Method and device for spatial filtering and feature enhancement of multi-channel electroencephalogram signals of smart glasses

Abstract

The present application discloses a method and device for spatial filtering and feature enhancement of multi-channel electroencephalogram signals of smart glasses. The method includes: taking the multi-channel electroencephalogram signals collected by the smart glasses as the original signals, generating corresponding electrode mapping relationships for analysis, determining the spatial structure and filtering scale, and generating a filter bank according to the filtering scale; separating the components of the original signals according to the filter bank, obtaining component information and corresponding signal sources, and constructing an independent component sequence according to the signal sources; performing spatial mapping on the independent component sequence, determining the main components, constructing a weight matrix for spatial filtering, generating a spatial feature sequence for spatial reconstruction, and generating enhanced features; performing multi-scale decomposition on the filter bank and the enhanced features, obtaining multi-scale features to construct a feature combination, obtaining corresponding time-frequency features for spatial mapping, obtaining interference patterns corresponding to the time-frequency features for feature enhancement, obtaining corresponding feature representations for spatial reconstruction, and obtaining spatially enhanced features.

Description

Technical Field

[0001] This application relates to the technical field of brain-computer interfaces, and particularly to a method and device for spatial filtering and feature enhancement of multi-channel electroencephalogram (EEG) signals in smart glasses. Background Art

[0002] The multi-channel EEG acquisition system in smart glasses can simultaneously acquire EEG signals at multiple positions, providing the possibility for spatial analysis of EEG signals. However, due to the limited number of acquisition electrodes in smart glasses and the limitation of the electrode distribution by the glasses structure, traditional spatial filtering methods are difficult to effectively apply. At the same time, problems such as spatial crosstalk between electrodes, correlation brought by the common reference electrode, and local electromyogram interference seriously affect the extraction of effective EEG features. Especially in real application scenarios, the movement, emotional changes, and cognitive activities of users will introduce complex interference patterns, resulting in a significant decline in signal quality.

[0003] Existing spatial filtering techniques are mainly designed for traditional EEG caps, requiring a large number of electrode covers and regular electrode distributions. When these methods are directly applied to smart glasses, problems such as over-smoothing, decreased spatial resolution, and suppression of target features often occur. In dynamic scenarios, due to the small changes in electrode positions and the influence of motion artifacts, the performance of spatial filtering will be further reduced. In addition, existing methods lack targeted consideration of the specific wearing positions and electrode distribution characteristics of smart glasses, and cannot establish effective electrode mapping relationships and multi-scale filtering strategies. At the same time, these methods often use unified parameter configurations when processing time-frequency features, making it difficult to adapt to the feature requirements of different cognitive tasks, and also unable to achieve dynamic enhancement and spatial reconstruction of target EEG features.

[0004] Therefore, there is an urgent need for a method to solve at least one of the above problems. Summary of the Invention

[0005] The embodiments of this application provide a method and device for spatial filtering and feature enhancement of multi-channel EEG signals in smart glasses, aiming to solve the problems that existing methods lack targeted consideration of the specific wearing positions and electrode distribution characteristics of smart glasses, and cannot establish effective electrode mapping relationships and multi-scale filtering strategies. At the same time, these methods often use unified parameter configurations when processing time-frequency features, making it difficult to adapt to the feature requirements of different cognitive tasks, and also unable to achieve dynamic enhancement and spatial reconstruction of target EEG features.

[0006] In a first aspect, the embodiments of this application provide a method for spatial filtering and feature enhancement of multi-channel EEG signals in smart glasses, including:

[0007] Taking the multi-channel EEG signals collected by the smart glasses as the original signals, establishing an electrode spatial position model according to the original signals, and generating an electrode mapping relationship according to the electrode spatial position model;

[0008] Perform topological analysis on the electrode mapping relationship, determine the spatial structure and the corresponding filtering scale, and generate a filter bank according to the filtering scale;

[0009] Separate the components of the original signal according to the filter bank, obtain component information, obtain the signal sources corresponding to the component information, and construct an independent component sequence according to the signal sources; the component information includes at least independent components, clustering components, and principal components;

[0010] Perform spatial mapping on the independent component sequence, determine the main components, construct a weight matrix according to the main components, perform spatial filtering according to the weight matrix to generate a spatial feature sequence, and perform spatial reconstruction on the spatial feature sequence to generate enhanced features;

[0011] Perform multi-scale decomposition on the filter bank and the enhanced features, obtain multi-scale features, construct a feature combination according to the multi-scale features, and obtain the time-frequency features corresponding to the feature combination;

[0012] Perform spatial mapping on the time-frequency features, obtain the interference patterns corresponding to the time-frequency features, perform feature enhancement according to the interference patterns, and obtain the feature representations corresponding to the interference patterns;

[0013] Perform spatial reconstruction according to the feature representations, perform feature fusion on the reconstruction results, and obtain spatially enhanced features.

[0014] In a second aspect, the present application further provides a spatial filtering and feature enhancement device, including:

[0015] A signal acquisition module, configured to use the multi-channel electroencephalogram signals collected by the smart glasses as the original signals, establish an electrode spatial position model according to the original signals, and generate an electrode mapping relationship according to the electrode spatial position model;

[0016] A topological analysis module, configured to perform topological analysis on the electrode mapping relationship, determine the spatial structure and the corresponding filtering scale, and generate a filter bank according to the filtering scale;

[0017] A component separation module, configured to separate the components of the original signal according to the filter bank, obtain component information, obtain the signal sources corresponding to the component information, and construct an independent component sequence according to the signal sources; the component information includes at least independent components, clustering components, and principal components;

[0018] A spatial mapping module, configured to perform spatial mapping on the independent component sequence, determine the main components, construct a weight matrix according to the main components, perform spatial filtering according to the weight matrix to generate a spatial feature sequence, and perform spatial reconstruction on the spatial feature sequence to generate enhanced features;

[0019] A scale decomposition module, configured to perform multi-scale decomposition on the filter bank and enhanced features to obtain multi-scale features, construct a feature combination according to the multi-scale features, and obtain time-frequency features corresponding to the feature combination;

[0020] A spatial mapping module, configured to perform spatial mapping on the time-frequency features to obtain interference patterns corresponding to the time-frequency features, perform feature enhancement according to the interference patterns, and obtain feature representations corresponding to the interference patterns;

[0021] A spatial reconstruction module, configured to perform spatial reconstruction according to the feature representations, perform feature fusion on the reconstruction results, and obtain spatially enhanced features.

[0022] In a third aspect, the present application further provides a computer device, including a processor and a memory, where the memory is used to store a computer program, and when the computer program is executed by the processor, it implements the method for spatial filtering and feature enhancement of multi-channel electroencephalogram signals of a smart glasses as described in the first aspect.

[0023] In a fourth aspect, the present application further provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, it implements the method for spatial filtering and feature enhancement of multi-channel electroencephalogram signals of a smart glasses as described in the first aspect.

[0024] Compared with the prior art, the present application at least has the following beneficial effects:

[0025] 1. This solution establishes a complete electrode mapping and multi-scale filtering system, overcoming the limitation of the limited number of electrodes in smart glasses. By establishing an electrode spatial position model, mapping relationship, performing topological analysis, and determining appropriate filtering scales, the optimal utilization of limited electrode signals is achieved.

[0026] 2. This solution constructs a complete processing link from signal decomposition to feature enhancement, effectively solving the problem of multi-source interference. By implementing spatial decomposition, component separation, and feature enhancement; and then through steps, performing interference pattern recognition and spatial template construction, the anti-interference ability of the method is significantly improved while ensuring the integrity of target features.

[0027] 3. This solution realizes the dynamic adaptability of feature extraction and enhancement, meeting the requirements of different application scenarios. By performing multi-scale decomposition and time-frequency analysis, and then through spatial reconstruction and feature fusion, precise responses to different cognitive tasks are achieved, ensuring the spatio-temporal consistency of enhanced features.

[0028] It should be understood that the above general description and subsequent detailed description are only exemplary and explanatory, and cannot limit the present application. Description of the Drawings

[0029] Figure 1 This is a schematic flowchart of the multi-channel EEG signal spatial filtering and feature enhancement method for the smart glasses shown in the embodiments of the present application;

[0030] Figure 2 This is a schematic structural diagram of the spatial filtering and feature enhancement device shown in the embodiments of the present application;

[0031] Figure 3 This is a schematic structural diagram of the computer device shown in the embodiments of the present application. Detailed implementation manners

[0032] In the following description, for the purpose of illustration rather than limitation, specific details such as specific system architectures, technologies, etc. are presented to thoroughly understand the embodiments of the present application. However, those skilled in the art should clearly understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present application.

[0033] It should be understood that when used in the specification and claims of the present application, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0034] It should also be understood that the term "and / or" as used in the specification and claims of the present application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0035] As used in the specification and claims of the present application, the term "if" can be interpreted as "when", "once", "in response to determining", or "in response to detecting" according to the context. Similarly, the phrase "if determined" or "if detecting [the described condition or event]" can be interpreted as meaning "once determined", "in response to determining", "once detecting [the described condition or event]", or "in response to detecting [the described condition or event]" according to the context.

[0036] In addition, in the description of the specification and claims of the present application, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0037] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that specific features, structures, or characteristics described in connection with that embodiment are included in one or more embodiments of this application. Thus, statements such as "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments", etc. that appear in different places in this specification do not necessarily all refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "comprising", "including", "having" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0038] The technical solutions of the embodiments of this application will be introduced below.

[0039] The multi-channel electroencephalogram (EEG) acquisition system in smart glasses can simultaneously acquire EEG signals at multiple positions, providing the possibility for spatial analysis of EEG signals. However, due to the limited number of acquisition electrodes in smart glasses and the limitation of the electrode distribution by the glasses structure, traditional spatial filtering methods are difficult to effectively apply. At the same time, problems such as spatial crosstalk between electrodes, correlation brought by the common reference electrode, and local electromyogram interference seriously affect the extraction of effective EEG features. Especially in real application scenarios, the movement, emotional changes, and cognitive activities of users will introduce complex interference patterns, resulting in a significant decline in signal quality.

[0040] Existing spatial filtering techniques are mainly designed for traditional EEG caps, requiring a large number of electrode covers and regular electrode distributions. When these methods are directly applied to smart glasses, problems such as over-smoothing, decreased spatial resolution, and suppression of target features often occur. In dynamic scenarios, due to the small changes in electrode positions and the influence of motion artifacts, the performance of spatial filtering will be further reduced. In addition, existing methods lack targeted consideration of the specific wearing positions and electrode distribution characteristics of smart glasses, and cannot establish effective electrode mapping relationships and multi-scale filtering strategies. At the same time, these methods often adopt unified parameter configurations when processing time-frequency features, making it difficult to adapt to the feature requirements of different cognitive tasks, and also unable to achieve dynamic enhancement and spatial reconstruction of target EEG features.

[0041] To solve the above problems, please refer to Figure 1 , Figure 1 is a schematic flow chart of a method for spatial filtering and feature enhancement of multi-channel EEG signals of smart glasses provided by an embodiment of this application. The method for spatial filtering and feature enhancement of multi-channel EEG signals of smart glasses in the embodiments of this application can be applied to computer devices, and such computer devices include but are not limited to devices such as smart phones, laptop computers, tablet computers, desktop computers, physical servers, and cloud servers. As Figure 1As shown, the method for spatial filtering and feature enhancement of multi-channel EEG signals of smart glasses in this embodiment includes steps S101 to S107, which are described in detail as follows:

[0042] Step S101, taking the multi-channel EEG signals collected by the smart glasses as original signals, establishing an electrode spatial position model according to the original signals, and generating an electrode mapping relationship according to the electrode spatial position model.

[0043] Specifically, the multi-channel EEG acquisition system of smart glasses acquires the user's EEG signals through a dry electrode array arranged at specific positions of the frame. The system is equipped with 12 acquisition electrodes, which are located at key positions in the forehead, temples and both sides of the head. The sampling frequency is set at 1000Hz to ensure the time resolution of the signal. Each acquisition channel is equipped with a dedicated signal conditioning circuit, including a preamplifier, a bandpass filter and a digital isolation module. The signal is amplified 5000 times and converted to 16 bits to form a multi-channel digital signal stream. During the acquisition process, the system monitors the contact impedance of each electrode in real time and ensures that it is stable below 5kΩ. The common mode interference is effectively suppressed through active shielding technology and differential amplifier circuits. The adaptive baseline correction algorithm is used to eliminate signal drift in real time, and the digital low-pass filter is used to suppress high-frequency noise and power frequency interference. The system also integrates a motion artifact detection module, which can mark and remove abnormal signal segments caused by blinking, chewing and other actions. For different wearing scenarios, an intelligent gain control strategy has been developed, which can automatically adjust the amplification factor according to the signal amplitude to ensure that the signal is always in the optimal acquisition range. At the same time, the system adopts a high-precision clock synchronization mechanism to ensure the precise alignment of multi-channel data sampling, and realizes real-time transmission and temporary storage of data through a cache management mechanism.

[0044] Based on the collected multi-channel EEG signals, the system establishes an accurate electrode spatial position model. The model uses the smart glasses frame as the standard reference system and establishes a spatial rectangular coordinate system with the center point of the nose bridge as the coordinate origin. In practical applications, the system integrates a micro triaxial acceleration sensor and an angle sensor to obtain the spatial attitude information of the glasses in real time. Through the sensor data fusion algorithm, the method can compensate for the position deviation caused by head movement and ensure the stability of the electrode spatial position. Considering the head shape differences of different users, the method uses a deformable grid model to describe the spatial distribution of the electrode array and dynamically deforms it according to the actual wearing situation. The position model also includes spatial constraints based on anatomical structures to ensure that the electrode position description corresponds to the actual brain region distribution. Through a two-layer position optimization strategy, the method first performs a rough alignment based on rigid body transformation, and then uses a local deformation algorithm to achieve fine matching, and finally obtains a high-precision representation of the electrode spatial distribution. For dynamic usage scenarios, a real-time position calibration mechanism is developed to dynamically update the electrode position information by analyzing the sensor data flow. The method also establishes a position error compensation model that can predict and correct the position deviation caused by glasses movement based on historical data. To improve the stability of the position model, a spatial filtering algorithm is introduced to smooth the original position data and effectively suppress the position jitter caused by sensor noise.

[0045] After obtaining the accurate spatial positions of the electrodes, the method constructs the mapping relationship between the electrodes through multi-level analysis. First, an initial connection relationship is established based on the geometric distance between the electrodes, and an adaptive weight function is used to calculate the spatial coupling strength. Considering the conductive characteristics of the scalp tissue, the method introduces a conduction model based on physiological characteristics to estimate the diffusion characteristics of signals in space. For the unique electrode distribution pattern of smart glasses, a special spatial correlation analysis method is developed. This method comprehensively considers the signal conduction path, tissue impedance distribution, and electrode contact characteristics to construct a more realistic electrode mapping relationship. For example, when the user is sitting still and working in an office environment, the method mainly focuses on the electrode mapping in the prefrontal region and improves the accuracy of signal detection by adjusting the weight configuration of the forehead electrodes; when the user is walking or in a mild exercise state, the method automatically switches to the exercise mode, focuses on the mapping relationship of the electrodes in the temple region, and suppresses the influence of motion artifacts by optimizing the mapping parameters; when the user is in a resting state, the method adopts a balanced mapping strategy and maintains similar weight distributions for all brain region electrodes. To cope with the dynamic measurement environment, the method uses a sliding window strategy to update the mapping weights in real time, and the window length is adaptively adjusted according to the signal stability. By analyzing the signal correlation and phase consistency between adjacent electrodes, the method can identify and suppress abnormal connections caused by local myoelectric interference or electrode loosening. During the construction of the mapping relationship, special attention is paid to the spatial constraints caused by the glasses structure. By calculating the relative position changes of the electrodes under different wearing conditions, a mapping optimization scheme considering physical limitations is established. The method realizes the multi-scale mapping analysis function, can describe the correlation characteristics between the electrodes at different spatial scales, and ensures a stable and reliable spatial mapping relationship in various actual usage scenarios.

[0046] Step S102: Perform a topological analysis on the electrode mapping relationship, determine the spatial structure and the corresponding filtering scale, and generate a filter bank according to the filtering scale.

[0047] Specifically, based on the electrode mapping relationship obtained in the previous step, the method first conducts an in-depth topological analysis to determine the spatial structure characteristics. By calculating the connection strength, connection density, and adjacency relationship of each electrode node, key electrode connection patterns are identified. Among them, the connection strength is measured by the signal correlation and phase synchronization between the electrodes, and the connection density reflects the tightness of the electrode interconnection within the local area.

[0048] In some embodiments, the topological analysis of the electrode mapping relationship to determine the spatial structure and the corresponding filtering scale includes: obtaining the connection strength, connection density, and adjacency relationship of each electrode node according to the electrode mapping relationship; determining the key electrode connection pattern according to the connection strength, connection density, and adjacency relationship; determining the spatial structure according to the key electrode connection pattern and constructing a connection feature matrix; and partitioning the connection feature matrix according to a multi-level clustering strategy to obtain the filtering scale.

[0049] For the electrode distributions in different functional regions such as the prefrontal lobe and temples, the average connection strength within the region and the connection density ratio between regions are calculated respectively. For example, when the user is concentrating on reading, the internal connection strength of the electrodes in the prefrontal lobe region reaches above 0.7, the connection density within the region is 85%, while the connection strength with other regions is below 0.3 and the connection density is only 30%; when the user is performing simple movements, the electrodes in the temporal lobe region show a significant asymmetric distribution. The average connection strength of the left electrode group is 0.65 and the connection density is 75%, which is about 30% higher than that of the right side. The method establishes a connection feature matrix to record the connection strength, directionality, and stability indexes of all electrode pairs, providing a quantitative basis for subsequent scale partitioning.

[0050] Exemplarily, the generation of the filter bank according to the filtering scale includes: constructing a parameter template according to the filtering scale; the parameter template at least includes kernel function parameters, spatial coverage, weight distribution, and directionality weight template; and generating the filter bank according to the parameter template.

[0051] It should be noted that in some embodiments, the construction of the parameter template according to the filtering scale includes: obtaining the average connection strength, inter-region connection density, and connection strength matrix corresponding to each electrode unit of the smart glasses; on a local scale, generating kernel function parameters according to the average connection strength; wherein, the higher the average connection strength, the smaller the spatial attenuation rate of the kernel function parameters; on a medium scale, adjusting the spatial coverage and weight distribution of the filter according to the inter-region connection density ratio; on a global scale, constructing a directionality weight template according to the eigenvector of the connection strength matrix; in the directionality weight template, the higher the weight coefficient in the direction with a larger eigenvalue of the eigenvector; and generating the parameter template according to the kernel function parameters, spatial coverage, weight distribution, and directionality weight template.

[0052] According to the obtained connection feature matrix, the method adopts a multi-level clustering strategy to divide the filtering scales. First, highly coupled electrode groups are identified based on a connection strength threshold (0.7) and a connection density threshold (75%), and are divided into local scale units. For example, two highly coupled local units are detected in the prefrontal region, containing 3 and 4 electrodes respectively. Then, functionally related local units are combined into medium-scale modules according to the connection density ratio between regions (the threshold is set to 50%). For example, two local units detected in the prefrontal region are merged into a medium-scale module because the connection density between them reaches 65%. Finally, based on the eigenvalue decomposition of the inter-region connection strength matrix, the global scale division is determined, and remote electrode pairs with a connection strength exceeding 0.3 are included in the same large-scale processing unit. This multi-level division based on connection features forms three spatial scale levels: local (2-4 cm), medium (4-6 cm), and global (>6 cm).

[0053] Based on the divided spatial scale levels, the method constructs a parameter template by combining specific connection features. At the local scale, the kernel function parameters are set according to the average connection strength within the electrode unit. The higher the connection strength, the smaller the spatial decay rate of the kernel function. For example, a Gaussian kernel with a decay rate of 0.3 is used for the highly coupled unit in the prefrontal region (connection strength 0.7), while a Gaussian kernel with a decay rate of 0.6 is used for the low-coupled unit in the temporal lobe (connection strength 0.5). At the medium scale, the spatial coverage range and weight distribution of the filter are adjusted based on the connection density ratio between regions. For example, the internal connection density of the prefrontal medium-scale module is 75%, and an effective coverage radius of 5 cm is set accordingly, and a central weight of 0.8 is adopted. At the global scale, a directional weight template is constructed using the eigenvectors of the connection strength matrix, and a higher weight coefficient is assigned to the direction with a larger eigenvalue. The obtained parameter template contains a complete parameter set such as spatial scale, kernel function type, and weight distribution.

[0054] Based on the parameter templates at each scale, the method generates corresponding filter banks. At the local scale level, each highly coupled unit is configured with a set of spatial filters that match the kernel function parameters. For example, the filter of the prefrontal unit uses a Gaussian kernel (decay rate 0.3) with a processing range of 3 cm, mainly used to extract local synchronization features; the filter of the temporal lobe unit uses a Gaussian kernel (decay rate 0.6) with a processing range of 2 cm, focusing on suppressing local interference. At the medium scale level, regional filters are generated based on the connection density distribution, and the coverage range is determined by the effective radius in the parameter template. For example, the filter of the prefrontal region has a coverage radius of 5 cm and a central weight of 0.8, used to extract regional cognitive features; the filter of the motion-related region has a coverage radius of 4 cm and a central weight of 0.7, used to monitor motion-related activities. At the global scale level, directional filters are constructed according to the connection strength eigenvector. For the direction with a larger eigenvalue (such as the front-back direction, eigenvalue 0.6), a response gain of 1.2 is configured, and for the direction with a smaller eigenvalue (such as the left-right direction, eigenvalue 0.3), a response gain of 0.8 is configured. This multi-level filter bank realizes the selective extraction of features at different spatial scales. In practical applications, the filter bank can adaptively adjust parameters according to the dynamic changes of connection features. For example, when it is detected that the connection strength in the prefrontal region drops from 0.7 to 0.5, the decay rate of the kernel function of the local filter is adjusted accordingly; when it is found that the eigenvalue of the connection feature in a certain direction increases significantly, the response gain of the filter in the corresponding direction is increased in a timely manner.

[0055] In step S103, the original signal is separated into components according to the filter bank to obtain component information, the signal sources corresponding to the component information are obtained, and an independent component sequence is constructed based on the signal sources; the component information includes at least independent components, clustering components, and principal components.

[0056] Specifically, in combination with the original signal and the filter bank, the method performs spatial decomposition. First, the original signal is segmented according to the sampling time window (2 seconds), and each segment is convolved with the multi-scale filter bank. At the local scale, a Gaussian kernel filter (attenuation rate 0.3 - 0.6) is used to spatially filter the signal, and 2 - 4 main spatial components are obtained for each local region. Among them, the energy proportion in the prefrontal region is the highest, reaching 65%, followed by the temporal lobe region at 35%, and the occipital lobe region at about 25%. At the medium scale, a regional filter (coverage radius 4 - 5 cm) is used to extract the regional activity characteristics, and 3 - 5 significant regional components are obtained. The energy proportion in the prefrontal-temporal connection region is 40%, and that in the parietal-occipital connection region is 30%. At the global scale, a directional filter (gain 0.8 - 1.2) is used to obtain the large-scale spatial pattern, and usually 2 - 3 stable global components are obtained, with energy proportions of 15%, 10%, and 5% respectively. The spatial power spectrum and phase spectrum are calculated for each scale level to obtain the energy distribution within the 0 - 30 Hz frequency band. In the resting state, the energy of the spatial components in the α band (8 - 13 Hz) is dominant, while in the cognitive task, the energy of the components in the θ band (4 - 7 Hz) increases significantly. The method statistically analyzes the decomposition results of 50 consecutive time windows and establishes a histogram of the energy distribution of the spatial components for setting the component separation threshold in the subsequent steps.

[0057] Based on the energy distribution characteristics obtained from the spatial decomposition, the method separates the components at each scale. At the local scale, first, the spatial components with an energy proportion exceeding 25% are selected as candidates, and the mutual information and correlation coefficient between these components are calculated. When the mutual information is lower than 0.2 and the correlation coefficient is less than 0.3, they are marked as independent components. When the correlation coefficient is between 0.3 - 0.6, the conditional mutual information is further calculated. If it is lower than 0.15, they are still considered independent components. For example, usually 3 independent components are separated in the prefrontal region, corresponding to the activity patterns of the left, middle, and right frontal lobes, with energy proportions of 35%, 20%, and 15% respectively. At the medium scale, feature clustering is performed based on the regional components with an energy proportion exceeding 30%. A two-step clustering strategy is adopted: first, a preliminary grouping is performed based on the spatial correlation degree (threshold 0.6), and then the phase synchronization index of the components within the group is calculated. When the synchronization index exceeds 0.5 and the duration is greater than 500 ms, it is confirmed as the same cluster. In this way, usually 2 stable clustering components are obtained in the prefrontal-temporal region, respectively reflecting the activity patterns related to cognitive processing and emotion regulation. For the global scale components, according to the previously obtained energy distribution, the components with an accumulated energy proportion reaching 80% are selected for principal component analysis, and the principal components with eigenvalues greater than 1 are extracted as the global signal features.

[0058] Use the separated independent components, clustering components, and principal components for signal source labeling. For local independent components, source localization is performed by combining their spatial positions and spectral characteristics. When the center of the spatial distribution of a component is located in the frontal region and the energy ratio in the θ frequency band exceeds 40%, it is labeled as an attention-related source; when it is located in the temporal lobe and the α frequency band energy is significant, it is labeled as a perceptual processing source. The method also calculates characteristic parameters such as source strength (based on energy ratio), source stability (based on time persistence), and source specificity (based on spatial distribution kurtosis). For medium-scale clustering components, analyze their spatial connection patterns and temporal synchronization characteristics. When components in multiple brain regions show stable phase synchronization (synchronization index > 0.7 and duration > 1 second), it is labeled as a functional network source, and topological parameters such as network centrality, clustering coefficient, and efficiency are calculated. For example, in a working memory task, the frontal-parietal working memory network (number of nodes 4 - 6, average path length 2.3) and the temporal-parietal attention network (number of nodes 3 - 5, clustering coefficient 0.65) can be identified. For global-scale principal components, label them according to the distribution characteristics of their spatial patterns and the temporal evolution law. When the principal component shows a spatially symmetric distribution between the front and back and exists stably in the resting state, it is labeled as the default network source; when it shows left-right asymmetry and is related to the task state, it is labeled as the cognitive control source.

[0059] Based on the marker-based signal source information, the method constructs multi-level independent component sequences. At the local level, first, the significance scores of each independent component are calculated, and the scores are weighted by three indicators: energy contribution rate (weight 0.4), temporal stability (weight 0.3), and spatial specificity (weight 0.3). Components with scores exceeding 0.7 are selected to construct local sequences, such as the prefrontal attention source (score 0.85) and the temporal auditory processing source (score 0.78). At the medium-scale level, network sequences are constructed based on the characteristic parameters of functional network sources. The overall scores of the networks are calculated, including connection strength (weight 0.35), spatial coverage (weight 0.35), and temporal stability (weight 0.3). Network sources with scores exceeding 0.65 are included in the sequence, such as the working memory network (score 0.82) and the attention network (score 0.73). At the global level, important components are screened according to the explained variance of the principal component source (threshold 10%) and temporal consistency (consistency index > 0.6), and a sequence reflecting the large-scale brain activity pattern is constructed. The method dynamically maintains these multi-level sequences. When significant changes in source features are detected (such as intensity change > 20% or spatial offset > 2 cm), the sequence update mechanism is triggered. For example, in an attention switching task, when the subject's attention shifts from visual stimuli to auditory stimuli, the method can capture the energy redistribution of the prefrontal attention source (the energy of the visual-related source decreases by 30%, and the energy of the auditory-related source increases by 25%) and the dynamic reorganization of the attention network (the visual-parietal connection weakens, and the auditory-parietal connection strengthens), and accordingly update the independent component sequences at each level.

[0060] Step S104: Perform spatial mapping on the independent component sequence to determine the main components, construct a weight matrix according to the main components, perform spatial filtering according to the weight matrix to generate a spatial feature sequence, and perform spatial reconstruction on the spatial feature sequence to generate enhanced features.

[0061] Specifically, perform spatial mapping on the obtained independent component sequences. At the local scale, map the independent components with scores exceeding 0.7 to the corresponding brain region space to establish an accurate spatial distribution map. For example, after mapping the prefrontal attention source (score 0.85), an activation center with an intensity of 0.8 is formed in the midline prefrontal region, showing a circular attenuation outward, covering an area of 5 cm²; the temporal lobe auditory processing source (score 0.78) shows a strongly activated area of 4 cm² (intensity 0.75) on the left and a secondary activated area of 3 cm² (intensity 0.65) on the right. At the medium scale, the mapping of the working memory network (score 0.82) forms a banded connection between the prefrontal lobe and the parietal lobe, with a width of 3 cm and a total coverage area of 12 cm², and the connection intensity shows a gradual change distribution of 0.7 - 0.5; the attention network (score 0.73) forms a circular distribution in the temporal-parietal region, covering an area of 10 cm², and the activation intensity at the main nodes remains above 0.6. At the global scale, the mapping of the default network source forms a symmetric distribution in the anterior and posterior brain regions, with the anterior covering 8 cm² and the posterior covering 12 cm², and the intensity distribution is 0.5 - 0.4. When the user performs a video viewing task, the method monitors a continuous activation of 6 cm² (intensity 0.7) in the occipital visual region, and forms a connection band with a width of 4 cm with the parietal attention region.

[0062] Based on the specific numerical characteristics of the spatial mapping, the method constructs a weight matrix. At the local scale, according to the 5 cm² coverage feature of the prefrontal attention source, a 5×5 local weight grid is set, and the weight at the central position is set to 0.85 (consistent with the mapping intensity), and it decays by 0.15 every 1 cm outward to form an accurate weight gradient; for the temporal lobe auditory source, based on the left-right asymmetric distribution characteristics (4 cm² and 3 cm²), 4×4 and 3×3 weight grids are set respectively, with the maximum weight on the left being 0.75 and on the right being 0.65. At the medium scale, the weight assignment of the working memory network follows the banded distribution characteristics. In the 12 cm² coverage area, the weight at the prefrontal lobe node is set to 0.7 (corresponding to the mapping intensity), the weight at the parietal lobe node is set to 0.6, and the weight on the connection band decreases linearly by 0.1 according to the 3 cm width; the attention network sets a polar coordinate form of weight distribution in the 10 cm² circular region, with the weight at the node position being 0.6, the ring spacing being 1 cm, and the weight decay step size being 0.1. At the global scale, according to the total coverage area of 20 cm², a large-scale weight field is constructed, and the corresponding weights are set for the anterior 8 cm² region and the posterior 12 cm² region according to the measured intensity range of 0.5 - 0.4.

[0063] Spatial filtering is performed on the signal using the constructed weight matrix. In the 5 cm² high-weight area of the prefrontal lobe (weight > 0.7), a 2 cm × 2 cm sliding window is used. After multiplying the window weights point by point with the matrix elements and summing them, the θ-wave features related to attention are retained; in the asymmetric area of the temporal lobe, a 2.5 cm × 2.5 cm window is used for the 4 cm² area on the left, and a 2 cm × 2 cm window is used for the 3 cm² area on the right to extract the features of auditory processing respectively. For the 12 cm² area covered by the working memory network, an anisotropic filtering window of 3 cm × 1 cm is used along the direction of the 3 cm-wide connection band, enhancing the information transfer features between nodes; for the 10 cm² area of the attention network, an adaptive window in polar coordinate form is used, and the window size varies between 2 - 4 cm along the circumferential position. In the 20 cm² covered area of the default network source, a 3 cm × 3 cm window is used for the front part and a 4 cm × 4 cm window is used for the rear part to achieve large-scale spatial smoothing. For example, when the user transitions from rest to a cognitive task, the method detects a 35% increase in the θ-wave energy in the prefrontal lobe area, and stable attention signature features are obtained through local fine filtering; when performing dual-task switching, the direction of the information flow on the connection band of the working memory network changes, and the method captures this dynamic change process through anisotropic filtering.

[0064] According to the spatial filtering results, the method generates a spatial feature sequence. For the 5 cm² high-weight area of the prefrontal lobe, each 2 cm² is divided into a feature unit, and three groups of features, namely energy distribution, phase consistency, and directionality, are extracted to form a 15-dimensional local feature vector; in the asymmetric area of the temporal lobe, a 12-dimensional feature vector is extracted from the 4 cm² area on the left, and a 9-dimensional feature vector is extracted from the 3 cm² area on the right. The two form a cross-hemisphere feature sequence through the position correspondence relationship. The 12 cm² covered area of the working memory network is evenly sampled along the connection band direction, and a set of feature parameters including node activity intensity, connection flow direction, and phase difference are extracted for each 1 cm² to form a 36-dimensional network feature sequence; in the 10 cm² area of the attention network, 8 sampling points are set on the ring structure, and 5-dimensional features are extracted for each point to obtain a 40-dimensional circular feature sequence. For the 20 cm² area of the default network source, wide-area synchrony, energy gradient, and phase coupling features are extracted from the 8 cm² front part and the 12 cm² rear part respectively, and synthesized to form a 100-dimensional global feature sequence. For example, in the video viewing task, the method extracts an 18-dimensional visual feature sequence from the 6 cm² activated area of the occipital lobe, and at the same time extracts a 12-dimensional attention regulation feature sequence on the 4 cm connection band with the parietal lobe. The two constitute the spatio-temporal feature expression of visual attention. The method performs real-time analysis on these feature sequences. When the local energy distribution deviation exceeds 25% or the network connection feature change exceeds 30%, the feature extraction parameters are adjusted in a timely manner to ensure the timeliness and reliability of the feature sequences.

[0065] In some embodiments, the spatial reconstruction of the spatial feature sequence to generate enhanced features includes: performing local projection on the spatial feature sequence to obtain a feature region; performing spatial reconstruction according to the feature region; and generating the enhanced features according to the reconstruction result corresponding to the spatial reconstruction.

[0066] Perform local projection on the generated spatial feature sequence. In the prefrontal region, perform energy-guided principal direction decomposition on local feature vectors to obtain the projection results of three feature axes. For example, when the user is performing a continuous reading task, the first feature axis reflects the energy distribution in the θ band, showing a central focus feature during the attention maintenance stage; when the attention shifts to different positions on the page, the energy center moves with the line of sight; when the attention is dispersed, the energy distribution shows an obvious diffusion trend. The second feature axis describes the propagation characteristics of phase consistency, showing an orderly propagation from the prefrontal to the parietal lobe during in-depth reading comprehension; when encountering difficult sentences, this propagation experiences a brief interruption and reorganization; while during mind wandering, it shows disordered phase changes. The third feature axis captures the lateral directional changes, especially showing a significant left-right alternating pattern during dual-task switching. In the temporal lobe region, perform asymmetric projection on the left and right hemisphere features. It is observed that the projection intensity of the left hemisphere is significantly higher than that of the right hemisphere in language processing tasks, and this asymmetry is most significant during native language reading and relatively weakened during foreign language reading; while during music appreciation, the projection intensity of the right hemisphere gradually increases, especially being more obvious in music segments with strong rhythm.

[0067] Based on the amplitude distribution of the local projection, the method marks the feature regions. In the prefrontal region, the regions are divided into three categories according to the projection amplitude: the main activity area, the secondary activity area, and the background area. For example, when concentrating on solving a math problem, the main activity area is concentrated in the midline position of the forehead, showing strong θ wave activity; when the problem difficulty gradually increases, the range of the secondary activity area gradually expands, reflecting the increase in cognitive load; when encountering a key breakthrough point, a brief intensity peak appears in the main activity area. During a long-term continuous calculation task, the secondary activity area will periodically alternate and enhance with the main activity area, showing the dynamic allocation process of cognitive resources. The marking in the temporal lobe region reflects the lateralization characteristics of auditory processing. In a speech recognition task, the activity intensity of the main feature area on the left side is significantly higher than that on the right side; when processing a multi-person conversation, the activity area on the left side will change dynamically with the switching of speakers; when processing environmental sounds, the difference in activity intensity between the two sides decreases, but shows stronger temporal synchronization. The marking of the working memory network shows task-related dynamic changes. In a serial memory task, the node regions in the prefrontal and parietal lobes show regular alternating activations; when the serial length increases, the duration of node activation extends and the intensity also increases accordingly; in a spatial working memory task, a persistent activation pattern is observed in the parietal lobe region, accompanied by periodic information exchange with the visual region.

[0068] Based on the characteristic distribution of the marked regions, the method performs spatial reconstruction. In complex cognitive tasks, corresponding reconstruction strategies are adopted based on the characteristics of the aforementioned three types of regions. High-density reconstruction is used for the main active region of the prefrontal lobe to capture the rapid changes in theta wave activity, and its reconstruction accuracy is dynamically adjusted according to the marked activity intensity. For example, in a multi-target tracking task, when the intensity peak appears in the main active region, the reconstruction accuracy is increased to the highest; the reconstruction accuracy of the secondary active region is proportional to the cognitive load level marked by it, and the sampling density is gradually increased when the load increases; low-density reconstruction is used for the background region to maintain computational efficiency. In an emotion processing task, the reconstruction parameters are set based on the marked results of temporal lobe asymmetry. The left main characteristic region adopts a higher reconstruction accuracy due to its stronger activity intensity, while the right side reduces the accuracy configuration accordingly. The reconstruction strategy of the working memory network is directly based on the previously marked node activity patterns. Time-varying reconstruction accuracy is adopted in the identified alternating activation regions, and smooth transition of accuracy is achieved on the connection pathways. For example, in an n-back task, the method dynamically adjusts the reconstruction density according to the marked node activity intensity, maintains a stable reconstruction accuracy during the information maintenance stage, and increases the local reconstruction accuracy at the information update time point to capture transient changes. The reconstruction process also takes into account the spatial characteristics of different brain regions, focusing on signal reconstruction in the depth direction in the prefrontal region and strengthening feature recovery in the horizontal direction in the temporal lobe region.

[0069] Using the reconstructed spatial distribution, the method generates enhanced features, and the degree of enhancement is directly dominated by the reconstruction accuracy. In the prefrontal region, an adaptive enhancement strategy is adopted for the main active area with high-precision reconstruction, and the enhancement intensity is proportional to the density of reconstruction points. For example, in driving fatigue monitoring, the method modulates the enhancement amplitude of theta wave features according to the reconstruction accuracy of the main active area. Regions with higher reconstruction accuracy (such as the mid-frontal region) obtain 35% feature enhancement, while the peripheral regions with lower reconstruction accuracy only provide 15% enhancement. In the classroom attention monitoring scenario, the degree of feature enhancement of the working memory network changes following the reconstructed spatial resolution, providing strong feature enhancement in the node regions with high-precision reconstruction, while adopting a gradual enhancement intensity in the transition regions to ensure the spatial continuity of features. For the language processing features in the temporal lobe, the method designs an enhancement strategy based on the asymmetric reconstruction results, and the high-precision reconstruction region in the left hemisphere obtains stronger feature enhancement, thus better highlighting the activity characteristics of the language-dominant hemisphere. In the multi-task processing scenario, the dynamic adjustment of feature enhancement also strictly follows the distribution of reconstruction accuracy. For example, when the reconstruction accuracy of the visual search task exceeds the threshold, the method correspondingly increases the feature enhancement intensity in the occipital region; when the reconstruction accuracy of the auditory task improves, the feature expression in the temporal lobe is strengthened. This adaptive enhancement mechanism based on reconstruction accuracy not only ensures the accuracy of feature enhancement but also maintains the structural integrity of spatial features. In the long-term working scenario, the method continuously monitors the reconstruction quality. When there is a significant change in the reconstruction accuracy of a certain region (such as the signal quality degradation caused by fatigue), the feature enhancement strategy of the corresponding region is adjusted in a timely manner to ensure the reliability of the enhanced features. For example, in a sustained cognitive task, if a decrease in the reconstruction accuracy of the prefrontal region is detected, the method will correspondingly reduce the feature enhancement intensity to avoid erroneously amplifying noise features.

[0070] Step S105: Perform multi-scale decomposition on the filter bank and the enhanced features to obtain multi-scale features, construct a feature combination based on the multi-scale features, and obtain the time-frequency features corresponding to the feature combination.

[0071] Specifically, using the filter banks obtained in step (local-scale Gaussian kernel filter, medium-scale band-pass filter, and global-scale low-pass filter) and the enhanced features obtained in step (35% enhancement of the signal in the main active area of the prefrontal lobe, 20% enhancement in the secondary active area, 40% enhancement in the left temporal lobe area and 25% enhancement in the right, 30% enhancement in the nodes of the working memory network), a multi-scale decomposition is performed. In the prefrontal lobe area, the enhanced main active area signal is processed using a Gaussian kernel filter (attenuation rate 0.3, window 2 cm) at the local scale; the enhanced secondary active area signal is processed using a band-pass filter (attenuation rate 0.5, window 4 cm) at the medium scale; and the background area signal is processed using a low-pass filter (attenuation rate 0.7, window 6 cm) at the global scale. In the concentration test, when attention is concentrated, the energy of the processed local component accounts for 65% in the central area and 15% in the peripheral area; when attention is dispersed, the central energy drops to 35% and the peripheral energy rises to 45%. For the temporal lobe area, the method applies a directional filter bank to the enhanced asymmetric signals of the left and right hemispheres respectively. The 40% enhanced left area decomposition shows 7-13 Hz activity related to language, and the 25% enhanced right area presents slow-wave activity of 3-7 Hz. On the working memory network, the 30% enhanced node signal is processed using a connection band filter (width 1.5 cm) to obtain a θ-band activity component of 4-7 Hz. Finally, feature components at three scales, namely local, medium, and global, are obtained, and each component contains spatio-temporal activity characteristics of a specific frequency band.

[0072] Perform hierarchical extraction on the above-mentioned multi-scale components. At the local scale, perform time-frequency analysis on the central component with an energy proportion of 65% and the peripheral component with an energy proportion of 15% to obtain attention state features, including the instantaneous intensity sequence of the central theta wave (0.5 - 2.5 μV) and the modulation sequence of the peripheral alpha wave (20% - 60%). Perform spatial clustering on the asymmetric components of the temporal lobe to obtain the features of 3 main language processing nodes on the left side (intensities are 1.8 μV, 1.5 μV, and 1.2 μV respectively) and the features of 2 secondary nodes on the right side (intensity is 1.0 μV and 0.8 μV). Perform connectivity analysis on the 4 - 7 Hz component of the working memory network to obtain the features of 4 information transfer pathways, and the connection strengths are 0.75, 0.68, 0.62, and 0.55 respectively. These features show significant differences in different tasks: when attention is concentrated, the central-peripheral intensity ratio remains above 3:1; when performing difficult tasks, the connection strengths of the pathways increase by 30% across the board; when performing language comprehension, the intensity difference between the left and right sides reaches 40%. In actual application scenarios, these features exhibit obvious task specificity: in the assessment of classroom concentration, the continuous exceeding of the central theta wave intensity of 2.0 μV indicates deep engagement; in language learning tasks, the sequential activation pattern of the left-side nodes reflects the degree of understanding; in memory tests, the dynamic changes in the connection strengths of the pathways indicate the memory load. Through these analyses, a complete set of hierarchical features is finally obtained, including the local attention hierarchical feature sequence (sampling rate 10 Hz), the regional language processing hierarchical features (spatial resolution 1 cm), and the network connection hierarchical features (update rate 2 Hz).

[0073] Based on the extracted hierarchical features, the method generates multi-scale feature representations. First, the central theta wave sequence (intensity 1.5 - 2.5 μV) and the peripheral alpha wave sequence (intensity 0.8 - 1.2 μV) are combined into an attention distribution map with a spatial resolution of 1 cm and a temporal resolution of 100 ms. Then, 5 nodes in the temporal lobe (3 on the left, 2 on the right) and their activity intensities are integrated into a language processing network, and the phase relationship between the nodes is characterized by a synchronization index (range 0.3 - 0.8). For the 4 pathways of working memory, the information flow direction and transfer efficiency are calculated to obtain a 4×4 connectivity matrix, and the matrix elements represent the immediate state of the pathways. These multi-scale features show dynamic changes during actual tasks: Classroom attention monitoring shows that when students are focused, the intensity at the center of the attention distribution map stabilizes above 2.0 μV, and the modulation of the peripheral alpha waves drops to the lowest; during reading comprehension, the left nodes of the language network are activated sequentially, and the synchronization index gradually increases to above 0.7; during the serial memory task, the working memory pathways are activated sequentially according to the stimulus sequence, and the connection strength increases with the increase in load. Through long-term data accumulation and statistical analysis, the feature distribution patterns are finally obtained: the attention features show a center-periphery gradient distribution (gradient coefficient 0.4 - 0.8), the language processing features show task-related lateralization (asymmetry index 0.3 - 0.6), and the working memory features show load-dependent network reorganization (reorganization coefficient 0.2 - 0.5).

[0074] According to the above distribution patterns, the method constructs a feature combination. A dynamic weight is configured for the local-scale features (central theta wave energy 2.0 μV, peripheral alpha wave energy 0.8 μV, gradient coefficient 0.4 - 0.8), and the weight increases to 0.6 when attention is concentrated and drops to 0.3 when it is dispersed. Lateralization weighting is performed on the medium-scale features (average 1.5 μV on the left temporal lobe, 0.9 μV on the right, asymmetry index 0.3 - 0.6), and the weight on the left increases to 0.5 during language tasks. A network-dependent weight is set for the global-scale features (average connection strength 0.65, reorganization coefficient 0.2 - 0.5), and the weight dynamically increases when the cognitive load increases. The method adjusts the combination parameters in real time according to the signal quality (signal-to-noise ratio threshold 8 dB) and task relevance (correlation coefficient threshold 0.7). Through this dynamic combination process, a complete feature combination is finally obtained: the local-scale combination (center-periphery energy distribution feature, phase modulation feature), the medium-scale combination (bilateral language processing node feature, functional connectivity feature), and the global-scale combination (large-scale network synchronization feature), and each combination contains the corresponding weight configuration and dynamic adjustment mechanism.

[0075] In some embodiments, obtaining the time-frequency features corresponding to the feature combination includes: performing time-domain analysis on the feature combination to obtain timing features; performing frequency-domain transformation on the timing features according to a preset multi-level transformation strategy to obtain frequency-domain statistical results; and constructing the time-frequency features based on the frequency-domain statistical results.

[0076] For the obtained feature combination, the method performs time-domain analysis. For different scale combinations, sliding window processing is respectively adopted, with a window length of 2 seconds and an overlap rate of 50%. For the local scale combination, a dual-core processing unit is designed: the central detection core (radius 2 cm) is used to extract the energy-time envelope of the θ wave (2.0 μV) in the central region, and the edge detection core (radius 4 cm) is used to obtain the suppression modulation feature of the α wave (0.8 μV) in the peripheral region. Both detection cores use a sampling rate of 100 Hz and an amplitude resolution of 0.1 μV. For the medium scale combination, a bilateral analysis module is constructed: the activation sequences are respectively extracted from the left language processing node (1.5 μV) and the right node (0.9 μV), with a time accuracy of 5 ms, and at the same time, the functional connection change between the nodes is calculated, and the connection update rate is 10 Hz. For the global scale combination, a network synchronization analyzer is used to track the large-scale network pattern (average connection strength 0.65), and the global phase coherence (resolution 2°) and network reorganization features (time resolution 50 ms) are calculated. For example, during the cognitive task switching, a rapid change process in which the energy of the central θ wave drops from 2.0 μV to 1.4 μV is observed at the local scale, and a compensatory enhancement appears in the peripheral α wave; at the medium scale, it can be seen that the activation sequences of the bilateral nodes are reorganized, and there is an adjustment period of 200 ms in the connection strength; at the global scale, a brief decrease and reconstruction process of the network synchronization degree are recorded. The method establishes a family of response curves for each type of feature, including an energy response curve (dynamic range 40 dB), a phase response curve (phase range -π to π), and a connection response curve (normalized scale 0-1). Through multi-dimensional analysis, a set of multi-dimensional timing features including timing, polarity, and spatial distribution information is finally obtained.

[0077] Perform a frequency-domain transformation on the above-mentioned multi-dimensional time-series features, adopting a multi-level transformation strategy. Perform a short-time Fourier transform on the energy time series (sampled at 100 Hz), select a 1-second transformation window to match the cognitive rhythm characteristics, use an improved Hanning window for the window function (the main lobe width is optimized to 1 Hz), and set the step size to 100 ms to ensure time-frequency resolution. The transformation result covers the range of 0.1 - 30 Hz, and overlapping FFT channels are set: the θ band 4 - 7 Hz (8 frequency points), the α band 8 - 13 Hz (10 frequency points), the β band 13 - 30 Hz (16 frequency points), and the overlapping rate of adjacent channels is 30%. Perform a Hilbert transform on the phase time series (sampled at 200 Hz), extract the instantaneous phase through an accurate phase unwrapping algorithm (error < 0.1°), and obtain the frequency modulation feature using an instantaneous frequency estimator with an adaptive threshold (time resolution 5 ms). The method also calculates the phase locking value (16×16 matrix), the group delay characteristic (frequency resolution 0.5 Hz), and the phase-amplitude coupling index. Perform a wavelet transform on the connection time series (updated at 10 Hz), select the Morlet basis wavelet (central frequency 5.5, bandwidth parameter 1.2), set the decomposition scale to an exponential sequence from 1 to 64, and configure an energy normalization and phase correction module at each scale. For example, during the cognitive task transition, the spectral analysis of the energy time series shows that the energy in the θ band drops from 45% to 25%, and the energy in the α band increases correspondingly. The phase time series reveals the phase reorganization process between the prefrontal lobe and the parietal lobe. Through these transformations and statistical analyses, the final frequency-domain statistical results are obtained: the full-band energy distribution map (frequency resolution 0.5 Hz, dynamic range 60 dB), the phase coupling matrix (64×64, confidence level 0.95), and the time-varying connection mode spectrum (number of modes 32, update rate 5 Hz).

[0078] According to the above frequency-domain statistical results, when constructing the time-frequency features, a multi-scale synthesis strategy is adopted. At the energy feature level, a high-precision time-frequency energy map is constructed. The time-axis resolution is inherited from the sliding window (100 ms), and the frequency axis is divided using a logarithmic scale (0.1 - 30 Hz is divided into 128 frequency points). An adaptive threshold technique is introduced during the spectrogram generation process. The noise background is determined through wavelet denoising and median estimation (signal-to-noise ratio threshold 10 dB), and the signal peak detection uses an improved maximum suppression algorithm (the suppression radius is adaptively adjusted with frequency, ranging from 0.5 - 2 Hz). At the phase feature level, a multi-level phase coupling map is established, including an instantaneous phase difference map (resolution 5°, update rate 10 Hz), a phase-locking value map (resolution 0.05, significance threshold p < 0.01), and a cross-frequency band phase coupling intensity map (frequency band interval 1 Hz). The method performs a Rayleigh statistical test on the phase data, eliminates insignificant phase relationships (p > 0.01), and retains stable phase patterns. At the network feature level, a dynamic connection time-frequency map is formed to describe the reorganization characteristics of the functional network in different frequency bands. Adaptive segmented processing is adopted (segment length 50 - 200 ms), and within each segment, the connection strength matrix (32×32) and network topology indicators (clustering coefficient, path length) are calculated. For example, during the cognitive state transition, the energy time-frequency map shows the energy migration process from theta waves to alpha waves, lasting approximately 200 ms; the phase coupling map captures a short-term synchronization enhancement in the beta band, maintaining for 100 - 150 ms; the network time-frequency map reflects the dynamic reorganization pattern of the large-scale functional network. The method configures a timestamp (accuracy 1 ms), a frequency marker (accuracy 0.1 Hz), an intensity value (dynamic range 60 dB), and a reliability index (confidence 0.95) for each feature. Through these processes, a complete set of time-frequency features is finally obtained, including: an energy time-frequency spectrogram (signal-to-noise ratio > 10 dB, 128 frequency points), phase time-frequency relationships (64 coupling modes), network time-frequency dynamics (32 connection modes), and all features have traceable and quantifiable statistical attributes.

[0079] Step S106, perform spatial mapping on the time-frequency features to obtain the interference patterns corresponding to the time-frequency features, and perform feature enhancement according to the interference patterns to obtain the feature representations corresponding to the interference patterns.

[0080] Specifically, spatial mapping is performed on the obtained time-frequency feature set (including the energy time-frequency spectrogram, phase time-frequency relationship, and network time-frequency dynamics). The method first maps the energy time-frequency spectrogram to the electrode space to construct a spatio-temporal-frequency three-dimensional graph, establishing signal representations in the three dimensions of time, space, and frequency. Then, spatial projection is performed on the phase time-frequency relationship to establish a phase propagation graph, describing the phase difference and propagation delay characteristics between electrodes. Spatial reconstruction is performed on the network time-frequency dynamics to form a spatio-temporal distribution map of functional connections, including connection strength and direction information. For example, in a visual attention task, the energy time-frequency features show a gamma-band response in the occipital region that varies with the intensity of visual stimuli; in a working memory task, the phase time-frequency features reflect the theta-band information transmission between the prefrontal and parietal lobes. Through these spatial mappings, the method identifies three main interference patterns: local high-frequency bursts (manifested as instantaneous abnormal enhancement of energy), phase jumps (manifested as sudden changes in phase relationships), and connection breaks (manifested as local damage to the network structure). These interferences exhibit specific spatio-temporal distribution characteristics in different cognitive tasks, ultimately forming a complete interference pattern map.

[0081] In some embodiments, the time-frequency features at least include an energy time-frequency spectrogram, a phase time-frequency relationship, and network time-frequency dynamics; the performing spatial mapping on the time-frequency features to obtain an interference pattern corresponding to the time-frequency features includes: mapping the energy time-frequency spectrogram to the electrode space to construct a spatio-temporal-frequency three-dimensional graph for establishing signal representations in the three dimensions of time, space, and frequency; performing spatial projection on the phase time-frequency relationship to establish a phase propagation graph for describing the phase difference and propagation delay characteristics between electrodes; performing spatial reconstruction on the network time-frequency dynamics to obtain a spatio-temporal distribution map of functional connections, the spatio-temporal distribution map including connection strength and direction information; generating an interference pattern map based on the spatio-temporal-frequency three-dimensional graph, the phase propagation graph, and the spatio-temporal distribution map to determine the interference pattern according to the interference pattern map.

[0082] Based on the identified interference pattern atlas, the method constructs spatial templates. For local high-frequency bursts, an adaptive suppression template is designed, and the suppression intensity is dynamically adjusted according to the interference amplitude, and the spatial coverage range adaptively changes according to the interference diffusion degree. For the phase jump region, a phase correction template is constructed, and the phase relationship of surrounding stable electrodes is used for interpolation reconstruction to achieve smooth transition of the phase. For connection breaks, a connection repair template is established, and pattern supplementation is carried out by combining historical connection patterns and surrounding complete connections. These templates are in the form of three-dimensional tensors, including spatial weights, frequency selectivity, and time modulation characteristics. For example, in the processing of electrooculogram artifacts, the suppression template forms a directional suppression field in the forehead area, and the intensity gradually decays with distance; in the processing of electromyogram interference, the correction template achieves selective suppression in the high-frequency band. During the cognitive task transition process, the repair template can effectively maintain the continuity of the functional network and ensure the stable extraction of signal features. The method also optimizes the template parameters for specific scenarios, such as enhancing the direction selectivity of the suppression template in the motion state and optimizing the time response characteristics of the connection repair template in social interactions.

[0083] The features are enhanced using the constructed spatial templates. The method adopts a two-layer processing strategy of template matching and feature reconstruction. In the first layer, the suppression template is convolved with the burst region to achieve precise suppression of interference while maintaining the integrity of the effective signal. In the second layer, the correction template is applied to restore the continuity of the signal, including smooth transition of the phase relationship and stable reconstruction of the functional connection. An adaptive threshold mechanism is introduced during the enhancement process, and the threshold parameter is dynamically adjusted according to the local signal-to-noise ratio to ensure the spatial consistency of the enhancement effect. In classroom concentration monitoring, this enhancement strategy can effectively handle various interferences: transient blink artifacts are significantly suppressed, signal drift caused by head movement is corrected, and the connection features of the attention network remain stable. In the driving fatigue monitoring scenario, the method can accurately separate the electromyogram interference generated by driving operations and retain the slow-wave features reflecting the fatigue state. In multi-person interaction tasks, the enhanced signal can clearly reflect the neural synchronization pattern between individuals, providing reliable data support for social cognition research.

[0084] Exemplarily, the obtaining of the feature representation corresponding to the interference pattern includes: constructing a spatial template according to the interference pattern atlas; the spatial template includes an adaptive suppression template, a phase correction template, and a connection repair template; the spatial template is subjected to feature enhancement according to a two-layer processing strategy of template matching and feature reconstruction to obtain the feature representation corresponding to the interference pattern.

[0085] Based on the enhancement result, the method generates a feature representation. First, a multi-level feature descriptor is constructed, which includes the time dimension, frequency dimension, and spatial dimension to achieve a comprehensive expression of signal features. Then, feature encoding is performed to convert the enhanced signal into a compact feature vector, and a sparse representation strategy is adopted to retain the main feature components. The method establishes a feature index structure to support fast retrieval based on time, frequency, and spatial positions. In practical applications, this feature representation exhibits excellent performance: in cognitive state monitoring, it can accurately depict the spatio-temporal features of attention transfer, including the spatial migration of frontal theta wave activity, the phase reorganization of parietal alpha waves, and the dynamic reconstruction of functional connectivity. In the emotion recognition task, it represents the dynamic features of emotion-related neural networks, such as the time-varying pattern of prefrontal asymmetry and the activation sequence of limbic methods. In the skill learning process, it captures the plastic changes of the motor control network, including the evolution of the cooperative pattern in the sensorimotor area and the regulatory features of cortical excitability. The finally obtained feature representation not only has a high signal-to-noise ratio and spatial resolution but also maintains the physiological meaning and temporal continuity of the signal.

[0086] Step S107: Perform spatial reconstruction based on the feature representation, perform feature fusion on the reconstruction result, and obtain spatially enhanced features.

[0087] Specifically, in combination with the obtained multi-level feature representation (including high signal-to-noise ratio energy features, phase time-frequency relationships, and network dynamic features), the method performs spatial reconstruction. First, a reconstruction mapping matrix is established to map the feature vector back to the electrode space and maintain the spatial continuity of the signal. For the frontal lobe region, a high-precision reconstruction strategy is adopted to restore the theta-band attention-related features through a local interpolation algorithm; for the temporal lobe region, a lateralized reconstruction method is used to maintain the left-right asymmetry of language processing features; for the parietal lobe region, a networked reconstruction is implemented to reconstruct the local connection pattern related to spatial tasks. The method configures a quality monitoring module for each reconstructed region to evaluate the reconstruction accuracy and spatial consistency in real time. For example, in attention monitoring, the reconstruction result shows the theta wave activity center in the prefrontal center and the surrounding alpha wave suppression ring, and the correlation of the spatial distribution with the original features reaches more than 0.85; in the language processing task, the left hemisphere dominant activation pattern is reconstructed, and the deviation of the activity intensity ratio from the original features is less than 10%. In the motor imagery task, the reconstruction process can accurately restore the mu rhythm suppression pattern in the sensorimotor area and maintain the functional connectivity features with the premotor area. Through multi-region collaborative reconstruction, a complete spatial activity map is obtained, including local feature distributions, inter-regional connection relationships, and large-scale network structures.

[0088] In some embodiments, spatial reconstruction is performed according to the feature representation, and feature fusion is performed on the reconstruction result to obtain spatial enhanced features, including: establishing a reconstruction mapping matrix, mapping the feature vector corresponding to the feature representation to the electrode space to obtain a spatial activity map; the spatial activity map includes local feature distribution, inter-region connection relationship, and large-scale network structure; performing feature fusion on the spatial activity map to obtain the spatial enhanced features.

[0089] Feature fusion is performed on the reconstructed spatial activity map, and the method designs a multi-level fusion strategy. At the spatial level, adaptive weighted combination is implemented for the reconstructed features of different brain regions, and the weight coefficients are dynamically adjusted according to signal reliability and task relevance. For example, in the attention assessment, the weight of the prefrontal theta wave activity is adjusted between 0.4 and 0.8 as the attention level changes; in the language understanding task, the weight of the left temporal lobe is increased to more than 0.6 according to the language processing requirements. At the frequency level, the method fuses the activity features of multiple frequency bands to establish cross-frequency interaction patterns, such as theta-alpha coupling strength, beta-gamma energy ratio, etc. At the time level, multi-scale sliding window analysis is used to achieve continuous integration of features. Short windows (200 ms) are used to capture transient changes, and long windows (2 s) are used to extract steady-state features. In practical applications, this multi-level fusion strategy exhibits excellent performance: in the working memory task, the method fuses the prefrontal theta wave, parietal alpha wave, and phase synchronization features between the two regions, accurately reflecting the dynamic changes of memory load; during the motor imagery process, the mu rhythm suppression in the sensorimotor area and the beta wave change in the premotor area are integrated to achieve early recognition of motor intention. Through this multi-dimensional feature fusion, a set of combined features reflecting the brain state is finally obtained, which not only retains the unique attributes of each component but also reflects the synergistic relationship between them.

[0090] Based on the combined features obtained by fusion, the method generates and outputs spatially enhanced features. First, the fused features are normalized to establish a unified cross-modal quantization standard, including amplitude normalization, phase standardization, and connection strength normalization. Then, feature enhancement is performed using an adaptive gain control strategy, where the gain coefficient is dynamically adjusted according to feature saliency and task relevance. The method sets region-specific enhancement parameters. The frontal lobe region focuses on enhancing theta-band features, the temporal lobe region strengthens the interaction between frequency bands, and the parietal lobe region highlights network connection features. In continuous monitoring scenarios, this spatial enhancement strategy shows good adaptability: in classroom attention monitoring, it can track changes in cognitive engagement in real time, with an accuracy of 90% in detecting attention dispersion; in driving fatigue monitoring, by enhancing slow-wave activity features, it can give an early warning of the fatigue state 30 seconds in advance; in multi-person collaborative tasks, it can accurately capture the neural synchronization patterns among team members, providing a quantitative basis for evaluating interaction efficiency. The method also optimizes the enhancement strategy for different application scenarios: during skill learning, it focuses on enhancing plasticity features reflecting learning progress, including changes in cortical excitability and functional network reorganization; in emotion regulation tasks, it highlights neural circuit features related to emotions, such as prefrontal asymmetry and limbic method activity; in creative thinking training, it enhances the dynamic interaction features of the default network and the executive control network. Through these targeted enhancement processes, the finally output spatially enhanced features not only maintain the physiological authenticity of the signals but also provide prominent task-related features, achieving precise representation and reliable evaluation of brain states.

[0091] The method provided has at least the following beneficial effects:

[0092] 1. This solution establishes a complete electrode mapping and multi-scale filtering system, overcoming the limitation of the limited number of electrodes in smart glasses. By establishing an electrode spatial position model, mapping relationship, and performing topological analysis to determine the appropriate filtering scale, the optimal utilization of limited electrode signals is achieved.

[0093] 2. This solution constructs a complete processing link from signal decomposition to feature enhancement, effectively solving the problem of multi-source interference. By implementing spatial decomposition, component separation, and feature enhancement; and then through steps, performing interference pattern recognition and spatial template construction, the anti-interference ability of the method is significantly improved while ensuring the integrity of target features.

[0094] 3. This solution realizes the dynamic adaptability of feature extraction and enhancement, meeting the requirements of different application scenarios. By performing multi-scale decomposition and time-frequency analysis, and then through spatial reconstruction and feature fusion, precise responses to different cognitive tasks are achieved, ensuring the spatio-temporal consistency of enhanced features.

[0095] To implement the multi-channel electroencephalogram (EEG) signal spatial filtering and feature enhancement method corresponding to the above method embodiments for the smart glasses, so as to achieve the corresponding functions and technical effects. Refer to Figure 2 , Figure 2 FIG. Figure 2 shows a structural block diagram of a spatial filtering and feature enhancement device 200 provided by an embodiment of the present application. For ease of description, only the parts related to this embodiment are shown. The spatial filtering and feature enhancement device 200 provided by the embodiment of the present application includes:

[0096] A signal acquisition module 201, configured to use the multi-channel EEG signals collected by the smart glasses as original signals, establish an electrode spatial position model according to the original signals, and generate an electrode mapping relationship according to the electrode spatial position model;

[0097] A topology analysis module 202, configured to perform topology analysis on the electrode mapping relationship, determine a spatial structure and a corresponding filtering scale, and generate a filter bank according to the filtering scale;

[0098] A component separation module 203, configured to perform component separation on the original signals according to the filter bank, obtain component information, obtain signal sources corresponding to the component information, and construct an independent component sequence according to the signal sources; the component information includes at least independent components, clustering components, and principal components;

[0099] A component determination module 204, configured to perform spatial mapping on the independent component sequence, determine main components, construct a weight matrix according to the main components, perform spatial filtering according to the weight matrix to generate a spatial feature sequence, and perform spatial reconstruction on the spatial feature sequence to generate enhanced features;

[0100] A scale decomposition module 205, configured to perform multi-scale decomposition on the filter bank and the enhanced features, obtain multi-scale features, construct a feature combination according to the multi-scale features, and obtain time-frequency features corresponding to the feature combination;

[0101] A spatial mapping module 206, configured to perform spatial mapping on the time-frequency features, obtain interference patterns corresponding to the time-frequency features, perform feature enhancement according to the interference patterns, and obtain feature representations corresponding to the interference patterns;

[0102] A spatial reconstruction module 207, configured to perform spatial reconstruction according to the feature representations, perform feature fusion on the reconstruction results, and obtain spatially enhanced features.

[0103] The above-mentioned spatial filtering and feature enhancement device 200 can implement the multi-channel EEG signal spatial filtering and feature enhancement method of the smart glasses in the above method embodiment. The optional items in the above method embodiment are also applicable to this embodiment, which will not be elaborated here. The remaining content of the embodiment of the present application can refer to the content of the above method embodiment and will not be repeated in this embodiment.

[0104] Figure 3 It is a schematic structural diagram of a computer device provided by an embodiment of the present application. As Figure 3 shown, the computer device 3 of this embodiment includes: at least one processor 30 ( Figure 3 only one is shown in the figure), a memory 31, and a computer program 32 stored in the memory 31 and executable on the at least one processor 30. When the processor 30 executes the computer program 32, it implements the steps in any of the above method embodiments.

[0105] The computer device 3 may be a computing device such as a smart phone, a tablet computer, a desktop computer, and a cloud server. The computer device may include but is not limited to the processor 30 and the memory 31. Those skilled in the art can understand that Figure 3 this is only an example of the computer device 3 and does not constitute a limitation on the computer device 3. It may include more or fewer components than shown in the figure, or combine some components, or different components. For example, it may also include input / output devices, network access devices, etc.

[0106] The so-called processor 30 may be a central processing unit (CPU). The processor 30 may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0107] In some embodiments, the memory 31 may be an internal storage unit of the computer device 3, such as the hard disk or memory of the computer device 3. In other embodiments, the memory 31 may also be an external storage device of the computer device 3, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the computer device 3. Further, the memory 31 may also include both the internal storage unit and the external storage device of the computer device 3. The memory 31 is used to store an operating system, application programs, a BootLoader, data, and other programs, such as the program code of the computer program. The memory 31 may also be used to temporarily store data that has been output or is to be output.

[0108] In addition, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0109] An embodiment of the present application provides a computer program product, and when the computer program product runs on a computer device, the computer device is caused to implement the steps in each of the above method embodiments when executed.

[0110] In several embodiments provided in the present application, it can be understood that each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, the program segment, or the part of code includes one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order from that marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved.

[0111] When the above-mentioned functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device to execute all or part of the steps of the methods described in various embodiments of this application. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0112] The specific embodiments described above further elaborate on the purpose, technical solution, and beneficial effects of this application. It should be understood that the above description is only the specific embodiments of this application and is not used to limit the protection scope of this application. It is particularly pointed out that for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of this application shall be included in the protection scope of this application.

Claims

1. An intelligent glasses multi-channel electroencephalogram signal spatial filtering and feature enhancement method, characterized in that, Including: Taking the multi-channel electroencephalogram signals collected by the smart glasses as the original signals, establishing an electrode spatial position model according to the original signals, and generating an electrode mapping relationship according to the electrode spatial position model; Performing topological analysis on the electrode mapping relationship, determining the spatial structure and the corresponding filtering scale, and generating a filter bank according to the filtering scale; Separating components of the original signals according to the filter bank, obtaining component information, obtaining the signal sources corresponding to the component information, and constructing an independent component sequence according to the signal sources; The component information at least includes independent components, clustering components, and principal components; Performing spatial mapping on the independent component sequence to determine the main components, including: at the local scale, mapping the independent components with a significance score exceeding 0.7 to the corresponding brain region space to establish an accurate spatial distribution map; at the medium scale, the mapping of the working memory network forms a banded connection between the prefrontal lobe and the parietal lobe, with a width of 3 cm and a total coverage area of 12 cm², and the connection strength shows a gradual change distribution of 0.7 - 0.5; at the global scale, the mapping of the default network source forms a symmetric distribution in the anterior and posterior brain regions, with an anterior coverage of 8 cm² and a posterior coverage of 12 cm², and the intensity distribution is 0.5 - 0.4; constructing a weight matrix according to the main components, including: at the local scale, according to the 5 cm² coverage feature of the prefrontal lobe attention source, a 5×5 local weight grid is set, the weight at the center position is set to 0.85, and it decays by 0.15 per 1 cm outward to form an accurate weight gradient; at the medium scale, the weight distribution of the working memory network follows the banded distribution feature. In the 12 cm² coverage area, the weight at the prefrontal lobe node is set to 0.7, the weight at the parietal lobe node is set to 0.6, and the weight on the connection band decreases linearly by 0.1 according to the 3 cm width; at the global scale, according to the total coverage area of 20 cm², a large-scale weight field is constructed, and the anterior 8 cm² area and the posterior 12 cm² area are respectively set with corresponding weights according to the measured intensity range of 0.5 - 0.4; performing spatial filtering according to the weight matrix to generate a spatial feature sequence, and performing spatial reconstruction on the spatial feature sequence to generate enhanced features; Performing multi-scale decomposition on the filter bank and the enhanced features to obtain multi-scale features, constructing a feature combination according to the multi-scale features, and obtaining the time-frequency features corresponding to the feature combination; Performing spatial mapping on the time-frequency features to obtain the interference patterns corresponding to the time-frequency features, performing feature enhancement according to the interference patterns, and obtaining the feature representations corresponding to the interference patterns; Performing spatial reconstruction according to the feature representations, and performing feature fusion on the reconstruction results to obtain spatially enhanced features.

2. The method according to claim 1, wherein The performing topological analysis on the electrode mapping relationship to determine the spatial structure and the corresponding filtering scale includes: Obtaining the connection strength, connection density, and adjacency relationship of each electrode node according to the electrode mapping relationship; Determining the key electrode connection pattern according to the connection strength, connection density, and adjacency relationship; Determining the spatial structure according to the key electrode connection pattern and constructing a connection feature matrix; Partition the connection feature matrix according to a multi-level clustering strategy to obtain the filtering scale.

3. The method according to claim 2, wherein Generating a filter bank according to the filtering scale includes: Constructing a parameter template according to the filtering scale; the parameter template includes at least kernel function parameters, spatial coverage, weight distribution, and directional weight templates; Generating a filter bank according to the parameter template.

4. The method according to claim 3, wherein Constructing a parameter template according to the filtering scale includes: Obtaining the average connection strength, inter-region connection density, and connection strength matrix corresponding to each electrode unit of the smart glasses; On a local scale, generating kernel function parameters according to the average connection strength; wherein, the higher the average connection strength, the smaller the spatial decay rate of the kernel function parameters; On a medium scale, adjusting the spatial coverage and weight distribution of the filter according to the ratio of inter-region connection densities; On a global scale, constructing a directional weight template according to the eigenvector of the connection strength matrix; in the directional weight template, the higher the weight coefficient in the direction with a larger eigenvalue of the eigenvector; Generating the parameter template according to the kernel function parameters, spatial coverage, weight distribution, and directional weight template.

5. The method according to claim 1, wherein Performing spatial reconstruction on the spatial feature sequence to generate enhanced features includes: Performing local projection on the spatial feature sequence to obtain a feature region; Performing spatial reconstruction according to the feature region; Generating the enhanced features according to the reconstruction result corresponding to the spatial reconstruction.

6. The method according to claim 1, characterized in that, Obtaining the time-frequency features corresponding to the feature combination includes: Performing time-domain analysis on the feature combination to obtain time-series features; Performing frequency-domain transformation on the time-series features according to a preset multi-level transformation strategy to obtain frequency-domain statistical results; Constructing the time-frequency features according to the frequency-domain statistical results.

7. The method according to claim 1, characterized in that The time-frequency features at least include an energy time-frequency map, a phase time-frequency relationship, and network time-frequency dynamics; mapping the time-frequency features in space to obtain the interference pattern corresponding to the time-frequency features includes: Mapping the energy time-frequency map to electrode space to construct a spatio-temporal-frequency three-dimensional map for establishing signal representation in three dimensions of time, space, and frequency; Performing spatial projection on the phase time-frequency relationship to establish a phase propagation map for describing the phase difference and propagation delay characteristics between electrodes; Performing spatial reconstruction on network time-frequency dynamics to obtain a spatio-temporal distribution map of functional connections, and the spatio-temporal distribution map includes connection strength and direction information; Generating an interference pattern map according to the spatio-temporal-frequency three-dimensional map, the phase propagation map, and the spatio-temporal distribution map to determine the interference pattern according to the interference pattern map.

8. The method according to claim 7, wherein Obtaining the feature representation corresponding to the interference pattern includes: Constructing a spatial template according to the interference pattern map; the spatial template includes an adaptive suppression template, a phase correction template, and a connection repair template; Performing feature enhancement on the spatial template according to a double-layer processing strategy of template matching and feature reconstruction to obtain the feature representation corresponding to the interference pattern.

9. The method according to claim 1, characterized in that Performing spatial reconstruction according to the feature representation, and performing feature fusion on the reconstruction result to obtain spatially enhanced features, including: Construct a reconstruction mapping matrix to map the feature vectors corresponding to the feature representations to the electrode space and obtain a spatial activity map; the spatial activity map includes local feature distributions, inter-region connection relationships, and large-scale network structures; Perform feature fusion on the spatial activity map to obtain the spatially enhanced features.

10. A computer device, characterized in that, It includes a memory and a processor; the memory is used to store computer programs; the processor is used to execute the computer programs and implement the method according to any one of claims 1 to 9 when executing the computer programs.

Images