Nerve coupling decoding method and system based on double electroencephalogram

Abstract

The invention discloses a nerve coupling decoding method based on double electroencephalogram. The method comprises the following steps: S1, importing multiple groups of EEG data synchronously recorded by double persons and carrying out data preprocessing; s2, extracting features from the preprocessed electroencephalogram data, and calculating brain function connectivity across individual electrodes; s3, carrying out phase lag index analysis in combination with the frequency domain characteristics of each frequency band, and capturing the dynamic coupling characteristics of the double-brain signals in task execution; nonlinear coupling analysis is carried out by combining a calculated value of cross-individual electrode connectivity, and a nonlinear coupling relation between brain intervals is further quantified; and S4, integrating a phase lag index analysis result with the frequency domain features of each frequency band and the nonlinear time domain features of the single brain signal to form a high-dimensional feature matrix, performing dimension reduction, taking the features after dimension reduction as input, constructing an SVR model and calculating a double brain nerve synchronization index, and realizing decoding and prediction of the nerve coupling degree of the double brain signal.

Description

Technical Field

[0001] The present invention relates to the technical fields of multi - person brain - computer interfaces and electroencephalogram hyperscanning technology, and particularly relates to a method and system for decoding neural coupling based on dual - person electroencephalogram. Background Art

[0002] In recent years, the research on the interaction and coupling mechanism between two brains has become an important direction in the fields of cognitive science and neuroscience. By studying the electroencephalogram (EEG) synchronization and dynamic cooperation mechanism between two subjects, the neural basis of multi - subject cooperation can be deeply understood. This has great application value in social behavior analysis, emotional resonance assessment, and clinical rehabilitation intervention. However, there are still technical bottlenecks in the existing technology for accurately decoding the dynamic signal patterns of two brains and capturing non - linear interaction characteristics. How to efficiently decode and model the complex dynamic patterns of two - brain signals has become a key challenge for the further development of two - brain research.

[0003] In the existing technology, toolboxes such as HyPyP and DEEP are widely used for inter - brain connectivity analysis and partial dynamic pattern analysis, but there are the following problems: (1) Limited decoding performance: The existing methods have insufficient decoding accuracy for complex two - brain tasks and high - dimensional brain signals, and it is particularly difficult to analyze non - linear, short - time, and local features in the synchronization pattern. (2) Weak dynamic coupling analysis ability: Most methods are based on static connectivity analysis (such as phase - locking value PLV), and it is difficult to capture the coupling characteristics that change with time and frequency. (3) Limited ability to model real - world scenarios: The existing technology mainly focuses on linear analysis and does not fully explore the potential of complex non - linear interactions in real - world scenarios. (4) Complicated operation and insufficient applicability: The tools mainly rely on script writing and command - line operations, which are difficult to meet the needs of non - computer - major users in neuroscience, psychology, clinical medicine, etc. to quickly get started, and limit the wide promotion and application of the technology.

[0004] Therefore, there is an urgent need to provide a new method and system for decoding neural coupling based on dual - person electroencephalogram to solve the above problems. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for decoding neural coupling based on dual - person electroencephalogram, which can improve the brain - coupling decoding accuracy and real - time processing ability under complex task conditions.

[0006] To solve the above - mentioned technical problem, one technical solution adopted by the present invention is: to provide a method for decoding neural coupling based on dual - person electroencephalogram, including the following steps: S1: Import multiple groups of EEG data synchronously recorded by two people and perform data pre - processing; S2: Extract features from the preprocessed EEG data in step S1, including the frequency-domain features of each frequency band of the single-person brain signal, the non-linear time-domain features, and calculate the brain functional connectivity across individual electrodes; S3: Conduct phase lag index analysis by combining the frequency-domain features of each frequency band to capture the dynamic coupling characteristics of the dual-brain signal during task execution; conduct non-linear coupling analysis by combining the calculated values of the connectivity across individual electrodes to further quantify the non-linear coupling relationship between brain regions; S4: Integrate the results of the phase lag index analysis in step S3 with the frequency-domain features of each frequency band and the non-linear time-domain features of the single-person brain signal extracted in step S2 to form a high-dimensional feature matrix and perform dimensionality reduction. Use the dimensionality-reduced features as input to construct an SVR model and calculate the dual-brain nerve synchronization index to achieve the decoding and prediction of the nerve coupling degree of the dual-person EEG signal.

[0007] In a preferred embodiment of the present invention, in step S1, the steps of preprocessing multiple groups of EEG data include: S101: Data import and format standardization: Check the channel consistency of the imported data, and correct the inconsistent channels; Set the sampling rate to the target frequency, and perform resampling if the sampling rates of the input data are different; S102: Band-pass filtering: Use a zero-phase FIR filter, and set the filtering frequency range to 1 - 45 Hz to remove extremely low-frequency drift, high-frequency artifacts, and power frequency interference; S103: Artifact removal: Use the FastICA algorithm to decompose the EEG into independent components; Automatically label the artifact components through ICLABEL, and remove the labeled artifact components; S104: Interpolation and reference reset: For the leads with artifacts, use the adjacent electrode signal interpolation method to repair the bad leads; Apply the method of calculating the mean for resetting, and the calculation formula is:

[0008] where is the signal of the i-th channel, is the signal of the i-th channel after averaging the surrounding adjacent electrodes, and N is the number of channels around the i-th channel.

[0009] In a preferred embodiment of the present invention, in step S2, the method for extracting the frequency-domain features of each frequency band of the single-person brain signal is: Through short-time Fourier transform, convert the time-domain signal into a frequency-domain signal, so as to extract the power spectral density PSD of different frequency bands:

[0010] where is the frequency-domain signal, T is the duration, and the frequency-domain features of the EEG signal are comprehensively characterized by extracting the power spectral density of different frequency bands.

[0011] In a preferred embodiment of the present invention, in step S2, multi-scale sample entropy (MSE) is used to quantify the complexity of the signal in the non-linear time-domain feature extraction, and the formula is as follows:

[0012] where m is the embedding dimension, r is the tolerance, and N is the number of sample points, is the frequency of the template vector pairs with a distance less than r under the embedding dimension m.

[0013] In a preferred embodiment of the present invention, in step S2, the steps of calculating the cross-individual electrode connectivity include: S211: The mutual information method is used to quantify the non-linear dependence of the electrode signals between two people, and the formula is as follows:

[0014] where, X , Y are two random variables, is the probability density of X, is the probability density of Y, is the joint probability density of X and Y; S212: The extracted non-linear time-domain features are input into the mutual information calculation, and the specific method is as follows: First, calculate the multi-scale sample entropy MSE of each electrode signal to obtain and ; Then, take and as inputs and calculate the mutual information between them:

[0015] In a preferred embodiment of the present invention, in step S3, the specific steps of combining the frequency-domain features of each frequency band for phase lag index analysis to capture the dynamic coupling characteristics of the dual-brain signals during task execution include: S301: Use the Hilbert transform to extract the instantaneous phase of the signal and calculate the phase lag index analysis PLI, and the formula is as follows:

[0016] where, represents the instantaneous phase difference between two signals; represents the time-averaging operation; is the sign function, taking the positive or negative sign of the phase difference; S302: Introduce the frequency-domain features of each frequency band extracted in step S2 into the PLI calculation, and select a specific frequency band for phase lag index analysis.

[0017] In a preferred embodiment of the present invention, in step S3, the specific steps of further quantifying the non-linear coupling relationship between brain regions by performing non-linear coupling analysis in combination with the calculated value of cross-individual electrode connectivity include: S311: Quantify the causal relationship between brain regions using Granger causality analysis and dynamic mutual regression models to reveal the dynamic interaction pattern of dual-brain signals during task execution. The model is as follows:

[0018] Where is the dependent variable at the current moment, is the historical value of the dependent variable Y lagged by i orders, is the historical value of the independent variable X lagged by j orders, is the error, , are the model coefficients, reflecting the weights of historical values on the current value, p, q is the lag order, optimized by the AIC / BIC criterion; S312: Introduce the calculated value of cross-individual electrode connectivity in step S2 into the Granger causality model.

[0019] In a preferred embodiment of the present invention, in step S4, the constructed SVR model is as follows:

[0020]

[0021] Where is the weight vector, controlling the model complexity, b is the bias term, is the non-linear function that maps the input X i to a high-dimensional space, is the true output corresponding to the i-th sample, , are the slack variables, allowing the prediction error to exceed the range, C is the regularization parameter, used to balance the model complexity ( ) and the training error ( ); is the insensitive loss parameter, allowing the error range between the predicted value and the true value.

[0022] In a preferred embodiment of the present invention, in step S4, the calculation process of the dual-brain nerve synchronization index NSI is as follows:

[0023] Where is the kernel function, and is the Lagrange multiplier and b is the model parameter; The kernel function uses the radial basis function, and the formula is as follows:

[0024] where γ is the kernel function parameter, which controls the locality of data mapping to the high-dimensional space. is the dimensionality-reduced feature vector of the i-th sample, is the dimensionality-reduced feature vector of the j-th sample.

[0025] To solve the above technical problems, another technical solution adopted by the present invention is: to provide a neural coupling decoding system based on dual electroencephalogram, including: A data acquisition and preprocessing module, which is used to import multiple groups of synchronously recorded EEG data of two persons and perform data preprocessing; An electroencephalogram feature extraction module, which is used to extract features from the electroencephalogram data preprocessed by the data acquisition and preprocessing module, including the frequency-domain features of each frequency band of the single-person brain signal, the non-linear time-domain features, and calculate the brain functional connectivity of cross-individual electrodes; A dynamic coupling analysis module, which is used to perform phase lag index analysis by combining the frequency-domain features of each frequency band extracted by the electroencephalogram feature extraction module to capture the dynamic coupling characteristics of the dual-brain signals during task execution; and perform non-linear coupling analysis by combining the calculated values of the cross-individual electrode connectivity calculated by the electroencephalogram feature extraction module to further quantify the non-linear coupling relationship between brain regions; A decoding and prediction module, which is used to integrate the phase lag index analysis results in the dynamic coupling analysis module with the frequency-domain features of each frequency band and the non-linear time-domain features of the single-person brain signal extracted by the electroencephalogram feature extraction module to form a high-dimensional feature matrix and perform dimensionality reduction, and use the dimensionality-reduced features as inputs to construct an SVR model and calculate the dual-brain nerve synchronization index to realize the decoding and prediction of the nerve coupling degree of the dual electroencephalogram signals.

[0026] The beneficial effects of the present invention are: (1) The present invention significantly improves the decoding accuracy of dual-brain signals and the non-linear modeling ability by accurately extracting frequency-domain and non-linear time-domain features and using the mutual information (MI) and Granger causality joint model to break through the limitations of traditional linear assumptions; (2) The present invention enhances the dynamic signal analysis ability by performing phase lag index analysis to extract dynamic phase coupling characteristics in real time and combining the dynamic mutual regression model to quantify the non-linear causal relationship between brain regions; (3) The present invention adopts a modular design and PCA dimensionality reduction technology. Compared with the prior art (such as the HyPyP toolbox and the DEEP framework), the computing efficiency is improved by more than 60%, making the present invention applicable to the fields of psychology (team collaboration analysis), clinical medicine (such as autism intervention), and education (such as attention collaborative assessment), and providing an efficient cross-scenario computing method, pipeline, and system for them. (4) The present invention realizes the efficient analysis and decoding of dual-brain signals, has extensive scientific research and application values. The proposed dual-brain interaction analysis framework has improvements in algorithm innovation, dynamic decoding, and non-linear modeling, etc., making up for the deficiencies of the traditional EEG signal decoding in the prior art, such as insufficient method accuracy in EEG feature extraction and complex brain coupling relationship analysis, more feature redundancy, and poor real-time performance, and providing an effective solution for social behavior research, emotional interaction analysis, and collaboration optimization, having important academic significance and practical application values. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a flowchart of the neural coupling decoding method based on dual-brain EEG of the present invention; Figure 2 is a scatter plot of the predicted values and actual values of the dual-brain neural synchronization index of the training set and the test set; Figure 3 is a comparative dynamic coupling modeling error distribution diagram of the method of the present invention and the traditional method; Figure 4 is a structural block diagram of the neural coupling decoding system based on dual-brain EEG. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] The following elaborates on the preferred embodiments of the present invention in conjunction with the drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making the protection scope of the present invention more clearly defined.

[0029] First, an explanation is given to the neural coupling degree mentioned in the present invention: The so-called neural coupling degree, that is, the EEG synchronization between multiple people, mainly refers to that when the brain performs a specific task, there is a mutual cooperation mechanism for each brain region of different individuals, and there may be a synchronous phenomenon when the regions with a higher degree of cooperation discharge. On the contrary, if the EEG synchronization between two brains is high, it can be considered that there is a high degree of cooperation between these two brains. Briefly speaking, the neural coupling degree refers to the common change of the electrical signals between different individual brain regions, and the EEG synchronization index is a quantization tool for this change.

[0030] Please refer to Figure 1 , the embodiments of the present invention include: A neural coupling decoding method based on dual-brain EEG, comprising the following steps: S1: Import multiple sets of EEG data recorded synchronously by two persons and perform data preprocessing. The specific steps are as follows: S101: Data import and format standardization. Import multiple sets of EEG data recorded synchronously by two persons (two people), supporting common EEG file formats (such as.set,.mat, csv, etc.). Check the channel consistency of the imported data (such as the number and position of channels), and correct the inconsistent channels. Set the sampling rate to the target frequency (such as 256 Hz), and perform resampling if the sampling rate of the input data is different.

[0031] S102: Band-pass filtering. Use a zero-phase FIR filter (implemented by the pop_eegfiltnew function in EEGLAB), and set the filtering frequency range to 1 - 50 Hz to remove extremely low-frequency drift and high-frequency artifacts. The filtering parameters are set as follows: Filter type: Hanning window FIR filter; Transition bandwidth: Automatically calculated (default value is 25% of the cut-off frequency); Filtering direction: Bidirectional filtering (forward + backward) to eliminate phase shift; Order: Automatically optimized according to the frequency response.

[0032] Filter design formula:

[0033] S103: Artifact removal. Use the FastICA algorithm to decompose the EEG into independent components. Automatically label artifact components such as eye movement and electromyogram through ICLABEL, and remove the labeled components.

[0034] S104: Interpolation and reference reset. For leads with artifacts, use the neighboring electrode signal interpolation method to repair the bad leads. Apply the method of taking the mean for resetting, and the calculation formula is:

[0035] where is the signal of the i-th channel, is the signal of the i-th channel after averaging the surrounding neighboring electrodes, and N is the number of channels around the i-th channel.

[0036] S2: Extract EEG features from the data preprocessed in step S1, including the frequency-domain features of each frequency band of a single signal, non-linear time-domain features, and calculate the cross-individual electrode connectivity. The specific steps are as follows: S201: Extraction of the frequency-domain features of each frequency band of a single signal: By means of the Short-Time Fourier Transform (STFT), the time-domain signal is converted into a frequency-domain signal, thereby extracting the power spectral density (PSD) of different frequency bands. STFT is a time-frequency analysis method that obtains the time-frequency distribution of the signal by sliding a window on the time axis and performing a Fourier transform on the signal within each window. The formula is as follows:

[0037] where is the frequency-domain signal and T is the duration. Through STFT, the power spectral density of the signal in different frequency bands can be obtained, thereby characterizing the frequency-domain characteristics of the electroencephalogram (EEG) signal.

[0038] The extracted frequency bands include: Delta (1 - 4 Hz): Related to deep sleep and the recovery process.

[0039] Theta (4 - 8 Hz): Related to light sleep, relaxation state, and memory processing.

[0040] Alpha (8 - 13 Hz): Related to relaxation, closed-eye state, and concentration.

[0041] Beta (13 - 30 Hz): Related to alertness, thinking, and active cognitive tasks.

[0042] By extracting the power spectral density of these frequency bands, the frequency-domain characteristics of the EEG signal can be comprehensively characterized, providing a basis for subsequent EEG signal analysis and decoding.

[0043] S202: Nonlinear time-domain feature extraction: EEG signals have complex nonlinear characteristics, and traditional linear analysis methods are difficult to comprehensively capture their dynamic changes. Therefore, in the examples of the present invention, multi-scale sample entropy (MSE) is used to quantify the complexity of EEG signals. MSE is a nonlinear analysis method for quantifying signal complexity that can evaluate the entropy value of a signal on multiple time scales. The formula is as follows:

[0044] where m is the embedding dimension, r is the tolerance, N is the number of sample points, is the frequency of template vector pairs with a distance less than r under the embedding dimension m. By calculating the sample entropy of the signal at different scales, MSE can effectively capture the nonlinear characteristics of EEG signals, reveal the nonlinear dynamic changes in the signals, and provide support for the modeling of complex EEG activities. It is of great significance for understanding the complex dynamic patterns in bilateral brain interactions. Especially when capturing nonlinear coupling characteristics, MSE can provide additional information to make up for the deficiencies of traditional linear analysis.

[0045] S203: Calculate cross - individual electrode connectivity: In the study of dual - brain interaction, the connectivity analysis of cross - individual electroencephalogram (EEG) signals is a key step. In the example of this invention, the mutual information (MI) method is used to quantify the non - linear dependence of electrode signals between two people. The specific steps are as follows: (1) Mutual information calculation: Mutual information is an index used to quantify the dependence between two random variables and can capture non - linear relationships. The formula is as follows:

[0046] where X , Y are two random variables (EEG signals of two people), is the probability density of X, is the probability density of Y, is the joint probability density of X and Y. By calculating the mutual information of electrode signals between two people, the non - linear dependence between their EEG signals can be quantitatively evaluated, revealing the cross - individual EEG connection characteristics.

[0047] (2) Incorporate non - linear time - domain features: To improve the accuracy of mutual information calculation, the non - linear time - domain features (MSE) extracted in step S202 are introduced into mutual information calculation. The specific method is as follows: First, calculate the multi - scale sample entropy (MSE) of each electrode signal to obtain and ; Then, take and as inputs and calculate the mutual information between them:

[0048] By incorporating non - linear time - domain features, mutual information calculation can more accurately quantify the non - linear dependence relationship between dual - brain signals and enhance the adaptability to complex task scenarios.

[0049] S3: Incorporate frequency - domain features of each frequency band for phase lag index analysis to capture the dynamic coupling characteristics of dual - brain signals during task execution; incorporate the calculated values of cross - individual electrode connectivity for non - linear coupling analysis to further quantify the non - linear coupling relationship between brain regions; the specific steps include: S301: Phase lag index analysis: The phase lag index is an important index for measuring the phase relationship between two signals and can reflect the dynamic coupling characteristics of dual - brain signals during task execution. In the example of this invention, the Hilbert transform is used to extract the instantaneous phase of the signal and calculate the phase lag index (PLI). The specific steps are as follows: (1)Hilbert Transform: Hilbert transform is a method used to extract the instantaneous phase of a signal. Through the Hilbert transform, the instantaneous phase of the signal can be obtained , and then the Phase Lag Index (PLI) can be calculated. The formula is as follows:

[0050] where represents the instantaneous phase difference between two signals; represents the time-averaging operation; is the sign function, taking the positive or negative sign of the phase difference; By statistically analyzing the asymmetry of the signs of the phase differences, PLI quantifies the stable phase lag relationship between signals. Its value ranges from [0, 1]. The closer PLI is to 1, the more stable the phase lag between the two signals (e.g., signal 2 always lags behind signal 1); PLI approaching 0 indicates that the phase relationship is random or there is no stable lag pattern.

[0051] (2)Combined with power spectrum features: To improve the accuracy of phase lag index analysis, the power spectrum features (PSD) extracted in step S201 are introduced into the PLI calculation. The specific method is as follows: First, the power spectral density (PSD) of the signal is extracted through STFT to obtain the power distribution of each frequency band; Then, according to the power spectrum features, a specific frequency band (such as the Alpha band) is selected for phase lag index analysis to more accurately capture the task-related dynamic coupling characteristics.

[0052] S302: Nonlinear coupling analysis: To further quantify the nonlinear coupling relationship between brain regions, the Granger causality analysis and the dynamic mutual regression model are adopted in the example of the present invention. Granger causality analysis is a statistical method used to quantify the causal relationship between two time series. Through Granger causality analysis, the causal relationship between brain regions can be quantified, and the dynamic interaction pattern of the dual-brain signals during task execution can be revealed. The model is defined as follows:

[0053] where is the dependent variable at the current moment (such as the signal of brain region B), is the historical value of the dependent variable Y (lagged by i orders), is the historical value of the independent variable X (lagged by j orders), is the error, , are the model coefficients, reflecting the weights of the historical values on the current value, p, q is the lag order, optimized by the AIC / BIC criterion.

[0054] To capture the non - linear coupling relationship, the mutual information feature Z (i.e., I(X;Y)) is introduced into the Granger causality model.

[0055] The mutual information feature Z is used as an additional input variable to extend the Granger causality model. The new model is defined as follows:

[0056] where, is the model coefficient of the mutual information feature Z, and r is the historical lag order of Z.

[0057] S4: Integrate the phase lag index analysis results in step S3 with the frequency - domain features and non - linear time - domain features of the single signals extracted in step S2 to form a high - dimensional feature matrix and perform dimensionality reduction. Use the dimensionality - reduced features as inputs to construct an SVR (Support Vector Regression) model and calculate the dual - brain neural synchronization index (NSI, Neural Synchronization Index, i.e., the predicted dual - brain coupling strength) to achieve the decoding and prediction of dual - brain EEG signals. The specific steps include: S401: Feature integration and dimensionality reduction In EEG signal analysis, the feature dimension is often high. Directly using these features will increase the computational complexity. Therefore, in this example of the present invention, principal component analysis (PCA) is used for feature dimensionality reduction. Integrate the dynamic coupling feature (PLI) calculated in S3 with the frequency - domain features and non - linear time - domain features obtained in S2 to form a high - dimensional feature matrix X. Perform dimensionality reduction on the high - dimensional feature matrix through PCA and retain the main feature components. The formula is as follows:

[0058] where W is the projection matrix, is the dimensionality - reduced feature vector, is the original high - dimensional feature matrix. Through PCA, the feature components with a cumulative contribution rate of up to 90% can be selected, thereby reducing the data dimension and improving the computational efficiency.

[0059] S402: Mapping of brain coupling index and task performance SVR is a regression model based on support vector machines that can handle non - linear relationships. Through SVR, a mapping relationship between the brain coupling index and task performance can be constructed to provide a high - precision prediction model for dual - brain interaction tasks. Use the dimensionality - reduced features as inputs to construct an SVR model. The formula is as follows:

[0060]

[0061] Among them, is the weight vector that controls the model complexity, b is the bias term, is the non-linear function that maps the input X i to a high-dimensional space, is the true output corresponding to the i-th sample, , is the slack variable that allows the prediction error to exceed the range. C is the regularization parameter used to balance the model complexity ( ) and the training error ( ). A larger C value enhances the fitting to the training data and may lead to overfitting; a smaller C value improves the generalization ability and may lead to underfitting. After optimization through Grid Search, C = 1.0 is selected.

[0062] is the insensitive loss parameter that allows the error range between the predicted value and the true value.

[0063] Finally, the double-brain nerve synchronization index NSI is calculated as follows. Its value range is a continuous scalar (e.g., 0 to 1), and a higher value indicates more significant dynamic coordination between the two brains:

[0064] Among them is the kernel function, and are the Lagrange multipliers, and b is the model parameter.

[0065] The radial basis function (RBF) is used as the kernel function, and the formula is as follows:

[0066] Among them, γ is the kernel function parameter that controls the locality of data mapping to the high-dimensional space. After optimization, γ = 0.1 is selected. is the dimensionality-reduced feature vector of the i-th sample, is the dimensionality-reduced feature vector of the j-th sample.

[0067] The present invention has developed a framework for dual-brain interaction analysis and decoding, which combines multi-level signal preprocessing, feature extraction, dynamic coupling analysis, and decoding prediction technologies, significantly improving the decoding efficiency and application effect. The specific description is as follows: (1) Precise frequency-domain and non-linear time-domain feature extraction The short-time Fourier transform (STFT) technology is used to calculate the power spectral density (PSD), enabling the precise extraction of frequency bands such as Delta, Theta, Alpha, and Beta, and comprehensively characterizing the frequency-domain features. Meanwhile, the multi-scale sample entropy (MSE) is used to analyze the signal complexity, capturing the non-linear characteristics of brain signals and providing support for modeling complex electroencephalogram activities.

[0068] (2) Cross-individual electroencephalogram connectivity analysis The present invention uses the mutual information method to quantitatively analyze the non-linear dependence of electrode signals between two individuals, revealing the cross-individual electroencephalogram connection characteristics from multiple dimensions. The mutual information (MI) algorithm is introduced to construct a coupled model of two-brain signals that adapts to complex non-linear dynamic patterns. Compared with the linear model, it significantly improves the modeling ability for non-linear electroencephalogram collaboration and the modeling ability for complex dynamic interactions.

[0069] (3) Dynamic coupling and non-linear causality analysis Based on the Hilbert transform, the phase lag index (PLI) is calculated to extract the instantaneous phase coupling characteristics between two individuals in real time. In addition, the non-linear coupling relationship between brain regions is quantified through Granger causality analysis and dynamic mutual regression models. Combining mutual information and Granger causality analysis breaks through the limitations of linear assumptions.

[0070] (4) Decoding and prediction performance optimization Feature dimensionality reduction is achieved through principal component analysis (PCA), and the feature components with a cumulative contribution rate of 90% are selected, thereby reducing the data dimension and improving the calculation efficiency. Meanwhile, the present invention uses a support vector regression (SVR) model to construct the mapping relationship between the coupling index and the task performance, optimizing the decoding performance based on kernel function technology and providing a high-precision prediction model for the two-brain interaction task.

[0071] Experiments are verified using the method described in the present invention based on the CoBra Lab two-person collaboration dataset (Müller et al., 2021). This dataset contains the electroencephalogram signals of 200 pairs of participants (400 people), covering three types of interaction scenarios: collaborative robot control, joint financial decision-making, and synchronous movement tasks. The age range of the participants is 18 - 60 years old (mean ± standard deviation: 32.5 ± 8.7 years old), and the gender distribution is balanced. The training set (180 pairs) and the test set (20 pairs) are divided using 10-fold cross-validation. After data preprocessing (ICA artifact removal, bad channel interpolation, average reference reset), 28 groups of effective data are retained. The core parameters of the experiment are set as follows: the short-time Fourier transform (STFT) window length is 256 ms, the Granger causality lag order is 5, the support vector regression (SVR) kernel function is the radial basis function (RBF), and the regularization parameter C = 1.0.

[0072] In terms of decoding accuracy, the performance of HyPyP (a traditional method based on PLV and linear regression) is 78.6%, that of DEEP (a comparative method based on dynamic mutual information) is 83.2%, while the method of the present invention (PLI combined with SVR) reaches 92.4%. Compared with HyPyP, this method has an improvement of 13.8%; compared with DEEP, the improvement is 9.2%. In the comparison of phase lag error (PLI-E), the error value of HyPyP is 0.15, and the method of the present invention is significantly optimized to 0.08, with the error reduced by 52.2%. It should be noted that the DEEP framework does not have a designed function for phase lag analysis, so no data is provided for this indicator. In terms of the root mean square error (RMSE) of dynamic modeling, the RMSE of HyPyP is 0.39, that of DEEP is 0.31, and the method of the present invention is further reduced to 0.21. Compared with HyPyP, the error is reduced by 45.8%; compared with DEEP, the error is reduced by 32.3%, as shown in Table 1.

[0073] Table 1 Comparative Quantitative Evaluation of the Performance of Neural Coupling Decoding Algorithms

[0074] In terms of computational efficiency, the present invention adopts a modular design and PCA dimensionality reduction technology, and the computational efficiency is improved by more than 60% compared with the prior art (such as the HyPyP toolbox and the DEEP framework). The specific comparison is as follows: HyPyP toolbox: Its Python-based serial processing architecture has an average processing time of 28.5 minutes per group on the same dataset (Dyad-EEG-2023, 300 groups of dual-channel EEG signals). Through modular pipeline design and parallel computing optimization, the present invention shortens the processing time to 11.4 minutes per group, with an efficiency improvement of 60%; DEEP framework: Its dynamic coupling analysis relies on a high-dimensional feature fully connected network without dimensionality reduction, resulting in a decoding task taking 15.2 seconds each time. Through PCA dimensionality reduction (retaining 90% of the variance contribution rate), the present invention reduces the feature dimension from 512 to 48, and the decoding time is reduced to 6.1 seconds, with an efficiency improvement of 59.8%.

[0075] The above optimizations make the present invention applicable to the fields of psychology (team collaboration analysis), clinical medicine (such as autism intervention), and education (such as collaborative attention assessment), providing an efficient cross-scenario solution.

[0076] The visualization results further verify the technical advantages of the present invention. Figure 2 Scatter plots of the predicted values and actual values of the dual-brain neural synchronization index NSI for the training set and the test set are shown, and the goodness of fit Reaches 0.89 and 0.82 respectively, indicating the efficient modeling ability of the model for the dual-brain dynamic coupling relationship. Figure 3This is the contrast dynamic coupling modeling error distribution diagram of the method of the present invention and traditional methods (HyPyP toolbox and DEEP framework). The errors of the method of the present invention are concentrated in the range of ±0.3 (accounting for 95%), as shown in green, while the errors of traditional methods are scattered within the range of ±0.6 (accounting for 80%), as shown in red, highlighting the advantages of Granger causality analysis and dynamic mutual regression model in capturing non-linear interaction characteristics.

[0077] Refer to Figure 4 In the example of the present invention, a neural coupling decoding system based on dual electroencephalogram is also provided, including a data acquisition and preprocessing module, an electroencephalogram feature extraction module, a dynamic coupling analysis module, and a decoding and prediction module.

[0078] The data acquisition and preprocessing module is used to import multiple groups of synchronously recorded EEG data of two persons and perform data preprocessing; The electroencephalogram feature extraction module is used to extract features from the electroencephalogram data preprocessed by the data acquisition and preprocessing module, including frequency domain features of each frequency band of single-person brain signals, non-linear time domain features, and calculate the brain functional connectivity of cross-individual electrodes; The dynamic coupling analysis module is used to perform phase lag index (PLI) analysis by combining the frequency domain features of each frequency band extracted by the electroencephalogram feature extraction module to capture the dynamic coupling characteristics of the dual brain signals during task execution; and perform non-linear coupling analysis by combining the calculated values of the cross-individual electrode connectivity calculated by the electroencephalogram feature extraction module to further quantify the non-linear coupling relationship between brain regions; The decoding and prediction module is used to integrate the phase lag index analysis results in the dynamic coupling analysis module with the frequency domain features of each frequency band of single-person brain signals (Delta, Theta, Alpha, Beta band power spectral density) and non-linear time domain features (multiscale sample entropy) extracted by the electroencephalogram feature extraction module to form a high-dimensional feature matrix and perform dimensionality reduction through principal component analysis (PCA), and use the dimensionality-reduced features as inputs to construct an SVR model and calculate the dual-brain nerve synchronization index to realize the decoding and prediction of the nerve coupling degree of dual electroencephalogram signals: Decoding: Quantify the nerve coupling degree from dual electroencephalogram signals, including the phase lag index (PLI) and non-linear dependence relationship (mutual information) between the two brains; Prediction: Based on the nerve coupling degree index, infer the execution efficiency or behavior performance of two persons in a collaborative task (such as task completion time, error rate).

[0079] The neural coupling decoding system based on dual - person electroencephalogram (EEG) can decode the neural coupling degree of dual - person synchronous EEG and visualize the inter - brain network connection. By analyzing the EEG signals generated by the cumulative synchronous activities of postsynaptic potentials in neurons in the human brain, performing time - frequency, network functional connectivity, etc. analyses on the EEG signals, calculating EEG synchrony indexes to evaluate the synchrony of EEG signals or dual - person neural activities, establishing an inter - brain functional network, researching different cognitive processes and behaviors, decoding the neural coupling degree of dual - person synchronous EEG, it has important significance for the research in the field of EEG hyperscanning.

[0080] A neural coupling decoding system based on dual - person EEG in this example can execute a neural coupling decoding method provided by the present invention, can execute any combination implementation steps of the method example, and has the corresponding functions and beneficial effects of the method.

[0081] The above are only the embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be similarly included in the patent protection scope of the present invention.

Claims

1. A neural coupling decoding method based on dual - person electroencephalogram, characterized in that, It includes the following steps: S1: Import multiple groups of EEG data recorded synchronously for two persons and perform data preprocessing; S2: Extract features from the EEG data preprocessed in step S1, including frequency-domain features of each frequency band of single-person brain signals, non-linear time-domain features, and calculate the brain functional connectivity across individual electrodes; S3: Conduct phase lag index analysis by combining the frequency-domain features of each frequency band to capture the dynamic coupling characteristics of the dual-brain signals during task execution; conduct non-linear coupling analysis by combining the calculated values of the connectivity across individual electrodes to further quantify the non-linear coupling relationship between brain regions; S4: Integrate the results of the phase lag index analysis in step S3 with the frequency-domain features of each frequency band and non-linear time-domain features of the single-person brain signals extracted in step S2 to form a high-dimensional feature matrix and perform dimensionality reduction. Use the dimensionality-reduced features as inputs to construct an SVR model and calculate the dual-brain nerve synchronization index to achieve the decoding and prediction of the nerve coupling degree of the dual-brain EEG signals.

2. The method for decoding neural coupling based on dual-EEG according to claim 1, wherein In step S1, the steps for preprocessing multiple groups of EEG data include: S101: Data import and format standardization: Check the channel consistency of the imported data, correct inconsistent channels; set the sampling rate to the target frequency, and perform resampling if the sampling rates of the input data are different; S102: Band-pass filtering: Use a zero-phase FIR filter, and set the filtering frequency range to 1 - 45 Hz to remove extremely low-frequency drift, high-frequency artifacts, and power frequency interference; S103: Artifact removal: Use the FastICA algorithm to decompose the EEG into independent components; automatically label the artifact components through ICLABEL, and remove the labeled artifact components; S104: Interpolation and reference reset: For the leads with artifacts, use the adjacent electrode signal interpolation method to repair the bad leads; apply the method of calculating the mean for resetting, and the calculation formula is: ， wherein is the signal of the i-th channel, is the signal obtained by averaging the surrounding adjacent electrodes of the i-th channel, and N is the number of channels around the i-th channel.

3. The method for decoding neural coupling based on dual-EEG according to claim 1, wherein In step S2, the method for extracting the frequency-domain features of each frequency band of the single-person brain signal is: Convert the time-domain signal into a frequency-domain signal through short-time Fourier transform, so as to extract the power spectral density PSD of different frequency bands: ， Among them is the frequency-domain signal, and T is the duration. By extracting the power spectral density of different frequency bands, the frequency-domain characteristics of the EEG signal can be comprehensively characterized.

4. The neuro-coupling decoding method based on dual-EEG according to claim 1, wherein In step S2, the extraction of non-linear time-domain features uses multi-scale sample entropy MSE to quantify the complexity of the signal, and the formula is as follows: ， where m is the embedding dimension, r is the tolerance, and N is the number of sample points, is the frequency of template vector pairs with a distance less than r under the embedding dimension m.

5. The neuro-coupling decoding method based on dual-EEG according to claim 1, wherein, In step S2, the steps for calculating the connectivity across individual electrodes include: S211: Use the mutual information method to quantify the non-linear dependence of the electrode signals between two persons, and the formula is as follows: ， Among them, X , Y are two random variables, is the probability density of X, is the probability density of Y, is the joint probability density of X and Y; S212: Enter the extracted non-linear time-domain features into the mutual information calculation, and the specific method is as follows: First, calculate the multi-scale sample entropy (MSE) of each electrode signal to obtain and ; Then, take and as inputs and calculate the mutual information between them: .

6. The neural coupling decoding method based on dual - person EEG according to claim 1, wherein, In step S3, the specific steps for conducting phase lag index analysis by combining the frequency-domain features of each frequency band to capture the dynamic coupling characteristics of the dual-brain signals during task execution include: S301: Use the Hilbert transform to extract the instantaneous phase of the signal and calculate the phase lag index analysis PLI, and the formula is as follows: , wherein, represents the instantaneous phase difference between two signals; represents time averaging operation; is the sign function, taking the positive or negative sign of the phase difference; S302: Introduce the frequency-domain features of each frequency band extracted in step S2 into the PLI calculation, and select specific frequency bands for phase lag index analysis.

7. The neural coupling decoding method based on dual - person EEG according to claim 1, wherein In step S3, the specific steps for conducting non-linear coupling analysis by combining the calculated values of the connectivity across individual electrodes to further quantify the non-linear coupling relationship between brain regions include: S311: Use Granger causality analysis and dynamic mutual regression model to quantify the causal relationship between brain regions, and reveal the dynamic interaction pattern of the dual-brain signals during task execution. The model is as follows: ， wherein, is the dependent variable at the current moment, is the historical value of the dependent variable Y lagged by i orders, is the historical value of the independent variable X lagged by j orders, is the error, , are the model coefficients, reflecting the weights of the historical values on the current value, p, q is the lag order, optimized by the AIC / BIC criterion; S312: Introduce the calculated values of cross-individual electrode connectivity in step S2 into the Granger causality model.

8. The neuro-coupling decoding method based on dual-EEG according to claim 1, characterized in that In step S4, the constructed SVR model is as follows: ， ， Among them, is the weight vector that controls the model complexity, b is the bias term, is the non-linear function that maps the input X i to a high-dimensional space, is the true output corresponding to the i-th sample, , is the slack variable that allows the prediction error to exceed the range, and C is the regularization parameter used to balance the model complexity ( ) and the training error ( ); is the insensitive loss parameter that allows the error range between the predicted value and the true value.

9. The method for decoding neural coupling based on dual - person EEG according to claim 1, wherein In step S4, the calculation process of the dual-brain neural synchronization index NSI is as follows: ， wherein is the kernel function, and are Lagrange multipliers, and b is a model parameter; The kernel function uses the radial basis function, and the formula is as follows: ， where γ is the kernel function parameter that controls the locality of data mapping to the high-dimensional space, is the dimensionality-reduced feature vector of the i-th sample, is the dimensionality-reduced feature vector of the j-th sample.

10. A neural coupling decoding system based on dual-EEG, characterized in that, including: A data acquisition and preprocessing module, which is used to import multiple groups of EEG data recorded synchronously by two people and perform data preprocessing; An EEG feature extraction module, which is used to extract features from the EEG data preprocessed by the data acquisition and preprocessing module, including the frequency-domain features of each frequency band of the single-brain signal, the non-linear time-domain features, and calculate the brain functional connectivity of cross-individual electrodes; A dynamic coupling analysis module, which is used to perform phase lag index analysis by combining the frequency-domain features of each frequency band extracted by the EEG feature extraction module to capture the dynamic coupling characteristics of the dual-brain signals during task execution; and perform non-linear coupling analysis by combining the calculated values of cross-individual electrode connectivity calculated by the EEG feature extraction module to further quantify the non-linear coupling relationship between brain regions; A decoding and prediction module, which is used to integrate the phase lag index analysis results in the dynamic coupling analysis module with the frequency-domain features of each frequency band of the single-brain signal and the non-linear time-domain features extracted by the EEG feature extraction module to form a high-dimensional feature matrix and perform dimensionality reduction. Taking the dimensionality-reduced features as input, construct an SVR model and calculate the dual-brain neural synchronization index to realize the decoding and prediction of the neural coupling degree of the dual-brain EEG signals.

Images