A method, device and equipment for removing vibration artifacts in electroencephalogram signals after fusion of vibration stimulation, and a brain-computer interface system

Abstract

The application discloses a kind of fusion vibration stimulation after the removal method, device and equipment of vibration artifact in electroencephalogram signal and brain-computer interface system, it is related to brain-computer interface signal processing field.The method comprises the following steps: obtaining the first electroencephalogram data under the state of no vibration stimulation and the second electroencephalogram data under the state of vibration stimulation;Based on the covariance matrix of the target frequency band of the first electroencephalogram data and the second electroencephalogram data, a spatial filter is constructed;Obtain the third electroencephalogram data collected in real time;The third electroencephalogram data is projected in space using the spatial filter to separate the vibration artifact.The application realizes the low-delay accurate separation of vibration artifact by constructing the double-state covariance model offline and executing the spatial projection online, significantly improves the signal quality and decoding performance of brain-computer interface system.

Description

Technical Field

[0001] This invention relates to the field of brain-computer interface signal processing, and specifically to a method, apparatus, device, and brain-computer interface system for removing vibration artifacts from electroencephalogram (EEG) signals after vibration stimulation. Background Technology

[0002] Brain-computer interface (BCI) technology, as a human-computer interaction technology that transmits thought information directly through brain electrical activity without relying on the peripheral nerves and muscles, can analyze the information carried in the brain electrical signals of patients with movement disorders. It has broad application potential in medical rehabilitation, neuroscience research, and other fields. Based on different signal acquisition methods, BCI technology is divided into three types: invasive, semi-invasive, and non-invasive. Among them, non-invasive BCI technology mainly uses techniques such as electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) to detect neural activity from the scalp surface or externally. It has advantages such as high safety, ease of operation, and low cost. With the development of signal processing and artificial intelligence algorithms, modern non-invasive systems can now meet the practical needs of most rehabilitation applications, and have even broader application prospects in the field of BCI rehabilitation.

[0003] In the field of rehabilitation medicine, BCI technology is primarily based on the principle of neuroplasticity in the motor cortex of the brain. It utilizes the characteristic neural signals generated during motor imagery (MI) to provide new rehabilitation training methods for patients with movement disorders. Motor imagery refers to the act of imagining limb movements without actually performing them. The neural pathways activated during MI are similar to those involved in actual movement. Through MI training, areas of the brain related to movement can be activated. However, the EEG signals generated by motor imagery are weak, non-stationary, and susceptible to interference, resulting in low signal recognition accuracy and difficulty meeting the reliability requirements of practical applications. Vibration stimulation, as a mature neurorehabilitation method, stimulates the limbs to generate sensory signals, which are then transmitted to the cerebral cortex and influence the excitability of the motor cortex through cortical connections. Introducing vibration stimulation technology into brain-computer interface systems can effectively compensate for the limitations of traditional single visual or auditory feedback modes and enhance the characteristic intensity of motor imagery signals. Studies have shown that brain-computer interface systems incorporating vibration stimulation achieve higher MI signal recognition accuracy compared to systems based solely on motor imagery, thus achieving better rehabilitation training results.

[0004] However, while vibration stimulation has positive effects, it can also introduce significant vibration artifacts into EEG signals. These artifacts typically manifest as periodic noise related to the frequency of the vibration device, severely interfering with the accurate acquisition and interpretation of EEG signals, leading to a decline in the decoding performance of the BCI system, and consequently affecting rehabilitation outcomes.

[0005] Currently, methods for removing vibration artifacts mainly include offline filtering, template matching, and adaptive filtering. Offline filtering methods rely on post-processing analysis, making it difficult to meet the real-time requirements of brain-computer interfaces (BCIs). Template matching methods require pre-constructing artifact templates, making it difficult to adapt to dynamic changes in vibration stimulation frequency and intensity. While adaptive filtering has some dynamic adjustment capabilities, its computational complexity and high hardware performance requirements limit its application in portable or low-power BCI devices. Furthermore, existing methods generally suffer from poor artifact removal performance, insufficient real-time performance, or poor algorithm robustness in scenarios involving vibration-infused motor imagery.

[0006] Therefore, there is an urgent need for a technology that can remove vibration artifacts online in real time, has high computational efficiency, and is highly adaptable, in order to improve the signal quality and application effect of brain-computer interface systems in motor imagery rehabilitation. Summary of the Invention

[0007] To address the problems of poor real-time performance, weak dynamic adaptability, and high computational complexity in existing vibration artifact removal methods, this application proposes a method, apparatus, device, and brain-computer interface system for removing vibration artifacts from EEG signals after vibration stimulation. This achieves low-latency and accurate separation of periodic vibration artifacts, significantly improving the signal quality and decoding performance of the brain-computer interface system.

[0008] The technical means employed in this invention are as follows:

[0009] A method for removing vibration artifacts from electroencephalogram (EEG) signals after vibration stimulation, comprising the following steps: Acquire first EEG data under no vibration stimulation state and second EEG data under vibration stimulation state; A spatial filter is constructed based on the covariance matrix of the first and second EEG data in the target frequency band. Acquire real-time third-party EEG data; The spatial filter is used to spatially project the third EEG data to remove vibration artifacts from the third EEG data.

[0010] Furthermore, based on the covariance matrix of the first EEG data and the second EEG data within the target frequency band, a spatial filter is constructed, including: The first EEG data and the second EEG data are respectively subjected to bandpass filtering within the target frequency band; The filtered first and second EEG data were divided into multiple data segments, and the data in the task state were extracted to obtain EEG data from multiple trials. Calculate the covariance matrix for each trial; The arithmetic mean of all single-trial covariance matrices corresponding to the first EEG data is used to obtain the first average covariance matrix, and the arithmetic mean of all single-trial covariance matrices corresponding to the second EEG data is used to obtain the second average covariance matrix. The spatial filter is constructed based on the first average covariance matrix and the second average covariance matrix.

[0011] Further, constructing the spatial filter based on the first average covariance matrix and the second average covariance matrix includes: Based on the first average covariance matrix and the second average covariance matrix, a dual covariance matrix model is constructed; The bicovariance matrix model is subjected to generalized eigenvalue decomposition to obtain the eigenvector matrix and eigenvalue vector; The spatial filter is generated based on the target feature vector in the feature vector matrix.

[0012] Furthermore, the generalized eigenvalue decomposition of the bicovariance matrix model is performed based on the following formula:

[0013] in It is the covariance matrix of the EEG signals subjected to vibration stimulation. It is the covariance matrix of the EEG signal without applied vibration stimulation. For spatial filters, It is an eigenvalue diagonal matrix.

[0014] Furthermore, the spatial filter is generated based on the following formula:

[0015]

[0016] in, The eigenvector matrix, for pseudo-inverse matrix This is the eigenvalue diagonal matrix with the first i diagonal elements set to zero. This is the optimal spatial filter. This is the spatial filter matrix corresponding to suppressing the first i components.

[0017] Further, acquiring real-time collected third EEG data, and using the spatial filter to spatially project the third EEG data to remove vibration artifacts in the third EEG data, includes: A time window is set to cache the real-time collected EEG data. When the length of the cached EEG data reaches the length threshold of the time window, the EEG data within the time window is used as the third EEG data. The spatial filter is triggered to perform the spatial projection on the third EEG data.

[0018] Furthermore, the duration of the time window is set to 3 to 5 seconds.

[0019] This invention also discloses a device for removing vibration artifacts from electroencephalogram (EEG) signals after vibration stimulation, used to implement the method described in any one of the above, comprising: The data acquisition module is used to acquire first EEG data under no vibration stimulation state, second EEG data under vibration stimulation state, and acquire third EEG data in real time. A filter construction module is used to construct a spatial filter based on the covariance matrix of the first EEG data and the second EEG data in the target frequency band. The real-time projection module is used to spatially project the third EEG data using the spatial filter in order to separate vibration artifacts in the third EEG data.

[0020] The present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in any of the preceding claims.

[0021] A brain-computer interface system, comprising: EEG acquisition equipment is used to collect EEG data from a target subject; A vibration stimulation device for applying vibration stimulation to the target object; The computer equipment described above is communicatively connected to both the EEG acquisition device and the vibration stimulation device.

[0022] Compared with the prior art, the present invention has the following advantages: This invention constructs a target frequency band covariance matrix under two states: no vibration stimulus and applied vibration stimulus. It then uses generalized eigenvalue decomposition to solve for the optimal spatial filter, transforming the vibration artifact separation problem into a mathematical joint diagonalization problem. This results in a spatial filter with clear physical meaning and energy-optimal characteristics. Furthermore, it employs an architecture combining offline calibration and online projection. The highly complex matrix decomposition operations are performed offline, while the online stage only requires low-overhead linear projection operations. This significantly reduces the real-time computational load while maintaining artifact removal accuracy.

[0023] This invention further incorporates mechanisms such as time window caching, data cleaning, and smooth stitching to enhance the robustness and signal continuity of the algorithm under non-stationary EEG signals and dynamic stimulation environments, ultimately achieving efficient, real-time, and accurate online removal of vibration artifacts suitable for portable brain-computer interface rehabilitation systems. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a timeline diagram of an offline experimental paradigm for motor imagery rehabilitation in an embodiment of the present invention.

[0026] Figure 2 This is a flowchart of vibration artifact removal and EEG signal processing in an embodiment of the present invention.

[0027] Figure 3 This refers to the position of lead electrode 64 in this embodiment of the invention. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] Example 1 like Figure 1 and Figure 2As shown in the figure, this invention provides a method for removing vibration artifacts in electroencephalogram (EEG) signals. This method achieves efficient separation of periodic vibration artifacts through an architecture combining offline calibration and online real-time processing. The overall process includes an offline calibration stage and an online processing stage. The specific implementation methods of each stage are described in detail below with reference to the accompanying drawings.

[0031] In the offline calibration phase, step S100 is first executed to acquire the first EEG data under the state of no vibration stimulation and the second EEG data under the state of applied vibration stimulation.

[0032] This step aims to acquire a benchmark dataset for constructing the spatial filter. In practical applications, the target object's resting or task-oriented EEG signals without the vibration stimulation device can be recorded using an EEG acquisition device as the first EEG data, while the EEG signals under the same conditions after the vibration stimulation device is turned on are recorded as the second EEG data.

[0033] In this embodiment, an STM32RCT6 was used to control two LRAs to generate vibrational stimulation, with 25Hz as the stimulation frequency. The stimulation was applied to the inner side of the wrist with a normalized amplitude of 2.8G. A 64-channel Boricon wireless surface multimodal acquisition system was used to acquire EEG signals. MATLAB software was used for programming and implementation of offline and online experiments. First, an offline experimental system based on the MI paradigm was built to conduct offline experiments and record EEG data from the motor and sensory cortices. Specifically, this included: A. The subject wore a 64-channel EEG electrode cap, such as Figure 3 The diagram shows the distribution of 64 electrodes. Measurement electrodes are placed at positions C3, C4, CP1, CP3, CPz, CP4, and CP2 (electrodes circled in red in the diagram). A grounding electrode is placed between Fz and FPz. A reference electrode is placed at position Cz in the top region. The sampling frequency of the EEG signal is 1000Hz. The EEG signal acquisition system is turned on.

[0034] B. Configure the MI experimental paradigm program using Psychtoolbox and write the program according to the prompts required in the predetermined process.

[0035] C. Conduct offline experiments with and without vibration stimulation, with three groups of experiments in each group. Collect and record EEG data from each channel throughout the experiment.

[0036] D. The collected EEG data were downsampled to a frequency of 250Hz. Data without vibration stimulation and data with vibration stimulation were extracted separately to obtain the dataset: and .

[0037] This offline experiment included three groups of experiments each with and without vibration stimulation under the MI task state. Each group of experiments consisted of 20 trials, with each trial including a 4-second rest period, a 1-second target indication time, and a 4-second motor imagery time. Throughout the experiment, participants were required to concentrate, keep their heads stable, minimize blinking, and perform the corresponding motor imagery task according to the prompts. EEG data were collected throughout the experiment for subsequent identification of spatial filters.

[0038] Next, step S200 is executed to construct a spatial filter based on the covariance matrix of the first EEG data and the second EEG data in the target frequency band.

[0039] This step generates an anti-artifact filter by exploiting the energy distribution differences in a specific frequency band under two states. Here, the target frequency band refers to any frequency band containing the main energy of the vibration artifact, and its specific range can be dynamically set according to the physical parameters of the vibration stimulus. For example, when the vibration stimulus frequency is 25Hz, the target frequency band can be set to a narrow band of 24Hz to 26Hz; if the vibration frequency changes, the target frequency band should also be adjusted accordingly to cover the new artifact center frequency.

[0040] To further improve the construction accuracy of the spatial filter, step S200 specifically includes the following sub-steps: First, step S201 is executed, where the first and second EEG data are bandpass filtered within the target frequency band. EEG signals inherently have extremely low signal-to-noise ratios and contain a large amount of electromyography (EMG), electrooculography (EOG), and environmental noise. If the covariance matrix is ​​directly calculated using broadband data, the weak spatial features of vibration artifacts are easily submerged by strong background noise, leading to the failure of subsequent generalized eigenvalue decomposition. By pre-filtering within the target frequency band, the signal is essentially pre-enhanced in the frequency domain, ensuring that the covariance matrix primarily reflects the spatial distribution differences between vibration artifacts and background EEG at specific frequencies. As a preferred embodiment of this application, [the following text is incomplete and requires further context to translate accurately]. and Perform a bandpass filter on [24, 26] to obtain the filtered dataset. and .

[0041] Then, step S202 is executed, dividing the filtered first and second EEG data into multiple data segments to extract the motor imagery data. In practice, the data segmentation can be based on the trial sequence of the motor imagery task, with each trial's task period data treated as an independent segment; alternatively, a sliding window of fixed time length can be used for continuous segmentation. This method is not dependent on specific experimental paradigm markers and has greater versatility. This example yields a dataset after extracting the motor imagery data. and

[0042] After completing the data partitioning process, step S203 is executed to calculate the covariance matrix for each trial. As a preferred embodiment of this application, the dataset for which vibration stimulation is applied... Data for each trial and datasets without vibration stimulation. Data for each trial The covariance matrix was calculated according to formulas (1) and (2), respectively. Then, the arithmetic mean of the covariance matrices for all trials with and without vibration stimulation was calculated to obtain two mean covariance matrices. and .

[0043] (1) (2) in, Indicates the first The mean vector of vibrational stimulus data applied in each trial, Indicates the first The mean vector of data from each trial without applied vibration stimulus.

[0044] Based on this, step S204 is executed to construct a dual covariance matrix model based on the first and second average covariance matrices. Subsequently, generalized eigenvalue decomposition was performed on the bicovariance matrix model. This yields the eigenvector matrix and eigenvalue vector. Specifically, formula (3) is used to calculate the average covariance matrix. and Perform generalized eigenvalue decomposition. eigenvector matrix, eigenvectors.

[0045] (3) To generate a usable spatial filter, step S205 is performed, generating the spatial filter based on the target eigenvectors in the eigenvector matrix. Specifically, let... It is a unit diagonal matrix. To set the first i diagonal elements to zero, calculate the spatial filter using formula (4). Obtain the optimal spatial filter using formula (5). .

[0046] (4) (5)

[0047]

[0048] in, The eigenvector matrix, for pseudo-inverse matrix To optimize variables, This is the eigenvalue diagonal matrix with the first i diagonal elements set to zero. This is the optimal spatial filter. This is the spatial filter matrix corresponding to suppressing the first i components.

[0049] The solution methods for formulas (4) and (5) are equivalent to: (6) After obtaining the optimal spatial filter through offline calibration, the process enters the online processing stage. Step S300 is executed to acquire the real-time acquired third EEG data. This third EEG data refers to the streaming data continuously output by the EEG acquisition device during actual application. This embodiment employs a time-window-based block processing mechanism. Specifically, a time window is set to cache the real-time acquired EEG data; when the length of the cached EEG data reaches the time window's length threshold, the EEG data within the time window is used as the third EEG data; the spatial filter is triggered to perform spatial projection on the third EEG data. For example, the duration of the time window can be set to 4 seconds. This range ensures both the minimum sample size required for matrix operations to maintain numerical stability and keeps the processing latency within an acceptable range for the user. A 4-second data cache is then stored. It processes the EEG signals in the cache every 4 seconds.

[0050] Finally, step S400 is executed, where a spatial filter is used to spatially project the third EEG data to separate vibration artifacts from the data. An optimal spatial filter is then used. Processing EEG signals According to formula (7), the electroencephalographic signal after vibration artifact separation is obtained.

[0051] (7) This invention employs an architecture design that combines offline heavy computation with online light projection, fundamentally solving the problems of high computational complexity and difficulty in real-time implementation of traditional adaptive filtering methods. This makes it possible to achieve high-precision vibration artifact removal on portable brain-computer interface devices. It should be understood that although this embodiment is described in detail using a motor imagery rehabilitation system as an example, this method is equally applicable to any scenario requiring real-time extraction of pure EEG signals under vibration stimulation, such as tactile feedback research and evaluation of neuromodulation effects. The scope of protection of this invention should not be limited to specific application areas.

[0052] Example 2 This embodiment provides a device for removing vibration artifacts from electroencephalogram (EEG) signals. As a virtualized product of the aforementioned method embodiment, this device achieves efficient processing of vibration artifacts through a modular architecture. The device includes a data acquisition module, a filter construction module, and a real-time projection module. Wherein: The data acquisition module is used to acquire first EEG data under no vibration stimulation, second EEG data under vibration stimulation, and third EEG data acquired in real time. Specifically, this module serves as the interface for the system to interact with the external physical world. In offline calibration mode, the data acquisition module is responsible for reading historically acquired first and second EEG data from local storage media or a database and loading them into memory for subsequent processing. In online operation mode, this module receives streaming data transmitted from the EEG acquisition device in real time as third EEG data via communication interfaces such as USB, Bluetooth, Wi-Fi, or fiber optics.

[0053] The filter construction module is used to construct a spatial filter based on the covariance matrix of the first and second EEG data within the target frequency band. Specifically, the internal processing logic of this module corresponds exactly to the offline calibration stage (step S200 and its sub-steps) described in Example 1. In actual engineering implementation, the filter construction module is usually triggered to execute once or multiple times during the calibration stage after system startup. This module first calls the digital signal processing library to perform target frequency bandpass filtering on the two sets of input offline data, and then segments and cleans the data according to a preset time window or trial mark. Next, the generalized eigenvalue decomposition solver integrated within the module performs calculations on the dual covariance matrix joint diagonalization model, and generates the optimal spatial filter coefficient matrix based on the eigenvalue sorting results. Once this coefficient matrix is ​​generated, it is written to shared memory or register file for the real-time projection module to call at any time.

[0054] The real-time projection module is used to spatially project the third EEG data using a spatial filter to separate vibration artifacts from the third EEG data. Specifically, this module is a resident execution unit during the online operation of the system, and its processing logic corresponds to the online processing stage (steps S300 and S400) in Embodiment 1. The real-time projection module continuously monitors the real-time data stream output by the data acquisition module. When the data length in the buffer reaches a preset time window threshold, it immediately reads the spatial filter coefficients generated by the filter construction module and performs matrix multiplication on the current data block.

[0055] Through the modular device architecture described above, this invention not only fully covers the technical features of the method embodiments, but also provides a flexible engineering implementation path for the technical solution. The data flow relationship between each module is clear, and the collaborative working mechanism is efficient, which not only ensures the accuracy and real-time performance of vibration artifact removal, but also endows the system with good scalability and maintainability, enabling it to adapt to the needs of different specifications of EEG acquisition equipment and various rehabilitation application scenarios.

[0056] Example 3 This embodiment provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the vibration artifact removal method in electroencephalogram (EEG) signals as described in Embodiment 1 or Embodiment 2. This computer device serves as the physical hardware carrier of the technical solution of this invention, providing the necessary computing power and data storage environment for the operation of the aforementioned algorithm, enabling the abstract signal processing flow to be transformed into concrete physical execution actions.

[0057] Example 4 This embodiment provides a brain-computer interface system that integrates the artifact removal method described in the previous embodiments into a complete rehabilitation training chain to solve the industry problem of system performance degradation caused by artifacts introduced by vibration stimulation. Specifically, the brain-computer interface system includes an EEG acquisition device, a vibration stimulation device, and a computer device as described in Embodiment 3. The computer device is communicatively connected to both the EEG acquisition device and the vibration stimulation device. This hardware architecture constitutes a neurorehabilitation platform with real-time perception, precise stimulation, and closed-loop feedback capabilities. The communication connection not only carries data transmission functions but also maintains microsecond-level time synchronization between devices, ensuring precise alignment of EEG signals and vibration stimulation events in the time domain, providing a reliable temporal reference for the subsequent construction of the dual-state covariance matrix and the application of spatial filters.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for removing vibration artifacts from electroencephalogram (EEG) signals after vibration stimulation, characterized in that, Includes the following steps: Acquire first EEG data under no vibration stimulation state and second EEG data under vibration stimulation state; A spatial filter is constructed based on the covariance matrix of the first and second EEG data in the target frequency band. Acquire real-time third-party EEG data; The spatial filter is used to spatially project the third EEG data to remove vibration artifacts from the third EEG data.

2. The method for removing vibration artifacts from EEG signals after fusion vibration stimulation according to claim 1, characterized in that, Based on the covariance matrix of the first and second EEG data within the target frequency band, a spatial filter is constructed, including: The first EEG data and the second EEG data are respectively subjected to bandpass filtering within the target frequency band; The filtered first and second EEG data were divided into multiple data segments, and the data in the task state were extracted to obtain EEG data from multiple trials. Calculate the covariance matrix for each trial; The arithmetic mean of all single-trial covariance matrices corresponding to the first EEG data is used to obtain the first average covariance matrix, and the arithmetic mean of all single-trial covariance matrices corresponding to the second EEG data is used to obtain the second average covariance matrix. The spatial filter is constructed based on the first average covariance matrix and the second average covariance matrix.

3. The method for removing vibration artifacts from EEG signals after fusion vibration stimulation according to claim 2, characterized in that, The spatial filter is constructed based on the first average covariance matrix and the second average covariance matrix, including: Based on the first average covariance matrix and the second average covariance matrix, a dual covariance matrix model is constructed; The bicovariance matrix model is subjected to generalized eigenvalue decomposition to obtain the eigenvector matrix and eigenvalue vector; The spatial filter is generated based on the target feature vector in the feature vector matrix.

4. The method for removing vibration artifacts from EEG signals after fusion vibration stimulation according to claim 3, characterized in that, The generalized eigenvalue decomposition of the bicovariance matrix model is performed based on the following formula: in It is the covariance matrix of the EEG signals subjected to vibration stimulation. It is the covariance matrix of the EEG signal without applied vibration stimulation. For spatial filters, It is an eigenvalue diagonal matrix.

5. The method for removing vibration artifacts from EEG signals after fusion vibration stimulation according to claim 4, characterized in that, The spatial filter is generated based on the following formula: in, The eigenvector matrix, for pseudo-inverse matrix This is the eigenvalue diagonal matrix with the first i diagonal elements set to zero. This is the optimal spatial filter. This is the spatial filter matrix corresponding to suppressing the first i components.

6. The method for removing vibration artifacts from EEG signals after fusion vibration stimulation according to claim 1, characterized in that, Acquiring real-time collected third EEG data, and spatially projecting the third EEG data using the spatial filter to remove vibration artifacts in the third EEG data, includes: A time window is set to cache the real-time collected EEG data. When the length of the cached EEG data reaches the length threshold of the time window, the EEG data within the time window is used as the third EEG data. The spatial filter is triggered to perform the spatial projection on the third EEG data.

7. The method for removing vibration artifacts from EEG signals after fused vibration stimulation according to claim 6, characterized in that, The duration of the time window is set to 3 to 5 seconds.

8. A device for removing vibration artifacts from electroencephalogram (EEG) signals after vibration stimulation, used to implement the method as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to acquire first EEG data under no vibration stimulation state, second EEG data under vibration stimulation state, and acquire third EEG data in real time. A filter construction module is used to construct a spatial filter based on the covariance matrix of the first EEG data and the second EEG data in the target frequency band. The real-time projection module is used to spatially project the third EEG data using the spatial filter in order to separate vibration artifacts in the third EEG data.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A brain-computer interface system, characterized in that, include: EEG acquisition equipment is used to collect EEG data from a target subject; A vibration stimulation device for applying vibration stimulation to the target object; The computer device as described in claim 9 is communicatively connected to both the electroencephalogram (EEG) acquisition device and the vibration stimulation device.

Images