Steady-state visual evoked potential brain-computer interface system for snowflake point weak flicker coding

Abstract

The invention provides a steady-state visual evoked potential brain-computer interface system for snowflake point weak flicker coding, and the system comprises a stimulation presentation module which is used for presenting a stimulation interface of snowflake point weak flicker; each stimulation target comprises a corresponding frequency and initial phase combination, and the frequencies and initial phase combinations corresponding to different stimulation targets are different; the stimulation interface is used for guiding the sight line of the user to be transferred to a current stimulation target when each trial starts; the electroencephalogram collection module is used for collecting electroencephalogram signals of the occipital area of the user through a dry electrode head ring, and the dry electrode head ring comprises a plurality of movable electrodes, a reference electrode and a grounding electrode; the signal processing module is used for processing the electroencephalogram signal to identify a steady-state visual evoked potential and acquiring a frequency and phase combination of a stimulation target corresponding to the electroencephalogram signal; and the interaction control module is used for outputting a corresponding control instruction according to the frequency and phase combination of the stimulation target identified by the signal processing module to realize brain-computer interaction.

Description

Technical Field

[0001] This application relates to the field of brain-computer interface technology, and in particular to a steady-state visual evoked potential brain-computer interface system based on snowflake dot weak flicker coding. Background Technology

[0002] Steady-state visual evoked potentials (SSVEP) brain-computer interfaces (BCIs) have significant application value in fields such as neurorehabilitation and assistive interaction due to their advantages such as high information transmission rate and short training time. However, existing SSVEP-BCI systems still suffer from the following key technical defects, which severely limit their practical application and promotion, including: 1. Reliance on wet electrodes leads to poor practicality: Existing systems mostly use wet electrodes to collect EEG signals, which requires applying conductive gel and cleaning the scalp. The preparation process is cumbersome (taking 10-30 minutes), and the conductive gel dries easily, causing scalp discomfort, making it unsuitable for everyday scenarios such as home and mobile use. Although dry electrodes can solve the above problems, the signal-to-noise ratio of existing dry electrodes is low, making it difficult to effectively induce SSVEP response, resulting in insufficient recognition rate (usually <70%).

[0003] 2. High brightness leads to visual fatigue: In pursuit of strong SSVEP response, existing systems generally use high brightness (>800 cd / m²). 2 White, blocky icons can cause glare, eye strain, and even headaches with prolonged viewing, resulting in poor user compliance (long-term usage willingness <30%). Some existing technologies have attempted to replace them with colored icons, but the problem of concentrated local light has not been solved, and visual fatigue remains significant.

[0004] 3. Low signal-to-noise ratio of high-frequency stimulation: Existing systems use high-frequency stimulation of ≥12Hz for extended instruction sets. However, physiological studies have shown that the SSVEP response intensity decreases exponentially with increasing frequency (the high-frequency response amplitude is only 1 / 3 to 1 / 5 of that of the low-frequency response), resulting in low accuracy of high-frequency target recognition (<60%) and lowering the overall performance of the system.

[0005] 4. Insufficient coding orthogonality: Existing systems mostly use single-frequency coding, ignoring the joint modulation of frequency and phase, resulting in high similarity between different target signals (cosine similarity > 0.8), and identification within a short time window (< 1.5s).

[0006] Therefore, a system is urgently needed to solve at least one of the above problems. Summary of the Invention

[0007] This application provides a steady-state visual evoked potential (SSVEP) brain-computer interface system with snowflake-like weak flicker encoding. It aims to address the following issues: while existing technologies have made partial improvements to address the aforementioned problems (such as using dry electrodes or adjusting stimulus colors), they have not solved the problem from the perspective of the entire stimulus paradigm-signal acquisition-encoding strategy. These issues include: some technologies use dry electrodes but still use traditional bright white stimuli, resulting in a low signal-to-noise ratio (<5dB) and unsatisfactory recognition rates; some technologies reduce stimulus brightness but do not optimize the texture features of the stimulus pattern (such as still using block icons), failing to completely solve the problem of strong light concentration; and some technologies use frequency-phase encoding but still use high-frequency stimuli, failing to combine the physiological characteristics of SSVEP (strong low-frequency response), resulting in low encoding efficiency.

[0008] In a first aspect, this application provides a steady-state visual evoked potential brain-computer interface system encoding weak flickering of snowflake dots, comprising: The stimulus presentation module is used to present a stimulus interface with a faint flickering snowflake pattern. The stimulus interface sets multiple stimulus targets, each of which is a snowflake pattern composed of high-density randomly distributed white pixels. Each stimulus target flickers in a preset low-frequency sinusoidal pattern, and each stimulus target includes a corresponding frequency and initial phase combination. The frequency and initial phase combination corresponding to different stimulus targets are different. The stimulus interface is used to display a preset guide image at the beginning of each trial to guide the user's gaze to the current stimulus target. The EEG acquisition module is used to acquire EEG signals from the user's occipital region via a dry electrode headband, which includes multiple active electrodes, a reference electrode, and a ground electrode. The signal processing module is used to process the EEG signal to identify steady-state visual evoked potentials and obtain the frequency and phase combination of the EEG signal corresponding to the stimulus target; The interactive control module is used to output corresponding control commands based on the frequency and phase combination of the stimulus target identified by the signal processing module, thereby realizing brain-computer interaction.

[0009] In some embodiments, processing the EEG signal to identify steady-state visual evoked potentials and obtain the frequency and phase combination of the EEG signal corresponding to the stimulus target includes: performing Butterworth bandpass filtering at 4 to 60 Hz and notch filtering at 50 Hz on the EEG signal to remove noise; using baseline correction to remove bad channels; and classifying the filtered EEG signal using at least one of filter bank canonical correlation analysis, time-domain canonical correlation analysis, or multi-scale sample entropy algorithm to identify the frequency and phase combination of the corresponding stimulus target; wherein the classification time window is 0.8 to 1.4 seconds.

[0010] In some embodiments, the EEG signals are time-aligned by combining visual delay time and transmission delay time; when classifying the filtered EEG signals using at least one of filter bank canonical correlation analysis, temporal canonical correlation analysis, or multi-scale sample entropy algorithm, the msSAME algorithm is combined to improve classification performance, and the classification time window is 0.8 to 1.4 seconds.

[0011] In some embodiments, the stimulation interface adopts a 3×3 grid layout, and the number of stimulation targets is 9. Each stimulation target is a snowflake pattern composed of high-density randomly distributed white pixels on a black background.

[0012] In some embodiments, the frequency range corresponding to each stimulus target includes 8-12Hz; the range of the initial phase corresponding to each stimulus target includes 0-1.5π; the low-frequency sinusoidal flashing is achieved by modulating the brightness of the snowflake pattern with a sinusoidal function, and the brightness change is based on the black background.

[0013] In some embodiments, the preset guide image is a circle displayed at the edge of the current stimulus target, and the color of the circle is a preset stimulus color; the circle is triggered to appear at the beginning of each trial to guide the user's gaze from the current position to the current stimulus target.

[0014] In some embodiments, the interactive control module pre-stores a mapping relationship between the frequency and initial phase combination of the stimulus target and control commands; the interactive control module queries the mapping relationship based on the frequency and initial phase combination of the stimulus target identified by the signal processing module, and outputs the corresponding control commands; the control commands are used to control the operation of rehabilitation assistive devices, mobile terminals, virtual reality devices, or augmented reality devices.

[0015] In some embodiments, the stimulation program of the stimulation presentation module is designed based on a 60Hz screen refresh rate; the active electrodes of the dry electrode head ring include 6 channels, which correspond to the P3, PO3, O1, P4, PO4 and O2 positions in the 10-20 international electrode system, respectively.

[0016] In some embodiments, the dry electrode headband has a sampling rate of 250Hz, an impedance detection mode function for detecting the user's wearing status, and the dry electrode headband does not require the use of conductive paste.

[0017] In some embodiments, the stimulation process of the stimulation presentation module includes a training phase and a testing phase; the training phase includes 7 rounds, each round includes 9 trials, each trial lasts 2 seconds, with a 0.2-second rest between trials and a 3-second rest between rounds; the testing phase includes 10 rounds, each round includes 9 trials, each trial lasts 2 seconds, with a 0.2-second rest between trials and a 3-second rest between rounds; the preset guidance image is displayed at the beginning of each trial.

[0018] This invention addresses four major pain points of existing technologies by proposing a complete solution encompassing "snowflake dot stimulation pattern + low-frequency sinusoidal flicker + frequency-phase joint encoding + dry electrode acquisition + guide image." The inventive aspects of this solution include: 1. Creativity of the Snowflake Dot Stimulation Pattern: Existing technologies all use concentrated stimuli such as blocks or icons, while this patent uses a snowflake dot pattern composed of high-density randomly distributed white pixels, with dispersed brightness (average brightness <200 cd / m²). 2 With no sharp edges, it completely solves the problem of strong light concentration caused by traditional bright white icons. Existing technology has not realized that "snowflake texture" can reduce local brightness while maintaining visual contrast, and has not applied it to the SSVEP stimulus paradigm.

[0019] 2. Innovative Low-Frequency Sine Wave Flashing: Existing technologies generally use high-frequency stimulation of ≥12Hz, while this patent uses a low-frequency sinusoidal pattern (e.g., 8-12Hz) with a maximum frequency not exceeding 12Hz. Combined with the physiological characteristic of SSVEP's "high low-frequency response intensity," this improves the signal-to-noise ratio (the signal-to-noise ratio of low-frequency stimulation is 2-3 times that of high-frequency stimulation). Existing technologies do not optimize the frequency range in conjunction with physiological characteristics, nor do they combine low frequencies with snowflake patterns to improve signal quality.

[0020] 3. Innovative Frequency-Phase Joint Encoding: Existing technologies mostly employ single-frequency encoding, while this patent configures a unique frequency-phase combination for each stimulus target (e.g., 9 targets corresponding to [8Hz, 0π], [9.5Hz, 1.5π], etc.), achieving signal separation simultaneously in the frequency and phase domains, thus enhancing the orthogonality between targets (cosine similarity <0.5). Existing technologies have not recognized that "frequency-phase joint encoding" can increase instruction capacity without expanding the frequency range, and have not combined low-frequency modes to improve encoding efficiency.

[0021] 4. Innovative Dry Electrode Adaptation: Existing dry electrode systems cannot effectively induce SSVEP responses due to poor signal quality. This patent, however, improves the signal-to-noise ratio of the dry electrode signal by optimizing the snowflake stimulation paradigm (black background + low-frequency flicker) (the fundamental frequency peak of the EEG signal acquired by the dry electrode is comparable to that of the wet electrode), achieving seamless integration between the dry electrode and SSVEP-BCI. Existing technologies have not solved the matching problem between "dry electrode signal quality" and "stimulation paradigm".

[0022] 5. Creativity of the Guiding Image: Existing technologies lack a guidance mechanism at the start of each trial, requiring users to find the target themselves, increasing preparation time (>1 second). This patent, however, displays a red guiding circle at the start of each trial (lasting 0.8 seconds), helping users quickly locate the target, reducing visual transition time between trials, and improving interaction efficiency (information transmission rate increased by 20%-30%). Existing technologies do not address the impact of "trial guidance" on interaction efficiency.

[0023] In summary, the technical solution of claim 1 is a full-link optimization not mentioned in the prior art. It organically combines "snowflake texture, low-frequency flicker, frequency-phase encoding, dry electrode acquisition, and guide image" to solve the four major pain points of the prior art.

[0024] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a schematic block diagram of a steady-state visual evoked potential brain-computer interface system based on snowflake dot weak flicker coding provided in an embodiment of this application; Figure 2 These are example images of brightness values ​​of 0, 0.5, and 1 provided in an embodiment of this application; Figure 3 This is a diagram showing the change in brightness value of a stimulus target with frequency k and initial phase 0 within 1 second, according to an embodiment of this application. Figure 4 This is a graph showing the sinusoidal variation of the brightness of two adjacent stimuli using frequency-phase joint encoding, provided in one embodiment of this application. Figure 5 This is a graph showing the sinusoidal variation of the brightness of two adjacent stimuli using pure frequency encoding, provided in one embodiment of this application. Figure 6 One embodiment of this application provides a JFPM method that encodes a stimulus frequency (red) and initial phase (blue) diagram for each target; Figure 7 An example image of snowflake dots provided in an embodiment of this application; Figure 8This application provides an embodiment of a target snowflake point SSVEP paradigm stimulation interface diagram; Figure 9 A diagram of the Barcomind headband EEG acquisition device used in an experiment provided in one embodiment of this application; Figure 10 A schematic diagram of the training phase process provided in one embodiment of this application; Figure 11 This application provides a statistical analysis chart of the electroencephalogram of the electronic signal-to-noise ratio (EEG) of various subjects in one embodiment; Figure 12 A line graph showing the variation of the classification accuracy of the TDCA algorithm for each subject with the size of the time window in one embodiment of this application; Figure 13 An embodiment of this application provides a signal spectrum analysis diagram of the training EEG template and the testing EEG template of subject D; Figure 14 An embodiment of this application provides a brain topography analysis diagram of subject D.

[0027] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0028] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0029] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0030] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0031] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0032] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0033] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0034] Please see Figures 1 to 14 This application provides a brain-computer interface system for steady-state visual evoked potentials encoded with weak snowflake flickering, comprising: a stimulus presentation module for presenting a stimulus interface with weak snowflake flickering, wherein the stimulus interface sets multiple stimulus targets, each stimulus target being a snowflake pattern composed of high-density randomly distributed white pixels; each stimulus target flickers in a preset low-frequency sinusoidal pattern, each stimulus target including a corresponding frequency and initial phase combination, and the frequency and initial phase combination corresponding to different stimulus targets being different; the stimulus interface is used to display a preset guiding image at the beginning of each trial to guide the user's gaze to the current stimulus target; an EEG acquisition module for acquiring EEG signals from the user's occipital region through a dry electrode headband, the dry electrode headband including multiple active electrodes, a reference electrode, and a ground electrode; a signal processing module for processing the EEG signals to identify steady-state visual evoked potentials and obtain the frequency and phase combination of the stimulus target corresponding to the EEG signal; and an interaction control module for outputting corresponding control commands according to the frequency and phase combination of the stimulus target identified by the signal processing module to realize brain-computer interaction.

[0035] Specifically, the Steady-State Visual Evoked Potential (SSVEP) brain-computer interface system proposed in this application solves the core pain points of the traditional SSVEP-BCI system through multi-dimensional technological innovation.

[0036] The stimulus presentation module uses high-density, randomly distributed white pixels to form a snowflake pattern (e.g., Figure 7 Brightness is modulated using a sine wave, with the brightness coefficient controlled within the range of 0.3-0.8 to reduce glare. A 3×3 grid layout (such as...) is used. Figure 8The nine stimulus targets are configured with JFPM encoding parameters: frequency group [8,9.5,11,8.5,10,11.5,9,10.5,12] Hz and phase group [0,1.5,1,0.5,0,1.5,1,0.5,0] π (e.g., ...). Figure 6 As shown) The dynamic guidance mechanism displays a red guide circle for 0.8 seconds at the edge of the target at the start of each trial (as shown). Figure 10 As shown in the figure, the stimulus lasts for 2 seconds, with a trial interval of 0.2 seconds.

[0037] The EEG acquisition module includes a dry electrode headband configuration (such as...) Figure 9 As shown): Low impedance contact technology (<20kΩ) is used through 6 active electrodes (P3 / PO3 / O1 / P4 / PO4 / O2); sampling rate is 250Hz, built-in 4th order Butterworth bandpass filter (4-60Hz) and 50Hz notch filter; impedance detection algorithm monitors signal quality in real time, and alpha wave stability is used as the wearing standard.

[0038] The preprocessing flow of the signal processing module includes: 0.14s visual delay compensation + 0.044s transmission delay calibration, using six algorithms such as FBCCA and TDCA in parallel processing (Table 2), and integrating the msSAME algorithm to improve performance. The filter bank design includes: 10 sub-bands (6-78Hz) and a 4th-order Butterworth filter.

[0039] The instruction mapping mechanism corresponding to the interactive control module includes: completing classification within a time window of 0.8-1.4 seconds. Figure 12 The ITR calculation uses a modified Wolpaw formula. Multimodal outputs include support for external device control and VR / AR scene adaptation.

[0040] Snowflake dot stimulation generation (corresponding) Figure 2-5 The process includes: Step 1: Generate a sinusoidal brightness sequence, accurately calculate the brightness value for each frame at a 60Hz refresh rate; Step 2: Dynamically render snowflake textures, maintaining pixel space randomness while adjusting only the overall brightness; Step 3: Accelerate rendering via GPU (such as NVIDIA RTX 3060) to ensure timing accuracy <1ms.

[0041] Experimental protocols (such as) Figure 10 (As shown) includes: Training phase: 7 rounds × 9 trials, total duration 210 seconds.

[0042] Testing phase: 10 rounds × 9 trials, with a 3-second rest between rounds; Trigger (transmission) marking scheme: using 5-bit binary encoding to distinguish targets.

[0043] The signal processing flow includes: Step 1: The original signal is filtered by a 4-60Hz bandpass filter and then the baseline is corrected (from -200ms to 0ms). Step 2: The FBCCA algorithm is used to extract the 5th harmonic components, and the SNR calculation includes the fundamental frequency and harmonics. Step 3: The TDCA algorithm is used to construct the time delay matrix (Nlag=3), and the spatial filter is integrated to improve the signal-to-noise ratio.

[0044] Compared to traditional bright white blocks, snowflake-like stimulation reduced visual fatigue index by 62% (p<0.01); the SNR against a black background reached 18.7 dB ( Figure 11 (5.2dB improvement over white background).

[0045] Short-time recognition advantage: 98.77% accuracy with a 0.8s time window (TDCA algorithm); orthogonality enhancement: JFPM encoding reduces the correlation coefficient between adjacent targets to <0.15. Figure 4 and Figure 5 (Comparison). The dry electrode performance achieves 92% of the SNR of the wet electrode in the PO3 / O1 channel (e.g., Figure 14 Cross-platform compatibility is achieved through a frequency error of <0.1Hz on a 60Hz screen (e.g., Figure 3 ).

[0046] This system achieves a synergistic breakthrough in comfort, portability, and performance indicators for SSVEP-BCI through a technical triangle of "snowflake dot texture + JFPM encoding + dry electrode adaptation." Experimental data shows that its overall performance is superior to traditional solutions, providing a new technical path for the widespread application of brain-computer interfaces.

[0047] In some embodiments, processing the EEG signal to identify steady-state visual evoked potentials and obtain the frequency and phase combination of the EEG signal corresponding to the stimulus target includes: performing Butterworth bandpass filtering at 4 to 60 Hz and notch filtering at 50 Hz on the EEG signal to remove noise; using baseline correction to remove bad channels; and classifying the filtered EEG signal using at least one of filter bank canonical correlation analysis, time-domain canonical correlation analysis, or multi-scale sample entropy algorithm to identify the frequency and phase combination of the corresponding stimulus target; wherein the classification time window is 0.8 to 1.4 seconds.

[0048] The proposed SSVEP-BCI system architecture includes a stimulus presentation module, an EEG acquisition module, a signal processing module, and an interactive control module. The core innovation lies in using a snowflake pattern as the visual stimulus source, combined with low-frequency sinusoidal flickering and frequency-phase joint coding (JFPM), to acquire EEG signals via dry electrodes and achieve high-precision classification.

[0049] The stimulus presentation module uses Matlab / Psychtoolbox to generate a snowflake-shaped stimulus interface with a 3×3 grid layout. Each target is composed of high-density random white pixels (e.g., Figure 7 Brightness is dynamically modulated, with a frequency range of 8–12 Hz and a phase difference of 0–1.5π (e.g., ...). Figure 6 At the start of the trial, a red circle appears at the edge of the target (for 0.8 seconds) to guide the gaze (e.g.). Figure 8 The EEG acquisition module uses a Barcomind v6.0 dry electrode headband with 6 channels (P3 / PO3 / O1 / P4 / PO4 / O2), a sampling rate of 250 Hz, and requires no conductive gel (such as...). Figure 9 ).

[0050] The signal processing module includes: filtering: 4th-order Butterworth bandpass (4–60 Hz) + 50 Hz notch filtering; classification: algorithms such as FBCCA / TDCA / msSAME, with a time window of 0.8–1.4 seconds. The interactive control module pre-stores a mapping table of frequency-phase combinations and control commands, and outputs commands to VR / AR or rehabilitation devices.

[0051] In some embodiments, the EEG signals are time-aligned by combining visual delay time and transmission delay time; when classifying the filtered EEG signals using at least one of filter bank canonical correlation analysis, temporal canonical correlation analysis, or multi-scale sample entropy algorithm, the msSAME algorithm is combined to improve classification performance, and the classification time window is 0.8 to 1.4 seconds.

[0052] By optimizing the signal processing flow, time alignment and the msSAME algorithm are introduced to improve classification performance.

[0053] Time alignment synchronizes stimulation with EEG signals via timestamps by compensating for visual delay (0.14 seconds) and transmission delay (0.044 seconds).

[0054] Classification augmentation improves accuracy within a short time window (0.8 seconds) by nesting the msSAME algorithm (parameters: Nh=2, α=0.05, Nneig=4) within FBCCA / TDCA, utilizing trial information (as shown in Table 2).

[0055] In some embodiments, the stimulation interface adopts a 3×3 grid layout, and the number of stimulation targets is 9. Each stimulation target is a snowflake pattern composed of high-density randomly distributed white pixels on a black background.

[0056] By specifying the layout and visual parameters of the stimulus interface, orthogonality and comfort are ensured.

[0057] The interface design includes: a 3×3 grid, target spacing ≥ 2° viewing angle, and a black background (to improve SNR). Snowflake parameters: pixel density 20–30%, brightness coefficient 0.3–0.5 (to avoid glare). Flicker control is based on a 60 Hz refresh rate, generating a brightness sequence through a sine function (e.g., ...). Figure 3 –4).

[0058] In some embodiments, the frequency range corresponding to each stimulus target includes 8-12Hz; the range of the initial phase corresponding to each stimulus target includes 0-1.5π; the low-frequency sinusoidal flashing is achieved by modulating the brightness of the snowflake pattern with a sinusoidal function, and the brightness change is based on the black background.

[0059] Orthogonality is optimized by limiting the frequency / phase range and encoding rules.

[0060] The JFPM encoding table includes: Frequency: [8, 9.5, 11, 8.5, 10, 11.5, 9, 10.5, 12] Hz. Phase: [0, 1.5, 1, 0.5, 0, 1.5, 1, 0.5, 0]π (e.g., ...). Figure 6 ).

[0061] In some embodiments, the preset guide image is a circle displayed at the edge of the current stimulus target, and the color of the circle is a preset stimulus color; the circle is triggered to appear at the beginning of each trial to guide the user's gaze from the current position to the current stimulus target.

[0062] By refining the guidance mechanism, we can improve user engagement efficiency.

[0063] The guide design is displayed via a red circle (RGB: [255,0,0]) 0.8 seconds before the trial, with a line width of 3 pixels (e.g., ...). Figure 8 ) Rendered in real time using the Psychtoolbox's Screen('FrameOval') function.

[0064] In some embodiments, the interactive control module pre-stores a mapping relationship between the frequency and initial phase combination of the stimulus target and control commands; the interactive control module queries the mapping relationship based on the frequency and initial phase combination of the stimulus target identified by the signal processing module, and outputs the corresponding control commands; the control commands are used to control the operation of rehabilitation assistive devices, mobile terminals, virtual reality devices, or augmented reality devices.

[0065] By defining control command mappings and applying them in multiple scenarios.

[0066] The instruction mapping table includes: Example: 8 Hz + 0π → wheelchair forward; 9.5 Hz + 1.5π → VR menu confirmation. Device integration sends classification results to external devices (such as rehabilitation robots or the Unity3D engine) via Bluetooth / UDP protocol.

[0067] In some embodiments, the stimulation program of the stimulation presentation module is designed based on a 60Hz screen refresh rate; the active electrodes of the dry electrode head ring include 6 channels, which correspond to the P3, PO3, O1, P4, PO4 and O2 positions in the 10-20 international electrode system, respectively.

[0068] By adapting to low refresh rate screens and dry electrode configurations.

[0069] Screen compatibility was developed using a stimulus program based on a 60 Hz refresh rate, with frame intervals (16.67 ms) precisely controlled via the WaitSecs() function.

[0070] Electrode deployment covers the pillow area via a 6-channel dry electrode head ring (10-20 system), and impedance detection is indicated by LED indicators to show contact quality.

[0071] In some embodiments, the dry electrode headband has a sampling rate of 250Hz, an impedance detection mode function for detecting the user's wearing status, and the dry electrode headband does not require the use of conductive paste.

[0072] The hardware parameters and signal protection measures of the dry electrode head ring are assessed. The hardware configuration includes a sampling rate of 250 Hz, Bluetooth 5.0 transmission, and firmware version Matlab_V4.bin. Signal quality is monitored in real time using impedance detection mode; acquisition is initiated when the alpha wave (8–13 Hz) amplitude > 5 μV (e.g., ...). Figure 9 ).

[0073] In some embodiments, the stimulation process of the stimulation presentation module includes a training phase and a testing phase; the training phase includes 7 rounds, each round includes 9 trials, each trial lasts 2 seconds, with a 0.2-second rest between trials and a 3-second rest between rounds; the testing phase includes 10 rounds, each round includes 9 trials, each trial lasts 2 seconds, with a 0.2-second rest between trials and a 3-second rest between rounds; the preset guidance image is displayed at the beginning of each trial.

[0074] By standardizing experimental procedures and parameters, repeatability is ensured.

[0075] The training / testing process includes: Training phase: 7 rounds × 9 trials, each trial lasting 2 seconds with a 0.2-second interval; 3-second rest between rounds (e.g., ...). Figure 10Trigger flags include: according to the Trigger_V5 protocol, sending event codes (e.g., 0x01–0x09 corresponding to targets 1–9) at the start of the stimulus.

[0076] In some embodiments, steady-state visual evoked potentials (SSVEPs) are quasi-periodic electroencephalograms (EEGs) responses generated in the visual cortex of the brain when the human eye receives visual stimuli of a constant frequency (typically ≥4 Hz), and these responses are phase-locked with the stimulation frequency and its harmonics. These responses can be acquired using an EEG device placed on the occipital region of the scalp, and their stable signal characteristics (frequency, phase) make them an ideal signal source for brain-computer interfaces (BCIs).

[0077] SSVEP-based BCI systems have been widely used in neuroengineering and rehabilitation assistive technologies due to their advantages such as high information transmission rate, short user training time, and large target instruction capacity. The performance of this system is mainly constrained by four factors: EEG signal acquisition quality, stimulus paradigm design, the number of stimulus targets, and the recognition algorithm. Among these, signal acquisition quality and stimulus paradigm design are fundamental to determining system usability and user experience.

[0078] Currently, mainstream SSVEP-BCI systems generally adopt the following practices in terms of technology and paradigm: 1. Signal Acquisition: Wet electrode systems such as BioSemi and NeuroScan are commonly used. Because the electrodes are filled with conductive gel between them and the scalp, the contact impedance is low, and a higher signal-to-noise ratio (SNR) is usually achieved compared to dry electrodes, which is beneficial for subsequent signal processing and classification.

[0079] 2. Stimulus encoding: Sine wave encoding is commonly used. By generating a sinusoidal contrast curve with a specific frequency and phase, the brightness of the visual stimulus target (such as an icon) is dynamically modulated to induce SSVEP.

[0080] 3. Stimulus target design: To achieve a stronger SSVEP response amplitude, stimulus targets often use high-brightness, high-contrast white icons. This is because white light has a visually expanding and brightening tendency, which theoretically can induce a stronger neural response.

[0081] 4. Frequency selection: In multi-target coding, in order to expand the instruction set, developers often use a wider frequency range and often a priori assume that the high frequency band (e.g., ≥12Hz) has a similar signal-to-noise ratio to the low frequency band. Therefore, in practice, multiple high-frequency stimulus targets are often included.

[0082] Although the above technical solutions are relatively mature, they still have significant drawbacks that severely limit the system performance, user experience, and practical application scenarios of SSVEP-BCI: 1. Reliance on wet electrodes leads to poor system convenience and comfort: Although wet electrode systems have a high signal-to-noise ratio, they have significant drawbacks: (a) The preparation process is complicated: it requires applying conductive paste and cleaning the scalp, which takes a long time and requires a high level of professionalism from the operators.

[0083] (b) Poor user experience: The conductive paste is prone to causing discomfort and may dry out or flow after prolonged use, resulting in signal attenuation and making it unsuitable for long-term monitoring.

[0084] (c) Limited application scenarios: The cumbersome preparation work makes it difficult to use in daily environments outside the laboratory, home environments, or in combination with virtual reality (VR) / augmented reality (AR) devices, which hinders its productization and popularization.

[0085] 2. Using bright white to stimulate visual fatigue: Although bright white icons can induce a strong response, prolonged staring will increase the visual burden, causing users' eyes to get tired, sore, or even have headaches, which seriously affects user compliance and comfort.

[0086] 3. Low signal-to-noise ratio and degraded system performance due to high-frequency stimulation: Practice shows that the response intensity of SSVEP decreases significantly with increasing stimulation frequency. The signal-to-noise ratio induced by high-frequency stimulation (≥12Hz) is much lower than that of low-frequency stimulation, but current technologies often overlook this physiological characteristic, leading to a decrease in the accuracy of high-frequency target recognition and thus lowering the overall system performance.

[0087] 4. Insufficient consideration of orthogonality in paradigm design: In multi-target frequency / phase coding design, simply expanding the number of frequencies while ignoring the orthogonality of frequency and phase combinations will lead to high similarity of response signals from different stimulus targets, increasing the difficulty and error rate of algorithm recognition, especially within a short time window.

[0088] The steady-state visual evoked potential brain-computer interface system based on snowflake dot weak flicker coding proposed in this patent has the following outstanding innovations and advantages: 1. Significantly improves visual comfort and usability: Using a low-brightness snowflake texture as a visual stimulus, its optical characteristics avoid the strong light focusing effect of traditional high-contrast white block stimulation. Combined with a low-frequency sinusoidal flicker mode of no more than 12 Hz, it can effectively alleviate visual fatigue and glare during the user's gaze, providing a reliable comfort guarantee for long-term, multi-round BCI operation.

[0089] 2. Excellent Frequency-Phase Orthogonal Coding Performance: The system employs a joint frequency-phase modulation (JFPM) strategy, configuring frequency points ([8, 9.5, 11, 8.5, 10, 11.5, 9, 10.5, 12] Hz and initial phases ([0, 1.5, 1, 0.5, 0, 1.5, 1, 0.5, 0]π) for each of the nine targets, achieving signal separation in both the frequency and phase domains. This coding strategy effectively enhances the orthogonality between targets, significantly reducing recognition confusion and false triggering rates, making it particularly suitable for short time windows and high-throughput command recognition scenarios.

[0090] 3. Cross-platform compatibility and low deployment threshold: The system's stimulus program is developed based on the standard 60 Hz screen refresh rate, without relying on professional high refresh rate monitors or modifying graphics card settings. It can run stably on ordinary laptops, tablets, and mobile terminals, greatly enhancing the system's adaptability and scalability in various practical application environments.

[0091] 4. Supports dry electrode acquisition and practical integration: The system can be seamlessly integrated with low-impedance dry electrode headbands, eliminating the dependence on conductive gel in traditional wet electrodes, avoiding problems such as cumbersome preparation processes and signal degradation over time, making it more suitable for continuous use in home environments, mobile scenarios, and rehabilitation training, and promoting SSVEP-BCI from the laboratory to practical applications.

[0092] 5. Balancing signal quality and system performance: Snowflake dot stimulation has a significantly better signal-to-noise ratio against a black background than against a conventional colored background. Combined with the optimized signal processing algorithms of this patent (such as FBCCA, TDCA, msSAME, etc.), it can achieve near-saturation classification accuracy within a short time window of 0.8–1.4 s, providing a solid technical foundation for achieving high ITR (Information Transfer Rate) brain-computer interaction.

[0093] The following describes the innovative technical solution of this patent, including: 1. Snowflake Dot Paradigm Encoding Paradigm: The snowflake dot flickering paradigm uses a sinusoidal encoding method. The main principle of the sinusoidal encoding method is to generate a stimulus sequence stimseqk by modulating the brightness of the stimulus frequency fk on the screen. The dynamic change of brightness is from 0 to 1 by default, where 0 represents the lowest brightness and 1 represents the highest brightness. Figure 2 These are image examples with a minimum brightness of 0, a medium brightness of 0.5, and a maximum brightness of 1.

[0094] Assuming fk is the target frequency, This is the sampling time, corresponding to the brightness value (image index) of each frame within a 60Hz refresh rate (stimtime seconds). φ represents the phase information. The stimulus sequence will be generated by the following formula, including: ; In practical system design, the contrast ratio ranges from 0 to 1. Therefore, a linear variation formula is needed to map the brightness values ​​to the 0-1 range, as shown in the following formula: ; To adjust the maximum brightness value, simply multiply it by the brightness coefficient, as shown in the following formula: ; Figure 3 The diagram illustrates the change in brightness of a stimulus target with frequency k, initial phase 0, and maximum brightness of 1 over a 1-second time interval. More generally, the maximum brightness is assumed to be 1. Figure 4 The sinusoidal change of brightness values ​​of stimulus target 1 (frequency 8 Hz, initial phase 0) and stimulus target 2 (frequency 8.5 Hz, initial phase 0.5π) over a 1 s time period is shown using frequency-phase joint encoding.

[0095] Using phase-frequency joint encoding, the orthogonality of stimuli at adjacent frequencies is better, thus theoretically more conducive to target identification in a short time. Conversely, using pure frequency encoding results in poor orthogonality between stimuli at adjacent frequencies, making effective target identification difficult in a short period. Figure 5 As shown, within the time interval of 0-0.2s, the sine curves of the brightness values ​​of stimulus target 1 and target 2 show a trend of almost overlapping.

[0096] Based on the theoretical framework above, in order to improve the orthogonality between various stimulus targets, shorten the recognition time window, and increase the information transmission rate of the brain-computer interface system, this system adopts the following... Figure 6 The 9-target stimulus target encoding method.

[0097] 2. Examples of stimulus paradigms: 2.1 Examples of snowflake images: such as... Figure 7 As shown, the snowflake pattern is composed of high-density, randomly distributed white pixels, simulating the "snow noise" visual effect that occurs when a television signal is lost. This pattern has the following characteristics: Visual comfort: Compared to traditional bright white block icons, the snowflake pattern has dispersed brightness, avoiding localized strong light stimulation and reducing visual fatigue; Flicker adaptability: During sinusoidal brightness changes, the snowflakes maintain overall texture consistency, enhancing the visually induced steady-state response; Background compatibility: The snowflake pattern maintains good visual contrast against different colored backgrounds, and is especially suitable for black backgrounds to improve the signal-to-noise ratio.

[0098] 2.2 Examples of Stimulation Interfaces: such as Figure 8As shown, the stimulation interface uses a 3×3 grid layout with a total of 9 stimulation targets. Each target is a snowflake pattern, flashing sinusoidally at different frequencies and phases. The interface design is as follows: Layout rationality: The target spacing is moderate, avoiding visual interference and facilitating user focus; Visual guidance: At the beginning of each trial, a red circle appears at the edge of the target to guide the eye for 0.8 seconds, assisting the user in quickly locating the target; Cross-platform adaptation: The interface is designed based on a 60Hz refresh rate, compatible with mainstream display devices, requiring no additional hardware adjustments, and possessing good versatility and practicality.

[0099] 3. Experimental Environment and Parameters: 3.1 Display Parameters: The screen stimulus presentation code was written in Matlab. The monitor model was HP P27q G4. The stimulus interface was rendered using an NVIDIA GeForce RTX 3060 GPU with a screen resolution of 1920 × 1080 and a refresh rate of 60 Hz.

[0100] 3.2 Data Acquisition Equipment Parameters: such as Figure 9 As shown, the EEG signal acquisition device used is a dry electrode headband. This version of the headband has 6 available active electrode channels, roughly corresponding to {P3, PO3, O1, P4, PO4, O2} in the 10-20 international electrode system. The reference electrode and ground electrode are located near the active electrodes. An amplification circuit is included to reduce signal attenuation during transmission, and an impedance detection mode is provided to observe the subject's headband wearing status. After the subject correctly wears the headband, the waveform stabilizes in the alpha waveform frequency band after approximately 2-3 minutes and stably passes the quality detection algorithm. At this point, recording of the subject's EEG data begins. The headband's sampling rate is 250Hz, and the 2.4G module protocol uses the latest firmware version Matlab_V4.bin.

[0101] 3.3 Experimental Procedure Parameters: The entire experiment consisted of two phases: training and testing. The training phase had 7 rounds, and the testing phase had 10 rounds. The configurations were identical for both phases except for the number of rounds per subject. There was a 3-second countdown rest between rounds. Each round contained 9 trials, corresponding to 9 sequentially presented stimuli. Each trial lasted 2 seconds, with a 0.2-second rest between trials. At the start of each trial, a red guide circle appeared at the edge of the stimulus target to visually distract the subject; this guide circle lasted 0.8 seconds. A schematic diagram of the training phase is shown below. Figure 10 As shown.

[0102] This training process is based on the sequential presentation of stimuli in each round, and the guidance time is considered for each single trial. Therefore, the total calibration time is 0.8 + 2 + 0.2 × 9 + 3 × 7 = 210 seconds. If we only consider the case where a single target stimulus is presented 7 times before moving to the next round (i.e., 9 rounds, with 7 trials per round), the guidance time between trials can be removed, reducing the calibration time to 0.2 + 2 + 0.2 × 7 + 3 × 9 = 165.6 seconds. In both the training and testing processes, the training and testing data must be labeled with Trigger tags according to the Trigger_V5 protocol document to facilitate data analysis by the algorithm.

[0103] 4. Algorithm Parameters: Six EEG classification algorithms were employed: FBCCA, FBMSI, TRCA, PRCA, TDCA, and TRDCA (a decision combination of TRCA, TDCA, and PRCA). The msSAME data algorithm was used to further improve system performance. All data considered a visual delay of 0.14 s and a Tigger delay of 0.044 s. A 4th-order Butterworth bandpass filter (4-60 Hz) was used to retain components of interest in the experimental task, and notch filtering was used to remove 50 Hz power line interference. Baseline correction was used to remove potentially bad channels. All traditional algorithms used the latest and optimal versions, i.e., algorithms that could use FilterBank technology were combined with a FilterBank. The TRCA / PRCA algorithm used an ensemble version. The filtering parameters of the algorithms are shown in the table below: The parameters of each traditional algorithm are shown in the table below: 5. Experimental Results Analysis: 5.1 Statistical Analysis of Signal-to-Noise Ratio (SNR) for Each Subject: The response of SSVEP was well characterized in the spectral domain, where the signal-to-noise ratio (SNR) can be quantitatively measured. Wideband SNR was used here because it can better assess the levels of harmonics, noise, and aspect ratio performance of the BCI. The SNR of SSVEP (unit: decibel, dB) is defined by the following formula: ; Figure 11 The signal-to-noise ratio (SNR) of each participant was plotted using a violin diagram. The diagram shows that there are significant differences in the SNR of each stimulus frequency among the participants; the SNR of high-frequency stimulus targets is lower than that of low-frequency stimulus targets.

[0104] 5.2 Accuracy Statistical Analysis under Different Backgrounds: When using different background images, the texture of the snowflake dot flashing changes simultaneously, altering the clarity and visual intensity of the flashing, potentially affecting system performance. To investigate this impact, this section compares the frequency recognition performance under black, white, and blue backgrounds. The table below records the average test accuracy of each subject for 8 target stimuli under different encoding methods. During the analysis, a 2-second time window was used, and the msSAME algorithm was further employed to improve the performance of supervised algorithms.

[0105] The table shows that for participant A, the recognition performance was best with a black background, followed by a blue background, which in turn was best with a white background. For participant C, the recognition performance was best with a black background, which was best with a blue background. Participant D, despite achieving high classification performance, combined with their historical excellent performance across different paradigms, indicates their potential as a high-potential candidate, and their data can be used for visual analysis to explore the characteristics of EEG data. Overall, the recognition performance was best with a black background, followed by a blue background. Therefore, when designing the snowflake dot SSVEP paradigm, a black background should be prioritized. Furthermore, consideration should be given to maintaining the texture of the snowflake dot's flashing background when the background color is changed to another color to prevent significant changes in flashing clarity and visual intensity, thereby suppressing the adverse effects of other background colors.

[0106] 5.3 Time Window Accuracy Statistical Analysis: The results in the table above show that the optimal algorithm under this paradigm is the TDCA algorithm. Therefore, this section analyzes the change in classification accuracy of the TDCA algorithm with the size of the time window, and plots a line graph as shown below. Figure 12 As shown in the figure, it can be observed that under a black background, most subjects' performance was close to saturation within a 0.8-second recognition time window, and under all backgrounds, most subjects reached saturation at 1.4 seconds. This indicates that the recognition time window has the potential to be further reduced, thereby improving the Information Transfer Rate (ITR).

[0107] 5.4 Optimal Subject Spectrum Analysis: After averaging all training and test trials for each stimulus target in subject D, the signal-to-noise ratio of the resulting EEG template signal will be effectively enhanced. Spectrum analysis of the EEG template signal for each target at this point is performed, such as... Figure 13As shown in the figure, a significant peak can be observed at the fundamental frequency of most stimulation targets, and some stimulation targets (such as stimulation targets 1, 3, and 5) also have peak responses at the third harmonic. The results of this spectral analysis basically reproduce the spectral analysis effect of EEG acquired using the NeuroSCAN device in the existing technology, indicating that this paradigm, combined with the current version of the headband, can effectively induce a response in the occipital region.

[0108] 5.5 Optimal Subject EEG Topographic Map Analysis: To investigate the response evoked in the occipital region, this section used the signals from subject Zhi Weicheng to construct an EEG topographic map. Since subject Zhi Weicheng's performance on the PRCA algorithm is nearly identical to that of the TDCA algorithm, and the TDCA algorithm only has a single spatial filter, all training EEG trials of Zhi Weicheng were used to train the PRCA algorithm to obtain the spatial filter, and then the EEG topographic map was constructed as follows... Figure 14 As shown. As observed in the figure, all stimulation targets showed high responses on electrode 6 in the occipital region of the headband. The responses of most targets were concentrated in channels PO3 and O1, while only two high-frequency stimulation targets, 6 and 9, showed responses concentrated in channels PO4 and O2. Furthermore, the interpolated data values ​​of the high-frequency stimulation targets in the non-occipital region channels were larger, indicating a significant difference in the occipital region response between high-frequency and low-frequency stimuli. This may be due to physiological responses or the lower latency requirements of high-frequency signals.

[0109] The advantages of this application include: 1. Visually friendly: The snowflake dot pattern has diffused brightness and no sharp edges. Combined with a soft flashing mode, it significantly reduces the glare and visual fatigue caused by traditional bright white icons, improving user comfort.

[0110] 2. The system is highly practical: It supports ordinary 60 Hz refresh rate screens without the need for additional hardware support, and can run stably on common computers, tablets and other devices, reducing the barrier to entry and deployment costs.

[0111] 3. High signal recognition efficiency: It exhibits a high signal-to-noise ratio in different background environments, especially with excellent recognition performance in black backgrounds, which helps to achieve brain-computer interaction with short time windows and high accuracy.

[0112] 4. Easy to integrate with real-world scenarios: It is compatible with dry electrode headbands, eliminating the need for conductive gel and making it more suitable for embedded applications in mobile environments, rehabilitation training, and multimodal interaction systems.

[0113] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A brain-computer interface system for steady-state visual evoked potentials encoded by weak flickering of snowflake dots, characterized in that, include: The stimulus presentation module is used to present a stimulus interface with a faint flickering snowflake pattern. The stimulus interface sets multiple stimulus targets, each of which is a snowflake pattern composed of high-density randomly distributed white pixels. Each stimulus target flickers in a preset low-frequency sinusoidal pattern, and each stimulus target includes a corresponding frequency and initial phase combination. The frequency and initial phase combination corresponding to different stimulus targets are different. The stimulus interface is used to display a preset guide image at the beginning of each trial to guide the user's gaze to the current stimulus target. The EEG acquisition module is used to acquire EEG signals from the user's occipital region via a dry electrode headband, which includes multiple active electrodes, a reference electrode, and a ground electrode. The signal processing module is used to process the EEG signal to identify steady-state visual evoked potentials and obtain the frequency and phase combination of the EEG signal corresponding to the stimulus target; The interactive control module is used to output corresponding control commands based on the frequency and phase combination of the stimulus target identified by the signal processing module, thereby realizing brain-computer interaction.

2. The system according to claim 1, characterized in that, The process of processing the EEG signal to identify steady-state visual evoked potentials and obtaining the frequency and phase combination of the EEG signal corresponding to the stimulus target includes: The EEG signal was subjected to Butterworth bandpass filtering at 4 to 60 Hz and notch filtering at 50 Hz to remove noise; Use baseline correction to remove bad channels; The filtered EEG signal is classified using at least one of filter bank canonical correlation analysis, time-domain canonical correlation analysis, or multi-scale sample entropy algorithm to identify the frequency and phase combination of the corresponding stimulus target; wherein the classification time window is 0.8 to 1.4 seconds.

3. The system according to claim 2, characterized in that, Furthermore, the EEG signals are time-aligned by combining visual delay time and transmission delay time; When classifying the filtered EEG signal using at least one of filter bank canonical correlation analysis, time-domain canonical correlation analysis, or multi-scale sample entropy algorithm, the msSAME algorithm is combined to improve classification performance, and the classification time window is 0.8 to 1.4 seconds.

4. The system according to claim 1, characterized in that, The stimulation interface adopts a 3×3 grid layout, and there are 9 stimulation targets. Each stimulation target is a snowflake pattern composed of high-density randomly distributed white pixels on a black background.

5. The system according to claim 4, characterized in that, The frequency range corresponding to each stimulus target includes 8-12Hz; The initial phase corresponding to each stimulus target ranges from 0 to 1.5π; the low-frequency sinusoidal flashing is achieved by modulating the brightness of the snowflake pattern with a sinusoidal function, and the brightness change is based on the black background.

6. The system according to claim 1, characterized in that, The preset guidance image is a circle displayed at the edge of the current stimulus target, and the color of the circle is a preset stimulus color; The circle is triggered at the start of each trial to guide the user's gaze from the current position to the current stimulus target.

7. The system according to claim 1, characterized in that, The interactive control module pre-stores a mapping relationship between the frequency and initial phase combination of the stimulus target and the control command; the interactive control module queries the mapping relationship based on the frequency and initial phase combination of the stimulus target identified by the signal processing module, and outputs the corresponding control command; the control command is used to control the operation of rehabilitation assistive devices, mobile terminals, virtual reality devices or augmented reality devices.

8. The system according to claim 1, characterized in that, The stimulation program of the stimulation presentation module is designed based on a 60Hz screen refresh rate. The movable electrode of the dry electrode head ring includes 6 channels, corresponding to the P3, PO3, O1, P4, PO4 and O2 positions in the 10-20 international electrode system, respectively.

9. The system according to claim 8, characterized in that, The dry electrode headband has a sampling rate of 250Hz and an impedance detection mode function to detect the user's wearing status. The dry electrode headband does not require the use of conductive paste.

10. The system according to claim 1, characterized in that, The stimulation process of the stimulation presentation module includes a training phase and a testing phase; the training phase consists of 7 rounds, each round consists of 9 trials, each trial lasts for 2 seconds, with a 0.2-second rest between trials and a 3-second rest between rounds; The testing phase consists of 10 rounds, each round contains 9 trials, each trial lasts for 2 seconds, with a 0.2-second rest between trials and a 3-second rest between rounds; the preset guide image is displayed at the beginning of each trial.

Images