Method and equipment for extracting and classifying multi-channel EEG signals from brain-computer intelligent headsets

Abstract

The present application relates to the field of brain-computer interface technology, and in particular to a method and device for extracting and classifying multi-channel EEG signal features of a brain-computer intelligent headset. The method comprises: obtaining channel identifiers corresponding to the multi-channel signals of the intelligent headset, performing component extraction and dividing the analysis domain, outputting component features for spatial positioning, obtaining functional areas to construct a regional network, and assigning weights through the regional network; outputting regional attribute data packets according to the weight assignment results corresponding to the assigned weights, establishing a structural model, and outputting a channel association network based on the root; performing hierarchical construction on the channel association network, defining feature hierarchies, performing scale fusion on the feature hierarchies to form feature groups, generating feature element sets, performing classification and combination to determine classification dimensions, obtaining corresponding classification tags for credibility assessment, establishing an evaluation system, obtaining classification results, and completing multi-channel EEG signal feature extraction and classification. Multi-scale feature decomposition and analysis domain division are achieved.

Description

Technical Field

[0001] The present application relates to the field of brain-computer interface technology, and in particular to a method and device for extracting and classifying multi-channel EEG signal features of a brain-computer intelligent headset. Background Art

[0002] The development of multi-channel EEG signal acquisition technology has provided a rich source of neural information for brain-computer interface applications. However, in the application scenario of portable brain-computer intelligent headphones, multi-channel EEG data faces unique technical challenges. First, the distribution of electrodes in the periauricular area is spatially restricted, and the number of channels is limited and densely distributed, resulting in severe inter-channel interference and information redundancy in the acquired signals. Second, positional deviations while wearing the headphones can cause changes in the electrode-skin contact impedance, affecting the stability of the signal quality. In addition, physiological noise such as electromyography and electrooculography during daily activities, as well as environmental electromagnetic interference, can significantly reduce the signal-to-noise ratio.

[0003] In terms of signal processing, existing technologies face several key issues. In terms of signal acquisition and spatial positioning, traditional methods lack specific consideration for the unique distribution of electrodes around the ear, making it difficult to accurately locate and separate the signal sources of each channel. In feature extraction, a single data processing strategy is often used, failing to fully integrate information in the time, frequency, and spatial domains, resulting in the extracted features lacking discriminability and stability. During functional area identification and feature classification, it is difficult to balance the real-time and accuracy of the algorithm, and there is a lack of a systematic feature fusion and screening mechanism.

[0004] Due to the portability requirements of smart headsets, the system must implement complex signal processing and feature analysis within limited computing resources. Existing methods, in particular, lack a comprehensive credibility assessment system for classification, making it difficult to quantify the reliability of classification results. This can easily lead to misjudgments in practical applications. Furthermore, individual differences and the diversity of usage environments pose significant challenges to feature extraction and classification. Currently, there is a lack of a systematic solution that can simultaneously optimize signal quality, enhance feature extraction, improve classification performance, and assess credibility.

[0005] Therefore, a method is urgently needed to solve at least one of the above problems. Summary of the Invention

[0006] The embodiments of the present application provide a method and device for extracting and classifying multi-channel EEG signals from brain-computer intelligent headphones. The method aims to solve the problem that due to the portability requirement of intelligent headphones, the system needs to implement complex signal processing and feature analysis under limited computing resources. In particular, in the classification and discrimination link, the existing methods lack a complete credibility assessment system, making it difficult to quantitatively assess the reliability of the classification results, which can easily lead to misjudgment in practical applications. At the same time, individual differences and the diversity of usage environments also bring huge challenges to feature extraction and classification. There is currently a lack of a systematic solution that can simultaneously solve the problems of signal quality optimization, feature extraction enhancement, classification performance improvement and credibility assessment.

[0007] In a first aspect, an embodiment of the present application provides a method for extracting and classifying multi-channel EEG signal features of a brain-computer intelligent headset, comprising:

[0008] Collecting multi-channel signals of the smart headset and determining channel distribution, obtaining channel identifiers corresponding to the channel distribution; performing component extraction on the channel identifiers and dividing analysis domains, and outputting component features based on the analysis domains;

[0009] spatially locating the component features to obtain functional regions, constructing a regional network based on the functional regions, and assigning weights via the regional network; outputting a regional attribute data packet based on a weight assignment result corresponding to the assigned weights, establishing a structural model based on the regional attribute data packet, and outputting a channel association network based on the structural model;

[0010] Performing hierarchical construction on the channel association network to define a feature hierarchy, performing scale fusion on the feature hierarchy to form a feature group, outputting fusion elements based on the feature group, and generating a feature element set based on the fusion elements; performing classification combination on the feature element set to determine a classification dimension, and obtaining a classification label corresponding to the classification dimension;

[0011] The classification labels are evaluated for credibility to establish an evaluation system, and discrimination labels are output according to the evaluation system; data integration is performed on the discrimination labels to obtain classification results, and multi-channel EEG signal feature extraction and classification are completed.

[0012] In a second aspect, the present application also provides a computer device comprising a processor and a memory, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the method for extracting and classifying multi-channel EEG signals of the brain-computer intelligent headset as described in the first aspect is implemented.

[0013] This method uses smart headphones to collect multi-channel EEG signals, determine channel distribution, and obtain channel identifiers. Components are extracted from the channel identifiers, and the analysis domain is divided to output component features. Component features are spatially localized, functional regions are identified, and a regional network is constructed. Dynamic weight assignment is used to generate regional attribute data packets, and then a structural model of the channel association network is established. The channel association network is hierarchically constructed, and after defining feature hierarchies, feature groups are formed through multi-scale fusion to generate feature feature sets. The feature feature sets are classified and combined, and classification dimensions and labels are determined. Discrimination labels are generated through credibility assessment, and finally, the data is integrated to output classification results. Through channel distribution and association network construction, the spatial correlation of multi-channel signals is addressed. Dynamic network weights are optimized based on functional regions to improve feature extraction efficiency. Multi-scale fusion techniques are combined to enhance the correlation of time-frequency-spatial domain features.

[0014] By spatially localizing and weighting multi-channel signals, noise interference is reduced and feature extraction accuracy is enhanced. Dynamic weight allocation and hierarchical feature fusion reduce redundant computation, making it suitable for real-time processing in embedded devices such as smart headsets. A credibility assessment system, combined with multi-dimensional classification labeling, enhances the robustness of EEG signal classification (e.g., intent recognition and disease diagnosis). This system is applicable to medical and consumer electronics applications such as brain-computer interfaces, neurofeedback training, and epilepsy warning.

[0015] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of a method for extracting and classifying multi-channel EEG signals from a brain-computer intelligent headset according to an embodiment of the present application;

[0017] Figure 2 This is a schematic diagram of the structure of a multi-channel EEG signal feature extraction and classification device for a brain-computer intelligent headset shown in an embodiment of the present application;

[0018] Figure 3 This is a schematic diagram of the structure of a computer device shown in an embodiment of the present application. DETAILED DESCRIPTION

[0019] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may 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 obscuring the description of the present application with unnecessary detail.

[0020] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0021] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0022] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0023] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0024] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0025] The technical solutions of the embodiments of this application are introduced below.

[0026] The development of multi-channel EEG signal acquisition technology has provided a rich source of neural information for brain-computer interface applications. However, in the application scenario of portable brain-computer intelligent headphones, multi-channel EEG data faces unique technical challenges. First, the distribution of electrodes in the periauricular area is spatially restricted, and the number of channels is limited and densely distributed, resulting in severe inter-channel interference and information redundancy in the acquired signals. Second, positional deviations while wearing the headphones can cause changes in the electrode-skin contact impedance, affecting the stability of the signal quality. In addition, physiological noise such as electromyography and electrooculography during daily activities, as well as environmental electromagnetic interference, can significantly reduce the signal-to-noise ratio.

[0027] In terms of signal processing, existing technologies face several key issues. In terms of signal acquisition and spatial positioning, traditional methods lack specific consideration for the unique distribution of electrodes around the ear, making it difficult to accurately locate and separate the signal sources of each channel. In feature extraction, a single data processing strategy is often used, failing to fully integrate information in the time, frequency, and spatial domains, resulting in the extracted features lacking discriminability and stability. During functional area identification and feature classification, it is difficult to balance the real-time and accuracy of the algorithm, and there is a lack of a systematic feature fusion and screening mechanism.

[0028] Due to the portability requirements of smart headsets, the system must implement complex signal processing and feature analysis within limited computing resources. Existing methods, in particular, lack a comprehensive credibility assessment system for classification, making it difficult to quantify the reliability of classification results. This can easily lead to misjudgments in practical applications. Furthermore, individual differences and the diversity of usage environments pose significant challenges to feature extraction and classification. Currently, there is a lack of a systematic solution that can simultaneously optimize signal quality, enhance feature extraction, improve classification performance, and assess credibility.

[0029] Please refer to Figure 1 , Figure 1 The flowchart of the method for extracting and classifying multi-channel EEG signals from a brain-computer intelligent headset provided in the embodiment of the present application is provided. The method for extracting and classifying multi-channel EEG signals from a brain-computer intelligent headset provided in the embodiment of the present application can be applied to computer devices, including but not limited to smartphones, laptops, tablet computers, desktop computers, physical servers, cloud servers and other devices. Figure 1 As shown, the multi-channel EEG signal feature extraction and classification method of the brain-computer intelligent headset of this embodiment includes steps S101 to S104, which are detailed as follows:

[0030] Step S101: collect multi-channel signals of the smart headset and determine the channel distribution, obtain channel identifiers corresponding to the channel distribution; perform component extraction on the channel identifiers and divide the analysis domains, and output component features based on the analysis domains.

[0031] Specifically, the smart headset collects multi-channel signals and determines the channel distribution. The headset is worn binaurally, with six dry electrode sensors on each side. Signal amplification is achieved through a built-in high-precision op amp, and the sampling rate is set within a range of 250-1000Hz. The channel distribution follows a modified international 10-20 system, focusing on temporal, parietal, and occipital regions. This layout fully utilizes the limited space around the ear. The electrode array adopts a radial distribution structure, with the central electrode positioned above the tragus and the remaining electrodes evenly distributed along the edges of the auricle. Each electrode is equipped with an independent elastic support structure to accommodate different ear shapes. To ensure signal acquisition quality, the system integrates an impedance measurement circuit to monitor electrode-skin contact in real time. When contact abnormalities are detected, a micro-mechanical structure automatically adjusts electrode pressure. The smart headset is equipped with a high-precision inertial measurement unit to monitor wearer position changes in real time. By mapping the spatial positions of the electrodes to anatomical landmarks on the scalp, the system can compensate for spatial errors caused by positional offset. This mapping uses cubic spline interpolation to ensure smooth spatial position reconstruction. The system utilizes an adaptive electrode pressure regulation mechanism, using a micro-pneumatic device to adjust the contact pressure between the electrode and the skin in real time. It automatically compensates for this pressure when abnormal contact impedance is detected. The smart earphones incorporate an integrated environmental noise monitoring module, which assesses the surrounding electromagnetic environment in real time and dynamically adjusts the gain and filtering parameters of the signal amplifier accordingly. The system establishes a comprehensive electrode performance evaluation system, including real-time monitoring of key indicators such as contact impedance, baseline drift, and noise level. Through these innovative designs, the system generates a spatial position matrix containing precise three-dimensional coordinate information, enabling accurate positioning and dynamic tracking of each electrode position.

[0032] In some embodiments, obtaining the channel identifier corresponding to the channel distribution includes: performing signal separation and constructing a filter chain according to the channel distribution; and outputting the channel identifier according to the filter chain.

[0033] Based on the acquired spatial position matrix, signal separation processing is performed and a filter chain is constructed. The signal separation algorithm introduces spatial constraints based on traditional independent component analysis, fully leveraging electrode position information to enhance separation performance. The algorithm's objective function includes information maximization, spatial regularization, and temporal correlation constraints. The optimization process comprehensively considers the statistical independence, spatial correlation, and temporal continuity of the signals. In practical implementation, the system employs a segmented processing strategy, processing 2-second data segments at a time with a 50% overlap between adjacent segments to ensure signal continuity. The filter chain employs an adaptive cascade structure, comprising three core components: a digital notch filter, a bandpass filter bank, and an adaptive filter. The digital notch filter utilizes a narrow-band stopband design, with its center frequency automatically adjusted to the local power frequency. The bandpass filter bank extracts EEG activity in the four frequency bands of δ, θ, α, and β, with the filter order determined by balancing signal quality and computational load. The adaptive filter dynamically tracks and eliminates non-stationary interference, with its parameters updated in real time using a recursive least squares algorithm. The system incorporates an intelligent bypass mechanism that automatically streamlines processing when signal quality is good, reducing system latency. To address motion artifacts, the system employs an adaptive filtering algorithm based on acceleration signals to effectively suppress signal interference caused by head movement. The system also establishes an inter-channel crosstalk compensation mechanism, estimating the coupling coefficients between channels and constructing a compensation matrix to eliminate crosstalk effects. This comprehensive signal processing system forms a multi-stage adaptive filter chain structure, achieving efficient separation and purification of multi-channel EEG signals.

[0034] Based on the constructed filter chain system, a channel identification encoding system is established for the processed signals. Channel identification utilizes a multi-level architecture, integrating spatial location, signal characteristics, and quality assessment information. Spatial location encoding uses a three-level structure of "orientation-level-number," directly inherited from the front-end spatial position matrix, ensuring precise correspondence between identification and actual electrode locations. Signal characteristic encoding encompasses two dimensions: dominant frequency band and signal type. The dominant frequency band is determined based on the output of the bandpass filter bank in the filter chain, while the signal type is identified through feature analysis of the filtered waveform. The system utilizes a dynamic weighting mechanism to automatically adjust the importance of different feature dimensions based on the quality assessment results of each stage of the filter chain. The quality assessment system fully utilizes status information from each processing step in the filter chain, performing a comprehensive evaluation based on five dimensions: signal-to-noise ratio, baseline stability, electromyographic contamination rate, signal integrity, and spatial consistency. Fuzzy logic is used to integrate these indicators. Channel identification utilizes a hierarchical storage structure, enabling fast retrieval and historical traceability. Identification data is compressed to reduce storage and transmission overhead. Based on this coding and management system, the system outputs standardized channel identification in the format of "orientation code-feature code-quality code", realizing the precise positioning and classification of EEG signals.

[0035] In some embodiments, outputting the component features based on the analysis domain includes: mapping the analysis domain into a transformation spectrum and performing feature decomposition; and outputting the component features based on the feature decomposition.

[0036] The system uses a selective processing strategy based on channel identification to extract components and partition the analysis domain for the channel identifications generated in the above steps. For high-quality channels such as "L23-AB-Q1," the system directly applies a fine 8-layer wavelet packet decomposition. For channels with lower-quality identifications, a coarse 4-layer decomposition is used to reduce the impact of noise. Component extraction utilizes the wavelet packet transform method, using the db4 wavelet basis for multi-layer decomposition to form a complete time-frequency analysis tree structure. In practical applications, for example, when users are engaged in concentration training, the system focuses on high-quality channels in the frontal and parietal regions, performing a fine decomposition of the alpha band (8-13Hz) in these channels. This can clearly capture the significant suppression of alpha wave energy when the user enters a state of concentration. During the decomposition process, the system uses a soft thresholding method to denoise the coefficients at each scale level, with the threshold parameters determined by maximum likelihood estimation. The analysis domain is partitioned using an adaptive segmentation strategy, which divides the data into subintervals with similar statistical characteristics based on the nonstationary nature of the signal. The system calculates local signal statistics, including mean, variance, skewness, and kurtosis, over a sliding time window. Significant changes in these statistical characteristics are marked as segmentation points. To improve the robustness of segmentation, a mechanism combining multiple statistics is introduced to ensure the reliability of segmentation points. In emotion recognition tasks, this segmentation strategy can accurately pinpoint the timing of emotional changes, as different emotional states often exhibit significantly different statistical properties. Through this adaptive decomposition and segmentation approach, the system establishes a hierarchical signal analysis structure, with each analysis domain containing detailed time-frequency features and statistical properties within a specific time window.

[0037] The obtained analysis domain, along with its inherent time-frequency features and statistical properties, is converted into a transformation map through nonlinear mapping, and feature decomposition is performed. The system selects an appropriate mapping strategy based on the characteristics of the analysis domain divided in the previous stage. The transformation map is constructed using kernel principal component analysis, using a radial basis kernel function to map the original feature space into a higher-dimensional feature space. During the mapping process, the kernel function parameters are optimized through cross-validation to balance feature extraction accuracy and computational complexity. For example, in the focus assessment process, the system divides one minute of EEG data into multiple analysis domains. Through nonlinear mapping of time-frequency features and statistical properties, the feature migration trajectory during changes in focus can be clearly observed, which is important for real-time assessment of the user's cognitive state. The system applies singular value decomposition to the transformation map to extract the main feature directions and corresponding eigenvalues. During the feature decomposition process, an energy threshold criterion is applied to retain feature components that contribute 95% of the cumulative energy. In sustained attention tests, this decomposition method effectively distinguishes EEG features during focused and distracted states, as focused EEG signals typically exhibit stronger time-frequency structure, reflected in the higher energy contribution of the main feature components in the decomposition results. At the same time, the system calculates the spatial distribution pattern of the eigenvectors and analyzes the contribution of signals from different brain regions. At this stage, the transformation map is decomposed into a series of orthogonal eigenvalues, each of which carries information about a specific neural activity pattern.

[0038] Based on the orthogonal feature components derived from eigendecomposition, the system constructs a multidimensional feature description system. First, time-domain features are extracted, including statistics such as mean amplitude, root mean square value, kurtosis, and skewness. Then, frequency-domain features are calculated, including parameters such as the energy distribution of each frequency band, spectral centroid, and bandwidth. Finally, joint time-frequency features are analyzed, including instantaneous frequency, spectral entropy, and modulation characteristics. In fatigue detection applications, time-domain features can reflect the overall intensity changes in brain activity, frequency-domain features can capture the typical increase in theta waves and decrease in alpha waves during fatigue, and joint time-frequency features can describe the dynamic process of this energy transfer. When fatigue is detected in the user, the system observes a significant increase in delta wave (0.5-4Hz) energy, accompanied by a decrease in alpha wave energy and an increase in spectral entropy. The system normalizes all extracted features to eliminate the effects of dimensional and scale differences. When processing motor imagery tasks, the normalized features clearly reflect the differences in EEG characteristics between different motor imagery modes. For example, the degree of suppression of the μ rhythm (8-13Hz) in the corresponding motor cortex regions is significantly different during left-hand and right-hand motor imagery. Through this complete feature extraction process, the system finally outputs a set of standardized component features, each of which is labeled with the corresponding physical meaning and reliability indicators.

[0039] Step S102: spatially locate the component features, obtain functional areas, construct a regional network based on the functional areas, and assign weights through the regional network; output a regional attribute data packet based on the weight assignment result corresponding to the assigned weights, establish a structural model based on the regional attribute data packet, and output a channel association network based on the structural model.

[0040] Specifically, when spatially localizing the component features output from the above steps, the system combines the physical meaning and reliability of each feature and employs an improved spatial filtering algorithm. First, based on the spatial distribution characteristics of the earphone's bilateral electrodes, a local spatial filter is constructed, focusing on signal source localization in the superficial layers of the temporal, parietal, and occipital lobes. During the localization process, the system utilizes the time-frequency and phase information in the component features and employs a spatial filtering method based on the Laplace operator to calculate the spatial projection of the signal source. The localization algorithm is optimized specifically for the unique electrode distribution pattern of the earphones, and the introduction of a distance weighting term improves the accuracy of localizing signal sources in the periauricular region. In a music appreciation task, this localization method can capture rhythmic processing activity in the temporal lobe. When the user is focused on listening, the system can detect increased synchronization in the alpha band of the superior temporal gyrus. The system continuously adjusts the spatial filtering parameters through an iterative optimization method until the residual error is less than a preset threshold. Ultimately, the spatial distribution characteristics of each signal component in the periauricular region are calculated. During the optimization process, the system dynamically adjusts the filtering parameters based on the feature's reliability, assigning greater processing weight to features with high reliability and adopting a more conservative processing strategy for features with low reliability. At the same time, the system also fully considers the physical meaning of each component feature to ensure that the spatial positioning results are consistent with neurophysiological principles.

[0041] The acquired spatial distribution features are used to identify functional regions and construct a regional network. Functional region identification employs a pattern matching approach, matching the localized signal features with standardized periauricular brain region functional templates. The system pre-establishes a library of functional region templates tailored to the characteristics of headphone acquisition, encompassing multiple functional network patterns reliably detected using periauricular electrodes, including auditory processing, attention networks, and emotion regulation. During the matching process, an adaptive weighting algorithm is employed to compensate for individual differences and ensure reliable identification. The system incorporates a spatial similarity calculation method, measuring the degree of match between signal features and the template using weighted cosine similarity, and sets an adaptive threshold for region segmentation. Regional network construction utilizes a local connectivity graph approach, using the identified functional regions as nodes and constructing edges based on signal correlation to form a network structure reflecting the functional connectivity of the periauricular regions. Connection strength is calculated using two metrics, mutual information and phase synchronization, to capture linear and nonlinear correlations between different regions. During focus training, the system can accurately identify activity in attention-related periauricular brain regions, such as the temporal lobe and posterior parietal lobe, and characterize the functional connectivity patterns between these regions, providing neural indicators for assessing a user's focus level. The system established a dynamic update mechanism to automatically adjust the matching parameters of the functional template according to the task type to improve the specificity of recognition. Through this local network construction method, the system ultimately obtained a functional area network structure based on the peri-auricular electrodes.

[0042] Weights are assigned based on a constructed functional region network. The system uses an improved local network analysis method to calculate the importance weight of each regional node. Weight assignment takes into account three key metrics: local connectivity strength, signal stability, and functional specificity. Local connectivity strength reflects the degree of signal synchronization between a brain region and other periauricular monitoring regions, quantified by calculating weighted connectivity and clustering coefficient. Signal stability characterizes the temporal consistency of the region's activity pattern, assessed by measuring the stability of signal variance and statistical properties. Functional specificity measures the region's selectivity in responding to specific cognitive tasks, determined by comparing activation differences across different task states. The system incorporates a multi-level weight fusion framework to assess regional importance at different time scales, including short-term changes (seconds), medium-term trends (minutes), and long-term patterns (sessions). In a music appreciation task, the system uses these metrics to identify temporal lobe regions that play a key role in the music processing network. For example, when a user listens to classical music, the system accurately calculates the weights of the superior temporal gyrus region in the music processing network. These weights reflect the degree to which the user processes various musical components, such as rhythm and melody. The system establishes a dynamic weight adjustment mechanism, updating the importance assessment of each area in real time based on task progress and user status changes. Through this weight calculation method based on the characteristics of headphone collection, the system ultimately determines the importance weight value of each detectable functional area.

[0043] Regional attributes are output based on assigned weights. The system employs a weight-based attribute grading mechanism. High-weight regions receive more detailed attribute descriptions, including functional type, activation pattern, connectivity characteristics, and other dimensions. Medium-weight regions record basic functional attributes and primary connectivity relationships. Low-weight regions retain only core functional labels. In focus training applications, this grading mechanism can highlight key task-related brain region characteristics. For example, when a user performs a sustained attention task, the system will focus on describing activation characteristics in attention-related brain regions such as the temporal lobe and posterior parietal lobe, including increased theta wave activity in these regions and enhanced functional coupling with other attention network nodes. The system also records changes in the degree of alpha wave suppression in these regions; these characteristics reflect the user's level of attention engagement. For less task-relevant brain regions, such as peripheral regions covered by limbic electrodes, only basic activity is recorded. During music listening, the system accurately characterizes the characteristics of each node in the auditory processing network based on weights. High-weighted temporal lobe regions record detailed rhythmic processing patterns, such as the time-frequency coupling of delta and gamma waves, which reflect the user's level of musical rhythm processing. Low-weighted peripheral regions, on the other hand, only record basic activation states. Through this weighted attribute allocation scheme, the system ultimately outputs a structured data package of regional attributes.

[0044] In some embodiments, establishing a structural model based on the regional attribute data packet includes: performing spatial mapping on the regional attribute data packet to construct a topological distribution map; detecting connection nodes based on the topological distribution map; and establishing the structural model based on the connection nodes.

[0045] The system uses an improved topological mapping algorithm to spatially map the regional attribute data packets output from the above steps. A specialized local spatial mapping strategy, tailored to the characteristics of headphone acquisition, projects the functional regions in the regional attribute data packets onto a two-dimensional plane, generating a topological map of the functional regions. This mapping process employs multidimensional scaling analysis to preserve the relative distances between functional regions in the original space. In a focus training scenario, this mapping method clearly visualizes the distribution patterns of attention-related regions, such as the temporal and parietal lobes, on a two-dimensional plane, as well as their spatial relationships with surrounding functional regions. The system dynamically adjusts mapping parameters using an iterative optimization strategy to minimize information loss during the projection process. To improve mapping accuracy, a locality-preserving constraint is introduced to ensure that adjacent functional regions maintain similar spatial relationships after projection. The system also establishes a mapping quality assessment mechanism, evaluating the mapping effect by calculating the degree of topological preservation before and after projection. In a music appreciation task, this mapping method accurately reflects the spatial distribution characteristics of each functional region in the auditory processing network, providing a reliable spatial basis for subsequent node detection. Through this precise spatial mapping process, the system ultimately obtains a complete topological map of the functional regions.

[0046] Based on the generated topological distribution map, the system conducts connection node detection. A multi-level node identification strategy is designed to gradually identify potential connection nodes from the local to the global level. At the local level, regions with significant connectivity patterns are identified by calculating the connection strength and information flow characteristics between functional regions. At the global level, node importance is assessed based on network topological characteristics. The node detection process focuses on three key indicators: signal transmission efficiency, connection stability, and functional selectivity. In a focus training application, the system accurately identifies key nodes in the temporal lobe responsible for attention regulation. These nodes often exhibit strong signal synchronization and stable connectivity patterns. The system also implements an adaptive threshold mechanism to dynamically adjust node screening criteria based on task characteristics. To improve detection reliability, a cross-validation mechanism is introduced to ensure the stability of node identification through repeated detection. In a music processing task, the system effectively identifies key functional nodes involved in rhythmic processing, which often exhibit significant phase synchronization. Through this systematic node detection approach, a set of connection nodes with core functions in the functional network was ultimately identified.

[0047] A structural model is constructed around the detected connected nodes. This model is constructed using a hierarchical and progressive strategy, first constructing a skeleton network based on core connected nodes, then gradually adding secondary connections and functional pathways to ultimately form a complete network structure. The model parameters are set taking into account the characteristics of headphone acquisition, quantitatively describing the strength and directionality of connections between nodes. The system incorporates a parameter optimization mechanism that adjusts model parameters by minimizing prediction error, while also incorporating a dynamic update strategy that enables the model to adjust its internal structure based on real-time data. During concentration training, the structural model clearly demonstrates the evolving state of a user's attention. For example, when a user maintains a high level of concentration during a lecture, the model captures significant suppression of alpha wave activity in the temporal lobe region and strong connectivity patterns with other attention-related nodes. Through continuous verification and optimization, the system ultimately establishes a functional network structural model that accurately reflects cognitive states.

[0048] Channel correlation analysis was conducted based on the established structural model. The system designed a precise channel mapping mechanism to establish a correspondence between functional nodes in the model and actual acquisition channels. This mapping process takes into account the spatial distribution, signal characteristics, and functional selectivity of the channels, ensuring accurate correspondence between the functional network and actual signal acquisition. The system employed a multi-validation strategy, selecting the optimal mapping relationship by comparing the performance of different mapping schemes, and established a weighting mechanism based on channel importance. In a concentration training scenario, the system focused on analyzing the correlation patterns between the T3 and T4 channels in the temporal lobe and the P7 and P8 channels in the parietal lobe. When the user maintains a focused state, these channels exhibit specific synergistic features: The T3 and T4 channels in the temporal lobe exhibit significant suppression of alpha waves (8-13Hz) and enhancement of beta waves (13-30Hz), while forming stable functional connections with the P7 and P8 channels in the parietal lobe. When attention levels decrease, these four channels experience a simultaneous increase in theta wave (4-7Hz) power, and phase synchronization between the channels decreases significantly. The system quantitatively calculates the correlation strength of each channel pair, integrating features such as time series correlation, frequency band energy ratio, and phase synchronization index into standardized correlation indices. These correlation indices are weighted and combined to form a multidimensional channel correlation matrix, whose elements represent the degree of functional coupling between channels. Ultimately, the system organizes these quantitative correlation features into a hierarchical network structure, outputting a complete channel functional correlation network.

[0049] Step S103: perform hierarchical construction on the channel association network, define feature levels, perform scale fusion on the feature levels to form feature groups, output fusion elements based on the feature groups, and generate feature element sets based on the fusion elements; perform classification combination on the feature element sets to determine classification dimensions, and obtain classification labels corresponding to the classification dimensions.

[0050] When constructing the hierarchical channel functional association network output from the above steps, the system first analyzes the functional connectivity strength and direction of each channel node in the network. Based on these correlation features, a hierarchical and progressive processing strategy is used to perform multi-scale decomposition, establishing a hierarchical structure in the time, frequency, and space dimensions. In the time dimension, a multi-scale sliding window is used to divide the signal into processing units of varying time lengths. In the frequency dimension, a wavelet transform is used to construct an eight-layer decomposition tree from detail to approximation. In the spatial dimension, a hierarchical network structure is formed from the local to the global level based on the functional connectivity strength between channels. In attention training applications, this hierarchical approach can simultaneously capture both rapidly changing transient attention features and slowly changing sustained attention features. For example, when a user is listening to a lecture in class, the system can simultaneously detect short-term attention shifts to sudden sounds (manifested by millisecond-level changes in beta waves) and sustained changes in overall concentration (manifested by minute-level alpha wave modulations). The system determines the time-frequency-spatial parameters of each layer through iterative optimization, ultimately constructing a feature hierarchy encompassing three levels: micro (milliseconds), meso (seconds), and macro (minutes).

[0051] Leveraging a constructed feature hierarchy, the system performs scale fusion on features at each level. For micro-level features, millisecond-level neuronal firing patterns and local field potential changes are primarily integrated; for meso-level features, second-level EEG rhythm changes and functional connectivity strength are integrated; and for macro-level features, minute-level cognitive state transitions and network topology evolution are integrated. The fusion process employs a multi-level cascade structure, starting with the finest scale and integrating feature information layer by layer. At each fusion node, the system combines features at different scales using a weighted summation approach, with weight coefficients optimized by maximizing mutual information. The system also introduces temporal consistency constraints to ensure the continuity of temporal information during feature fusion. To address the differences in temporal resolution between features at different scales, an adaptive interpolation algorithm is designed for temporal alignment. When fusing micro-level transient features and macro-level trend features, the system employs wavelet reconstruction to maintain signal integrity. Through this adaptive fusion mechanism, the system obtains a multi-scale feature representation, ranging from fine-grained neural activity to coarse-grained cognitive states.

[0052] Based on the fused multi-scale feature representation, the system establishes a feature organization framework. For millisecond-level micro-features, the system extracts the synchronization and causal relationships of neuronal population activity; for second-level meso-level features, it analyzes the energy distribution and phase coupling of EEG rhythms; and for minute-level macro-features, it focuses on the dynamic reorganization patterns of functional networks. Feature organization adopts a tree-like structure, with top-level nodes representing global cognitive states, middle-level nodes corresponding to different cognitive sub-functions, and bottom-level nodes retaining detailed time-frequency characteristics. In the context of attention monitoring, this organizational structure can accurately characterize changes in a user's cognitive state. For example, during a 45-minute online course, the system can organize and record a student's instantaneous attention fluctuations (reflected by rapid changes in beta wave energy), sustained concentration levels (reflected by the degree of alpha wave suppression), and overall learning engagement (demonstrated by the coordinated changes in energy across different frequency bands). Through the systematic organization of features at multiple scales, a multi-level structured feature set is ultimately formed.

[0053] The resulting structured feature groups are then fused to generate elements. The system first evaluates the discriminative power of each feature in the feature group across different cognitive tasks, including micro-level neuronal firing patterns, meso-level rhythmic characteristics, and macro-level network topology. Then, the most representative feature combinations are selected using the information gain criterion, while maintaining a balance between features across different timescales. A hierarchical feature evaluation mechanism is designed to screen key feature indicators at each timescale. At the micro-level, the system focuses on instantaneous spectral characteristics and phase synchronization; at the meso-level, it extracts energy trends and modulation relationships between frequency bands; and at the macro-level, it emphasizes the stability of functional connectivity patterns. The system dynamically adjusts feature combinations based on the characteristics of different cognitive tasks. For example, in sustained attention tasks, the system prioritizes the dynamic balance between alpha wave suppression and theta wave enhancement; in working memory tasks, it prioritizes the relationship between gamma wave activity and memory load. These selected and combined features are then standardized, ultimately outputting a set of fused elements encompassing cognitive features across multiple timescales.

[0054] In some embodiments, generating a feature element set based on the fusion elements includes: performing feature screening based on the fusion elements to extract key elements; integrating the key elements into a feature system for classifying features based on the feature system; and outputting the feature element set based on feature classification results corresponding to the classified features.

[0055] When performing feature screening on the fused elements output from the above steps (which contain multi-dimensional feature information and feature data at different time scales), the system established a multi-level evaluation index system. This index system encompasses four dimensions: feature information content, stability, noise immunity, and discriminability, with multiple quantitative evaluation parameters assigned to each dimension. The system employs an entropy-based feature evaluation method to calculate the information gain ratio for each fused element and select feature subsets with low information redundancy. In the focus assessment task, the feature screening process focuses on significant features such as alpha wave suppression and beta wave enhancement in the temporal lobe. For example, when a user maintains focus during an important meeting, the system can accurately identify key indicators reflecting attention levels from complex time-frequency features, such as the energy ratio of prefrontal theta waves to temporal alpha waves and the strength of phase synchronization between different brain regions. Furthermore, the system can identify changes in the user's cognitive engagement when listening to key information, as evidenced by dynamic modulation of beta wave energy and transient increases in gamma wave activity. Through repeated verification and optimization, the system ultimately identified a set of highly discriminative key elements encompassing time-frequency, spatial, and cognitive features.

[0056] Based on the identified key elements, the system constructs a feature integration framework. This integration process employs a hierarchical organizational structure tailored to the properties of each key element, grouping time-frequency, spatial, and cognitive features according to functional attributes and timescales. For time-frequency features, the system establishes a multiscale representation, ranging from instantaneous states to long-term trends; for spatial features, a multilevel description, from local channels to global networks; and for cognitive features, a functional hierarchy, from basic perception to advanced cognition. During the feature integration process, the system incorporates an adaptive weighting mechanism to dynamically adjust the contribution of features to the overall system based on their importance and reliability. For example, in assessing sustained attention, the system focuses on the stability of alpha wave suppression and the periodic changes in theta wave activity, as these features often more accurately reflect the user's cognitive engagement. The system can flexibly adjust feature combinations based on the characteristics of different cognitive tasks. During focus training, if fluctuations in user attention are detected, the system automatically adjusts feature weights, increasing the weighting of short-term attention indicators. Through this dynamic integration strategy, the system ultimately forms an adaptive feature system encompassing multiscale representations of time-frequency features, multilevel descriptions of spatial features, and a functional hierarchy of cognitive features.

[0057] Feature classification is performed using a constructed adaptive feature system. Based on the established time scale, spatial hierarchy, and cognitive function framework within the adaptive feature system, a multi-level classification strategy is employed. Initial classification is performed based on the physical properties of the features, such as classifying time-domain features, frequency-domain features, and joint time-frequency features within the multi-scale representation of time-frequency features into different categories. Features are then further subdivided based on their functional significance, such as assigning attention-related features within the multi-level representation of spatial features and cognitive load features within the cognitive feature functional hierarchy into corresponding functional categories. The system comprehensively considers the correlation and complementarity between features during the classification process, optimizing the classification using cluster analysis. In the context of professional reading comprehension, the classification system accurately captures the user's cognitive processing characteristics. For example, when a user is reading professional literature, the system can simultaneously monitor sustained attention level (via alpha wave suppression within the multi-scale representation of time-frequency features), information processing load (via the theta / beta ratio within the multi-scale representation of time-frequency features), and depth of comprehension (via gamma wave activity patterns within the cognitive feature functional hierarchy), and automatically classify these features into corresponding cognitive function groups. Through this systematic classification process, a hierarchical feature classification system including physical attribute classification and functional attribute classification was finally formed.

[0058] Based on the established feature classification system, the system outputs standardized feature elements. For each classification level in the hierarchical feature classification system, the system has designed a standardized feature output format, including information such as the feature's basic attributes, numerical range, timestamp, and reliability indicators. For each feature category in the hierarchical feature classification system, the system generates a detailed description document, including the feature's physiological significance, calculation method, and typical application scenarios. In professional meeting focus assessments, the system can output a complete attention feature data package. When a user participates in an important two-hour meeting, the system can track and record the complete cognitive state transition: From 9:00 AM to 10:00 AM, the user maintains a high level of focus, as evidenced by sustained alpha wave suppression (10-12 Hz) and stable beta wave activity (15-18 Hz) in the physical attribute classification of the hierarchical feature classification system. From 10:00 AM to 10:30 AM, attention fluctuates, as evidenced by a brief increase in theta wave activity (4-7 Hz) and intermittent decreases in alpha wave suppression in the functional attribute classification of the hierarchical feature classification system. From 10:30 AM to 11:00 AM, attention gradually recovers after a short break, as evidenced by a gradual optimization of the ratio of prefrontal theta waves to temporal alpha waves in both the physical and functional attribute classifications. The system organizes these time-series feature data in a standardized format, forming a feature element dataset encompassing the complete cognitive process. Each feature in the hierarchical feature classification system is labeled with a clear functional category and reliability score. This refined feature extraction and organization scheme ultimately outputs a multidimensional feature element dataset encompassing both physical and cognitive features.

[0059] In some embodiments, obtaining the classification mark corresponding to the classification dimension includes: converting the classification dimension into a classification framework; aggregating the results according to the classification framework; and outputting the classification mark according to the result aggregation content corresponding to the aggregation result.

[0060] When performing classification and combination on the multidimensional feature datasets output by the above steps, the system first establishes a feature combination strategy. This combination process utilizes a hierarchical combination scheme based on the physical and cognitive attributes within the multidimensional feature dataset. At the bottom level, the system combines features within the multidimensional feature dataset that share the same timescale and similar spatial distribution. For example, the energy, phase, and time-varying characteristics of alpha waves within the same channel are combined into alpha activity features, and the energy ratio, phase coupling, and modulation characteristics of theta and beta waves are combined into rhythmic interaction features. At the middle level, functionally related features within the multidimensional feature dataset are integrated across different timescales. For example, millisecond-level gamma wave bursts, second-level alpha wave suppression, and minute-level theta wave trends are combined into attention features. At the top level, features across functional regions within the multidimensional feature dataset are integrated, such as integrating activity features from multiple brain regions, such as the temporal and parietal lobes, to form cognitive network features. Feature relevance thresholds and information redundancy constraints are set during the combination process to ensure that the combined features maintain information integrity while avoiding redundancy. In the focus assessment scenario, the system combines features such as alpha wave suppression, theta / beta energy ratio, and prefrontal-temporal functional connectivity strength from a multidimensional feature dataset to form a feature combination reflecting attention level. To adapt to the dynamic changes in different cognitive states, the system also establishes an adaptive adjustment mechanism for feature combinations, dynamically adjusting the combination strategy based on signal quality and task requirements. Through multi-level feature combination optimization, the system ultimately determined a set of classification dimensions including basic feature combinations, functional feature combinations, and network feature combinations.

[0061] Based on the identified classification dimensions, a systematic classification framework is constructed. This framework employs a multi-level structure for different types of feature combinations, encompassing a temporal dimension corresponding to basic feature combinations, a spatial dimension corresponding to network feature combinations, and a functional dimension corresponding to functional feature combinations. The temporal dimension reflects the dynamic characteristics of basic feature combinations and is divided into three time scales: transient response (1-100ms), short-term change (0.1-10s), and long-term trend (10-600s). Each time scale is assigned a corresponding feature extraction window and update period. The spatial dimension describes the spatial distribution characteristics of network feature combinations, including local activity (single-channel features), regional linkage (adjacent channel feature combinations), and global coordination (multi-channel network features). Detailed feature extraction ranges and spatial filtering parameters are defined for each spatial scale. The functional dimension characterizes the cognitive processing properties of functional feature combinations, encompassing basic perception (signal strength and stability), attention regulation (alpha wave suppression and beta wave enhancement), and higher-level cognition (gamma wave burst frequency and theta wave modulation depth). Each functional level is assigned a corresponding feature threshold and state discrimination criteria. In meeting engagement analysis, the system comprehensively characterizes changes in a user's cognitive state through combinations of basic, functional, and network features. For example, basic features reflect fluctuations in attention, as evidenced by the duration and recovery rate of alpha wave suppression; network features reveal activation patterns in attention networks, as evidenced by dynamic changes in the strength of prefrontal-temporal functional connectivity; and functional features reveal the depth of cognitive processing, as characterized by trends in the theta / beta ratio and patterns of gamma wave bursts. By constructing this refined classification framework, the system generates a standardized set of classifications.

[0062] Based on the established classification framework, the system aggregates results. This aggregation process uses a hierarchical approach. First, features are aggregated within the time dimension of the classification framework. This involves calculating statistics such as the temporal mean, variance, and skewness, extracting periodic patterns and mutational signatures. Next, features are aggregated within the spatial dimension of the classification framework, calculating spatial features such as inter-channel correlation coefficients, phase synchronization indices, and functional connectivity strength. Finally, features are aggregated within the functional dimension of the classification framework, calculating functional metrics such as the combined energy characteristics of different frequency bands and cross-band coupling. The aggregation process employs an adaptive weighting strategy, dynamically adjusting the weights of features across each dimension based on the signal-to-noise ratio, stability, and discriminability of the features within the classification framework. The system also establishes mapping relationships between dimensions within the classification framework, calculating statistics such as mutual information and conditional entropy between dimensions, and analyzing the interactions between dimensional features. In a professional conference focus analysis scenario, the system tracks changes in attendees' cognitive states in real time. While users participate in important meeting discussions, the system simultaneously monitors instantaneous cognitive load, sustained attention level, and information integration, and aggregates and integrates these features across the three dimensions of the classification framework to produce a multi-dimensional summary of results.

[0063] Based on the resulting multidimensional results, the system outputs classification tags. The tagging system adopts a multi-level structure, including basic state tags, abnormal pattern tags, and comprehensive evaluation tags in the multidimensional results summary. Basic state tags include specific quantitative indicators in the multidimensional results summary, such as alpha wave suppression (0-100%), theta / beta ratio (0.5-2.0), and prefrontal gamma wave burst frequency (0-10Hz). Abnormal pattern tags include state indicators such as attention fluctuation warning and cognitive fatigue warning in the multidimensional results summary. Each warning has corresponding trigger conditions and credibility assessment rules. Comprehensive evaluation tags integrate state information from multiple dimensions in the multidimensional results summary to form a holistic description of cognitive status. In the ongoing meeting engagement assessment, the system tracks cognitive state changes: At the beginning of the meeting, the multidimensional summary results show that alpha suppression remains above 80%, the theta / beta ratio stabilizes around 0.8, and the prefrontal gamma burst frequency is 8Hz, marking it as "highly focused." In the middle of the meeting, the multidimensional summary results show that alpha suppression drops to 60%, the theta / beta ratio rises to 1.2, and the gamma burst frequency drops to 5Hz, marking it as "fluctuating attention." In the latter part of the multidimensional summary results, alpha suppression falls below 40%, the theta / beta ratio exceeds 1.5, and the gamma burst frequency falls below 3Hz, marking it as "cognitive fatigue." The system generates these detailed classification labels for the multidimensional summary data in each time window, forming a complete cognitive state assessment sequence.

[0064] Step S104: perform credibility evaluation on the classification labels to establish an evaluation system, and output discrimination labels based on the evaluation system; perform data integration on the discrimination labels to obtain classification results, and complete multi-channel EEG signal feature extraction and classification.

[0065] Specifically, the system constructs a hierarchical evaluation system to assess the credibility of the multi-level classification labels (including basic state labels, abnormal pattern labels, and comprehensive evaluation labels) output from the above steps. This evaluation system quantifies the credibility of the multi-level classification labels along three dimensions: signal quality, feature stability, and state consistency. Signal quality assessment involves calculating the signal-to-noise ratio, performing artifact detection, and performing baseline drift analysis on the basic state labels within the multi-level classification labels, assigning a data quality score for each time window. Feature stability assessment involves performing temporal variance analysis, trend testing, and mutation detection on the abnormal pattern labels within the multi-level classification labels, assessing the temporal continuity of the features. State consistency assessment evaluates the internal consistency of the state judgment by calculating the mutual information and conditional entropy between the different dimensional features of the comprehensive evaluation labels within the multi-level classification labels. Specific evaluation metrics include: signal-to-noise ratio, artifact ratio, and baseline stability in the signal quality dimension; coefficient of variation, trend coefficient, and mutation index in the feature stability dimension; and mutual information, conditional entropy, and correlation coefficient in the state consistency dimension. In the concentration monitoring scenario, the system quantitatively assesses the credibility of each label within the multi-level classification labels. For example, when the system detects a "highly focused" state within the multi-level classification label, it simultaneously calculates the signal quality score of the alpha wave suppression feature, the energy stability index, and the consistency coefficient with other attention-related features to comprehensively form a credibility score for this state label. Through this multi-dimensional credibility assessment, the system has established a complete credibility assessment system.

[0066] Based on the established evaluation system, the system divides threshold intervals. Different threshold levels are set for the evaluation indicators of different dimensions in the credibility assessment system. The signal quality dimension is divided into three intervals: high quality (signal-to-noise ratio >20dB, artifact ratio <5%, baseline drift <1μV / min), medium quality (10-20dB, 5-15%, 1-3μV / min), and low quality (<10dB, >15%, >3μV / min). The feature stability dimension is divided into three levels: stable (coefficient of variation <0.1, trend coefficient <0.05), fluctuating (0.1-0.3, 0.05-0.15), and unstable (>0.3, >0.15). The state consistency dimension is divided into three levels: highly consistent (mutual information >0.8, conditional entropy <0.2), partially consistent (0.5-0.8, 0.2-0.5), and inconsistent (<0.5, >0.5). The system also incorporates an adaptive threshold adjustment mechanism, dynamically optimizing threshold parameters based on task type and environmental conditions. For meeting focus assessment, the system dynamically adjusts the thresholds within the credibility assessment system to accommodate the characteristics of different meeting phases. For example, during important presentations, the threshold for state consistency within the credibility assessment system is raised to ensure reliable state judgments. During breaks and discussions, the threshold for signal quality within the credibility assessment system is appropriately relaxed to accommodate greater levels of motion interference. Through this flexible threshold division strategy, the system forms a comprehensive multi-level threshold interval system.

[0067] In some embodiments, outputting the discrimination label according to the evaluation system includes: dividing the evaluation system into threshold intervals to output calibration results according to the threshold intervals; and outputting the discrimination label according to the calibration results.

[0068] The classification and labeling results are calibrated based on a multi-level threshold interval system. The calibration process employs a multi-validation strategy. First, calibration is performed within each of the three evaluation dimensions of the multi-level threshold interval system: signal quality, feature stability, and state consistency. The confidence level of each dimension is determined. Cross-validation is then performed across the dimensions of the multi-level threshold interval system to analyze the consistency of the evaluation results across these dimensions. Finally, a weighted fusion approach is used to obtain a comprehensive calibration result. The system incorporates a calibration compensation mechanism. When the evaluation metric of a dimension within the multi-level threshold interval system approaches the threshold limit (e.g., signal-to-noise ratio between 19-21dB), corrections are made by referencing the evaluation results of other dimensions within the multi-level threshold interval system. In the context of attention assessment, the calibration process can identify potential misclassifications. For example, when a sudden and significant drop in attention level is detected, the system comprehensively analyzes the calibration results of the three dimensions of signal quality, feature stability, and state consistency within the multi-level threshold interval system to determine whether the change is a true fluctuation in attention or a misclassification due to external interference. Through this rigorous calibration process, the system finally obtained a set of verified signal quality dimension scores, feature stability dimension scores and state consistency dimension scores, as well as a comprehensive rating that combines these three dimensions.

[0069] Based on the obtained signal quality, feature stability, and state consistency scores, as well as the comprehensive rating, the system generates and outputs discriminant labels. Label generation employs a hierarchical structure, including a basic state label based on the signal quality score, a credibility level label based on the feature stability score, and a comprehensive evaluation label based on the state consistency score and the comprehensive rating. The basic state label describes the specific cognitive state category; the credibility level label indicates the reliability level of the state judgment; and the comprehensive evaluation label combines state information and credibility information to provide a complete state description. In meeting focus monitoring, the discriminant labels output by the system contain rich state information. For example, when the system identifies a state of "high concentration," it generates the following labels: "State category: high concentration (alpha wave suppression >80%, theta / beta <0.8); Credibility: Grade A (Signal quality dimension score: 9.2 / 10, Feature stability dimension score: 0.95, State consistency dimension score: 0.92); Timestamp: 10:30:45; Duration: 15 minutes; Associated features: Prefrontal beta wave enhancement (+60%), temporal lobe alpha wave suppression (-75%)." The system generates such detailed discriminant labels for each time window of data, maintaining consistency in the label format, achieving an organic combination of accurate cognitive state identification and credibility assessment. These basic state labels, credibility level labels, and comprehensive assessment labels not only contain rich state information but also come with complete credibility assessment results, forming a comprehensive set of discriminant labels.

[0070] In some embodiments, performing data integration on the discriminant labels to obtain a classification result includes: performing data integration on the discriminant labels to construct an output structure; and obtaining the classification result according to the output structure.

[0071] When integrating the discriminant labels output from the above steps, the system establishes a multi-level integration strategy. The integration process first groups the discriminant labels by time series, setting a basic time window of 1 minute (for analyzing instantaneous state changes), a medium time window of 5 minutes (for analyzing state stability), and a long-term time window of 15 minutes (for analyzing state trends). For the basic time window, the system calculates the statistical distribution and transition probabilities of the basic state labels, such as the frequency and duration of the "highly focused" state. For the medium time window, the system analyzes the changing trends and stability characteristics of the basic state label sequence, including state retention and transition frequency. For the long-term time window, the system combines comprehensive evaluation labels to extract periodicity and long-range correlations in state evolution, such as the regularity of attention fluctuations. The system also integrates the credibility level labels of each time window to establish credibility-weighted state statistics. For the concentration assessment, the system performs multi-scale integration of continuous discriminant labels, such as calculating the weighted persistence ratio of the "highly focused" state within 15 minutes (weighted by the credibility level labels), the number of valid state transitions (excluding low-confidence fluctuations), and the average credibility level of the state discriminant obtained based on the comprehensive evaluation labels. Through this multi-scale data integration, the system constructs a hierarchical output structure that includes time series features (state duration, transition time), state features (distribution features of each state, transition features) and credibility features (credibility distribution at each time scale).

[0072] Based on the temporal, state, and credibility features in the constructed hierarchical output structure, the system devises a conversion strategy to transform the integrated data into a standardized categorized sequence. This conversion process employs a three-level sequence structure: the first-level sequence describes the temporal evolution of states based on temporal features, including state category, duration, transition moment, and instantaneous credibility, specifically recording each minute's state label and credibility score. The second-level sequence characterizes the stability and reliability of states based on state features, integrating metrics such as state distribution, average credibility, and state transition frequency over a 5-minute period. The third-level sequence characterizes the long-term evolution of states based on credibility features, including 15-minute trend indicators, periodicity parameters, and cumulative credibility. The system defines a standard data format and update rules based on credibility features for each level of sequence. In a meeting attention analysis, the first-level sequence uses temporal features to record each minute's attention state and its credibility; the second-level sequence uses state features to describe attention stability and credibility changes over a 5-minute period; and the third-level sequence uses credibility features to reflect the 15-minute attention trend and overall reliability. Through this hierarchical transformation strategy, the system generates a set of classification sequences containing short-term, medium-term and long-term state evolution characteristics.

[0073] The system integrates features based on the short-, medium-, and long-term state evolution characteristics of the generated classification sequence. This integration process employs a credibility-weighted adaptive mechanism, dynamically adjusting the weights of features in the feature integration based on the reliability and relevance of different levels within the classification sequence. For the first-level sequence within the classification sequence, the system extracts state transition timeliness metrics (e.g., state switching delay) and accuracy metrics (e.g., false positive rate), with weights determined by the instantaneous credibility of the classification sequence. For the second-level sequence within the classification sequence, it analyzes state stability metrics (e.g., fluctuation frequency) and predictability metrics (e.g., trend consistency), with weights determined by the average credibility of the classification sequence. For the third-level sequence within the classification sequence, it extracts long-term trends (e.g., fatigue accumulation rate) and regularity features (e.g., attention span), with weights determined by the cumulative credibility of the classification sequence. A feature conflict resolution mechanism is designed to arbitrate inconsistencies between features at different levels within the classification sequence by evaluating the credibility and contextual information of each level. In attention monitoring scenarios, the system can effectively integrate attention features from different time scales within the classification sequence. For example, when the first-level sequence in the classification sequence shows a sudden drop in attention, the system will combine the stability characteristics of the second-level sequence in the classification sequence and the trend characteristics of the third-level sequence in the classification sequence to comprehensively judge the nature and possible duration of this fluctuation. Through this multi-level feature integration, the system forms a feature integration that includes the dimensions of time, state, and credibility.

[0074] Exemplarily, obtaining the classification result according to the output structure includes: converting the output structure into a classification sequence, obtaining an integrated feature according to the classification sequence; and outputting the classification result according to the integrated feature.

[0075] Based on the resulting feature integration, the system outputs classification results. This classification generation fully utilizes information from various dimensions within the feature integration, including state classification (current state category, state duration, state stability score, key physiological indicators), trend analysis (short-term trend prediction, medium-term change direction, long-term evolution pattern), and credibility assessment (state credibility, trend credibility, overall assessment credibility). In the focus assessment application, the system's real-time classification output includes: state classification based on feature integration (attention level: 85%, duration: 10 minutes, stability: 0.92, alpha wave suppression: -75%, beta wave enhancement: +45%); trend analysis based on feature integration (short-term: stable, medium-term: slight decline (-5% / hour), long-term: periodic fluctuation (T = 90 minutes)); and credibility indicators based on feature integration (state: 0.95, trend: 0.88, overall: 0.91). The system generates normalized classification results for each time window and dynamically updates them based on the credibility scores from the feature integration. When a significant change in the state of feature integration is detected (e.g., a change in alpha wave suppression exceeding 20% and a confidence level greater than 0.9), the classification result is immediately updated. Through this dynamic classification mechanism based on feature integration, the system ultimately outputs classification results with high temporal resolution and high confidence.

[0076] The provided method has the following beneficial effects:

[0077] 1. An adaptive signal processing system was established. Through a multi-level acquisition mechanism and processing strategy, the inter-channel interference, myoelectric interference and environmental noise in the periauricular area were suppressed. On this basis, multi-scale feature decomposition and analysis domain division were realized, thereby improving the signal quality and stability while enhancing the ability to extract the time-frequency characteristics of EEG signals.

[0078] 2 A complete functional recognition framework was constructed, and the functional connectivity of the periauricular brain region was accurately characterized using spatial mapping technology and regional network structure. The feature organization structure was optimized through a multi-level feature fusion mechanism and adaptive screening strategy, significantly improving the accuracy of feature expression and system stability.

[0079] 3. A systematic classification evaluation mechanism has been formed, which organically combines classification combination optimization with multi-dimensional credibility evaluation. Through dynamic result integration strategy and real-time update mechanism, the accuracy and timeliness of classification labeling are improved, and the reliability and adaptability of the system in daily application environment are enhanced.

[0080] In order to implement the above method embodiment corresponding to the brain-computer intelligent headset multi-channel EEG signal feature extraction and classification method, to achieve the corresponding functions and technical effects. Figure 2 , Figure 2 The following is a block diagram of a multi-channel EEG signal feature extraction and classification device 200 for a brain-computer intelligent headset provided in an embodiment of the present application. For ease of explanation, only the parts relevant to this embodiment are shown. The multi-channel EEG signal feature extraction and classification device 200 for a brain-computer intelligent headset provided in an embodiment of the present application includes:

[0081] The distribution determination unit 201 is configured to collect multi-channel signals of the smart headset and determine channel distribution, obtain channel identifiers corresponding to the channel distribution, perform component extraction on the channel identifiers, divide the analysis domains, and output component features based on the analysis domains.

[0082] The network output unit 202 is configured to spatially locate the component features, obtain functional regions, construct a regional network based on the functional regions, and assign weights via the regional network; output a regional attribute data packet based on a weight assignment result corresponding to the assigned weights; establish a structural model based on the regional attribute data packet, and output a channel association network based on the structural model;

[0083] The tag acquisition unit 203 is configured to perform hierarchical construction on the channel association network, define a feature hierarchy, perform scale fusion on the feature hierarchy to form a feature group, output fusion elements based on the feature group, generate a feature element set based on the fusion elements, perform classification combination on the feature element set to determine a classification dimension, and obtain a classification tag corresponding to the classification dimension;

[0084] The classification completion unit 204 is used to perform credibility evaluation on the classification mark to establish an evaluation system and output a discrimination label according to the evaluation system; perform data integration on the discrimination label to obtain classification results and complete multi-channel EEG signal feature extraction and classification.

[0085] The above-mentioned brain-computer intelligent headset multi-channel EEG signal feature extraction and classification device 200 can implement the brain-computer intelligent headset multi-channel EEG signal feature extraction and classification method of the above-mentioned method embodiment. The optional options in the above-mentioned method embodiment are also applicable to this embodiment and will not be described in detail here. The rest of the contents of the embodiment of this application can refer to the contents of the above-mentioned method embodiment and will not be repeated in this embodiment.

[0086] Figure 3 This is a schematic diagram of the structure of a computer device provided in one embodiment of the present application. Figure 3 As shown, the computer device 3 of this embodiment includes: at least one processor 30 ( Figure 3Only 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, wherein the processor 30 implements the steps of any of the above method embodiments when executing the computer program 32.

[0087] The computer device 3 may be a computing device such as a smart phone, a tablet computer, a desktop computer, or a cloud server. The computer device may include but is not limited to a processor 30 and a memory 31. It will be understood by those skilled in the art that Figure 3 This is merely an example of the computer device 3 and does not constitute a limitation on the computer device 3 . The computer device 3 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device 3 may also include input and output devices, network access devices, etc.

[0088] The processor 30 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor.

[0089] In some embodiments, the memory 31 may be an internal storage unit of the computer device 3, such as a hard drive 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 drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the computer device 3. Furthermore, the memory 31 may include both an internal storage unit of the computer device 3 and an external storage device. The memory 31 is used to store an operating system, application programs, a boot loader, 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 about to be output.

[0090] In addition, an embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0091] An embodiment of the present application provides a computer program product. When the computer program product is run on a computer device, the computer device implements the steps in the above-mentioned various method embodiments when executing the computer program product.

[0092] In several embodiments provided in the present application, it is understood that each box in the flow chart or block diagram can represent a part of a module, program segment or code, and the part of the module, program segment or code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which depends on the functions involved.

[0093] If the 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 the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0094] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this application by those skilled in the art should be included within the scope of protection of this application.

Claims

1. A method for extracting and classifying multi-channel EEG signals from a brain-computer intelligent headset, characterized in that: include: Collect multi-channel signals of the smart headset and determine the channel distribution, and obtain channel identifiers corresponding to the channel distribution; Performing component extraction on the channel identifier and dividing the analysis domain, and outputting component features based on the analysis domain; spatially locating the component features to obtain functional regions, constructing a regional network based on the functional regions, and assigning weights via the regional network; outputting a regional attribute data packet based on a weight assignment result corresponding to the assigned weights, establishing a structural model based on the regional attribute data packet, and outputting a channel association network based on the structural model; Performing hierarchical construction on the channel association network to define a feature hierarchy, performing scale fusion on the feature hierarchy to form a feature group, outputting fusion elements based on the feature group, and generating a feature element set based on the fusion elements; performing classification combination on the feature element set to determine a classification dimension, and obtaining a classification label corresponding to the classification dimension; Performing credibility evaluation on the classification mark to establish an evaluation system, and outputting a discriminant label according to the evaluation system; The discriminant labels are integrated to obtain classification results, and multi-channel EEG signal feature extraction and classification are completed.

2. The method according to claim 1, characterized in that The obtaining of the channel identifier corresponding to the channel distribution includes: Perform signal separation and construct a filter chain according to the channel distribution; The channel identification is output according to the filter chain.

3. The method according to claim 1, characterized in that The outputting component features based on the analysis domain includes: Mapping the analysis domain into a transformation map and performing eigendecomposition; The component features are output based on the eigendecomposition.

4. The method according to claim 1, wherein The step of establishing a structural model according to the regional attribute data packet includes: Implementing spatial mapping on the regional attribute data packets to construct a topological distribution map; Detecting connected nodes based on the topological distribution graph; The structural model is established according to the connection nodes.

5. The method according to claim 1, wherein Generating a feature element set according to the fused elements includes: Perform feature screening based on the fusion elements to extract key elements; integrating the key elements into a feature system for classifying features according to the feature system; The feature element set is output based on the feature classification result corresponding to the classification feature.

6. The method according to claim 1, characterized in that The obtaining of the classification mark corresponding to the classification dimension includes: Converting the classification dimensions into a classification framework; Aggregate the results according to the described classification framework; Output a classification mark according to the result summary content corresponding to the summary result.

7. The method according to claim 1, characterized in that Outputting a discrimination label according to the evaluation system includes: Dividing the evaluation system into threshold intervals to output calibration results according to the threshold intervals; The discrimination label is output according to the calibration result.

8. The method according to claim 1, characterized in that The step of integrating the discriminant labels to obtain classification results includes: Performing data integration on the discriminant labels to construct an output structure; The classification result is obtained according to the output structure.

9. The method according to claim 8, characterized in that The obtaining the classification result according to the output structure includes: Converting the output structure into a classification sequence, and obtaining integrated features according to the classification sequence; The classification result is output according to the integrated features.

10. A computer device, characterized in that: The method comprises a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement the method according to any one of claims 1 to 8 when executing the computer program.

Images