A MEI-TID Framework under Manifold-Euclidean Cross Space

Abstract

The present invention discloses a MEI-TID framework under a manifold-Euclidean cross space for accurately decoding the topological-implicit EEG representation of EEG signals, comprising the following steps: S1, aggregating sub-topological features by constructing an adaptive cluster center strategy on the manifold space; S2, decoding the topological-implicit representation of EEG signals in the manifold-Euclidean cross space; S3, establishing an adaptive calibration mechanism and performing decoding. The MEI-TID framework proposed by the present invention realizes high-precision decoding of such topological-implicit representations of EEG in the manifold-Euclidean cross space. The framework captures the unknown implicit topological relationships and representations of EEG signals through a dynamic cluster center strategy on the manifold space, a manifold-Euclidean cross space structure, and an adaptive calibration mechanism.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal processing, and particularly relates to a MEI-TID framework in a manifold-Euclidean cross space for accurately decoding the topological-implicit EEG representation of EEG signals. Background Art

[0002] The research on electroencephalogram (EEG) signal decoding builds a bridge for the instruction transmission between the brain and devices. Accurately and efficiently decoding EEG plays a decisive role in the field of brain-computer interface (BCI). For example, EEG signals can be used to monitor the mental fatigue states of vehicle drivers and drone operators, and can also be used to develop devices for regulating emotions so as to cure patients with mental illnesses. The MI motor imagery system based on EEG signals can also conduct rehabilitation training for stroke patients. However, the prerequisite for these tasks to be achieved is to accurately decode EEG signals. Since EEG signals are collected from multiple channels corresponding to various regions of the brain, containing both brain spatial topological information and temporal information, and having characteristics such as high time resolution, strong randomness, non-stationarity, non-linearity, and non-Gaussian processes, it has always been a difficult task and a huge challenge in this field to accurately decode EEG signals and convert complex and subtle neuron interaction information into effective discriminative information.

[0003] To address this challenge, EEG decoding technologies based on deep learning (DL) models have become the most popular focus and have achieved impressive results in aspects such as fatigue detection, emotion recognition, and motor imagery classification. Their research mainly focuses on these aspects: (1) Some research tends to extract discriminative features from the perspective of spatio-temporal-frequency multi-dimensional feature fusion and improve the decoding effect by iteratively optimizing the feature extractor; (2) Some research starts from the prior knowledge of neurophysiology and designs encoders-decoders by separately or comprehensively considering the local fine-grained features and global coarse-grained features of the brain to attempt to capture the spatial topological pattern representation of EEG signals; (3) When considering the decoding differences of individual differences, some research considers the problem that the feature differences between the source domain and the target domain are large and proposes a transfer learning scheme to narrow the differences so as to improve the generalization ability of the model and the stability of decoding. It is found that these studies are all carried out in the Euclidean space. However, the flat nature of the Euclidean space cannot truly represent the complexity of EEG signals, especially their complex brain topological representation, resulting in generally unsatisfactory decoding effects.

[0004] Recently, with the discovery that Geometric Awareness Learning (GAL) is beneficial to improving the robustness of EEG signal decoding, it has attracted much attention. Some studies have found that combining DL and GAL to design a decoder can improve the decoding performance of EEG signals. And it is worth noting that Riemannian geometry (RG), as a non-Euclidean geometry in GAL, has shown impressive decoding effects in BCI tasks. For example, Pan Y T, Chou J L, Wei C S. MAtt: A manifold attention network for EEG decoding[J]. Advances in Neural Information Processing Systems, 2022, 35: 31116-31129. proposed a manifold attention network to map EEG signals into a manifold space so as to better represent the spatio-temporal information of EEG signals, and it has been effectively verified in multiple BCI tasks. Kobler R, Hirayama J, Zhao Q, et al. SPD domain-specific batch normalization to crack interpretable unsupervised domain adaptation in EEG[J]. Advances in Neural Information Processing Systems, 2022, 35: 6219-6235 proposed a new geometric deep learning block to keep the symmetric positive definite (SPD) manifold features unchanged to alleviate the unsupervised domain adaptation problem. In addition, Yi K, Wang Y, Ren K, et al. Learning Topology-Agnostic EEG Representations with Geometry-Aware Modeling[J]. Advances in Neural Information Processing Systems, 2024, 36 introduced the geometric awareness idea to design a multi-stage pre-training strategy to communicate the spatial information between the electrode channel structures, and thus obtained unified topological mapping features and also achieved better decoding results.In addition, the method proposed by Suh Y J and Kim B H in "Riemannian embedding banks for common spatial patterns with EEG-based SPD neural networks" [C] / / Proceedings of the AAAI Conference on Artificial Intelligence. 2021, 35(1): 854-862, which learns the common spatial patterns of multiple subdomains in the Riemannian manifold space to extract EEG neurodynamics features, also achieved a certain improvement in the decoding results.

[0005] These GAL-based studies have improved the decoding effect of EEG signals to a certain extent, but they ignore some important EEG features that cannot be fully extracted in the manifold space. At the same time, neurobiology shows that there are some important implicit information in the extremely complex EEG topology, and different topologies will lead to differences in important discriminative representations in specific cognitive task states, which increases the difficulty of accurately decoding EEG under different specific cognitive tasks. Summary of the Invention

[0006] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a MEI-TID framework in the manifold-Euclidean cross space that can effectively extract the implicit topological relationship of EEG signals, improve the decoding accuracy, and achieve the efficient decoding of topological-implicit EEG representations under different BCI tasks.

[0007] The purpose of the present invention is achieved by the following technical solutions: A MEI-TID framework in the manifold-Euclidean cross space for accurately decoding the topological-implicit EEG representation of EEG signals, comprising the following steps:

[0008] S1. Aggregate sub-topological features by constructing an adaptive cluster center strategy on the manifold space;

[0009] S2. Decode the topological-implicit representation of EEG signals in the manifold-Euclidean cross space;

[0010] S3. Establish an adaptive calibration mechanism and perform decoding.

[0011] The beneficial effects of the present invention are:

[0012] (1) Based on the significant differences in individual EEGs, a dynamic cluster center strategy on the manifold space is constructed to adaptively aggregate sub-topologies, which can represent the topological information of the brain.

[0013] (2) Perform cross - operations on the sub - topological structure after the reconstruction of the joint manifold space and the sub - topological representation of the Euclidean space, and propose an adaptive calibration mechanism to unify the spatio - temporal EEG representation of the sub - topology.

[0014] (3) MEI - TID can effectively extract the implicit topological relationship of EEG signals, improve the decoding accuracy, and achieve efficient decoding of the topological - implicit EEG representation under different BCI tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a data - processing flowchart of the MEI - TID framework in the manifold - Euclidean cross - space of the present invention;

[0016] Figure 2 It is a schematic structural diagram of the MEI - TID framework of the present invention;

[0017] Figure 3 It is the clustering effect of Subject 8 - 10 in Dataset I;

[0018] Figure 4 It is the topological - agnostic representation captured by MEI - TID on three subjects in Dataset I. DETAILED DESCRIPTION OF THE INVENTION

[0019] The present invention proposes a new framework for decoding EEG signals in the manifold - Euclidean cross - space, called MEI - TID. Different from previous studies, we construct a dynamic cluster - center strategy for EEG signals on the manifold space, and adaptively aggregate into K sub - topologies according to node - centrality for different individual differences. In addition, according to previous research findings, the complex relationship between electrode channels makes it difficult to fully extract the EEG signal features in the Euclidean space. Therefore, in this paper, we map these K sub - topological structures to the Riemannian manifold space to reconstruct the high - dimensional topological structure representation. Next, we jointly construct a graph mechanism operator with the sub - topological representation in the Euclidean space, so as to realize the decoding of the topological - implicit EEG representation in the manifold - Euclidean cross - space. Finally, due to the significant differences in the sub - topologies of interest presented by the cognitive states at different time points, we perform adaptive calibration on these sub - topologies to obtain a unified spatio - temporal EEG representation. We verify the proposed MEI - TID on four EEG benchmark datasets, and the results show that the decoding performance of MEI - TID is better than the existing state - of - the - art technologies. Moreover, the analysis of MEI - TID reveals its excellent ability to explore the unknown implicit relationships of EEG and capture the representational differences under different cognitive states.

[0020] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0021] As Figure 1As shown in the figure, a MEI-TID framework under a manifold-Euclidean cross space of the present invention is used for accurately decoding the topological-implicit EEG representation of EEG signals, including the following steps:

[0022] S1. Aggregate sub-topological features by constructing an adaptive cluster center strategy on the manifold space; including the following steps:

[0023] S11. Perform band-pass filtering on the original data X = {X 1 , X 2 ,..., X N} in the EEG dataset to obtain five frequency bands: Delta, Theta, Alpha, Beta, and Gamma, and their frequency bands are 0.5 - 4 Hz, 4 - 8 Hz, 8 - 12 Hz, 12 - 30 Hz, and 30 - 50 Hz respectively;

[0024] S12. For the data in each frequency band, use differential entropy (DE) X f ∈ R N×C×B and Pearson correlation coefficient (PCC) X P ∈ R N×B×C×C to represent channel and structure information; where N, C, and B respectively represent the number of samples, the number of channels, and the number of frequency bands, and R represents the real number field;

[0025] S13. Since it is difficult to fully capture the complexity of the brain topological information represented between electrode channels in the Euclidean space, the calculated structure information X P is transformed from the Euclidean space to the SPD manifold space for clustering; it should be noted that all electrode channels are potential region centers of interest, so the affinity propagation algorithm is introduced to construct a dynamic cluster center strategy on the manifold space. Even for samples under the same label, as time changes, their structure information will show some unknown transformations, so the cluster centers shown by different samples must be different.

[0026] For X P containing N samples, map it to the SPD manifold space according to the structure information corresponding to each frequency band under a single sample. In the present invention, B is used as the feature embedding, and B is divided into B1, B2,.., B5; based on B m calculate the covariance matrix, so as to obtain a series of covariance matrices and denote them as CM B1 , CM B2 ,.., CM B5 :

[0027]

[0028] where C i and Cj represent the i-th and j-th channels respectively, and are the feature vectors of the i-th and j-th channels respectively, and n represents the number of elements in the vector. are respectively and the l-th element in; Cov(·) represents the covariance matrix, m = 1,..,5;

[0029] S14. Reconstruct the high-dimensional topological structure representation in the manifold space by tracking normalization and assigning learning factors to the diagonal elements in the matrix, denoted as

[0030]

[0031] where tr() represents the trace of the matrix, and μ and E I represent the learning factor and the identity matrix respectively;

[0032] S15. For a specific object C i and other objects C j , obtain the messages from C i to C j and the messages from C j to C i , indicating the fitness of C j to serve as the cluster center of C i and the willingness of C i to select C j as the cluster center; Denote these two degrees as and

[0033] Adopt the log-Euclidean metric as the cluster center clustering basis of C i and C j in the manifold space, and its calculation formula is:

[0034]

[0035] where ||·|| F represents the Frobenius norm, and U i , U j are the matrices obtained by performing singular value decomposition on respectively;

[0036] The log-Euclidean mean is expressed as:

[0037]

[0038] After adding the constrained weighted metric, w lThe ability to obtain the weighted log-Euclidean mean is as follows:

[0039]

[0040] Based on Iterative update And

[0041] For Iterate in the following manner:

[0042]

[0043] Where Represents the willingness of other objects except object C j To select C as the cluster center in the current state, i And

[0044] Represents the basis for clustering other objects and C j As cluster centers in the manifold space; i

[0045] Represents C i Is more likely to become the cluster center than C j ;

[0046] Iterate in the following manner

[0047]

[0048] To avoid oscillations generated during the iteration process, a decay coefficient λ is introduced, and the iterative calculation process is updated as:

[0049]

[0050] By the above process to dynamically cluster the cluster centers of different samples, the purpose of adaptively aggregating the features at different time points into K sub-topologies based on the differences of different individuals is achieved, and the corresponding sub-topology structures And the sub-topology characterization set Are shown as follows:

[0051]

[0052] S2. After completing the dynamic cluster clustering in the manifold space, it is then necessary to decode the topological-implicit representation of the EEG signal in the manifold-Euclidean cross space; the specific implementation method is: Based on X PTopological structure, develop a preliminary cognition-driven module. The design concept of this module is based on the original features of the sub-topological structures of each sample, randomly reconstruct the corresponding hidden topological structure Ω, and input them into a basic convolutional layer for spatial filtering respectively, and then preliminarily obtain the essential cognition-driven structure E p and the hidden cognition-driven structure H p ;

[0053]

[0054] Note that the purpose of this operation is to map different regions to a similar space. Then, map the two to the Riemannian manifold space to reconstruct the cognition-driven structure:

[0055]

[0056] represents manifold calculation;

[0057] Fuse the reconstructed cognition-driven structure through the following formula:

[0058]

[0059] Then, map them to a similar space through the Sigmoid function according to the following formula to abstract the stable essential cognition-driven structure and the hidden cognition-driven structure:

[0060]

[0061] θ [K] represents the linear layer parameter, σ sig () represents the Sigmoid function, and FC() represents the fully connected layer;

[0062] The cognition-driven structure is a unidirectionally connected structure. In order to reduce the influence of redundant features on the decoding result, we develop a unidirectional driving mechanism here to sparse its connectivity. The present invention develops a unidirectional driving mechanism to sparse its connectivity, and the specific calculation process is shown in the following formula:

[0063]

[0064] where and represent the weight parameters, represents the bias vector;

[0065] Then, through the graph mechanism operator, update and calculate the sub-topological structure on the manifold space, and then obtain the sub-topological representation on the manifold space, and finally realize the acquisition of the essential cognition-driven representation and the hidden cognition-driven representation across spaces:

[0066]

[0067] Among them, Graph θ () represents a graph mechanism operator, and θ is the parameter of the graph mechanism operator; θ s represents the Fourier domain parameter, and T s (.) represents the s-order Chebyshev polynomial, and W f represents the weighting parameter.

[0068] S3. Establish an adaptive calibration mechanism and perform decoding; the specific implementation method is: after obtaining the essential cognitive-driven representation and the hidden cognitive-driven representation across spaces of each sub-topology, first fuse them to obtain the unified representation of the sub-topology:

[0069]

[0070] Then, according to the following formula, randomly generate K N calibration parameters for K N sub-topology structures to obtain the unified EEG spatio-temporal representation:

[0071]

[0072] Concat() represents the concatenation operation, and W [K] represents the calibration parameter;

[0073] Next, input UR f into the decoder to obtain the result:

[0074] UR f = FC 1×class (σ sig (BN(FC 1×C (UR f ; θ C ))) ; θ class )

[0075] FC 1×class represents that the parameter setting of the fully connected layer is the final number of classes class, BN() represents the normalization function, and FC 1×C represents that the parameter setting of the fully connected layer is the number of electrode channels C.

[0076] The above steps of the present invention are implemented through three modules, as Figure 2As shown in the figure. In the figure, module (a) Dynamicclustercenterstrategy represents the dynamic clustering center strategy, module (b) Manifold euclidean hybridspace structure represents the manifold Euclidean hybrid space structure, and module (c) Adaptive calibrationmechanism represents the adaptive calibration mechanism. The three modules respectively correspond to the three steps of the present invention. In module (a), from left to right are: Input: Topologies for multipledimensions; Compression of a singlesubject means compressing into a single subject, that is, the operation of band-pass filtering in the present invention to obtain five frequency bands; Topologies on manifold space is the topology on the manifold space, indicating that the calculated structural information X P is transformed from the Euclidean space to the SPD manifold space for clustering; finally, the clustering result is obtained. In module (b), Topologies ofK N brain regions are the topologies of K N brain regions, that is, the K N sub-topological structures generated in the present invention; randomizedgeneration of latent topolpgies means randomly generating hidden topological structures; latent represents the hidden topological structure, and intrinsic represents the original topological structure; Unidirectional congnitive driven is the unidirectional cognitive drive; Reconstructed topologies→similar space→Riemannian manifold space means mapping the reconstructed topologies to a similar space and then to the Riemannian manifold space; to obtain the final cognitive connectivity structure (Final cognitive connectivity structure) (In module (c), Paramter Sharing represents parameter sharing; Joint marifold euclidean space for cross space featurerepresentations represents the joint manifold-Euclidean space for cross-space feature representation; Random generationof K N calibration parametes for K N subtopologies means generating random calibration parameters for the K NRandom generation of K sub-topologies N calibration parameters; Unified representatios of fusion subtopologies represent the unified representation of fusion subtopologies; the final input UR f into the decoder to obtain the classification result.

[0077] Generally speaking, the forward calculation process of the framework proposed by the present invention is shown in the following table.

[0078]

[0079]

[0080] The above model of the present invention can be trained through the process of back-updating parameters by LOSS. The main training parameters are the topological representation in Euclidean space and the essential driving structure and hidden driving structure in the manifold space. Denote the loss of this framework as Then for X N of K p sub-topologies and Ω, the corresponding training parameters W P and W Ω The update process is as follows:

[0081]

[0082] It can be further expanded as:

[0083]

[0084] It should be noted that and are SPD matrices, denoted as Therefore Their update can be expressed as:

[0085]

[0086] Further summarize the update rule of the SPD matrix as:

[0087]

[0088] where l represents the learning rate. Then the topological representation weights as shown below can be obtained:

[0089]

[0090] where

[0091] Next, the effect of the present invention is further verified through experiments.

[0092] In this embodiment, four EEG datasets are used to verify the decoding performance of the proposed MEI-TID, denoted as Dataset I - Dataset IV respectively. Based on these four datasets, for different cognitive tasks, the state-of-the-art decoding methods are compared with the proposed MEI-TID. Among them, the SEED (Wei-Long Zheng, and Bao-Liang Lu, Investigating Critical Frequency Bands and Channels for EEG-based Emotion Recognition with Deep Neural Networks, accepted by IEEE Transactions on Autonomous Mental Development (IEEE TAMD) 7(3):162 - 175, 2015) and SEED-IV (Wei-Long Zheng, Wei Liu, Yifei Lu, Bao-Liang Lu, and Andrzej Cichocki, EmotionMeter: A Multimodal Framework for Recognizing Human Emotions. IEEE Transactions on Cybernetics, Volume: 49, Issue: 3, March 2019, Pages: 1110 - 1122, DOI: 10.1109 / TCYB.2018.2797176) datasets are used to evaluate the decoding performance of MEI-TID for valence emotions and discrete emotions across sessions; the self-made fatigue dataset is used to evaluate the decoding performance of MEI-TID for fatigue (load) detection tasks across sessions; the Shu (Ma J, Yang B, Qiu W, et al. A large EEG dataset for studying cross-session variability in motor imagery brain-computer interface[J]. Scientific Data, 2022, 9(1):531) dataset is used to evaluate the decoding performance of MEI-TID for the same-kind cross-session motor imagery tasks.

[0093] (1) Framework training

[0094] Considering the actual situations of different cognitive task datasets, a series of experiments were conducted on a cross-session basis to compare the decoding performance of MEI-TID and other state-of-the-art methods. When training these frameworks and models, a part of the subjects' data is needed for training, and then a part of the data is reserved as the test set. Therefore, for the datasets used in this paper, each time one session is used as the test set, and the remaining sessions are used as the training set to train the decoder. Our experiments were implemented based on the Pytorch framework. In addition to Accuracy (Acc), we also used the standard deviation (STD) to measure the volatility and stability of the accuracy metric results. It should be emphasized that due to the relatively significant differences in the distribution of the subjects' data characteristics for each dataset, the convergence of MEI-TID varies across different datasets.

[0095] (2) Comparison of decoding performance: MEI-TID was compared with the current state-of-the-art methods, and the results are shown in Tables 1 and 2. The results indicate that MEI-TID demonstrates significant advantages in decoding emotional and motor imagery tasks, highlighting the superiority of this framework in terms of both accuracy and stability. We also noticed that the decoding stability of MEI-TID on the self-made fatigue dataset (STD = 11.55%) is slightly weaker than that of SFTNet (STD = 9.23%) and LAG (STD = 10.17%), but there is no difference in magnitude overall. Generally speaking, the classification accuracy of MEI-TID is on average about 3% - 4% higher than that of other methods. Moreover, it is worth noting that most studies only develop decoders for EEG signals of single cognitive tasks. In contrast, the MEI-TID proposed in this paper has the ability to achieve high-accuracy decoding for EEG signals of multiple cognitive tasks and can be applied to different types of EEG data.

[0096] Table 1 Comparison of the decoding performance of MEI-TID and other methods on Dataset I and Dataset II

[0097]

[0098] Table 2 Comparison of the decoding performance of MEI-TID and other methods on Dataset III and Dataset IV

[0099]

[0100]

[0101] (3) Ablation experiments: In this section, ablation experiments are conducted to explore the rationality of each component of the MEI-TID framework and the contribution of each part to the final decoding result. As shown in Table 3, first, a comparative study on adaptive clustering was carried out, and it was found that the decoding effect after the adaptive clustering process was significantly better, which indicates that to a certain extent, the adaptive clustering process captures the brain topological information represented by the interaction between electrode channels and increases the important feature information of the decoder. Secondly, we compared the decoding results of the conventional Euclidean space and the proposed manifold-Euclidean cross space. The experimental results show that cross-learning in the two spaces can update the Euclidean space features while capturing the complex manifold space structure, ultimately significantly improving the decoding ability of the framework.

[0102] Table 3 Ablation study of MEI-TID on four datasets.

[0103]

[0104] NAC: non-adaptive clustering; AC: adaptive clustering; ES: Euclidean space; MECS: manifold-Euclidean cross space.

[0105] (4) Interpretability experiments: Further interpret the MEI-TID framework and experimental results by combining neuroscience knowledge. Figure 3 The adaptive clustering effects of three subjects in Dataset I are presented. Among them, (a) is the clustering effect of Subject 8, (b) is the clustering effect of Subject 9, and (c) is the clustering effect of Subject 10; in the figure, the left figure is the clustering result figure, and the right figure is the corresponding channel distribution figure.

[0106] It can be found that there are similarities and differences between the clustering results in the manifold space and the prior knowledge of neuroscience. According to the clustering distribution, the following interesting patterns are found: (1) The EEG signal feature distributions of the electrodes in the prefrontal region are highly similar to those of the electrodes such as Cz that extend along the central line to the central region. Compared with the common research on the division of brain functional regions, the present invention can discover the strong connectivity and cooperation between the prefrontal region and the anterior half of the central line region from here. (2) Generally, the cluster centers of the clustering results show the characteristics of left-right hemisphere symmetry, which corresponds to the conclusion that the left and right hemispheres of the brain work together during the emotion induction process (Zhang G, Yu M, Liu Y J, et al. SparseDGCNN: Recognizing emotion from multichannel EEG signals[J]. IEEE Transactions on Affective Computing, 2021, 14(1): 537-548), and it is particularly symmetrical in the anterior half of the brain. (3) This study finds that the distributions of the clustering results of different subjects in the frontal, central, and temporal lobe regions are basically the same. On the contrary, there are significant differences in the clustering results of these three subjects in the parietal and occipital lobe regions, indicating that the perceptual and visual impacts received by the brain during the emotion induction process vary from person to person. Therefore, this may also be one of the reasons for the differences between subjects. Here, we can find that the dynamic cluster center strategy can effectively distinguish the differences while obtaining the common characteristics among subjects.

[0107] Figure 4 It presents the unknown implicit relationships and representations under emotion induction of three subjects in Dataset I captured by the MEI-TID framework. Figure (a) is for Subject 8, Figure (b) is for Subject 9, and Figure (c) is for Subject 10. From left to right in the figure are neutral, positive, and negative. It can be found that when the brain is not significantly stimulated, strong responses are presented in the right parietal lobe (TP8, P8, PO8), occipital lobe (Oz, O2), and central region (C1, Cz, FCz).

[0108] Inspired by neurobiology, the present invention proposes a MEI-TID framework in a manifold-Euclidean cross space to achieve precise decoding of topological-implicit EEG representations of EEG signals. By constructing an adaptive cluster center strategy on the manifold space to aggregate sub-topological features, and then jointly reconstructing the sub-topological structure driven unidirectionally on the manifold space and the sub-topological representation in the Euclidean space, through the development of graph mechanism operators and an adaptive calibration mechanism, the sub-topological structures are finally unified to obtain the EEG spatio-temporal representation. Experimental results on four datasets show that the decoding performance of MEI-TID is better than the current state-of-the-art decoding techniques. Ablation experiments also prove that each module plays a positive role in the overall decoding ability of the framework and can improve the decoding accuracy of the MEI-TID framework. And through interpretability experiments, it is revealed that the unknown implicit cooperation patterns between different electrode channels and the differences in cognitive state representations between individuals will affect the decoding effect to a certain extent, and it is again proved that the MEI-TID framework proposed by the present invention is reasonable. In short, this study provides new ideas for EEG decoding research and can provide decoding support for further exploring the mysteries of neurobiology. In-depth research on the in-depth principle of the framework-neural mechanism of the mutual cooperation between the manifold-Euclidean cross space to improve the decoding result is future work.

[0109] Those of ordinary skill in the art will realize that the embodiments described herein are for the purpose of assisting the reader in understanding the principles of the present invention and should be understood that the scope of protection of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific deformations and combinations without departing from the essence of the present invention based on the technical revelations disclosed in the present invention, and these deformations and combinations are still within the scope of protection of the present invention.

Claims

1. A MEI-TID framework in a manifold-Euclidean intersection space for accurate decoding of topological-implicit EEG representations of EEG signals, characterized in that: The following steps are involved: S1, aggregating sub-topological features by constructing an adaptive cluster center strategy on the manifold space; The steps include: S11, the original data X in the data set D = {X 1 ,X 2 ,...,X N }Bandpass filtering is performed to obtain five frequency bands: Delta, Theta, Alpha, Beta, and Gamma, whose frequency bands are 0.5-4Hz, 4-8Hz, 8-12Hz, 12-30Hz, and 30-50Hz respectively; S12, for each frequency band data, use differential entropy X f ∈R N×C×B and Pearson correlation coefficient X p ∈ R N×B×C×C Represents channel and structure information; where N, C, and B represent the number of samples, channels, and frequency bands, respectively; S13, X P Transform from Euclidean space to SPD manifold space for clustering: Use B as feature embedding and divide B into B1, B2, .., B5; Based on B m Compute the covariance matrix: Among them C i and C j denote the i-th and j-th channels respectively, and are the feature vectors of the i-th and j-th channels respectively, n represents the number of elements in the vector, They are and The lth element in; Cov(·) represents the covariance matrix, m＝1,..,5； S14. Reconstructing the high-dimensional topological structure representation on the manifold space is achieved by tracking normalization and assigning learning factors to the diagonal elements in the matrix, denoted as where tr() represents the trace of the matrix, μ and E I denote the learning factor and the identity matrix respectively; S15. For a specific object C that is dynamically initialized i and other objects C j , respectively obtain C i to C j Message and C j to C i The message indicates that C j Can be used as C i The fitness of cluster centers and C i Select C j The willingness to be the cluster center; these two degrees are respectively and Using logarithmic-Euclidean metric as C i and C j The cluster center clustering basis on the manifold space is calculated as follows: where ||·|| F represents the Frobenius norm, U i , U j are respectively The matrix obtained by singular value decomposition; The log-Euclidean mean is expressed as: After adding the constrained weighted measure w l The weighted log-Euclidean mean can be obtained as: based on Iterative Updates and for Iterate as follows: in Indicates that in the current state, except for object C j For other objects, select C i The willingness to be the cluster center, Indicates that except object C j In addition, other objects and C i As the basis for clustering cluster centers on manifold space; Represents C i Compared to C j More likely to become a cluster center; Iterate as follows In order to avoid oscillation during the iteration process, the attenuation coefficient λ is introduced, and the iterative calculation process is updated as follows: By dynamically clustering the cluster centers of different samples through the above process, the purpose of adaptively aggregating the features at different time points into K sub-topologies based on the differences between different individuals is achieved, and the corresponding sub-topology structure and sub-topology representation set are obtained, as shown in the following formula: S2. Topological-implicit representation of EEG signals decoded in manifold-Euclidean cross space; S3. Establish an adaptive calibration mechanism and perform decoding.

2. According to the MEI-TID framework in the manifold-Euclidean intersection space of claim 1, the specific implementation method of step S2 is: based on the original features of the sub-topological structure of each sample, the corresponding hidden topological structure Ω is randomly reconstructed, and each is input into a basic convolutional layer for spatial filtering, and then the essential cognitive driving structure E is preliminarily obtained. p and hidden cognitive driving structure H p ; Then, we map the two into the Riemann manifold space to reconstruct the cognitive driving structure: Represents manifold computation; The reconstructed cognitive driving structure is integrated through the following formula: Then, according to the following formula, they are mapped to the similarity space through the Sigmoid function to abstract the stable essential cognitive driving structure and the hidden cognitive driving structure: definition Represents the calculation process; θ [K] represents the linear layer parameters; A unidirectional driving mechanism is developed to sparse its connectivity. The specific calculation process is shown in the following formula: in and represents the weight parameter, represents the bias vector; Then, through the graph mechanism operator, the sub-topological structure on the manifold space is updated and calculated, and then the sub-topological representation on the manifold space is obtained, and finally the essential cognitive driven representation and hidden cognitive driven representation across space are obtained: where θ represents the Fourier domain parameter, T s (.) represents the s-order Chebyshev polynomial, W f represents the weighting parameter.

3. According to the MEI-TID framework in the manifold-Euclidean intersection space of claim 2, the specific implementation method of step S3 is: after obtaining the essential cognitive driven representation and the hidden cognitive driven representation of each sub-topology across the space, first merge them to obtain the sub-topology unified representation: Then K is calculated according to the following formula N The sub-topology structure is randomly generated K N calibration parameters to obtain a unified EEG spatiotemporal representation: Concat() represents the concatenation operation. [K] represents the calibration parameters; Next, UR f Feed this into the decoder to get the result: UR f =FC 1×class (s sig (BN(FC 1×C (UR f ;θ C )));θ class ) FC 1×class Indicates that the parameters of the fully connected layer are set to the final number of categories class, BN() represents the normalization function, FC 1 ×C The parameter representing the fully connected layer is set to the number of electrode channels C.

Images