Brain-computer interface imagination marker design method, device, equipment and storage medium

Abstract

The application provides a brain-computer interface imagination marker design method, device, equipment and storage medium, and relates to the technical field of signal processing. A plurality of semantic stimuli and corresponding electroencephalogram signals of each semantic stimulus are acquired; the electroencephalogram signals corresponding to each semantic stimulus are encoded and processed to obtain the hidden encoding of each semantic stimulus; based on the hidden encoding of each semantic stimulus, a target semantic stimulus set is determined from all semantic stimuli, the target semantic stimulus set includes a plurality of target semantic stimuli, and the electroencephalogram signals corresponding to the plurality of target semantic stimuli have the largest classification boundary; each target semantic stimulus is taken as a brain-computer interface imagination marker, and a preset brain-computer interface task is performed, so that the decoding accuracy of the brain-computer interface task is improved, and the interaction demand is met.

Description

Technical Field

[0001] This invention relates to the field of signal processing technology, and more specifically, to a method, apparatus, device, and storage medium for designing imaginary markers for brain-computer interfaces. Background Technology

[0002] Brain-computer interface (BCI) is a technology that uses electroencephalogram (EEG) signals to control external devices, providing a new way for people with or without disabilities to interact. In BCI, imagined markers refer to the concepts or objects that the test subject imagines while performing a task; these can serve as characteristics of the EEG signals to guide the movement or operation of the external device.

[0003] In traditional brain-computer interfaces (BCIs), commonly used visual markers include limb movements (such as movements of the left or right hand), movements of the target device (such as movements of a robotic arm), and other specific commands. These visual markers are usually selected based on experience or intuition and are limited in number. BCIs that use these visual markers as commands can only provide limited control options and cannot meet complex or diverse interaction needs. Currently, there is a lack of a design method for visual markers that maximize the discriminative power of EEG signals. Summary of the Invention

[0004] This invention provides a method, apparatus, device, and storage medium for designing brain-computer interface imaginary markers, which can maximize the distinguishability of brain electrical signals and meet interactive requirements.

[0005] The technical solution of this invention can be implemented as follows:

[0006] In a first aspect, the present invention provides a method for designing imaginary markers for brain-computer interfaces, the method comprising:

[0007] Acquire multiple semantic stimuli and the corresponding EEG signals for each semantic stimulus;

[0008] The EEG signal corresponding to each semantic stimulus is encoded to obtain the hidden code of each semantic stimulus;

[0009] Based on the hidden code of each of the semantic stimuli, a target semantic stimulus set is determined from all the semantic stimuli. The target semantic stimulus set includes multiple target semantic stimuli, and the EEG signals corresponding to the multiple target semantic stimuli have the largest classification boundary.

[0010] Each of the target semantic stimuli is used as a brain-computer interface imaginative marker to execute a preset brain-computer interface task.

[0011] Optionally, the step of encoding the EEG signal corresponding to each semantic stimulus to obtain the hidden code of each semantic stimulus includes:

[0012] For each of the semantic stimuli, each EEG signal corresponding to the semantic stimulus is input into a pre-trained neural signal encoder to obtain the hidden code of each EEG signal;

[0013] The mean of the hidden codes of all the EEG signals is used as the hidden code of the semantic stimulus.

[0014] Optionally, the implicit encoding of the semantic stimulus satisfies the following formula:

[0015]

[0016] Where, x i For the semantic stimulus, y is a set including each EEG signal corresponding to the semantic stimulus. j Let m be the j-th EEG signal corresponding to the semantic stimulus, m be the number of EEG signals corresponding to the semantic stimulus, E be the pre-trained neural signal encoder, and h be the number of EEG signals corresponding to the semantic stimulus. i This is the implicit encoding of semantic stimuli.

[0017] Optionally, multiple target semantic stimuli in the target semantic stimulus set satisfy the following formula:

[0018]

[0019]

[0020] Where Ω is the set of target semantic stimuli, x i Let x be the i-th target semantic stimulus in the target semantic stimulus set. j h is the j-th target semantic stimulus in the target semantic stimulus set. i h is the hidden code of the i-th target semantic stimulus in the target semantic stimulus set. j Let S(x) be the hidden code of the j-th target semantic stimulus in the target semantic stimulus set. i ,x j ) is the target semantic stimulus x i The corresponding EEG signal and the target semantic stimulus x j The similarity of the corresponding EEG signals, F θ This refers to the classification boundary of EEG signals corresponding to multiple target semantic stimuli in the target semantic stimulus set.

[0021] Optionally, the step of determining a target semantic stimulus set from all the semantic stimuli based on the implicit encoding of each of the semantic stimuli includes:

[0022] Create an empty set to be diffused;

[0023] Based on the hidden encoding of each semantic stimulus, the set to be diffused is subjected to multiple diffusion processes to obtain the target semantic stimulus set, wherein each diffusion process adds one of the semantic stimuli to the set to be diffused.

[0024] Optionally, the step of performing multiple diffusion processes on the set to be diffused based on the implicit encoding of each semantic stimulus to obtain the target semantic stimulus set includes:

[0025] During each diffusion process, the semantic stimulus added to the set to be diffused during the previous diffusion process is used as the first semantic stimulus, each semantic stimulus already added to the set to be diffused is used as the second semantic stimulus, and each semantic stimulus not added to the set to be diffused is used as the third semantic stimulus.

[0026] For each of the third semantic stimuli, the classification boundary corresponding to the third semantic stimulus is calculated based on the hidden code of the third semantic stimulus and the hidden code of each of the second semantic stimuli, thus obtaining the classification boundary corresponding to each of the third semantic stimuli.

[0027] Based on the classification boundary of the first semantic stimulus and each of the third semantic stimuli, it is determined whether a diffusible third semantic stimulus can be identified from all the third semantic stimuli, and the diffusible third semantic stimulus satisfies the first semantic stimulus under a preset condition.

[0028] If so, the diffusible third semantic stimulus is added to the set to be diffused, and the next diffusion process continues;

[0029] If not, the diffusion process ends, and the set to be diffused obtained after the previous diffusion process is taken as the target semantic stimulus set.

[0030] Optionally, the first semantic stimulus, the second semantic stimulus, and the diffusible third semantic stimulus satisfy the following formula:

[0031]

[0032]

[0033]

[0034]

[0035]

[0036] Where, x t For the first semantic stimulus, x i For the i-th second semantic stimulus in the set to be diffused, x t-1For the diffuseable third semantic stimulus, β t For the preset noise term scaling factor, ∈ t For noise terms, Let F(x) be a normal distribution, I be the identity matrix, Ω be the set to be diffused, and F(x) be the set to be diffused. t-1 ,Ω) represents the classification boundary corresponding to the diffusible third semantic stimulus. h is the logarithmic gradient of the classification boundary corresponding to the diffusible third semantic stimulus. t-1 h is the hidden encoding of the diffuseable third semantic stimulus. i S(x) is the hidden code of the i-th second semantic stimulus in the set to be diffused. t-1 ,x i The similarity is denoted as the EEG signal corresponding to the diffusible third semantic stimulus and the EEG signal corresponding to the i-th second semantic stimulus in the set to be diffused.

[0037] In a second aspect, the present invention provides a brain-computer interface imaginative marker design device, the device comprising:

[0038] The acquisition module is used to acquire multiple semantic stimuli and the corresponding EEG signal for each semantic stimulus;

[0039] The encoding processing module is used to encode the EEG signal corresponding to each semantic stimulus to obtain the hidden code of each semantic stimulus;

[0040] The determination module is used to determine a target semantic stimulus set from all the semantic stimuli based on the hidden code of each semantic stimulus, the target semantic stimulus set including multiple target semantic stimuli, and the EEG signals corresponding to the multiple target semantic stimuli having the largest classification boundary;

[0041] The task execution module is used to use each of the target semantic stimuli as a brain-computer interface imaginative marker to execute a preset brain-computer interface task.

[0042] Thirdly, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the brain-computer interface imaginary marker design method as described in the first aspect above.

[0043] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the brain-computer interface imaginary marker design method as described in the first aspect above.

[0044] Compared to existing technologies, the brain-computer interface (BCI) visualization marker design method provided by this invention acquires multiple semantic stimuli and their corresponding EEG signals; encodes the EEG signals corresponding to each semantic stimulus to obtain a hidden code for each semantic stimulus; based on the hidden code of each semantic stimulus, a target semantic stimulus set is determined from all semantic stimuli, the target semantic stimulus set including multiple target semantic stimuli, and the EEG signals corresponding to the multiple target semantic stimuli having the largest classification boundary; each target semantic stimulus is used as a BCI visualization marker to execute a preset BCI task. Because this invention uses the hidden code of each semantic stimulus to select multiple target semantic stimuli with the largest classification boundary from all semantic stimuli as BCI visualization markers to execute preset BCI tasks, it improves the decoding accuracy of BCI tasks and meets interaction requirements. Attached Figure Description

[0045] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart illustrating a brain-computer interface imaginative marker design method provided in an embodiment of the present invention;

[0047] Figure 2 An example diagram of a semantic stimulus provided in an embodiment of the present invention;

[0048] Figure 3 An example diagram of an encoded electroencephalogram (EEG) signal provided in an embodiment of the present invention;

[0049] Figure 4 An example diagram illustrating the determination of a target semantic stimulus set provided in an embodiment of the present invention;

[0050] Figure 5 A functional unit block diagram of a brain-computer interface imaginary marker design device provided in an embodiment of the present invention;

[0051] Figure 6 This is a schematic block diagram of an electronic device provided in an embodiment of the present invention.

[0052] Icons: 100 - Brain-computer interface imaginative marker design device; 101 - Acquisition module; 102 - Encoding processing module; 103 - Determination module; 104 - Task execution module; 200 - Electronic device; 210 - Memory; 220 - Processor. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0054] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0055] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0056] Furthermore, the terms "first" and "second" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0057] It should be noted that, where there is no conflict, the features in the embodiments of the present invention can be combined with each other.

[0058] The imagery markers used in traditional brain-computer interfaces have two main drawbacks: first, the number of commands is limited by the imagery markers; and second, the EEG signals corresponding to the selected commands may be difficult to decode.

[0059] On the one hand, due to the limited variety of imaginary markers such as limb movement or target device movement, brain-computer interfaces that use these imaginary markers as commands can only provide limited control options and cannot meet complex or diverse interaction needs.

[0060] On the other hand, since there are significant individual differences and non-stationarity in the EEG signals between different test subjects (hereinafter referred to as subjects) or the same subject at different time points, brain-computer interfaces that use these imaginary markers as commands also require a lot of training or calibration to improve the decoding accuracy of EEG signals, but this will increase the burden and cost on users.

[0061] In view of this, embodiments of the present invention provide a method for designing imaginary markers for brain-computer interfaces, which will be described in detail below.

[0062] Please refer to Figure 1The brain-computer interface imaginary marker design method includes steps S101 to S104.

[0063] S101, acquire multiple semantic stimuli and the corresponding EEG signals for each semantic stimulus.

[0064] Semantic stimuli include visual stimuli, auditory stimuli, and textual stimuli. The EEG signals corresponding to semantic stimuli refer to the EEG signals generated when the subject receives semantic stimuli.

[0065] Figure 2 Three visual stimuli were shown: pictures of "tiramisu," "kitten," and "light bulb."

[0066] Figure 2 Three textual stimuli were also shown: “Turin doesn’t love me, and never will,” “Cas stares at the sky,” and “The king must have an heir.”

[0067] like Figure 3 As shown, different semantic stimuli were presented to the subjects, and EEG signals were collected from the subjects when they received the stimuli using an EEG device. Considering the low signal-to-noise ratio of EEG signals and the differences between individuals, multiple EEG signals could be collected for the same semantic stimulus, meaning that there are multiple EEG signals corresponding to any given semantic stimulus.

[0068] S102, the EEG signal corresponding to each semantic stimulus is encoded to obtain the hidden code of each semantic stimulus.

[0069] As one possible implementation, step S102 can be implemented as follows:

[0070] S102-1, for each semantic stimulus, each EEG signal corresponding to the semantic stimulus is input into a pre-trained neural signal encoder to obtain the hidden code of each EEG signal.

[0071] The neural signal encoder can be EEGnet, which can be trained by collecting a large number of neural signals and using supervised learning. That is, the category of semantic stimulus is used as the learning target, and the neural signals are used to predict the semantic stimulus in order to obtain the loss function for learning.

[0072] Since collecting large amounts of neural signals would increase costs, it is also possible to train on any neural signal without using any supervisory information by deliberately masking the neural signal and predicting the complete signal.

[0073] Other types of machine learning models can also be used to obtain neural signal encoders, such as Support Vector Machines, Decision Trees, or Linear Regression, as long as they can predict the subject's EEG signals based on the feature vectors of semantic stimuli and reconstruct the feature vectors of semantic stimuli from the EEG signals.

[0074] By using a pre-trained neural signal encoder, each EEG signal can be encoded into a low-dimensional latent space to obtain the latent code of the EEG signal.

[0075] S102-2, the mean of the hidden codes of all EEG signals is used as the hidden code of the semantic stimulus.

[0076] Since any given semantic stimulus corresponds to multiple EEG signals, the hidden code of that semantic stimulus can be the average of the hidden codes of all the EEG signals corresponding to that semantic stimulus.

[0077] Understandably, for any semantic stimulus, the implicit encoding of that semantic stimulus satisfies the following formula:

[0078]

[0079] Where, x i For semantic stimuli, y is a set including each EEG signal corresponding to a semantic stimulus. j Let m be the j-th EEG signal corresponding to the semantic stimulus, m be the number of EEG signals corresponding to the semantic stimulus, E be the pre-trained neural signal encoder, and h be the number of EEG signals corresponding to the semantic stimulus. i This is the implicit encoding of semantic stimuli.

[0080] S103, based on the implicit encoding of each semantic stimulus, determine a target semantic stimulus set from all semantic stimuli.

[0081] The target semantic stimulus set includes multiple target semantic stimuli, and the EEG signals corresponding to multiple target semantic stimuli have the largest classification boundary.

[0082] In this embodiment of the invention, multiple target semantic stimuli in the target semantic stimulus set satisfy the following formula:

[0083]

[0084]

[0085] Where Ω is the set of target semantic stimuli, x i Let x be the i-th target semantic stimulus in the target semantic stimulus set. jh is the j-th target semantic stimulus in the target semantic stimulus set. i h is the hidden code of the i-th target semantic stimulus in the target semantic stimulus set. j Let S(x) be the hidden code of the j-th target semantic stimulus in the target semantic stimulus set. i ,x j ) is the target semantic stimulus x i The corresponding EEG signal and the target semantic stimulus x j The similarity of the corresponding EEG signals, F θ This refers to the classification boundary of EEG signals corresponding to multiple target semantic stimuli in the target semantic stimulus set.

[0086] As one possible implementation, step S103 may include sub-steps S103-1 to S103-2.

[0087] S103-1, Create an empty set to be diffused.

[0088] S103-2, based on the implicit encoding of each semantic stimulus, perform multiple diffusion processes on the set to be diffused to obtain the target semantic stimulus set.

[0089] In each diffusion process, a semantic stimulus is added to the set to be diffused.

[0090] Optionally, step S103-2 can be implemented as follows:

[0091] S103-2a, during each diffusion process, the semantic stimulus added to the set to be diffused during the previous diffusion process is taken as the first semantic stimulus, each semantic stimulus already added to the set to be diffused is taken as the second semantic stimulus, and each semantic stimulus not added to the set to be diffused is taken as the third semantic stimulus.

[0092] In each diffusion process, all semantic stimuli are divided into three categories: the first category consists of semantic stimuli added to the set to be diffused in the previous diffusion process; the second category consists of each semantic stimulus that has been added to the set to be diffused before the current diffusion process; and the third category consists of each semantic stimulus that has not yet been added to the set to be diffused.

[0093] Understandably, during each diffusion process, the semantic stimuli to be added to the diffusion set are determined from all the third semantic stimuli present during that diffusion process.

[0094] S103-2b, for each third semantic stimulus, calculate the classification boundary corresponding to the third semantic stimulus based on the hidden code of the third semantic stimulus and the hidden code of each second semantic stimulus, and obtain the classification boundary corresponding to each third semantic stimulus.

[0095] Specifically, by using the hidden code of the third semantic stimulus and the hidden code of any second semantic stimulus, the similarity between the EEG signal corresponding to the third semantic stimulus and the EEG signal corresponding to the second semantic stimulus can be calculated.

[0096] By using the similarity between the EEG signal corresponding to the third semantic stimulus and the EEG signal corresponding to each second semantic stimulus, the classification boundary corresponding to the third semantic stimulus can be obtained.

[0097] In a possible implementation, the classification boundary corresponding to any third semantic stimulus can be the sum of the similarities between the EEG signal corresponding to the third semantic stimulus and the EEG signal corresponding to each second semantic stimulus.

[0098] S103-2c, based on the classification boundaries of the first semantic stimulus and each third semantic stimulus, determine whether a diffusible third semantic stimulus can be identified from all third semantic stimuli.

[0099] Among them, the diffusible third semantic stimulus and the first semantic stimulus satisfy the preset conditions.

[0100] If a diffusible third semantic stimulus exists among all third semantic stimuli, proceed to step S103-2d; if no diffusible third semantic stimulus exists among all third semantic stimuli, proceed to step S103-2c.

[0101] S103-2d, add the diffusible third semantic stimulus to the set to be diffused, and continue the next diffusion process.

[0102] S103-2e, end the diffusion process and use the set to be diffused obtained after the previous diffusion process as the target semantic stimulus set.

[0103] In this embodiment of the invention, during each diffusion process, the first semantic stimulus, the second semantic stimulus, and the diffusible third semantic stimulus satisfy the following formula:

[0104]

[0105]

[0106]

[0107]

[0108]

[0109] Where, x t As the first semantic stimulus, x i Let x be the i-th second semantic stimulus in the set to be diffused. t-1 For a diffusible third semantic stimulus, βt Ω is the preset noise term scaling factor, F(x) is the set to be diffused, and F(x) is the noise term scaling factor. t-1 ,Ω) represents the classification boundary corresponding to the diffusible third semantic stimulus. Let ∈ be the logarithmic gradient of the classification boundary corresponding to the diffusible third semantic stimulus. t For noise terms, It follows a normal distribution, where I is the identity matrix, i.e., ∈ t Follow the mean The normal distribution, h t-1 For the latent encoding of a diffusible third semantic stimulus, h i S(x) is the hidden code of the i-th second semantic stimulus in the set to be diffused. t-1 ,x i ) represents the similarity between the EEG signal corresponding to the diffusible third semantic stimulus and the EEG signal corresponding to the i-th second semantic stimulus in the set to be diffused.

[0110] Using the above formula, we can first randomly select a semantic stimulus as the initial element of the set to be diffused Ω, and then gradually determine the semantic stimulus that is furthest away from the semantic stimuli in the set to be diffused Ω, that is, from any semantic stimulus x t Initially, the set to be diffused Ω is updated step by step according to the conditional probability distribution, and finally the target semantic stimulus set is obtained.

[0111] S104 uses each target semantic stimulus as a brain-computer interface imaginative marker to execute a preset brain-computer interface task.

[0112] Figure 2 The decoded targets are also shown, including hands, grasping, and feet from left to right.

[0113] like Figure 4 As shown, when subjects performed preset brain-computer interface tasks, they were asked to imagine these brain-computer interface visual symbols, each corresponding to a decoding target, namely a hand, grasp, or foot.

[0114] Understandably, since the EEG signals corresponding to multiple target semantic stimuli in the target semantic stimulus set have the largest classification boundary, using each target semantic stimulus as a brain-computer interface (BCI) visualization marker to execute a pre-set BCI task can make the task instructions clearer, easier for the subject to understand and execute, thereby improving the decoding accuracy of the BCI task.

[0115] The embodiments of the present invention can also flexibly expand the number and types of instructions. That is, by adding new semantic stimuli and the corresponding EEG signals of the new semantic stimuli, a target semantic stimulus set can be determined from all semantic stimuli using the above method. Then, each target semantic stimulus in the target semantic stimulus set is used as a brain-computer interface imagination marker to execute a preset brain-computer interface task.

[0116] The embodiments of the present invention can also use other types of search algorithms to determine the target semantic stimulus set, such as reinforcement learning, collaborative filtering, or Bayesian optimization, that is, as long as the target semantic stimulus that can generate the maximum classification boundary of the EEG signal can be selected according to the individual differences of the subjects and the task requirements.

[0117] In order to perform the corresponding steps in the above method embodiments and various possible implementations, an implementation of a brain-computer interface imaginative marker design device 100 is given below.

[0118] Please refer to Figure 5 The brain-computer interface imaginative marker design device 100 includes an acquisition module 101, an encoding processing module 102, a determination module 103, and a task execution module 104.

[0119] The acquisition module 101 is used to acquire multiple semantic stimuli and the corresponding EEG signal for each semantic stimulus.

[0120] The encoding processing module 102 is used to encode the EEG signal corresponding to each semantic stimulus to obtain the hidden code of each semantic stimulus.

[0121] The determination module 103 is used to determine a target semantic stimulus set from all semantic stimuli based on the hidden code of each semantic stimulus. The target semantic stimulus set includes multiple target semantic stimuli, and the EEG signals corresponding to the multiple target semantic stimuli have the largest classification boundary.

[0122] The task execution module 104 is used to use each target semantic stimulus as a brain-computer interface imaginative marker to execute a preset brain-computer interface task.

[0123] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the brain-computer interface imaginative marker design device 100 described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0124] Furthermore, this embodiment of the invention also provides an electronic device 200, please refer to... Figure 6 The electronic device 200 may include a memory 210 and a processor 220.

[0125] The processor 220 can be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits used to control the program execution of the brain-computer interface imaginary marker design method provided in the above method embodiments.

[0126] The memory 210 may be a ROM or other type of static storage device capable of storing static information and instructions, RAM or other type of dynamic storage device capable of storing information and instructions, or an electrically erasable programmable-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 210 may exist independently and be connected to the processor 220 via a communication bus. The memory 210 may also be integrated with the processor 220. The memory 210 is used to store machine-executable instructions for executing the scheme of this application. The processor 220 is used to execute the machine-executable instructions stored in the memory 210 to implement the above-described method embodiments.

[0127] This invention also provides a computer-readable storage medium containing a computer program, which, when executed, can be used to perform related operations in the brain-computer interface imaginary marker design method provided in the above-described method embodiments.

[0128] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for designing imaginary markers for brain-computer interfaces, characterized in that, The method includes: Acquire multiple semantic stimuli and the corresponding EEG signals for each semantic stimulus; For each of the semantic stimuli, each EEG signal corresponding to the semantic stimulus is input into a pre-trained neural signal encoder to obtain the hidden code of each EEG signal; The mean of the hidden codes of all the EEG signals is taken as the hidden code of the semantic stimulus; the hidden code of the semantic stimulus satisfies the following formula: in, For the semantic stimulus, It is a set including each EEG signal corresponding to the semantic stimulus. The first corresponding to the semantic stimulus One EEG signal, The number of EEG signals corresponding to the semantic stimulus. For a pre-trained neural signal encoder, This is the implicit encoding of semantic stimuli; Create an empty set to be diffused; During each diffusion process, the semantic stimulus added to the set to be diffused during the previous diffusion process is used as the first semantic stimulus, each semantic stimulus already added to the set to be diffused is used as the second semantic stimulus, and each semantic stimulus not added to the set to be diffused is used as the third semantic stimulus. For each of the third semantic stimuli, the classification boundary corresponding to the third semantic stimulus is calculated based on the hidden code of the third semantic stimulus and the hidden code of each of the second semantic stimuli, thus obtaining the classification boundary corresponding to each of the third semantic stimuli. Based on the classification boundary of the first semantic stimulus and each of the third semantic stimuli, it is determined whether a diffusible third semantic stimulus can be identified from all the third semantic stimuli, and the diffusible third semantic stimulus satisfies the first semantic stimulus under a preset condition. If so, the diffusible third semantic stimulus is added to the set to be diffused, and the next diffusion process continues; If not, the diffusion process ends, and the set to be diffused obtained after the previous diffusion process is taken as the target semantic stimulus set; the target semantic stimulus set includes multiple target semantic stimuli, and the EEG signals corresponding to the multiple target semantic stimuli have the largest classification boundary; Each of the target semantic stimuli is used as a brain-computer interface imaginative marker to execute a preset brain-computer interface task.

2. The method as described in claim 1, characterized in that, The multiple target semantic stimuli in the target semantic stimulus set satisfy the following formula: in, For the target semantic stimulus set, For the first in the target semantic stimulus set A target semantic stimulus For the first in the target semantic stimulus set A target semantic stimulus For the first in the target semantic stimulus set The implicit encoding of a target semantic stimulus For the first in the target semantic stimulus set The implicit encoding of a target semantic stimulus For target semantic stimulus Corresponding EEG signals and target semantic stimuli The similarity of the corresponding EEG signals, This refers to the classification boundary of EEG signals corresponding to multiple target semantic stimuli in the target semantic stimulus set.

3. The method as described in claim 1, characterized in that, The first semantic stimulus, the second semantic stimulus, and the diffusible third semantic stimulus satisfy the following formula: in, For the first semantic stimulus, For the first in the set to be diffused The second semantic stimulus, For the diffuseable third semantic stimulus, This is the preset noise term scaling factor. For noise terms, It follows a normal distribution. It is the identity matrix. For the set to be diffused, The classification boundary corresponding to the diffusible third semantic stimulus. The logarithmic gradient of the classification boundary corresponding to the diffusible third semantic stimulus. The hidden encoding of the diffuseable third semantic stimulus, For the first in the set to be diffused The hidden encoding of the second semantic stimulus The EEG signal corresponding to the diffusible third semantic stimulus and the first in the set to be diffused. The similarity of the EEG signals corresponding to the second semantic stimulus.

4. A brain-computer interface imaginative marker design device, characterized in that, The device includes: The acquisition module is used to acquire multiple semantic stimuli and the corresponding EEG signal for each semantic stimulus; The encoding processing module is used to input each EEG signal corresponding to each semantic stimulus into a pre-trained neural signal encoder to obtain the hidden code of each EEG signal; the mean of the hidden codes of all EEG signals is used as the hidden code of the semantic stimulus; the hidden code of the semantic stimulus satisfies the following formula: in, For the semantic stimulus, It is a set including each EEG signal corresponding to the semantic stimulus. The semantic stimulus corresponding to the first One EEG signal, The number of EEG signals corresponding to the semantic stimulus. For a pre-trained neural signal encoder, This is the implicit encoding of semantic stimuli; A determination module is used to create an empty set to be diffused. During each diffusion process, the semantic stimuli added to the set to be diffused in the previous diffusion process are used as the first semantic stimulus, each semantic stimulus already added to the set is used as the second semantic stimulus, and each semantic stimulus not added to the set is used as the third semantic stimulus. For each third semantic stimulus, the classification boundary corresponding to the third semantic stimulus is calculated based on the hidden code of the third semantic stimulus and the hidden code of each second semantic stimulus, thus obtaining the classification boundary corresponding to each third semantic stimulus. Based on the first semantic stimulus and the classification boundary of each third semantic stimulus, it is determined whether a diffuseable third semantic stimulus can be determined from all the third semantic stimuli, and whether the diffuseable third semantic stimulus satisfies a preset condition with the first semantic stimulus. If yes, the diffuseable third semantic stimulus is added to the set to be diffused, and the next diffusion process continues. If no, the diffusion process ends, and the set to be diffused obtained after the previous diffusion process is used as the target semantic stimulus set. The target semantic stimulus set includes multiple target semantic stimuli, and the EEG signals corresponding to the multiple target semantic stimuli have the largest classification boundary. The task execution module is used to use each of the target semantic stimuli as a brain-computer interface imaginative marker to execute a preset brain-computer interface task.

5. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, which, when executed by the processor, implements the brain-computer interface imaginary marker design method as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the brain-computer interface imaginary marker design method as described in any one of claims 1 to 3.

Images