right, it will be mapped to the affected-side hand. At the same time, the game software will interpret it as an obstacle avoidance command, controlling the bird to fine-tune horizontally to avoid obstacles.

[0043] Obstacle crossing and scoring mechanism: The bird needs to cross various obstacles while flying to the right. If the bird successfully avoids an obstacle, the player will receive a one-point bonus. If the bird collides with an obstacle, the game will sound a warning, indicating that the game has failed. At this time, the player can choose to restart the game or view the score statistics and operation record analysis of this game in order to summarize experience and lessons and improve hand movement control strategies.

[0044] Levels and difficulty increase: In the first level, there are fewer obstacles and they move slowly; in the later levels, obstacles will be more densely distributed on the bird's flight path, and new action prompts will appear, such as clenching a fist to indicate that the bird is accelerating, and opening a palm to indicate slowing down. As the level progresses, the player needs to control the hand robot's movements more accurately to make the bird pass through obstacles smoothly. This helps to gradually improve the hand coordination ability and ability of the hand robot operator to deal with complex situations.

[0045] Game feedback and evaluation: During the game, real-time feedback is provided to players through visual, auditory and tactile means. For example, when the bird successfully avoids an obstacle, a point-adding special effect will be displayed on the screen, and a cheerful sound effect will be played; when the bird collides with an obstacle, in addition to the above-mentioned prompt sound, the game screen will have a brief flash or explosion effect to enhance the player's perception of the operation results. At the same time, the player's current score and relevant data of hand movements, such as the speed, amplitude, and frequency of hand movements, are displayed in real time on the game interface, allowing players to understand their game status and operation performance at any time.

[0046] Training evaluation and report: After each game, a detailed training evaluation report is generated for the player. The report includes the total score of the game, level completion status, analysis of operation errors (such as misjudgment of obstacle position, hand movement delay, etc.), hand movement stability and coordination evaluation, etc. Through the accumulation and analysis of multiple game training data, players can also be provided with long-term training trend charts to show the player's improvement or deficiency in hand movement control ability. These evaluation reports and analysis data can help players adjust training plans and methods in a targeted manner and improve the effect of hand robot mirror training.

[0047] Through the above-designed hand rehabilitation robot mirror training method with a bird crossing obstacles as the game interface, players can control the bird to cross obstacles by continuously operating the hand rehabilitation robot in an interesting game environment, and gradually improve the coordination, accuracy and reaction speed of hand movements, thereby achieving the goal of efficient hand rehabilitation robot mirror training. Example 2

[0048] Based on Example 1, this embodiment further provides a virtual interactive hand mirror rehabilitation training system based on myoelectric control, including: The electromyographic signal acquisition module is configured to collect real-time electromyographic signal data of the patient's healthy hand.

[0049] The feature extraction module is configured to extract features from the real-time electromyographic signal data of the healthy hand of the patient to obtain time domain features and frequency domain features.

[0050] The gesture classification module is configured to take the time domain features and the frequency domain features as inputs and obtain the gesture classification results based on the output of the convolutional neural network model.

[0051] The training interaction module is configured to perform mirror training on the affected hand using a hand rehabilitation robot according to the gesture classification results, and to achieve interaction between the virtual game and the patient's real movements through the game software.

[0052] 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 virtual interactive hand mirror rehabilitation training method based on electromyography control, characterized in that: include: Collect real-time electromyographic signal data of the patient's healthy hand; The real-time electromyographic signal data of the patient's healthy hand is extracted to obtain time domain features and frequency domain features; Taking the time domain features and frequency domain features as input, the gesture classification results are obtained based on the output of the convolutional neural network model; According to the gesture classification results, a hand rehabilitation robot is used to perform mirror training on the patient's affected hand, and the interaction between the virtual game and the patient's real movements is achieved through game software.

2. The virtual interactive hand mirror rehabilitation training method based on myoelectric control according to claim 1 is characterized in that: The real-time electromyographic signal data of the healthy hand of the patient is collected by a MYO electromyographic collection armband, which includes a nine-degree-of-freedom motion sensor and 8 dry electrodes, specifically including: Eight dry electrodes were evenly placed on the patient's healthy forearm; The electromyographic signal of the patient's healthy hand forearm is collected by dry electrodes at a sampling frequency of 200 Hz and a baud rate of 1 Mbps, and transmitted to a nine-degree-of-freedom motion sensor. The real-time electromyographic signal data of the patient's healthy hand is obtained through the output of the nine-degree-of-freedom motion sensor.

3. The virtual interactive hand mirror rehabilitation training method based on myoelectric control according to claim 1 is characterized in that: The feature extraction of the real-time electromyographic signal of the healthy hand of the patient to obtain time domain features and frequency domain features includes: Performing signal waveform processing on the real-time electromyographic signal of the patient's healthy hand, wherein the signal waveform processing includes using a sliding window to divide the real-time electromyographic signal of the patient's healthy hand into windows to reduce signal spectrum leakage; Set the bandpass filter to 10~500Hz to remove interference and noise and obtain the processed EMG signal data; The processed electromyographic signal data is subjected to feature extraction to obtain time domain features and frequency domain features.

4. The virtual interactive hand mirror rehabilitation training method based on myoelectric control according to claim 3 is characterized in that: The time domain features include the average value of the electromyographic signal voltage, the standard deviation of the electromyographic signal voltage, and the waveform length of the electromyographic signal; the frequency domain features include the peak frequency of the spectrum of the electromyographic signal voltage and the frequency bandwidth change of the electromyographic signal voltage; Among them, the average value of the electromyographic signal voltage is used to reflect the overall changes in muscle tension and activation state; the standard deviation of the electromyographic signal voltage is used to reflect the magnitude of muscle contraction force; and the waveform length of the electromyographic signal is used to reflect the complexity of muscle movement.

5. The virtual interactive hand mirror rehabilitation training method based on myoelectric control according to claim 1, characterized in that: The convolutional neural network model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer connected in sequence; The input layer includes 6 neurons, and its activation function adopts the ReLU activation function. The convolution kernel size of the convolution layer is 3×3, the step size is 1, and the convolution kernel includes 50 neurons. The pooling layer adopts a 2×2 pooling window, which is used to reduce the dimension of the data. The fully connected layer adopts 6 neurons, and its activation function adopts the Softmax activation function. The output layer is used to output the gesture probability distribution, and the gesture category corresponding to the maximum value of the gesture probability distribution is the gesture classification result.

6. The virtual interactive hand mirror rehabilitation training method based on myoelectric control according to claim 5 is characterized in that: The method also includes pre-training a convolutional neural network model, wherein the pre-training method includes: S1, collecting the electromyographic signals generated when the patient's healthy hand makes different gestures, and obtaining electromyographic signal data corresponding to different gestures; S2, extracting time domain features and frequency domain features of the electromyographic signal data corresponding to different gestures to obtain feature data sets corresponding to different gestures; S3, dividing the feature data sets corresponding to different gestures into training set data and test set data; S4. Use the training set data as input to pre-train the convolutional neural network model. During the training process, calculate the cross entropy loss function and use the Adam optimizer to adjust the model weights according to the loss function. S5. Use the test set data as input to test the convolutional neural network model and calculate the evaluation index. If the evaluation index is higher than the preset value, the pre-trained convolutional neural network model is obtained. If the accuracy of the gesture classification result output by the convolutional neural network model is lower than the preset value, repeat S4 to S5 until the pre-trained convolutional neural network model is obtained.

7. The virtual interactive hand mirror rehabilitation training method based on myoelectric control according to claim 6, characterized in that: The expression of the cross entropy loss function is: ； in, represents the cross entropy loss function value, is the number of data samples, Indicates The gesture probability distribution of data samples, Indicates The true gesture category of the data samples, The value of is 0 or 1; The evaluation index is the harmonic mean, which is calculated by taking the harmonic mean of the precision and the recall.

8. The virtual interactive hand mirror rehabilitation training method based on myoelectric control according to claim 6, characterized in that: There are 6 types of gestures, which are single thumb bending, pointed shape, tripod shape, cylindrical shape, half grip and full grip; The myoelectric signals generated when the patient's healthy hand makes different gestures are collected to obtain the myoelectric signal data corresponding to different gestures through the MYO myoelectric acquisition armband. The acquisition parameters include collecting each gesture 10 times, each time with an interval of 10 seconds, and stopping for 3 minutes before changing to the next gesture.

9. The virtual interactive hand mirror rehabilitation training method based on myoelectric control according to claim 1, characterized in that: The method uses a hand rehabilitation robot to perform mirror training on the affected hand of the patient according to the gesture classification result, and realizes the interaction between the virtual game and the patient's real action through the game software, including: The gesture classification results are transmitted to the hand rehabilitation robot and the game software respectively through the host computer; While the hand rehabilitation robot performs mirror training on the affected hand, the game software makes corresponding actions based on the gesture classification results, realizing the interaction between the virtual game and the patient's real actions.

10. A virtual interactive hand mirror rehabilitation training system based on electromyography control, characterized in that: include: The electromyographic signal acquisition module is configured to collect real-time electromyographic signal data of the patient's healthy hand; A feature extraction module is configured to extract features from the real-time electromyographic signal data of the healthy hand of the patient to obtain time domain features and frequency domain features; A gesture classification module is configured to take the time domain features and the frequency domain features as inputs and obtain a gesture classification result based on the output of the convolutional neural network model; The training interaction module is configured to perform mirror training on the affected hand using a hand rehabilitation robot according to the gesture classification results, and to achieve interaction between the virtual game and the patient's real movements through the game software.

Images