Electroencephalogram data processing method and system, readable storage medium and computer

Abstract

The invention provides an electroencephalogram data processing method and system, a readable storage medium and a computer, and the method comprises the steps: generating an SSVEP potential signal through a preset signal stimulator, and carrying out the signal processing of the SSVEP potential signal, so as to obtain corresponding electroencephalogram data; performing data preprocessing on the electroencephalogram data to obtain corresponding preprocessed data, constructing a preliminary data processing model, and introducing jump connection into the preliminary data processing model to obtain a corresponding deep learning model; and performing data processing on the preprocessed data through a deep learning model to obtain a corresponding data identification result, and converting the identification result into a corresponding control signal to realize control of the external equipment. According to the method, the local feature extraction capability of the CNN and the long-distance dependence modeling capability of the Transform are combined, and different types of features are processed through a multi-branch design and a weighted fusion mechanism.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an electroencephalogram (EEG) data processing method, system, readable storage medium, and computer. Background Art

[0002] Most current SSVEP-based brain-computer interfaces (BCIs) rely on traditional machine learning algorithms, single convolutional neural networks, or a single Transformer architecture. Traditional machine learning algorithms have good generalization capabilities, but are slow to recognize signals. Typically, approximately four seconds of EEG signal acquisition are required to produce a good recognition result. This results in significant latency when controlling external devices with SSVEP. Traditional convolutional neural networks (CNNs) perform well in processing images and sequential data, but struggle to capture long-range dependencies. The Transformer architecture, while effective at processing long-range dependencies in sequential data through its self-attention mechanism, may not be as effective as CNNs when processing non-sequential data. Therefore, both single convolutional neural networks and Transformer architectures have drawbacks. SSVEP signals represent the brain's sustained response to visual stimuli of a specific frequency, manifesting as stable rhythms corresponding to the stimulus frequency in the electroencephalogram (EEG). These rhythms can exhibit complex interrelationships across different parts of the signal, potentially spanning long time intervals. Summary of the Invention

[0003] Based on this, the purpose of the present invention is to provide an EEG data processing method, system, readable storage medium and computer to at least solve the deficiencies in the above-mentioned technology.

[0004] The present invention provides a method for processing EEG data, comprising: Generating an SSVEP potential signal through a preset signal stimulator, and performing signal processing on the SSVEP potential signal to obtain corresponding EEG data; Performing data preprocessing on the EEG data to obtain corresponding preprocessed data, constructing a preliminary data processing model, and introducing skip connections into the preliminary data processing model to obtain a corresponding deep learning model; The pre-processed data is processed by the deep learning model to obtain a corresponding data recognition result, and the recognition result is converted into a corresponding control signal to realize the control of the external device.

[0005] Furthermore, the step of preprocessing the EEG data to obtain corresponding preprocessed data includes: Downsampling the EEG data to reduce the frequency of the EEG data to 250 Hz; The down-sampled EEG data is filtered to obtain corresponding pre-processed data, wherein the filtering frequency range of the filtering processing includes 3-14 Hz, 9-26 Hz, 14-38 Hz, and 19-50 Hz.

[0006] Furthermore, the steps of constructing a preliminary data processing model and introducing skip connections into the preliminary data processing model to obtain a corresponding deep learning model include: Constructing a convolutional module, wherein the convolutional module consists of three convolutional layers, and employing skip connections after the second and third convolutional layers to enhance features; Construct a coding module, an adaptive feature fusion module and a fully connected layer, and construct a deep learning model based on the convolution module, the coding module, the adaptive feature fusion module and the fully connected layer.

[0007] Furthermore, the first convolutional layer and the second convolutional layer in the convolutional module are both composed of a convolution kernel, a batch normalization layer, a GeLu activation function and a Dropout layer, and the third convolutional layer is composed of a convolution kernel and a batch normalization layer. The convolution kernel size of the first convolutional layer is (channel, 1), the convolution kernel size of the second convolutional layer is (1, K), and the convolution kernel size of the third convolutional layer is (1, 5), where channel represents the number of EEG data channels, and K represents the number of data points in different time windows.

[0008] Furthermore, the step of processing the preprocessed data using the deep learning model to obtain corresponding data recognition results includes: Performing convolution processing on the preprocessed data using the convolution module of the deep learning model to obtain corresponding feature data; Performing feature extraction on the feature data using the encoding module of the deep learning model to obtain a number of global features; The adaptive feature fusion module of the deep learning model is used to perform weighted summation on the global features, and data compression is performed on the features after weighted summation. The data compression results are input into the fully connected layer of the deep learning model for flattening and full connection processing to obtain the corresponding data recognition results.

[0009] The present invention also provides an EEG data processing system, comprising: A signal processing module is used to generate an SSVEP potential signal through a preset signal stimulator and perform signal processing on the SSVEP potential signal to obtain corresponding EEG data; A data preprocessing module is used to perform data preprocessing on the EEG data to obtain corresponding preprocessed data, construct a preliminary data processing model, and introduce skip connections into the preliminary data processing model to obtain a corresponding deep learning model; A data recognition module is used to process the preprocessed data through the deep learning model to obtain corresponding data recognition results, and convert the recognition results into corresponding control signals to achieve control of external devices.

[0010] Furthermore, the data preprocessing module includes: a downsampling processing unit, configured to perform downsampling processing on the EEG data so as to reduce the frequency of the EEG data to 250 Hz; The filtering processing unit is used to filter the downsampled EEG data to obtain corresponding preprocessed data, wherein the filtering frequency range of the filtering processing includes 3-14Hz, 9-26Hz, 14-38Hz, and 19-50Hz.

[0011] Furthermore, the data preprocessing module includes: A module construction unit, configured to construct a convolution module, wherein the convolution module is composed of three convolution layers, and features are enhanced by using jump connections after the second and third convolution layers; A model construction unit is used to construct a coding module, an adaptive feature fusion module and a fully connected layer, and to construct a deep learning model based on the convolution module, the coding module, the adaptive feature fusion module and the fully connected layer.

[0012] Furthermore, the data identification module includes: A convolution processing unit, configured to perform convolution processing on the preprocessed data using a convolution module of the deep learning model to obtain corresponding feature data; A feature extraction unit, configured to extract features from the feature data using the encoding module of the deep learning model to obtain a plurality of global features; A data recognition unit is used to use the adaptive feature fusion module of the deep learning model to perform weighted summation on each of the global features, and to perform data compression on the features after the weighted summation, and to input the data compression result into the fully connected layer of the deep learning model for flattening and full connection processing to obtain the corresponding data recognition result.

[0013] The present invention also provides a readable storage medium on which a computer program is stored. When the program is executed by a processor, the above-mentioned EEG data processing method is implemented.

[0014] The present invention also proposes a computer, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned EEG data processing method when executing the computer program.

[0015] The EEG data processing method, system, readable storage medium and computer of the present invention construct a preliminary data processing model and introduce jump connections into the preliminary data processing model to construct a corresponding deep learning model. The deep learning model combining the convolutional neural network CNN and the Transformer architecture is used to realize the classification of steady-state visual evoked potential (SSVEP) signals. The position encoding module is used to enhance the ability to capture temporal features, and the Transformer encoder is used to effectively capture long-distance dependencies. The jump connections in the convolution process reduce the gradient vanishing problem and improve information flow. The features of different filtering ranges are dynamically weighted and fused, and dropout layers are added in multiple places to reduce overfitting and improve generalization ability. The model combines the local feature extraction capability of CNN and the long-distance dependency modeling capability of Transformer, and processes different types of features through multi-branch design and weighted fusion mechanism. It performs well in the field of brain-computer interface (BCI), especially in SSVEP tasks that require fast and accurate identification of visual evoked potentials. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of the EEG data processing method in the first embodiment of the present invention; Figure 2 The electrode distribution diagram of the data test in the first embodiment of the present invention; Figure 3 is a structural block diagram of an EEG data processing system in a second embodiment of the present invention; Figure 4 FIG. 4 is a structural block diagram of a computer in a third embodiment of the present invention.

[0017] The following specific embodiments will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0018] To facilitate understanding of the present invention, the present invention will be described more fully below with reference to the accompanying drawings. The drawings illustrate several embodiments of the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the present invention.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one skilled in the art to which this invention pertains. The terms used in this specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] Example 1 See also Figure 1 , which shows the EEG data processing method in the first embodiment of the present invention, and specifically includes steps S101 to S105: S101, generating an SSVEP potential signal through a preset signal stimulator, and performing signal processing on the SSVEP potential signal to obtain corresponding EEG data; During specific implementation, a preset signal stimulator is used to induce an SSVEP potential signal, and the signal is collected and amplified using an electroencephalogram (EEG) acquisition device, mainly collecting data from the occipital lobe area of ​​the brain. In this embodiment, a wireless EEG device is used to collect EEG data, and the collected data is transmitted via a WIFI module.

[0021] S102, performing data preprocessing on the EEG data to obtain corresponding preprocessed data, constructing a preliminary data processing model, and introducing skip connections into the preliminary data processing model to obtain a corresponding deep learning model; Furthermore, the step S102 specifically includes steps S1021 and S1022: S1021, downsampling the EEG data to reduce the frequency of the EEG data to 250 Hz; S1022 , filtering the down-sampled EEG data to obtain corresponding pre-processed data, wherein the filtering frequency range of the filtering includes 3-14 Hz, 9-26 Hz, 14-38 Hz, and 19-50 Hz.

[0022] In the specific implementation, the collected EEG data was downsampled, reducing the original 1000Hz to 250Hz, using decimation. The data was filtered using a Butterworth bandpass filter, creating a 6th-order Butterworth bandpass filter. The higher the order, the steeper the transition between the passband and the stopband. The filter frequency range consists of four bands: 3-14Hz, 9-26Hz, 14-38Hz, and 19-50Hz. After filtering in four different frequency bands, four different EEG data were output. This multi-band design can capture SSVEP features in different frequency ranges.

[0023] Furthermore, the step S102 further includes steps S1121-S1122: S1121, constructing a convolution module, wherein the convolution module is composed of three convolution layers, and skip connections are used after the second and third convolution layers to enhance features; S1122, construct a coding module, an adaptive feature fusion module and a fully connected layer, and construct a deep learning model based on the convolution module, the coding module, the adaptive feature fusion module and the fully connected layer.

[0024] In a specific implementation, a convolution module is constructed, wherein the first and second convolutional layers in the convolutional module are each composed of a convolution kernel, a batch normalization layer, a GeLu activation function, and a dropout layer. The third convolutional layer is composed of a convolution kernel and a batch normalization layer. The convolution kernel size of the first convolutional layer is (channel, 1), the convolution kernel size of the second convolutional layer is (1, K), and the convolution kernel size of the third convolutional layer is (1, 5). Wherein, channel represents the number of EEG data channels, and K represents the number of data points in different time windows. For example, if 0.1s of EEG data is input with a 250Hz sampling rate, then K is 0.1*250=25. The information after the first convolution layer passes through a connected convolution and is then added to the output of the third convolution layer using a skip connection to retain more information.

[0025] In this embodiment, skip connections are introduced in the convolutional blocks to reduce the vanishing gradient problem and improve the flow of information in the network. The output of the first convolution is added to the output of the third convolution after a single convolution. This can be expressed as follows:

[0026] in, is the output of the first convolution, is the first convolution, is the third convolution. The jump connection process is Figure 2 In the figure, part of the data flow on the Conv layer module.

[0027] Furthermore, an encoding module, an adaptive feature fusion module, and a fully connected layer are constructed, and a deep learning model is constructed based on the convolution module, the encoding module, the adaptive feature fusion module, and the fully connected layer. The encoding module includes a position encoder, a normalization layer, and a Transformer encoder layer. The Transformer Encoder processes the data in the following steps: a. Data reshaping: Convert the feature dimensions of the convolution output to the Transformer input format; b. Input normalization: Apply LayerNorm to normalize the input to stabilize training; c. Positional encoding: Adding a positional encoder enables the model to perceive sequence position information; d. Attention mask generation: This implements a local attention mechanism, where each position can only focus on itself and a limited number of subsequent positions. e. Transformer encoding: The prepared input and mask are passed to the Transformer encoder, which contains a self-attention mechanism and a feedforward network; f. Data reshape back to convolutional format: Reshape the Transformer output back to convolutional format.

[0028] In this embodiment, the adaptive feature fusion module uses trainable weights to perform weighted fusion on features to better integrate features extracted under different filtering ranges.

[0029] The GELU activation function is used in the model to improve the nonlinear expression ability of the model. The model uses layer normalization after multiple convolutional layers to accelerate the training process and improve the stability of the model. The model adds dropout layers in multiple places to reduce overfitting and improve the generalization ability of the model.

[0030] S103, performing data processing on the preprocessed data through the deep learning model to obtain a corresponding data recognition result, and converting the recognition result into a corresponding control signal to realize the control of the external device.

[0031] Furthermore, the step S103 specifically includes steps S1031 to S1033: S1031, using the convolution module of the deep learning model to perform convolution processing on the preprocessed data to obtain corresponding feature data; S1032, performing feature extraction on the feature data using the encoding module of the deep learning model to obtain a plurality of global features; S1033, using the adaptive feature fusion module of the deep learning model to perform weighted summation on each of the global features, and performing data compression on the features after weighted summation, and inputting the data compression result into the fully connected layer of the deep learning model for flattening and full connection processing to obtain the corresponding data recognition result.

[0032] In specific implementation, the convolution module is used to perform convolution processing on the preprocessed data to obtain the corresponding feature data. The processed features are input into the Transformer encoding module. The Transformer module is mainly composed of a position encoder and a Transformer encoder. First, the EEG features after the convolution module are positionally encoded to add position information to the features. Then, they are input into a Transformer encoder and the self-attention mechanism in the Transformer encoder is used to further extract the EEG global features. After processing by the Transformer encoding module, four feature branches are output. The four branch features come from the EEG data obtained by filtering in four different frequency bands in the data preprocessing part. During the training process, the convolution module and the Transformer module share weights.

[0033] Furthermore, the four feature branches are fused using an adaptive feature fusion module. During training, this module automatically updates the weight coefficients of the different branches through backpropagation to learn the optimal feature fusion weights. This allows the model to automatically learn the importance of each branch without manually setting fixed weights. This approach allows the model to dynamically adjust its focus on different feature branches during training. The fused features are then subjected to a convolution with a kernel size of (1, K / 5) to further compress and integrate channel information, reducing the number of channels to 32.

[0034] Specifically, the fully connected layer is used to output the final recognition result. The fully connected layer includes a flattening operation and a fully connected operation. The flattening operation flattens the features into a "flattened" process, converts the features into a one-dimensional space, and then performs a fully connected operation to output 4 classification features. Finally, the softmax function is used to output the category with the highest probability.

[0035] In this embodiment, when merging features from different branches, the learned weight coefficients are used to adaptively perform weighted summation to better integrate information from different branches. The feature weighted fusion process can be expressed as:

[0036] in, It is the output after the Transformer Encoder Features, The weight is calculated by the softmax function: , is a learnable parameter.

[0037] Using softmax ensures that the sum of all weights is 1, which allows the model to automatically learn the importance of each branch without manually setting fixed weights. This approach allows the model to dynamically adjust the attention paid to different feature branches during training. The fully connected layer is used to output the final recognition result. The fully connected layer includes a flattening operation and a fully connected operation. The flattening operation flattens the features into a "flattened" process, converts the features into a one-dimensional space, and then performs a fully connected operation to output 4 classification features. Finally, the softmax function is used to output the category with the highest probability.

[0038] After the above processing, the original EEG signal outputs the recognition result, which is converted into a control signal to control the external device.

[0039] In this example, the dataset used a four-target SSVEP design, with flicker frequencies of 5.45 Hz, 6.67 Hz, 8.57 Hz, and 12 Hz, and displayed in four locations (up, down, left, and right). Each trial consisted of a 4-second target attention phase, a 4-second stimulus flicker phase, and a 2-second rest phase. Each stimulus target was displayed 25 times. Therefore, the offline EEG data consisted of 100 trials (4 directions × 25 times). The electrode distribution is as follows: Figure 2 ; Using the data of ten subjects, the model was trained in a time window of 0.1s-1s. The final results are shown in Table 1: Table 1: Recognition accuracy of ten subjects in different time windows

[0040] Experimental results show that the present invention achieves better recognition performance within a shorter time window.

[0041] In summary, the EEG data processing method in the above-mentioned embodiment of the present invention constructs a preliminary data processing model and introduces jump connections into the preliminary data processing model to construct a corresponding deep learning model. The deep learning model combining the convolutional neural network CNN and the Transformer architecture is used to realize the classification of steady-state visual evoked potential (SSVEP) signals. The position encoding module is used to enhance the ability to capture temporal features, and the Transformer encoder is used to effectively capture long-distance dependencies. The jump connections in the convolution process reduce the gradient vanishing problem and improve the information flow. The features of different filtering ranges are fused by dynamic weighting, and dropout layers are added in multiple places to reduce overfitting and improve generalization ability. The model combines the local feature extraction capability of CNN and the long-distance dependency modeling capability of Transformer, and processes different types of features through multi-branch design and weighted fusion mechanism. It performs well in the field of brain-computer interface (BCI), especially in SSVEP tasks that require fast and accurate identification of visual evoked potentials.

[0042] Example 2 Another aspect of the present invention is to provide an EEG data processing system. Figure 3 , which shows an EEG data processing system in a second embodiment of the present invention, the system includes: The signal processing module 11 is used to generate an SSVEP potential signal through a preset signal stimulator and perform signal processing on the SSVEP potential signal to obtain corresponding EEG data; A data preprocessing module 12 is used to perform data preprocessing on the EEG data to obtain corresponding preprocessed data, construct a preliminary data processing model, and introduce skip connections into the preliminary data processing model to obtain a corresponding deep learning model; Furthermore, the data preprocessing module 12 includes: a downsampling processing unit, configured to perform downsampling processing on the EEG data so as to reduce the frequency of the EEG data to 250 Hz; The filtering processing unit is used to filter the downsampled EEG data to obtain corresponding preprocessed data, wherein the filtering frequency range of the filtering processing includes 3-14Hz, 9-26Hz, 14-38Hz, and 19-50Hz.

[0043] Furthermore, the data preprocessing module 12 includes: A module construction unit, configured to construct a convolution module, wherein the convolution module is composed of three convolution layers, and features are enhanced by using jump connections after the second and third convolution layers; A model construction unit is used to construct a coding module, an adaptive feature fusion module and a fully connected layer, and to construct a deep learning model based on the convolution module, the coding module, the adaptive feature fusion module and the fully connected layer.

[0044] The data recognition module 13 is used to process the pre-processed data through the deep learning model to obtain a corresponding data recognition result, and convert the recognition result into a corresponding control signal to realize the control of the external device.

[0045] Furthermore, the data identification module 13 includes: A convolution processing unit, configured to perform convolution processing on the preprocessed data using a convolution module of the deep learning model to obtain corresponding feature data; A feature extraction unit, configured to extract features from the feature data using the encoding module of the deep learning model to obtain a plurality of global features; A data recognition unit is used to use the adaptive feature fusion module of the deep learning model to perform weighted summation on each of the global features, and to perform data compression on the features after the weighted summation, and to input the data compression result into the fully connected layer of the deep learning model for flattening and full connection processing to obtain the corresponding data recognition result.

[0046] The functions or operation steps implemented when the above modules and units are executed are substantially the same as those in the above method embodiments and will not be repeated here.

[0047] The EEG data processing system provided in the embodiment of the present invention has the same implementation principle and technical effects as those in the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the system embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0048] Example 3 The present invention also provides a computer, see Figure 4 , shown is a computer in the third embodiment of the present invention, including a memory 10, a processor 20, and a computer program 30 stored in the memory 10 and executable on the processor 20. When the processor 20 executes the computer program 30, the above-mentioned EEG data processing method is implemented.

[0049] The memory 10 includes at least one type of readable storage medium, including flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 10 may be an internal storage unit of a computer, such as the computer's hard disk. In other embodiments, the memory 10 may also be an external storage device, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. Furthermore, the memory 10 may include both an internal storage unit of the computer and an external storage device. The memory 10 can be used not only to store application software installed in the computer and various types of data, but also to temporarily store data that has been output or is about to be output.

[0050] Among them, in some embodiments, the processor 20 can be an electronic control unit (Electronic Control Unit, abbreviated as ECU, also known as a vehicle computer), a central processing unit (CPU), a controller, a microcontroller, a microprocessor or other data processing chip, used to run the program code stored in the memory 10 or process data, such as executing access restriction programs.

[0051] It should be pointed out that Figure 4 The structure shown does not constitute a limitation of the computer. In other embodiments, the computer may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0052] An embodiment of the present invention further provides a readable storage medium having a computer program stored thereon, which implements the above-mentioned EEG data processing method when executed by a processor.

[0053] Those skilled in the art will appreciate that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device), or in conjunction with such instruction execution system, apparatus, or device. For purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by an instruction execution system, apparatus, or device, or in conjunction with such instruction execution system, apparatus, or device.

[0054] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0055] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the aforementioned embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one or a combination of the following technologies known in the art may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0056] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0057] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for processing electroencephalogram data, characterized in that: include: Generating an SSVEP potential signal through a preset signal stimulator, and performing signal processing on the SSVEP potential signal to obtain corresponding EEG data; Performing data preprocessing on the EEG data to obtain corresponding preprocessed data, constructing a preliminary data processing model, and introducing skip connections into the preliminary data processing model to obtain a corresponding deep learning model; The pre-processed data is processed by the deep learning model to obtain a corresponding data recognition result, and the recognition result is converted into a corresponding control signal to realize the control of the external device.

2. The method for processing EEG data according to claim 1, wherein: The step of preprocessing the EEG data to obtain corresponding preprocessed data includes: Downsampling the EEG data to reduce the frequency of the EEG data to 250 Hz; The down-sampled EEG data is filtered to obtain corresponding pre-processed data, wherein the filtering frequency range of the filtering processing includes 3-14 Hz, 9-26 Hz, 14-38 Hz, and 19-50 Hz.

3. The method for processing EEG data according to claim 1, wherein: The steps of constructing a preliminary data processing model and introducing skip connections into the preliminary data processing model to obtain a corresponding deep learning model include: Constructing a convolutional module, wherein the convolutional module consists of three convolutional layers, and employing skip connections after the second and third convolutional layers to enhance features; Construct a coding module, an adaptive feature fusion module and a fully connected layer, and construct a deep learning model based on the convolution module, the coding module, the adaptive feature fusion module and the fully connected layer.

4. The method for processing EEG data according to claim 3, wherein: The first convolutional layer and the second convolutional layer in the convolutional module are both composed of a convolution kernel, a batch normalization layer, a GeLu activation function and a Dropout layer. The third convolutional layer is composed of a convolution kernel and a batch normalization layer. The convolution kernel size of the first convolutional layer is (channel, 1), the convolution kernel size of the second convolutional layer is (1, K), and the convolution kernel size of the third convolutional layer is (1, 5), where channel represents the number of EEG data channels and K represents the number of data points in different time windows.

5. The method for processing EEG data according to claim 1, wherein: The step of processing the pre-processed data by using the deep learning model to obtain a corresponding data recognition result includes: Performing convolution processing on the preprocessed data using the convolution module of the deep learning model to obtain corresponding feature data; Performing feature extraction on the feature data using the encoding module of the deep learning model to obtain a number of global features; The adaptive feature fusion module of the deep learning model is used to perform weighted summation on the global features, and data compression is performed on the features after weighted summation. The data compression results are input into the fully connected layer of the deep learning model for flattening and full connection processing to obtain the corresponding data recognition results.

6. An electroencephalogram data processing system, characterized in that: include: A signal processing module is used to generate an SSVEP potential signal through a preset signal stimulator and perform signal processing on the SSVEP potential signal to obtain corresponding EEG data; A data preprocessing module is used to perform data preprocessing on the EEG data to obtain corresponding preprocessed data, construct a preliminary data processing model, and introduce skip connections into the preliminary data processing model to obtain a corresponding deep learning model; A data recognition module is used to process the preprocessed data through the deep learning model to obtain corresponding data recognition results, and convert the recognition results into corresponding control signals to achieve control of external devices.

7. The EEG data processing system according to claim 6, characterized in that: The data preprocessing module includes: a downsampling processing unit, configured to perform downsampling processing on the EEG data so as to reduce the frequency of the EEG data to 250 Hz; The filtering processing unit is used to filter the downsampled EEG data to obtain corresponding preprocessed data, wherein the filtering frequency range of the filtering processing includes 3-14Hz, 9-26Hz, 14-38Hz, and 19-50Hz.

8. The EEG data processing system according to claim 6, characterized in that: The data preprocessing module includes: A module construction unit, configured to construct a convolution module, wherein the convolution module is composed of three convolution layers, and features are enhanced by using jump connections after the second and third convolution layers; A model construction unit is used to construct a coding module, an adaptive feature fusion module and a fully connected layer, and to construct a deep learning model based on the convolution module, the coding module, the adaptive feature fusion module and the fully connected layer.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the EEG data processing method according to any one of claims 1 to 6 is implemented.

10. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the EEG data processing method according to any one of claims 1 to 6 is implemented.

Images