Dynamic brain network clustering method and system based on nonlinear electroencephalogram signal feature fusion

Abstract

The invention provides a dynamic brain network clustering method and system based on nonlinear electroencephalogram signal feature fusion, and belongs to the crossing field of neuroscience and information technology. The method specifically comprises the following steps: data acquisition and processing: acquiring electroencephalogram signals of a subject by using a plurality of electrodes; constructing a dynamic cognitive network: dividing the multi-dimensional electroencephalogram signals of each frequency band by adopting a sliding window, extracting two groups of electroencephalogram signals of different channels of the same frequency band in a time window at the same moment to obtain a phase lock value, constructing a dynamic function brain network and obtaining a multi-frequency-band multi-channel dynamic function connection matrix; dynamic network clustering: taking each connection matrix as an independent cluster, and combining two clusters with the shortest distance to form a new cluster to realize dynamic network clustering to form an incomplete binary tree; and aiming at the incomplete binary tree, accessing from top to bottom from the topmost cluster to merge and reject the clusters, so that the clusters are in a set range interval, and finally, dynamic brain network clustering is realized.

Description

Technical Field

[0001] The present invention proposes a dynamic brain network clustering method and system based on nonlinear electroencephalogram signal feature fusion, which is suitable for constructing brain network state classification in the process of continuous tasks and belongs to the intersection of neuroscience and information technology. Background Art

[0002] EEG signals are the key to brain network analysis. By recording the regional interactions of brain nerve electricity under different states, the working mechanism of the brain can be effectively revealed. Extracting accurate and reliable EEG signals is the basis for building brain networks. Given the high complexity and nonlinearity of EEG signals, traditional Fourier transform-based decomposition methods are difficult to capture the dynamics of the nervous system, such as phase synchronization and chaos. Based on this, the interaction between multiple brain regions is also nonlinear. Empirical mode decomposition is often used for nonlinear signal extraction, but it will cause modal aliasing effects.

[0003] The brain's functional networks are divided into resting-state and task-state networks. Current brain network research focuses on static research in the resting state or on specific tasks. Considering that high-order cognitive activities are mostly derived from the highly dynamic and complex nature of brain networks, some scholars have used sliding window correlation to reveal the connectivity of disease resting-state networks, while others have used recursive dynamic connections to study schizophrenia subnetworks. The sliding window method is the mainstream of current dynamic brain network research, and faces the dilemma of selecting window length and window resolution. In order to accurately quantify the dynamic characteristics of neural rhythms and smooth out the effects of noise, the present invention proposes to set the window length to 5 to 10 times the corresponding neural rhythm period, and the window overlap rate is set to 30% to avoid the truncation effect caused by a smaller overlap rate and the state aliasing caused by a larger overlap rate.

[0004] Brain network clustering analysis is an important method for analyzing brain functional connectivity patterns. During the cognitive process, the brain will have different connectivity patterns, which can show high stability and consistency. In the study of dynamic functional connectivity of diseases, researchers used K-means clustering to identify subtypes of Parkinson's disease patients. Other researchers clustered dynamic brain networks during the onset of psychosis to distinguish the specific connectivity patterns of patients with major depression and schizophrenia at different stages of onset. However, traditional clustering methods have the following dilemmas: the k-means algorithm using Manchester distance or Euclidean distance as distance measurement lacks practical physiological support, has large errors, and makes the results uninterpretable. Moreover, as the data dimension increases, the "dimensionality disaster" problem will cause the algorithm efficiency to be extremely reduced. On the other hand, spectral clustering methods and density-based spatial clustering of applications with noise (DBSCAN) need to define matrix similarity and will also face the "dimensionality disaster". Although the current dimensionality reduction algorithm can greatly reduce the data dimension, it is easy to cause different degrees of functional connection loss and damage the data quality. Moreover, in the research on setting thresholds to focus on strong connections, there is still no unified conclusion on how to control the thresholds in the clustering process and eliminate false connections. Summary of the invention

[0005] In view of this, the present invention provides a dynamic brain network clustering method and system based on nonlinear EEG signal feature fusion, which adopts masked empirical mode decomposition technology to accurately extract EEG signal components of different frequency bands by adding mask signals.

[0006] The technical solution for implementing the present invention is as follows:

[0007] In the first aspect, the present invention provides a dynamic brain network clustering method based on nonlinear EEG signal feature fusion, the specific process is:

[0008] Data acquisition and processing: Use multiple electrodes to collect the EEG signals of the subjects; pre-process the collected EEG signals; use the masked empirical mode decomposition method to extract the EEG signal frequency band for the EEG signals collected by each electrode;

[0009] Dynamic cognitive network construction: Use sliding windows to divide the multidimensional EEG signals of each frequency band, extract two groups of EEG signals of different channels in the same frequency band within the same time window, calculate the instantaneous phase time series of EEG signals between two channels and obtain the phase locking value; use the channels for collecting data as network nodes and the phase locking values ​​as the connecting edges between nodes to form a dynamic functional brain network and obtain a multi-band and multi-channel dynamic functional connection matrix.

[0010] Dynamic network clustering: Each connection matrix is ​​regarded as an independent cluster. Dynamic network clustering is achieved by merging the two clusters with the shortest distance into a new cluster to form an incomplete binary tree. For the incomplete binary tree, starting from the top-level cluster, access from top to bottom, merge and eliminate clusters so that the clusters are within the set range, and finally dynamic brain network clustering is achieved.

[0011] Furthermore, when the present invention uses multiple electrodes to collect EEG signals of a subject, the subject is in a specific emotional state or is in task learning.

[0012] Furthermore, the preprocessing of the present invention includes: eliminating the error caused by the measurement of the original reference electrode pair; performing filtering processing using a bandpass filter; and separating the interference of the eye movement signal.

[0013] Furthermore, the present invention adopts the masked empirical mode decomposition method to extract the EEG signal frequency band of interest, and the specific process is as follows:

[0014] 1. Select the mask signal m(t,θ) according to the frequency characteristics of the collected EEG signal;

[0015] 2. Superimpose the mask signal with the original EEG signal x(t) to form a new signal x m (t) = x(t) + a·m(t,θ), where a is the amplitude coefficient of the mask signal;

[0016] 3. For the new signal x superimposed with the mask signal m (t) Apply empirical mode decomposition to obtain a series of intrinsic mode functions IMF. The specific process is as follows:

[0017] (1) Determine the new signal x m (t) All local maxima and minima, local interpolation between extreme points, connecting all points to construct the upper envelope e up (t) and the lower envelope e down (t);

[0018] (2) Calculate the average value of the envelope n(t) = (e up (t)+e down (t)) / 2, calculate the deviation of the new signal from the average value d(t) = x m (t)-n(t);

[0019] (3) Determine whether d(t) satisfies the IMF condition. If so, treat d(t) as an IMF. If not, treat it as a new signal and return to step (1) until the condition is met. Each IMF should meet two basic conditions: in the entire data set, the number of extreme points and the number of zero-crossing points must be equal or differ by at most one; at any point, the local average must be zero.

[0020] 4. Repeat the above process until the signal x m (t) is completely decomposed into a series of IMFs and a residual term;

[0021] 5. The corresponding IMFs of each group after decomposition are averaged to obtain the final series of IMFs, which are the EEG signals in different frequency bands.

[0022] Furthermore, when the sliding window is used to divide the multidimensional EEG signal of each frequency band in the present invention, the sliding window is set to 5-10 times the maximum period of the selected frequency band signal, and the window overlap rate is set to 30%.

[0023] Furthermore, the phase lock value of the present invention is calculated according to the following formula:

[0024]

[0025] Among them, Δφ rel (t n ) indicates that the signal X and the signal Y are n The phase difference at the moment, N represents the total number of sample points within the signal duration.

[0026] Furthermore, the distance between the two clusters A and B of the present invention is:

[0027]

[0028] Among them, A i With B j is a subcluster of clusters A and B, n A With n B represents the number of subclusters, and Finger Cluster A i With B j The number of matrices contained in , N A and N B represents the number of matrices contained in clusters A and B, d(A i ,B j ) represents subcluster A i With B j The distance between.

[0029] Furthermore, the present invention assumes that P and Q are two m×m adjacent connection matrices, and calculates the connection strength similarity d between P and Q strength for:

[0030]

[0031] in, Refers to the number of connections shared by P and Q, p(i,j) and q(i,j) are the i-th row and j-th column elements in the corresponding P and Q matrices.

[0032] Furthermore, the present invention assumes that P and Q are two adjacent connection matrices of m×m, and calculates the connection structure similarity d between P and Q. structure for:

[0033]

[0034] Among them, e p (i,j) and e q (i,j) is the corresponding E P and E q The element in row i and column j in E P and E q Indicates a connection unique to P and Q.

[0035] In a first aspect, the present invention provides a dynamic brain network clustering system based on nonlinear EEG signal feature fusion, comprising:

[0036] The data acquisition and processing module includes multiple electrodes for collecting the EEG signals of the subjects; preprocessing the collected EEG signals; and extracting the EEG signal frequency band using the mask empirical mode decomposition method for the EEG signals collected by each electrode;

[0037] The dynamic cognitive network construction module uses a sliding window to divide the multi-dimensional EEG signals of each frequency band, extracts two groups of EEG signals of different channels in the same frequency band within the time window at the same time, calculates the instantaneous phase time series of the EEG signals between two channels and obtains the phase locking value; the channels for collecting data are used as network nodes, and the phase locking values ​​are used as the connecting edges between nodes to form a dynamic functional brain network and obtain a multi-band and multi-channel dynamic functional connection matrix;

[0038] The dynamic network clustering module is used to treat each connection matrix as an independent cluster, and realize dynamic network clustering by merging the two clusters with the shortest distance to form a new cluster, forming an incomplete binary tree; for the incomplete binary tree, starting from the top-level cluster, accessing from top to bottom, merging and eliminating clusters, so that the clusters are within the set range, and finally realizing dynamic brain network clustering. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 This is a flow chart of the dynamic brain network clustering method based on nonlinear EEG signal feature fusion;

[0041] Figure 2 It is the flow chart of clustering algorithm;

[0042] Figure 3 This is the flow chart of matrix distance calculation;

[0043] Figure 4 This is the flow chart of clustering post-processing algorithm;

[0044] Figure 5 The distances between each point in the six clusters formed by the two clustering methods and the cluster center; (a) k-means algorithm, (b) innovative algorithm.

[0045] Figure 6 The distances between the matrices within the six clusters formed by the two clustering methods; (a) k-means algorithm, (b) innovative algorithm.

[0046] Figure 7 The network topology of the six clusters formed by the two clustering methods; (a) the clustering coefficient of each cluster under the k-means algorithm; (b) the clustering coefficient of each cluster under the innovative algorithm; (c) the global efficiency of each cluster under the k-means algorithm; (d) the global efficiency of each cluster under the innovative algorithm. DETAILED DESCRIPTION

[0047] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0048] It should be noted that the following embodiments and features in the embodiments may be combined with each other in the absence of conflict; and, based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making any creative work are within the scope of protection of the present disclosure.

[0049] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein may be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein may be implemented independently of any other aspect, and two or more of these aspects may be combined in various ways. For example, any number of aspects described herein may be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein may be used to implement this device and / or practice this method.

[0050] In order to fully construct a dynamic brain network, the present application embodiment proposes a dynamic brain network clustering method based on nonlinear EEG signal feature fusion. Steps 1 to 3 are data acquisition and preprocessing modules, steps 4 to 5 are dynamic cognitive network construction modules, and steps 6 to 8 are dynamic network clustering modules. Figure 1 The detailed process is as follows:

[0051] Step 1: Use multiple electrodes to collect the subject's EEG signal with a sampling frequency of fs.

[0052] In this embodiment, EEG signal data of subjects in a specific emotional state induced by stimulation materials such as pictures and video clips or EEG signal data of subjects accepting different learning tasks are collected, and the sampling frequency is fs.

[0053] Step 2: Preprocess the collected EEG signals to eliminate measurement errors and interference.

[0054] The preprocessing in this embodiment is specifically as follows: using the average value of the whole-brain electrodes as a reference to eliminate the error caused by the measurement of the original reference electrode pair; using a 50Hz notch filter to eliminate power frequency interference, and using a 0.5Hz to 150Hz bandpass filter to extract the neural signals of the corresponding frequency band; finally, using independent component analysis technology to compare the separated independent components with the recorded eye movement signals to separate the interference of the eye movement signals.

[0055] Step 3: For the EEG signals collected by each electrode, the masked empirical mode decomposition method is used to extract the EEG signal frequency bands of interest, which usually include the following frequency bands: delta (0.5-4Hz), theta (4-8Hz), alpha (8-12Hz), beta (12-30Hz), low-gamma (30-75Hz), high-gamma (75-150Hz). The specific steps are as follows:

[0056] 1. Select the mask signal m(t,θ) according to the frequency characteristics of the collected EEG signal.

[0057] 2. Superimpose the mask signal with the original EEG signal x(t) to form a new signal x m (t)=x(t)+a·m(t,θ), where a is the amplitude coefficient of the mask signal and θ is the phase.

[0058] 3. For the new signal x superimposed with the mask signal m (t) Apply empirical mode decomposition to obtain a series of intrinsic mode functions IMF. The specific process is as follows:

[0059] (1) Determine the new signal x m(t) All local maxima and minima, local interpolation between extreme points, connecting all points to construct the upper envelope e up (t) and the lower envelope e down (t).

[0060] (2) Calculate the average value of the envelope n(t) = (e up (t)+e down (t)) / 2, calculate the deviation of the new signal from the average value d(t) = x m (t)-n(t).

[0061] (3) Determine whether d(t) satisfies the IMF conditions. If so, treat d(t) as an IMF. If not, treat it as a new signal and return to step (1) until the conditions are met. Each IMF should meet two basic conditions: in the entire data set, the number of extreme points and the number of zero-crossing points must be equal or differ by at most one; at any point, the local average must be zero.

[0062] 4. Repeat the above process until the signal x m (t) is completely decomposed into a series of IMFs and a residual term.

[0063] 5. For all mask signals, the IMF corresponding to each group of decomposition numbers is averaged to obtain the final series of IMFs, which are EEG signals in different frequency bands.

[0064] Step 4: Use a sliding window to divide the multi-dimensional EEG signal in each frequency band.

[0065] In this embodiment, to ensure that each window can accurately describe the corresponding frequency information, the window length is determined by the frequency band of interest and is set to 5-10 times the maximum period of the signal in the frequency band. The window overlap rate is set to 30% to avoid the truncation effect caused by a smaller overlap rate and the state aliasing caused by a larger overlap rate.

[0066] Step 5: extract two groups of EEG signals of different channels in the same frequency band within the same time window, calculate the instantaneous phase time series of the EEG signals between the two channels, and calculate the phase locking value based on the EEG sequence according to the following formula:

[0067]

[0068] The phase lock value is a quantitative value used to measure phase synchronization, which is used to indicate the degree of synchronization between two different channel EEG signals within a certain period of time. rel (t n )=Δφ x (t n )-Δφ y (tn ), indicating that the signal X and the signal Y are n The phase difference at the moment, N is the total number of sample points in the signal duration. The value range of PLV is 0 to 1. The larger the value, the stronger the synchronization between the two signals. The minimum PLV value is 0, indicating that the two signals are completely out of sync. The maximum PLV value is 1, indicating that the two signals are completely phase synchronized.

[0069] The channels for collecting data are selected as network nodes, and the phase-locked values ​​are selected as the connecting edges between network nodes, thereby constructing a multi-band and multi-channel dynamic functional connection matrix to form a dynamic functional brain network; assuming the number of channels is m, the number of time windows is L, and the dynamic functional brain network with m nodes and L time windows constitutes an m×m×L dynamic functional connection matrix.

[0070] Step 6: For any set of multi-channel dynamic functional connectivity matrices, dynamic network clustering based on the “new distance” is performed. The clustering algorithm flow chart is shown in Figure 2 , the specific steps are as follows:

[0071] 1. A dynamic functional brain network with m nodes and L time windows constitutes an m×m×L dynamic functional connection matrix. These L adjacent connection matrices are regarded as L independent clusters, and the distance between the two clusters is calculated. The specific calculation method of the distance between two clusters A and B is as follows:

[0072]

[0073] Among them, A i With B j is a subcluster of clusters A and B (i.e., a non-empty subset of a cluster. For the initial L independent clusters, its subcluster is itself), n A With n B represents the number of subclusters, and Finger Cluster A i With B j The number of matrices contained in A and N B represents the number of matrices contained in clusters A and B, d(A i ,B j ) represents the sub-cluster A of clusters A and B i With B j The distance between them, when n A =n B = 1, clusters A and B each have only one subcluster, indicating that the subclusters of the two clusters have not been merged. Since the initial L clusters at the bottom are L adjacent connection matrices, d represents the distance between the two matrices at this time, such as Figure 3 As shown, the calculation method is as follows:

[0074] (1) Let P and Q be two m×m adjacent connection matrices, and their corresponding binary matrices are P * and Q * , E P and E q Indicates the connection between P and Q: E p =P∩(1-Q * ), E q =Q∩(1-P * ), C pq Indicates the connection where P and Q coexist: C pq =P∩Q

[0075] (2) Calculate the connection strength similarity d between P and Q structure for:

[0076]

[0077] Among them, N Cpq Refers to the number of connections shared by P and Q, where p(i,j) and q(i,j) are the elements in the i-th row and j-th column of the corresponding P and Q matrices. Connection strength similarity d strength The value range of is [0,1]. The smaller the value, the more similar the connection strength of the two matrices is.

[0078] (3) Calculate and define the connection structure similarity d between P and Q structure for:

[0079]

[0080] Among them, e p (i,j) and e q (i,j) is the corresponding E P and E q The element in the i-th row and j-th column has a connection structure similarity d structure The value range is [0,1]. The smaller the value, the more similar the connection structures of the two matrices are.

[0081] (4) The distance d(P,Q) between two matrices is calculated by the following formula:

[0082]

[0083] The value range of matrix distance d(P,Q) is [0,1]. The smaller the value, the more similar the two matrices are.

[0084] 2. Merge the two clusters with the shortest distance to form a new cluster. The new cluster is regarded as the "root" of the two merged clusters. The two clusters are also called subclusters of the new cluster, and the distance between the clusters is updated.

[0085] 3. Determine whether there is only one cluster left (i.e., the final cluster). If so, clustering ends. If not, return 1. After clustering ends, a non-complete binary tree is formed, with the final cluster at the top and the initial L independent clusters at the bottom.

[0086] Step 7: For the incomplete binary tree formed in step 6, start from the topmost cluster and visit from top to bottom to merge and remove clusters so that the clusters are within the set range. Figure 4 As shown, the specific steps are as follows:

[0087] 1. Set the range of clusters to control the structure and size of clusters, starting from the root node (final cluster) and accessing from top to bottom, where C(i,j) represents the jth cluster of the i-th layer, C(i,j) sp represents the subcluster of C(i,j), p=1,2.

[0088] 2. Sort the clusters in the i-th layer from large to small, and start visiting from the first unvisited cluster: Determine whether the size of the current cluster C(i,j) meets the requirements. If it is larger than the preset range, go to step 3. If it is smaller than the preset range, go to step 4. If it meets the range, traverse the next cluster and go to step 6.

[0089] 3. Visit the subcluster C(i,j) of cluster C(i,j) s1 , go to step 2.

[0090] 4. Calculate the two subclusters C(i,j+1) of cluster C(i,j) and the cluster C(i,j+1) at the same level s1 and C(i,j+1) s2 The distance between them is calculated, and the cluster is merged with the sub-cluster closest to it. If the size of the two clusters combined meets the requirements, the sub-cluster size is modified, and the upper layer is returned to eliminate the merged cluster. If the size of the cluster is larger than the range, go to step 5.

[0091] 5. Visit the two subclusters C(i+1,k) of the closest subcluster (let it be C(i+1,k)) s1 and C(i+1,k) s2 , calculate the cluster C(i,j) and sub-cluster C(i+1,k) that need to be merged s1 and C(i+1,k) s2 If the cluster size meets the requirement, the sub-cluster size is modified, and the upper layer is returned to eliminate the clusters that need to be merged. If the cluster size is larger than the range, step 5 is entered again.

[0092] 6. Determine whether the clusters of the current layer have been traversed. If so, proceed to step 7.

[0093] 7. Determine whether the current layer is the top layer. If so, complete the clustering post-processing. If not, return to the previous layer and go to step 2.

[0094] After executing steps 1 to 7, a clustering result of a set of dynamic functional connection matrices can be obtained. The result can intuitively show the hierarchical relationship between each category and has a strong classification accuracy.

[0095] The present invention can be used for brain network research in cognitive behavioral tasks or identification and classification of neurological diseases. In cognitive behavioral tasks, accurate emotion classification can have great application value in education, traffic safety and other aspects. For example, in education, long-term negative emotions will affect students' interpersonal communication and learning efficiency. Through emotion classification and recognition, students' negative emotional state can be discovered earlier; the anger generated by drivers during driving can easily cause road rage, which seriously affects the life safety of drivers and other traffic participants. Through emotion classification and recognition, this phenomenon can be discovered in time and accidents can be prevented. In the identification and classification of neurological diseases, this method can provide a new and effective method for cognitive assessment of patients with depression and Alzheimer's disease, provide potential for early diagnosis and early treatment of neurological diseases, and provide a new perspective for the development of early diagnostic tools for neurological diseases.

[0096] Examples:

[0097] Taking the self-built dataset for the study of the mechanism of emotion's influence on cognition as an example, in the experiment, 5 subjects carried out a reading task after receiving positive emotional stimulation, and clustered 485 time windows in the alpha frequency band, and compared with the k-means method based on Manchester distance commonly used in the same field. Figure 5-7 As shown, the following findings were found:

[0098] (1) The present invention can comprehensively consider the brain's physiological structure connections and connection strength, and form clusters with a clear hierarchical structure and controllable size range through clustering.

[0099] (2) From the perspective of the distance from each point in the cluster to the cluster center, the method selected by the present invention can effectively keep the structural distance close and optimize the cluster connection strength distance better.

[0100] (3) From the perspective of the distances between each pair of points in the cluster, the clusters formed by the method selected by the present invention are more compact.

[0101] (4) From the perspective of brain network topology research and analysis, the clusters formed by the present invention are more compact in terms of global efficiency and clustering coefficient, indicating that the functional structure within the cluster is more consistent.

[0102] The above results indicate that the algorithm and system provided by the present invention can be effectively applied to the classification of dynamic functional networks in cognitive behavior or neurology, and provide a reliable tool for revealing human cognitive activities and brain network research.

[0103] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A dynamic brain network clustering method based on nonlinear EEG signal feature fusion, characterized in that: The specific process is: Data acquisition and processing: Use multiple electrodes to collect the EEG signals of the subjects; pre-process the collected EEG signals; use the masked empirical mode decomposition method to extract the EEG signal frequency band for the EEG signals collected by each electrode; Dynamic cognitive network construction: Use sliding windows to divide the multidimensional EEG signals of each frequency band, extract two groups of EEG signals of different channels in the same frequency band within the same time window, calculate the instantaneous phase time series of EEG signals between two channels and obtain the phase locking value; use the channels for collecting data as network nodes and the phase locking values ​​as the connecting edges between nodes to form a dynamic functional brain network and obtain a multi-band and multi-channel dynamic functional connection matrix. Dynamic network clustering: Each connection matrix is ​​regarded as an independent cluster. Dynamic network clustering is achieved by merging the two clusters with the shortest distance into a new cluster to form an incomplete binary tree. For the incomplete binary tree, starting from the top-level cluster, access from top to bottom, merge and eliminate clusters so that the clusters are within the set range, and finally dynamic brain network clustering is achieved.

2. The dynamic brain network clustering method based on nonlinear EEG signal feature fusion according to claim 1, characterized in that: When multiple electrodes are used to collect the subject's EEG signals, the subject is in a specific emotional state or is in task learning.

3. The dynamic brain network clustering method based on nonlinear EEG signal feature fusion according to claim 1 is characterized in that: The preprocessing includes: eliminating the error caused by the original reference electrode pair measurement; performing filtering processing using a bandpass filter; and separating the interference of the eye movement signal.

4. The dynamic brain network clustering method based on nonlinear EEG signal feature fusion according to claim 1, characterized in that: The masked empirical mode decomposition method is used to extract the EEG signal frequency band of interest. The specific process is as follows:

1. Select the mask signal m(t,θ) according to the frequency characteristics of the collected EEG signal; 2. Superimpose the mask signal with the original EEG signal x(t) to form a new signal x m (t) = x(t) + a·m(t,θ), where a is the amplitude coefficient of the mask signal; 3. For the new signal x superimposed with the mask signal m (t) Apply empirical mode decomposition to obtain a series of intrinsic mode functions IMF. The specific process is as follows: (1) Determine the new signal x m (t) All local maxima and minima, local interpolation between extreme points, connecting all points to construct the upper envelope e up (t) and the lower envelope e down (t); (2) Calculate the average value of the envelope n(t) = (e up (t)+e down (t)) / 2, calculate the deviation of the new signal from the average value d(t) = x m (t)-n(t); (3) Determine whether d(t) satisfies the IMF condition. If so, treat d(t) as an IMF. If not, treat it as a new signal and return to step (1) until the condition is met. Each IMF should meet two basic conditions: in the entire data set, the number of extreme points and the number of zero-crossing points must be equal or differ by at most one; at any point, the local average must be zero.

4. Repeat the above process until the signal x m (t) is completely decomposed into a series of IMFs and a residual term; 5. The corresponding IMFs of each group after decomposition are averaged to obtain the final series of IMFs, which are the EEG signals in different frequency bands.

5. The dynamic brain network clustering method based on nonlinear EEG signal feature fusion according to claim 1, characterized in that: When the sliding window is used to divide the multidimensional EEG signal of each frequency band, the sliding window is set to 5-10 times the maximum period of the selected frequency band signal, and the window overlap rate is set to 30%.

6. The dynamic brain network clustering method based on nonlinear EEG signal feature fusion according to claim 1, characterized in that: The phase lock value is calculated according to the following formula: Among them, Δφ rel (t n ) indicates that the signal X and the signal Y are n The phase difference at the moment, N represents the total number of sample points within the signal duration.

7. The dynamic brain network clustering method based on nonlinear EEG signal feature fusion according to claim 1, characterized in that: The distance between two clusters A and B is: Among them, A i With B j is a subcluster of clusters A and B, n A With n B represents the number of subclusters, and Finger Cluster A i With B j The number of matrices contained in , N A and N B represents the number of matrices contained in clusters A and B, d(A i ,B j ) represents subcluster A i With B j The distance between i ,B j )Based on A i With B j The distance between the matrices contained in is calculated, and the distance d(P,Q) between two matrices P and Q is calculated by the following formula: Among them, d strength is the connection strength similarity between matrices P and Q, d structure is the connection structure similarity between P and Q.

8. The dynamic brain network clustering method based on nonlinear EEG signal feature fusion according to claim 7, characterized in that: Assume P and Q are two m×m adjacent connection matrices, and calculate the connection strength similarity d between P and Q strength for: in, Refers to the number of connections shared by P and Q, p(i,j) and q(i,j) are the i-th row and j-th column elements in the corresponding P and Q matrices.

9. The dynamic brain network clustering method based on nonlinear EEG signal feature fusion according to claim 7, characterized in that: Assume P and Q are two m×m adjacent connection matrices, and calculate the connection structure similarity d between P and Q structure for: Among them, e p (i,j) and e q (i,j) is the corresponding E P and E q The element in row i and column j in E P and E q Indicates a connection unique to P and Q.

10. A dynamic brain network clustering system based on nonlinear EEG signal feature fusion, characterized in that: include: Data acquisition and processing module, dynamic cognitive network construction module and dynamic network clustering module The data acquisition and processing module includes a plurality of electrodes for collecting the EEG signals of the subject; Preprocessing the collected EEG signals; For the EEG signals collected by each electrode, the masked empirical mode decomposition method is used to extract the EEG signal frequency band; The dynamic cognitive network construction module uses a sliding window to divide the multi-dimensional EEG signals of each frequency band, extracts two groups of EEG signals of different channels in the same frequency band within the time window at the same time, calculates the instantaneous phase time series of the EEG signals between two channels and obtains the phase locking value; the channels for collecting data are used as network nodes, and the phase locking values ​​are used as the connecting edges between nodes to form a dynamic functional brain network and obtain a multi-band and multi-channel dynamic functional connection matrix; The dynamic network clustering module is used to treat each connection matrix as an independent cluster, and realize dynamic network clustering by merging the two clusters with the shortest distance to form a new cluster, forming an incomplete binary tree; for the incomplete binary tree, starting from the top-level cluster, accessing from top to bottom, merging and eliminating clusters, so that the clusters are within the set range, and finally realizing dynamic brain network clustering.

Images