Multi-person cooperative motor imagery brain-computer interaction method and system for group brain calculation

Abstract

The invention discloses a group brain calculation-oriented multi-person cooperative motor imagery brain-computer interaction method and system, and the method comprises the steps: (1) building a cooperative motor imagery task through compounding a single motor imagery task, and enabling a plurality of users to wear electroencephalogram caps at the same time to collect data; (2) enhancing the motor imagery performance of the user through a multi-person collaborative motor imagery task; (3) preprocessing the electroencephalogram signal data; (4) electroencephalogram signal feature extraction and classification; 5) performing group brain analysis and motor imagery performance analysis during user cooperation; and (6) multi-user collaborative experiment training. The motor imagery performance of the user is enhanced in a multi-person cooperation mode, and electroencephalogram signals of multiple persons are collected in real time, processed and converted into control signals so as to cooperatively control external equipment. According to the method, the classification accuracy of the single-person electroencephalogram motor imagery of the user is effectively improved.

Description

Technical Field

[0001] The present invention relates to the fields of brain-computer interaction, motor imagery, and more particularly to a multi-person collaborative motor imagery brain-computer interaction method and system for group brain computing. Background Art

[0002] Motor imagery is a type of experimental paradigm in brain-computer interaction, which means that users can actively control objects to perform corresponding actions when imagining. When the imagined action occurs, this process can activate the user's motor nervous system and strengthen the associative mechanism of sensory simulation. As the research deepens, researchers have found that motor imagery brain-computer interaction can not only rebuild a bridge for people with motor disabilities to communicate with the outside world and train and restore some human functions, but also bring new breakthroughs to healthy people in different fields (such as mechanical control, leisure and entertainment).

[0003] Existing research on brain-computer interaction faces two problems: (1) Brain-computer interaction designed with a single user as the core is difficult to adapt to complex interactive environments, and the method of requiring a single user to repeat a large number of experiments to improve the performance of brain-computer interaction will also cause many troubles to the user. (2) Multi-person brain-computer interaction based on motor imagery paradigm only focuses on a large number of test users completing the same task in the same batch, while ignoring the collaboration between test users and how motor imagery in a collaborative form affects the performance of individual users. At present, research on collaborative brain-computer interaction revolves around the model of a large number of users performing the same task. At the algorithm level, the EEG data of each user is fused at the feature layer and sent to the classifier, or an integrated classifier is used to fused and predict the classification results of multiple users. How to allow multiple test users to perform different types of motor imagery tasks at the same time and decode the coupling between the test users' brains is one of the important technical problems in this field. Summary of the invention

[0004] The present invention aims to overcome the above-mentioned deficiencies of the prior art and provide a multi-person collaborative motor imagery brain-computer interaction method and system for group brain computing.

[0005] The present invention provides a multi-person collaborative motor imagery brain-computer interaction method for group brain computing to enable multiple users to collaboratively perform complex motor imagery tasks to enhance the motor imagery performance of a single user. It also provides multi-user brain synchronization analysis and multi-brain coupling analysis to verify the collaboration between users, and provides a neural network model for classification to verify the accuracy of users' motor imagery execution.

[0006] The present invention provides a multi-person collaborative motor imagery brain-computer interaction method for group brain computing, which mainly includes the following steps:

[0007] (1) A collaborative motor imagery task is established by combining single motor imagery tasks. Multiple users wear EEG caps to collect data at the same time. The traditional four-category motor imagery task is combined and multiple users are required to complete the collaborative motor imagery task in real-time cooperation. Each user performs a part of the collaborative task separately.

[0008] (2) Enhance the user's motor imagination performance through multi-person collaborative motor imagination tasks; through the designed collaborative motor imagination tasks, multiple users can observe a motor imagination stimulus source at the same time to perform motor imagination tasks, and enhance the user's motor imagination ability and performance through collaboration; single-person task-multi-person collaborative task training-single-person task mode; users first determine the baseline in the single-person motor imagination task, and use the 2D CNN-LSTM model to measure the baseline classification accuracy; then let the users experience the collaborative task, allowing users to complete collaborative motor imagination under the guidance of visual stimulation to enhance their motor imagination ability.

[0009] (3) Preprocessing of EEG signal data: removing noise and other irrelevant components in EEG data through electrode positioning, filtering, downsampling, segmentation, re-referencing, independent component analysis, etc.

[0010] (4) EEG signal feature extraction and classification: extract motor imagery features from the preprocessed data and classify motor imagery using a 2D CNN-LSTM network model.

[0011] For a single subject, the preprocessed EEG signal can be used for the following calculations: calculation of inter-brain synchronization, construction of a single brain functional network, and input into the preprocessed 2D CNN-LSTM model for single-person motor imagery classification prediction;

[0012] For multiple subjects, the preprocessed EEG signals of the subjects can be used for the following calculations in addition to the calculation process for a single subject: calculation of inter-brain synchronization between multiple subjects, calculating the inter-brain synchronization of each subject relative to other subjects among multiple subjects;

[0013] (5) Group brain analysis and motor imagery performance analysis during user collaboration; use inter-brain synchronization calculation and functional brain network reconstruction to analyze the coherence of EEG signals between different users during multi-user collaboration in real time, and use the 2D CNN-LSTM network to output the motor imagery classification results and calculate the motor imagery classification accuracy to analyze the motor imagery performance of a single user and verify the effect of the collaborative training experiment.

[0014] The CNN part of the 2D CNN-LSTM model described in step (2) includes two convolution blocks, each of which contains two convolution layers followed by one pooling layer. Each convolution layer uses the ReLU activation function to reduce training time and improve performance. In the first convolution block, the number of convolution kernels in each convolution layer is 32, and in the second convolution block, the number of convolution kernels in each convolution layer is 16. In order to capture the difference in details, the convolution kernel size of the first convolution block is 3, and the convolution kernel size of the second convolution block is 2. All convolution kernel steps are set to 1, and "Same" is selected in the Padding function. To ensure that the edge position information of the input signal is extracted as much as possible; after every 2 convolutional layers, the model adds a pooling layer to reduce the dimension of the features. The pool kernel size of each pooling layer is 2, and the step size is also 2; in addition, the Dropout method is added to each pooling layer. By setting some hidden layer nodes to 0, some data are randomly discarded, so that the model does not rely too much on local features and reduces overfitting; all Dropout coefficients are 0.3, which means that each node in the hidden layer has a 30% chance of being turned into 0; the preprocessed EEG features are classified, and finally the corresponding prediction labels are output.

[0015] The present invention also provides a multi-person collaborative motor imagery brain-computer interaction system for group brain computing, which includes the following modules in sequence:

[0016] (1) a motor imagery task combination module, used to set a combined motor imagery task;

[0017] (2) Multi-person collaborative motor imagery module, which presents the combined motor imagery task and simultaneously collects motor imagery EEG signals of multiple people;

[0018] (3) an EEG signal preprocessing module, which preprocesses the collected or stored EEG signals;

[0019] (4) Feature extraction and classification module: extract features from the preprocessed EEG data for input and training of the 2D CNN-LSTM classification model.

[0020] (5) Multi-user collaborative group brain analysis module, which uses inter-brain synchronization calculation and functional brain network reconstruction to analyze the coherence of EEG signals between different users during multi-user collaboration in real time;

[0021] The present invention also provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the inter-brain synchronization calculation and functional brain network construction of the present invention.

[0022] The present invention further provides a computing device, comprising a memory and a processor, wherein the memory stores executable codes, and when the processor executes the executable codes, the method of the present invention is implemented.

[0023] The advantage of the present invention is that by analyzing the coupling between different areas of the users' brains (parietal lobe, frontal lobe, central area) during collaborative tasks, decoding is performed through the indicator of inter-brain synchronization, revealing the way of information exchange between different areas, explaining why multi-person collaborative motion imagery tasks have higher classification accuracy than single-person motion imagery tasks, and by constructing the users' functional brain networks, explaining why multi-person collaborative motion tasks can be used as a training method to help single users better complete motion imagery. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 It is a schematic diagram of the process of the present invention.

[0026] Figure 2 It is a schematic diagram of the 2D CNN-LSTM classification model of the present invention.

[0027] Figure 3 Schematic diagram of the method used to calculate brain-to-brain synchronization.

[0028] Figure 4 Schematic diagram of the method used to construct brain networks.

[0029] Figure 5 It is a basic logical structure diagram of the system of the present invention. DETAILED DESCRIPTION

[0030] The following will be combined with the accompanying drawings and embodiments to provide a detailed and complete description of a multi-person collaborative motor imagery brain-computer interaction system for group brain computing of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments, and cannot be understood as limitations on the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0031] Example 1

[0032] The embodiment of the present invention provides a multi-person collaborative motor imagery brain-computer interaction method for group brain computing. After training through the multi-person collaborative motor imagery brain-computer interaction method, participants can effectively improve their motor imagery performance and enhance the quality of motor imagery EEG signals. This can be specifically reflected by the classification accuracy of the 2D CNN-LSTM network. Figure 1 , describe the specific steps:

[0033] (1) Combining single motor imagery tasks into a composite motor imagery task;

[0034] This method combines different types of motor imagery into a composite motor imagery stimulus paradigm, and then displays it through the robot, such as showing upper limb motor imagery through left and right hand movements, and showing tongue motor imagery through head rotation. In a multi-person collaborative motor imagery experiment, multiple users face a stimulus source (robot) at the same time, and observe the movements of the corresponding parts of the robot according to the single type of motor imagery they need to perform, so as to perform their own subtasks.

[0035] (2) Enhance users’ motor imagery performance through multi-person collaborative motor imagery tasks;

[0036] Participants first performed a single type of motor imagery, such as common upper limb motor imagery, tongue motor imagery, etc. After the execution, the 2D CNN-LSTM model was used to classify and calculate the classification accuracy;

[0037] This step is mainly used to determine the baseline, that is, the original performance level of the user's motor imagery. Generally speaking, the EEG signals collected by EEG electrodes are mixed with a lot of noise data. In order to solve this problem, through preprocessing, noise interference is reduced and the signal-to-noise ratio is improved, while the amount of data and computational load are reduced to improve the classification accuracy. Finally, the processed signal is divided into several time series segments of equal length to match the input dimension of the classification model and input into the 2D CNN-LSTM model. In this model, the CNN part includes 2 convolution blocks, each of which contains 2 convolution layers and followed by 1 pooling layer. Each convolution layer uses the ReLU activation function to reduce training time and improve performance. In the first convolution block, the number of convolution kernels in each convolution layer is 32, and in the second convolution block, the number of convolution kernels in each convolution layer is 16. In order to capture the difference in details, the convolution kernel size of the first convolution block is 3, and the convolution kernel size of the second convolution block is 2. All convolution kernel steps are set to 1, and "Same" is selected in the padding function to ensure The edge position information of the input signal is extracted as much as possible; after every two convolutional layers, the model adds a pooling layer to reduce the dimension of the features. The pool kernel size of each pooling layer is 2, and the step size is also 2; in addition, the Dropout method is added to each pooling layer to randomly discard part of the data by setting some hidden layer nodes to 0, so that the model does not rely too much on local features and reduces overfitting; all Dropout coefficients are 0.3, which means that each node in the hidden layer has a 30% chance of being turned into 0; the preprocessed EEG features are classified and the corresponding prediction labels are output.

[0038] After determining the baseline, multiple users are asked to observe a motor imagery stimulus simultaneously to perform motor imagery tasks, thereby enhancing the users' motor imagery abilities and performances through collaboration.

[0039] (3) Preprocessing EEG data;

[0040] Noise and other irrelevant components in EEG data are removed through filtering, independent component analysis, re-referencing, etc.; including electrode positioning, filtering, downsampling, segmentation, re-referencing, and independent component analysis.

[0041] (4) EEG signal feature extraction and classification;

[0042] Extract motor imagery features from the preprocessed data and classify motor imagery using a 2D CNN-LSTM network model;

[0043] For a single subject, the preprocessed EEG signal can be used for the following calculations: calculation of inter-brain synchronization, construction of a single brain functional network, and input into the preprocessed 2D CNN-LSTM model for single-person motor imagery classification prediction;

[0044] For multiple subjects, the preprocessed EEG signals of the subjects can be used for the following calculations in addition to the calculation process for a single subject: calculation of inter-brain synchronization between multiple subjects, calculating the inter-brain synchronization of each subject relative to other subjects among multiple subjects;

[0045] (5) Group brain analysis and motor imagery performance analysis during user collaboration;

[0046] (51) Calculate inter-brain synchronization: calculate the phase-locking value of the electrode pair and the inter-brain phase-locking value matrix;

[0047] (511) Calculating the phase-locked value of the electrode pair;

[0048] When analyzing the inter-brain synchronization between multiple users, according to Figure 3 In the order in , first calculate the phase-locked value between the brain electrode pairs. Assume that there is a time series EEG signal x(t) and y(t) corresponding to an electrode X and Y in each of the two individuals. After filtering in the specified frequency band, an 8-13 Hz bandpass filter is used here to filter the EEG signal in the α frequency band, and the Hilbert Transform is performed to align the instantaneous phase:

[0049]

[0050] in The imaginary part of the EEG signal x(t) is obtained by Hilbert transform (H function), which decomposes x(t) into two parts: phase and amplitude. At this time, the analytical signal corresponding to the EEG signal x(t) can be obtained.

[0051]

[0052] Where a(t) is the instantaneous amplitude, Φ x (t) is the instantaneous phase, which can be obtained by the arctan function in the complex plane, and its calculation formula is:

[0053]

[0054] After obtaining the instantaneous phase Φ of the EEG signals x(t) and y(t) x (t) and φ y (t) After that, the phase-locking value can be calculated for these two electrodes:

[0055]

[0056] Where N represents the Nth trial. Therefore, when calculating the phase-locking value, the instantaneous phase differences of multiple trials are often summed and averaged. The phase-locking value ranges from "0" to "1", "1" indicates that the phases of the two sets of EEG signals are completely synchronized, and "0" indicates that there is no phase synchronization between the two sets of EEG signals.

[0057] (512) Calculate the inter-brain phase-locking value matrix;

[0058] After collecting the EEG signals corresponding to the 8 electrodes of the two subjects X and Y in the same group, they were preprocessed to reduce the noise in the data. First, the EEG signals were bandpass filtered using a 0.5-100Hz bandpass filter, and the 50Hz power frequency interference was removed by a 50Hz notch filter. Then, the sampling rate was reduced from 1000Hz to 250Hz using the downsampling method, and all electrodes were averaged and rereferenced. Finally, independent component analysis was used to remove the electrooculogram and electromyography noise data. Then, the EEG data was divided into multiple data segments with a length of 10 seconds (the same as the duration of a single trial) to extract the corresponding features. All data should be checked before and after preprocessing to ensure that all necessary EEG data have been collected and there are no invalid or missing data segments. The task state data and resting state data were extracted separately, and the sample matrix size of the task state data was 8 channels × (250Hz × 6s) × 60 trials, and the sample matrix size of the resting state data was 8 channels × (250Hz × 4s) × 60 trials. Since the EEG signal communication mechanism between subjects in the collaborative motor imagery task is not yet fully understood, it is impossible to directly calculate the phase-locking value for a specific frequency band. The solution of this experiment is to filter these sample matrices through notch filters for the five common frequency bands of EEG signals (δ, θ, α, β, γ), and calculate the phase-locking values ​​in turn. Finally, the phase-locking value matrices with a size of 8 channels × 8 channels × 60 trials are obtained, and the sum and average processing is performed on the trial dimension to obtain the average task state phase-locking value matrix and the average resting state phase-locking value matrix.

[0059] (52) Construction of brain functional network

[0060] Figure 5 The process shown is the process of building a brain functional network, which mainly includes the following steps:

[0061] (521) Collecting EEG signals and preprocessing them (filtering, dividing into 5 frequency bands, and denoising);

[0062] (522) Define "nodes" and quantify the functional connectivity of EEG signals between any two channels. Since EEG reflects the characteristics of neural activity in the brain, the cerebral cortex area covered by each EEG electrode is regarded as a node. The definition of "edges" represents the quantification of the functional connectivity strength between a pair of specific nodes, reflecting the strength of functionally related information interaction in the brain network;

[0063] (523) Since the connections of brain networks often present complex characteristics and contain many weakly connected edges, in order to remove these weak connections in the brain network and at the same time ensure the small-world property of the brain network itself, after calculating the functional connections between different channel pairs, it is necessary to select an appropriate threshold and perform a binarization operation, and then construct a functional network based on the binarization results;

[0064] (524) Based on the constructed functional network, various topological properties in the complex brain network are explored, such as node degree, shortest path length, and small-world property.

[0065] Finally, after the multi-user collaborative motor imagery task, the participants performed their own single-class motor imagery again. After the execution, the 2D CNN-LSTM model was used to classify and calculate the classification accuracy, which was compared with the baseline to determine the degree of motor imagery enhancement. At the same time, the inter-brain synchronization results and brain network construction results were used to observe and explain the degree of collaboration between participants.

[0066] Example 2

[0067] See also Figure 5 , a schematic diagram of the structure of a multi-person collaborative motor imagery brain-computer interaction system for group brain computing provided by an embodiment of the present invention. For the sake of convenience, only the parts related to the embodiment of the present invention are shown. The functional modules / units in the system can be hardware modules / units and software modules / units, mainly including:

[0068] (1) a motor imagery task combination module, used to set a combined motor imagery task;

[0069] (2) Multi-person collaborative motor imagery module, which presents the combined motor imagery task and simultaneously collects motor imagery EEG signals of multiple people;

[0070] (3) an EEG signal preprocessing module, which preprocesses the collected or stored EEG signals;

[0071] (4) Feature extraction and classification module: extract features from the preprocessed EEG data for input and training of the 2D CNN-LSTM classification model.

[0072] (5) Multi-user collaborative group brain analysis module, which uses inter-brain synchronization calculation and functional brain network reconstruction to analyze the coherence of EEG signals between different users during multi-user collaboration in real time;

[0073] Example 3

[0074] A computer-readable storage medium, characterized in that a program is stored thereon, and when the program is executed by a processor, the classification method described in Example 1 is implemented.

[0075] Example 4

[0076] A computing device includes a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, the method described in Example 1 is implemented.

[0077] In the embodiments of the present invention, each module can be integrated into one, or can be deployed separately, or can be further split into multiple sub-modules. Each module can be distributed in the system of the embodiment according to the description of the embodiment, or can be changed accordingly and located in one or more systems different from the embodiments of the present invention.

[0078] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the embodiments of the present invention are not limited to any specific combination of hardware and software.

[0079] The above disclosure is only a specific embodiment of the present invention, but the present invention is not limited thereto. Those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Obviously, these changes and modifications should all fall within the scope of protection required by the present invention.

Claims

1. A multi-person collaborative motor imagery brain-computer interaction method for group brain computing, characterized in that: The following steps are involved: (1) Combining single motor imagery tasks into a composite motor imagery task; A collaborative motor imagery task is established by combining single motor imagery tasks, with multiple users wearing EEG caps to collect data at the same time; the traditional four-category motor imagery task is merged, requiring multiple users to complete the collaborative motor imagery task in real-time collaboration, with each user performing a part of the collaborative task individually; (2) Enhance users’ motor imagery performance through multi-person collaborative motor imagery tasks; Through the designed collaborative motor imagery task, multiple users can observe a motor imagery stimulus source at the same time to perform the motor imagery task, thus enhancing the motor imagery ability and performance of users in a collaborative way; Single-player task-multi-player collaborative task training-single-player task mode; users first determine the baseline in the single-player motor imagery task, and use the 2D CNN-LSTM model to measure the baseline classification accuracy; Then let the users experience the collaborative task and complete the collaborative motor imagery under the guidance of visual stimulation to enhance their motor imagery ability; (3) Preprocessing EEG signal data; Remove noise and other irrelevant components in EEG data through filtering, independent component analysis, and re-referencing; Includes electrode localization, filtering, downsampling, segmentation, re-referencing, and independent component analysis; (4) EEG signal feature extraction and classification; Extract motor imagery features from the preprocessed data and classify motor imagery using a 2D CNN-LSTM network model; For a single subject, the preprocessed EEG signal can be used for the following calculations: calculation of inter-brain synchronization, construction of a single brain functional network, and input into the preprocessed 2D CNN-LSTM model for single-person motor imagery classification prediction; For multiple subjects, the preprocessed EEG signals of the subjects can be used for the following calculations in addition to the calculation process for a single subject: calculation of inter-brain synchronization between multiple subjects, calculating the inter-brain synchronization of each subject relative to other subjects among multiple subjects; (5) Group brain analysis and motor imagery performance analysis during user collaboration; Inter-brain synchronization calculation and functional brain network reconstruction are used to analyze the coherence of EEG signals between different users during multi-user collaboration in real time. The 2D CNN-LSTM network outputs the motor imagery classification results and calculates the motor imagery classification accuracy to analyze the motor imagery performance of a single user and verify the effect of the collaborative training experiment.

2. The multi-person collaborative motor imagery brain-computer interaction method for group brain computing according to claim 1 is characterized in that: The CNN part of the 2D CNN-LSTM model described in step (2) includes two convolution blocks, each of which contains two convolution layers followed by one pooling layer. Each convolution layer uses the ReLU activation function to reduce training time and improve performance. In the first convolution block, the number of convolution kernels in each convolution layer is 32, and in the second convolution block, the number of convolution kernels in each convolution layer is 16. In order to capture the difference in details, the convolution kernel size of the first convolution block is 3, and the convolution kernel size of the second convolution block is 2. The step size of all convolution kernels is set to 1, and "Same" is selected in the Padding function. To ensure that the edge position information of the input signal is extracted as much as possible; after every 2 convolutional layers, the model adds a pooling layer to reduce the dimension of the features. The pool kernel size of each pooling layer is 2, and the step size is also 2; in addition, the Dropout method is added to each pooling layer. By setting some hidden layer nodes to 0, some data are randomly discarded, so that the model does not rely too much on local features and reduces overfitting; all Dropout coefficients are 0.3, which means that each node in the hidden layer has a 30% chance of being turned into 0; the preprocessed EEG features are classified, and finally the corresponding prediction labels are output.

3. The multi-person collaborative motor imagery brain-computer interaction method for group brain computing according to claim 1 is characterized in that: Step (4) specifically includes: (41) Calculate inter-brain synchronization: calculate the phase-locking value of the electrode pair and the inter-brain phase-locking value matrix; (42) Constructing functional brain networks.

4. The multi-person collaborative motor imagery brain-computer interaction method for group brain computing as claimed in claim 3, characterized in that: Step (41) comprises: (411) Calculate the phase-locked value of the electrode pair: Assume that there are two individuals each with a time series EEG signal x(t) and y(t) corresponding to an electrode X and Y. After filtering in the specified frequency band, an 8-13 Hz bandpass filter is used here to obtain the EEG signal in the α frequency band, and the Hilbert Transform is performed to align them to calculate the instantaneous phase: in The imaginary part of the EEG signal x(t) is obtained by Hilbert transforming it. By decomposing x(t) into two parts, phase and amplitude, we can get the analytical signal corresponding to the EEG signal x(t). Where a(t) is the instantaneous amplitude, Φ x (t) is the instantaneous phase, which can be obtained by the arctan function in the complex plane, and its calculation formula is: After obtaining the instantaneous phase φ of the EEG signals x(t) and y(t) x (t) and Φ y (t) After that, the phase-locking value can be calculated for these two electrodes: Where N represents the Nth trial. Therefore, when calculating the phase-locking value, the instantaneous phase differences of multiple trials are summed and averaged. The phase-locking value ranges from "0" to "1", "1" indicates that the phases of the two sets of EEG signals are completely synchronized, and "0" indicates that there is no phase synchronization between the two sets of EEG signals. (412) Calculate the inter-brain phase-locking value matrix: After collecting the EEG signals corresponding to the 8 electrodes of the two subjects X and Y in the same group, they were preprocessed to reduce the noise in the data. First, the EEG signals were band-pass filtered using a 0.5-100Hz band-pass filter, and the 50Hz power frequency interference was removed by a 50Hz notch filter. Then, the sampling rate was reduced from 1000Hz to 250Hz using the downsampling method, and all electrodes were averaged and re-referenced. Finally, independent component analysis was used to remove the electrooculogram and electromyography noise data. Then, the EEG data was divided into multiple data segments with the same single trial duration to extract the corresponding features. All data needed to be checked before and after preprocessing to ensure that all necessary EEG data were collected and there was no invalid or missing data. According to the segment; the task-state data and resting-state data were extracted respectively, and the sample matrix size of the task-state data was 8 channels × (250 Hz × 6s) × 60 trials, and the sample matrix size of the resting-state data was 8 channels × (250 Hz × 4s) × 60 trials; because the EEG signal communication mechanism between subjects in the collaborative motor imagery task is not yet fully understood, it is impossible to directly calculate the phase-locking value for a specific frequency band; these sample matrices were filtered through notch filters for 5 common frequency bands of EEG signals, and the phase-locking values ​​were calculated in turn, and finally the phase-locking value matrices of 8 channels × 8 channels × 60 trials were obtained, and the sum and average processing was performed on the trial dimension to obtain the average task-state phase-locking value matrix and the average resting-state phase-locking value matrix.

5. The multi-person collaborative motor imagery brain-computer interaction method for group brain computing as claimed in claim 3, characterized in that: Step (42) comprises: (421) Collecting EEG signals and preprocessing them, including filtering and dividing them into five frequency bands; (422) "Nodes" are defined and the functional connections between any two channels of EEG signals are quantified. Since EEG reflects the characteristics of neural activity in the brain, the cerebral cortex area covered by each EEG electrode is regarded as a node. The definition of "edges" represents the quantification of the functional connection strength between a pair of specific nodes, reflecting the strength of functional-related information interaction in the brain network. (423) Since the connections of brain networks often present complex characteristics and contain many weakly connected edges, in order to remove these weak connections in the brain network and at the same time ensure the small-world property of the brain network itself, after calculating the functional connections between different channel pairs, it is necessary to select an appropriate threshold and perform a binarization operation, and then construct a functional network based on the binarization results; (424) Based on the constructed functional network, various topological properties in the complex brain network are explored, including node degree, shortest path length, and small-world properties.

6. A multi-person collaborative motor imagery brain-computer interaction system for group brain computing, characterized by include: A motor imagery task combination module, used to set a combined motor imagery task; The multi-person collaborative motor imagery module presents the combined motor imagery task and simultaneously collects motor imagery EEG signals of multiple people; An EEG signal preprocessing module is used to preprocess the collected or stored EEG signals; Feature extraction and classification module, which extracts features from preprocessed EEG data for input and training of 2D CNN-LSTM classification model; The multi-user collaborative group brain analysis module uses inter-brain synchronization calculation and functional brain network reconstruction to analyze the coherence of EEG signals between different users during multi-user collaboration in real time.

7. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, the multi-person collaborative motor imagery brain-computer interface system described in any one of claims 1-5 is implemented.

8. A computing device comprising a memory and a processor, wherein: The memory stores executable codes, and when the processor executes the executable codes, the method according to any one of claims 1 to 5 is implemented.

Images