A two-dimensional cognitive state diagnosis and intervention method based on electroencephalogram and eye movement

Abstract

The application belongs to the technical field of brain-computer interface, and particularly relates to a brain electric signal-eye movement two-dimensional cognitive state diagnosis and intervention method, which comprises the following steps: collecting brain electric signal-eye movement mixed signals, decoupling the brain electric signal-eye movement mixed signals to obtain brain electric signals and eye movement signals; calculating a cognitive concentration index based on the brain electric signals; calculating a visual activity index based on the eye movement signals; and obtaining an intervention strategy according to the cognitive concentration index and the visual activity index. The application calculates the cognitive concentration index based on the brain electric signals, calculates the visual activity index based on the eye movement signals, and calculates the cognitive state by combining the cognitive concentration index and the visual activity index, so that the cognitive state which is extremely easy to confuse can be objectively distinguished, and the "cause level" diagnosis of the state of concentrating on distraction can be realized.

Description

Technical Field

[0001] This invention belongs to the field of brain-computer interface technology, and in particular relates to a method for diagnosing and intervening in two-dimensional cognitive states using EEG and eye-tracking. Background Technology

[0002] With the development of human-computer interaction technology, real-time detection and intervention of users' cognitive states have become crucial for improving work efficiency, optimizing learning experiences, and ensuring safety in special operations. Existing attention assessment systems either rely solely on electroencephalogram (EEG) signals to analyze the overall brain activity or simply track visual focus using eye trackers, which have many limitations.

[0003] Assessment methods based solely on electroencephalogram (EEG) signals struggle to distinguish between cognitive states exhibiting similar EEG patterns but fundamentally different characteristics. For instance, during periods of "deep focus" and "mental fatigue," a user's head and eye movements may be relatively still. Analyzing only EEG signals could easily misclassify the latter as a low level of focus, failing to recognize its true nature as "fatigue." Similarly, methods based solely on eye tracking cannot differentiate between "active visual exploration" and "aimless, restless scanning," as both involve rapid eye movements.

[0004] Furthermore, existing intervention strategies are often one-size-fits-all. When a system detects user distraction, it typically provides a uniform alert, such as an audio or visual warning. However, the underlying causes of distraction are diverse, stemming from internal wandering thoughts or external visual disturbances. Blindly intervening without considering the root cause not only has limited effectiveness but may even cause unnecessary disruption to the user.

[0005] Therefore, there is an urgent need in this field for a technical solution that can accurately distinguish the underlying causes of different states of focused distraction and implement targeted interventions accordingly, in order to solve the problems of vague diagnosis and blind intervention in existing technologies. Summary of the Invention

[0006] To address the aforementioned problems in the prior art, this invention employs a two-dimensional cognitive state diagnosis and intervention method based on electroencephalography (EEG) and eye movement (EEG-EM) tracking, comprising:

[0007] S1. Acquire mixed EEG-eye movement signals, decouple the mixed EEG-eye movement signals from their sources, and obtain EEG signals and eye movement signals;

[0008] S2. Calculate cognitive attention index based on EEG signals;

[0009] S3. Calculate visual activity index based on eye movement signals;

[0010] S4. Obtain intervention strategies based on cognitive attention index and visual activity index.

[0011] Beneficial effects:

[0012] 1. This invention calculates cognitive focus indices based on EEG signals and visual activity indices based on eye movement signals. Combining these indices to calculate cognitive state allows for the objective differentiation of easily confused cognitive states, achieving a "causal" diagnosis of focused but distracted states. 2. Existing cognitive assessments often rely solely on single time-domain energy features, leading to misjudgments. This invention integrates time-domain focus indices and spatial-domain synchronization indices to accurately quantify a user's cognitive focus. The time-domain focus index reflects the brain's working focus rhythm, while the spatial-domain synchronization index reflects the spatial-domain synchronization efficiency of the left and right hemispheres. Their complementary integration overcomes the limitations of single features, thus comprehensively characterizing the focus state. 3. This invention uses multi-source feature fusion of total eye movement energy, dual-channel correlation, and signal complexity entropy to calculate visual activity indices. This fusion significantly improves the robustness of recognizing complex eye movement patterns and successfully maps physiological eye movement features to cognitive-level visual activity. Attached Figure Description

[0013] Figure 1 A flowchart illustrating a two-dimensional cognitive state diagnosis and intervention method based on EEG-eye movement provided in this embodiment of the invention;

[0014] Figure 2 This is a schematic diagram of a two-dimensional cognitive state coordinate system provided in an embodiment of the present invention;

[0015] Figure 3 This is a schematic diagram of a two-dimensional cognitive state diagnosis and intervention system based on EEG-eye movement provided in an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] like Figure 1 As shown, this invention employs a two-dimensional cognitive state diagnosis and intervention method based on electroencephalography (EEG) and eye movement (EEG-EM) tracking, comprising:

[0018] S1. Use an integrated acquisition device to acquire mixed EEG-eye movement signals, and decouple the mixed EEG-eye movement signals from their sources to obtain EEG signals and eye movement signals;

[0019] In this example, the user wears a headband-type wearable acquisition device to acquire multi-channel observation signals in real time through an integrated multi-electrode array (international 10-20 system). Since the electric field generated by the source signals (brain neural activity, eye movements, etc.) propagates on the scalp, the signal recorded by each acquisition channel is actually a linear superposition of multiple source signals at that electrode location. To separate pure brain activity from these EEG signals contaminated by eye movement artifacts, this invention employs a two-step hybrid signal source decoupling algorithm to decouple the EEG-eye movement mixed signals, specifically including:

[0020] S11. Based on the independent component analysis method, the EEG-eye movement mixed signal is separated into its source signals;

[0021] Collected Channel observation signal (i.e., mixed EEG-eye movement signals) are generated by An unknown independent source signal (Containing pure EEG components, eye-tracking components, and other noise sources) processed by an unknown linear mixing matrix It is a mixture. Its mathematical model can be expressed as: ,in yes The observed signals from each channel. This invention utilizes the ICA algorithm (Independent Component Analysis) to solve for an unmixing matrix. (make) ), and then based on the unmixing matrix From the observed signal The best estimate of the recovered source signal ;

[0022] S12. Decompose the source signal into multiple independent source components to obtain a set of source components;

[0023] Through optimization By leveraging the statistical independence of its components, the system can decompose the source signal into a set of independent source components, such as... .in, These are non-eye-movement artifacts. These are components of eye-movement artifacts.

[0024] S13. Identify eye-movement artifacts and non-eye-movement artifacts in the source component set, and combine the eye-movement artifacts to obtain the eye-movement artifact component set. The non-eye-tracking artifact components are combined to obtain the non-eye-tracking artifact component set. ;

[0025] Identifying eye-movement artifacts and non-eye-movement artifacts involves: using existing identification methods based on the spatial topological and temporal-frequency characteristics of each source component to automatically or semi-automatically identify which source components primarily represent eye-movement artifacts (such as strong blinks or large salivations). These identified eye-movement components are then categorized into an eye-movement artifact component set. The remaining source components are non-eye-movement artifact components, which are categorized into the non-eye-movement artifact component set. .

[0026] S14. Assemble the eye-tracking artifact components. As the independent variable, the non-eye-tracking artifact component set The various non-eye movement artifacts in Each non-eye-tracking artifact component was established as the dependent variable. The regression model; solving for each non-eye-tracking artifact component. The regression model yields the various non-eye-tracking artifact components. The residuals of the regression model, i.e., the components of each non-eye-tracking artifact. Corresponding EEG components;

[0027] After identifying the independent components representing eye-movement artifacts, this invention employs a fine-tuning method combining component removal and signal regression to eliminate eye-movement effects from the raw EEG signal. Regression model: ,in, Non-eye movement artifact component set The i-th non-eye-movement artifact component in; For the first The regression coefficients of the non-eye-tracking artifact components are calculated using the least squares method, and the closed-form solution of the regression coefficients can be expressed as: This process calculates the cross-covariance between the eye movement signal and the current EEG signal to obtain the optimal weights; It is the residual after regression, after obtaining the regression coefficients. Then, substitute it into the original equation, subtract the linear component of the eye-tracking artifact directly from the original signal, and what remains is the residual: The residual This is considered to be a purer EEG component after removing the part that is linearly correlated with eye movement artifacts.

[0028] S15. Reconstruct EEG signals based on the residuals of the regression models of each non-eye movement artifact component; reconstruct eye movement signals based on the set of eye movement artifact components.

[0029] Reconstructing EEG signals involves: building the source signal matrix of the residuals of a regression model that contains only non-eye movement artifact components. That is, among the independent source components obtained by decomposition, the row vectors corresponding to the non-eye-tracking artifact components are reset to the residuals (EEG components) of the regression model of the non-eye-tracking artifact components, and the row vectors corresponding to the remaining eye-tracking artifact components are set to zero. Then, the unmixing matrix is ​​used... inverse matrix , source signal matrix The formula for calculating the space of the reverse projection back skin sensor is as follows: The result of this calculation This refers to the reconstructed, pure EEG signal.

[0030] Reconstructing the eye-tracking signal involves: building a source signal matrix that contains only eye-tracking artifacts. That is, among the independent source components obtained by decomposition, the set of eye-movement artifact components is retained. Then, the row vectors corresponding to the remaining non-eye-tracking artifact components are set to zero. Subsequently, the unmixing matrix is ​​used... inverse matrix , source signal matrix The formula for calculating the space of the reverse projection back skin sensor is as follows: The result of this calculation This is the reconstructed, pure eye movement signal.

[0031] Through the above processing, this invention first uses ICA to mathematically separate the source signals (EEG and eye movements) mixed in multiple channels; then, through identification, elimination, and regression methods, it accurately removes the influence of eye movement artifacts on the EEG components. Finally, all the corrected pure EEG components are reconstructed to obtain multi-channel pure EEG signals for subsequent cognitive attention calculation.

[0032] S2. Calculate cognitive attention index based on EEG signals;

[0033] Based on the purified EEG signals separated in step S1, this invention proposes the Spatio-Temporal Synchronization Index (STSI) to accurately quantify a user's cognitive focus. This index integrates temporal rhythmic features reflecting the brain's level of focus with spatial synchronization features reflecting the brain's coordination efficiency, thereby comprehensively characterizing the state of focus.

[0034] STSI is a weighted combination of two sub-indicators: the Time-Domain Focus Index (TFI) and the Spatial-Domain Synchronization Index (SSI), which is a cognitive focus index. ,in , The weights are preset and satisfy the following conditions: ,as well as , The time-domain focus index of EEG signals. This is the spatial domain synchronization index of EEG signals.

[0035] Due to the time domain focus index The corresponding energy rhythm characteristics are relatively stable among individuals, while the spatial domain synchronization index It is easily affected by slight fluctuations in volumetric conduction or electrode impedance, therefore, in practical applications, it is endowed with... Higher weighting. In a preferred embodiment of the invention, The range of values ​​is set to , The range of values ​​is set to In the preferred embodiment, the optimal value is =0.6, =0.4.

[0036] Temporal focus index of EEG signals The calculation process includes:

[0037] Step S201: Perform time-frequency analysis on the EEG signal to obtain the average power spectral density of each frequency band;

[0038] Time-frequency analysis of EEG signals from each channel is performed, such as by using short-time Fourier transform or wavelet transform, to calculate the average power spectral density (PSD) of each key frequency band within a specific time window.

[0039] Step S202: Extract based on average power spectral density band, bands and Average energy of the band , ;

[0040] Extraction is positively correlated with concentration. band ( Average energy And related to relaxed or absent-minded states. band ( )and band ( Average energy and .

[0041] Step S203, Calculation The average energy of the band and The ratio of the average total energy of the bands This reflects the brain's transition from a relaxed to a focused state;

[0042] Step S204, Comparison value Normalization is performed to obtain the time-domain focus index. .

[0043] Spatial synchronization index of EEG signals It doesn't just focus on energy changes in individual brain regions, but rather quantifies the degree of synchronization and coordination of neural activity in the left and right hemispheres during focused tasks. The calculation process includes:

[0044] Step S211: Calculate the EEG signal of the left brain region from the EEG signal. and right brain region EEG signals ;

[0045] Specifically, this includes: extracting EEG signals from various channels located in the left and right hemispheres of the brain from the EEG signal; and performing a spatially weighted average of the extracted EEG signals from all channels in the left hemisphere to obtain the left hemisphere EEG signal. Spatially weighted averages were performed on the extracted EEG signals from all channels located in the right brain region to obtain the EEG signal of the right brain region. At least one channel representing the left hemisphere and at least one channel representing the right hemisphere are extracted from the electroencephalogram (EEG) signal. When extracting a single channel, the EEG signal of that single channel is used as the corresponding EEG signal for the left hemisphere. Or brainwave signals from the right brain region ...

[0046] Step S212: Analyze the EEG signals of the left brain region. and right brain region EEG signals Bandpass filtering was performed separately to obtain the EEG signal of the left brain region. and right brain region EEG signals Related to higher cognitive activities Signal components of the band and ;

[0047] Step S213: Analyze the signal components. and Perform Hilbert transforms on each to obtain the instantaneous phase. and ;

[0048] Hilbert Transform The definition of is: ; where P and V represent Cauchy principal values.

[0049] Step S214: Calculate the instantaneous phase and phase difference ;

[0050] Step S215: Based on the phase difference Calculate the phase-locked value, i.e., the spatial domain synchronization index.

[0051] SSI is defined as a measure of phase difference consistency within a given time window, also known as the Phase Locking Value (PLV). The formula for calculating PLV is: Where N is the number of sampling points within the time window, and when SSI is close to 1, it indicates that the two channels are... High synchronization of neural activity across wavebands indicates high efficiency in the coordinated work of the left and right hemispheres, corresponding to a high level of focus. When SSI approaches 0, it indicates that the phase relationship between the two channels is random and the coordination between the left and right hemispheres is poor.

[0052] Existing cognitive assessments often rely solely on single temporal energy features, which can easily lead to misjudgments. This invention proposes a complementary fusion of temporal fission synchronization index (STSI) reflecting brain rhythms and spatial fusion synchronization index (SSI) reflecting the efficiency of left and right hemisphere collaboration. This combination overcomes the limitations of single features and is an indispensable foundation for constructing subsequent "two-dimensional cognitive state diagnosis."

[0053] S3. Calculate visual activity index based on eye movement signals;

[0054] Based on the purified eye movement signal separated in step S1, this invention proposes the Dual-channel Ocular Dynamic Index (DODI) to quantify visual activity. This index assesses the overall level of visual activity by analyzing the energy, synchronicity, and complexity of the dual-channel eye movement signal.

[0055] Visual activity index :

[0056]

[0057] in, The weights are preset and satisfy the following conditions: , , ,as well as .

[0058] Weight The specific values ​​were determined through visual task calibration experiments combined with optimization algorithms. The specific acquisition process was as follows: During the calibration phase, eye movement signals were recorded when users performed high visual activity tasks (such as rapid reading and target searching) and low visual activity tasks (such as resting with eyes closed and fixing the fixation point); the corresponding total eye movement energy, dual-channel correlation, and signal complexity entropy features were extracted; subsequently, machine learning algorithms (such as support vector machines or random forests) were used to evaluate the contribution of each feature to visual state classification, or a grid search method was used to determine the optimal ratio of the three weights with the goal of maximizing classification accuracy. Because the total eye movement energy... It most directly reflects the macroscopic activity intensity of the visual system, therefore it is given a relatively high weight; dual-channel correlation Used to distinguish eye-tracking patterns, assigned a moderate weight; signal complexity entropy. Used to identify eye movement patterns as an aid in fine-tuning weights. In a preferred embodiment of the invention, The range of values ​​is , The range of values ​​is , The range of values ​​is In the best instance, the optimal value is =0.45, =0.35, =0.2.

[0059] Total Ocular Energy (TOE) reflects the total intensity of eye muscle activity, including blinking, saccadic eye movements, and other eye movement behaviors. (TOE is the total ocular energy of an eye movement signal.) The calculation process includes:

[0060] Extract the signal from each channel c located around the left eye in the eye movement signal. (Channels Fp1, AF3, F7, etc.) and signals from each channel c located around the right eye. (Channels such as Fp2, AF4, and F8);

[0061] Signals from all channels c located around the left eye Spatially weighted averaging was performed to obtain the comprehensive signal from the left eye movement signal, and the signals from all channels c located around the right eye were analyzed. Spatially weighted averages were performed to obtain the comprehensive right eye movement signal;

[0062] Calculate the root mean square of the combined left and right eye movement signals respectively. The root mean square values ​​of the combined left and right eye movement signals are added together, and the sum is normalized to obtain the total eye movement energy. To comprehensively assess the overall eye movement amplitude, a high TOE value means a large eye movement amplitude or high frequency, which directly corresponds to high visual activity.

[0063] In a preferred embodiment, only the Fp1 and Fp2 channels of the international 10-20 standard system are extracted from the eye-tracking signal. and signal As a combination of left-side eye movement signals and right-side eye movement signals.

[0064] Bichannel correlation (BCC) cleverly utilizes the differences in phase relationships between the left and right periocular channels for different eye movement types to indirectly distinguish eye movement patterns. (BCC is related to the bichannel correlation of eye movement signals.) The calculation process includes: calculating the combined left eye movement signal within a time window. and integrated right eye movement signals Pearson correlation coefficient between , It is covariance. It integrates the signals from the left eye movement. and integrated right eye movement signals The standard deviation; calculate the reciprocal of the Pearson correlation coefficient or... The reciprocal of the Pearson correlation coefficient or Normalization is performed to obtain the two-channel correlation. .

[0065] The range of BCC is When BCC approaches When the signal is in phase, it indicates that the two channel signals are highly positively correlated (in phase), which is mainly caused by blinking and vertical eye movements; when the BCC approaches... When the signal is close to a certain value, it indicates that the two channel signals are highly negatively correlated (out of phase), which is mainly caused by horizontal eye movements; when the BCC approaches a certain value... When the value is zero, it indicates that the two channel signals have no significant correlation or the modes are mixed. Therefore, the absolute value of BCC is used instead of the direct value of BCC. The reciprocal or As a characteristic of activity level, because The closer Both represent a relatively regular eye movement pattern, and The closer This indicates a mixture of horizontal, vertical, and blinking eye movements, resulting in a more chaotic pattern.

[0066] Signal complexity entropy (SCE) is used to quantify the temporal complexity of eye movement signals to distinguish between smooth fixation and rapidly changing saccades or blinks. The calculation process includes:

[0067] Comprehensive left eye movement signals and integrated right eye movement signals The average signal is then taken to obtain the comprehensive eye movement signal. Calculate integrated eye movement signals The sample entropy is obtained by normalizing the sample entropy to obtain the signal complexity entropy. .

[0068] Sample entropy is an existing nonlinear dynamic parameter that measures the predictability of a time series. The higher the pattern repeatability in the sequence, the lower the entropy value. A low SampEn value corresponds to a smooth, slow-changing signal, such as prolonged fixation; a high SampEn value corresponds to a signal full of rapid, irregular jumps, such as continuous saccades and blinks. Therefore, SCE directly takes the value of sample entropy. ,in, It is the embedded dimension. It is a similarity tolerance.

[0069] Embedded Dimension and similarity tolerance The value is pre-set based on the dynamic characteristics of the eye-tracking signal. Among them, the embedding dimension... The preferred value range is 1 or 2. In the preferred embodiment of the present invention, the optimal value is... This method aims to effectively capture short-term sequence patterns of eye movement signals while ensuring computational efficiency. The similarity tolerance *r* is used to set the distance threshold for pattern matching, in order to overcome the absolute differences in eye movement amplitude among different individuals. The value of is related to the standard deviation (SD) of the integrated eye-tracking signal. The preferred value range is... In the preferred embodiment of the present invention, the optimal value is... By using the above parameter settings, high-frequency measurement noise can be filtered out to the maximum extent, while irregular jump behavior in eye-tracking sequences can be accurately quantified.

[0070] The combination of three metrics—total eye-tracking energy, dual-channel correlation, and signal complexity entropy—produces a non-linear synergistic effect. A single [metric]... While it can reflect the intensity of eye movements, it cannot distinguish whether this is due to normal reading saccades or random, distracted looking; and combining... It can further analyze the direction pattern of eye movements, combined with This effectively distinguishes between smooth visual focus and chaotic, wandering gaze. The present invention, by endowing... By employing appropriate weighting, multi-source features are fused across three dimensions of eye movement: intensity, pattern, and regularity. This combination not only significantly improves the robustness of recognizing complex eye movement patterns but also successfully maps physiological eye movement features to visual activity at the cognitive level, laying a data foundation for accurate subsequent assessment of complex cognitive states in the four quadrants.

[0071] S4. Obtain intervention strategies based on cognitive attention index and visual activity index.

[0072] like Figure 2 As shown, the intervention strategies obtained based on the cognitive attention index and visual activity index include: constructing a two-dimensional cognitive state coordinate system with the visual activity index as the horizontal axis, the cognitive attention index as the vertical axis, and (0.5, 0.5) as the origin; each of the four quadrants of the cognitive state coordinate system has a corresponding intervention strategy; mapping the cognitive attention index and visual activity index to the cognitive state coordinate system yields the corresponding quadrants and their intervention strategies.

[0073] The first quadrant represents the active exploratory cognitive state: high cognitive focus and high visual activity. This corresponds to a highly active brain while the eyes are actively searching or reading, representing an ideal state for efficient learning and information acquisition.

[0074] The second quadrant represents a state of deep cognitive focus: high cognitive focus and low visual activity. This corresponds to a highly active brain, but with the eyes steadily fixed on a single point or a very small area, a sign of deep thinking and problem-solving.

[0075] The third quadrant represents a state of mental fatigue and cognitive exhaustion: low cognitive focus and low visual activity. This corresponds to low brain activity and a significant reduction in eye movement, typically manifesting as a blank stare, slow reaction time, and drowsiness.

[0076] The fourth quadrant represents a state of agitated and distracted cognition: low cognitive focus and high visual activity. This corresponds to a user's inability to concentrate on the current task, while their eyes are aimlessly and chaotically scanning, manifesting as restlessness and wandering.

[0077] The system automatically triggers highly targeted intervention strategies based on the quadrant in which the user's state point falls:

[0078] If the user's state remains consistently stable in the first or second quadrant, a protective silence strategy can be implemented. For example, incoming calls or application message notifications can be automatically muted or delayed to protect the user's valuable productive work state from being interrupted by external information.

[0079] If the user's state falls into the third quadrant, a physiological arousal strategy is executed. For example, a strong vibration is emitted through a connected haptic device (such as a smart bracelet), or a short, high-frequency warning sound is played through a speaker to increase the user's physiological alertness level, and a visual prompt "suggest taking a short rest" can be superimposed.

[0080] If the user's state falls into the fourth quadrant, a task restructuring or motivational incentive strategy is implemented. For example, in cognitive training software, the system can pause the current tedious task and automatically switch to a short, fun interactive game to recapture the user's attention; or in a work scenario, a reward window can pop up to reignite the user's intrinsic motivation.

[0081] During the execution of the method of the present invention, the system continuously executes steps S1 to S4 in a loop to form a closed loop, continuously diagnosing and adaptively intervening in the user's cognitive state until the monitoring task ends.

[0082] like Figure 3 As shown, this embodiment of the invention also provides a system for implementing the above method. This system can exist as a standalone device or be integrated as a software module into electronic devices such as computers, tablets, and smart glasses. The system includes:

[0083] The signal acquisition and decoupling module is used to acquire mixed EEG-eye movement signals, decouple the mixed EEG-eye movement signals from their sources, and obtain EEG signals and eye movement signals.

[0084] A two-dimensional feature extraction module is used to calculate cognitive attention index based on EEG signals and visual activity index based on eye movement signals.

[0085] The state diagnosis and targeted intervention module obtains intervention strategies based on cognitive attention index and visual activity index.

[0086] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for diagnosing and intervening in two-dimensional cognitive states using EEG-eye tracking, characterized in that, include: S1. Acquire mixed EEG-eye movement signals, decouple the mixed EEG-eye movement signals from their sources, and obtain EEG signals and eye movement signals; S2. Calculate cognitive attention index based on EEG signals; S3. Calculate visual activity index based on eye movement signals; S4. Obtain intervention strategies based on cognitive attention index and visual activity index.

2. The EEG-eye-tracking two-dimensional cognitive state diagnosis and intervention method according to claim 1, characterized in that, Signal source decoupling for mixed EEG-eye movement signals includes: S11. Based on the independent component analysis method, the EEG-eye movement mixed signal is separated into its source signals; S12. Decompose the source signal into multiple independent source components to obtain a set of source components; S13. Identify eye-movement artifacts and non-eye-movement artifacts in the source component set, combine the eye-movement artifacts to obtain the eye-movement artifact component set, and combine the non-eye-movement artifacts to obtain the non-eye-movement artifact component set. S14. Using the set of eye-tracking artifact components as the independent variable and each non-eye-tracking artifact component in the set of non-eye-tracking artifact components as the dependent variable, establish a regression model for each non-eye-tracking artifact component; solve the regression model for each non-eye-tracking artifact component to obtain the residuals of the regression model for each non-eye-tracking artifact component. S15. Reconstruct EEG signals based on the residuals of the regression models of each non-eye movement artifact component; reconstruct eye movement signals based on the set of eye movement artifact components.

3. The EEG-eye-tracking two-dimensional cognitive state diagnosis and intervention method according to claim 1, characterized in that, Cognitive Attention Index : in , For the preset weights, The time-domain focus index of EEG signals. This is the spatial domain synchronization index of EEG signals.

4. The EEG-eye-tracking two-dimensional cognitive state diagnosis and intervention method according to claim 3, characterized in that, The calculation process of the temporal domain focus index of EEG signals includes: Time-frequency analysis of EEG signals was performed to obtain the average power spectral density of each frequency band; Extracted from average power spectral density band, bands and Average energy of the band , ; calculate The average energy of the band and The ratio of the average total energy of the bands ; Comparison value Normalization is performed to obtain the time-domain focus index. .

5. The EEG-eye-tracking two-dimensional cognitive state diagnosis and intervention method according to claim 3, characterized in that, The calculation process of the spatial domain synchronization index of EEG signals includes: Extracting EEG signals from various channels located in the left and right hemispheres of the brain from the EEG signals; The EEG signals of the left brain region are obtained by averaging the EEG signals of all channels located in the left brain region. The average of the EEG signals from all channels located in the right brain region is used to obtain the EEG signal of the right brain region. ; EEG signals from the left brain region and right brain region EEG signals Bandpass filtering was performed separately to obtain the EEG signal of the left brain region. and right brain region EEG signals of Signal components of the band and ; For signal components and Perform Hilbert transforms on each to obtain the instantaneous phase. and ; Calculate instantaneous phase and phase difference ; Based on phase difference Calculate the phase-locked value, i.e., the spatial domain synchronization index.

6. The EEG-eye-tracking two-dimensional cognitive state diagnosis and intervention method according to claim 1, characterized in that, Visual activity index : in, For the preset weights, This represents the total eye movement energy of the eye movement signal. The correlation between the two channels of eye movement signals. The signal complexity entropy of the eye-tracking signal.

7. The EEG-eye-tracking two-dimensional cognitive state diagnosis and intervention method according to claim 6, characterized in that, Total eye movement energy of eye movement signals The calculation process includes: Extract the signal from each channel c located around the left eye in the eye movement signal. and the signal of each channel c located around the right eye ; Signals from all channels c located around the left eye Averaging yields the comprehensive left-side eye movement signal; signals from all channels c located around the right eye are then analyzed. The average signal was then obtained to synthesize the right eye movement signal. Calculate the root mean square (RMS) of the combined left and right eye movement signals separately. Add the RMS of the combined left and right eye movement signals together, and normalize the sum to obtain the total eye movement energy. .

8. The EEG-eye-tracking two-dimensional cognitive state diagnosis and intervention method according to claim 7, characterized in that, Dual-channel correlation of eye movement signals The calculation process includes: The Pearson correlation coefficient between the combined left and right eye movement signals was calculated. The reciprocal of the Pearson correlation coefficient was then calculated and normalized to obtain the two-channel correlation. .

9. The EEG-eye-tracking two-dimensional cognitive state diagnosis and intervention method according to claim 7, characterized in that, Signal complexity entropy of eye movement signals The calculation process includes: The composite eye movement signal is obtained by averaging the combined left and right eye movement signals. Calculate integrated eye movement signals The sample entropy is obtained by normalizing the sample entropy to obtain the signal complexity entropy. .

10. The EEG-eye-tracking two-dimensional cognitive state diagnosis and intervention method according to claim 1, characterized in that, Intervention strategies based on cognitive attention and visual activity indices include: constructing a two-dimensional cognitive state coordinate system with visual activity index as the horizontal axis, cognitive attention index as the vertical axis, and (0.5, 0.5) as the origin; each of the four quadrants of the cognitive state coordinate system has a corresponding intervention strategy; mapping the cognitive attention index and visual activity index to the cognitive state coordinate system yields the corresponding quadrants and their intervention strategies.

Images