Data processing method of brain wave glasses

Abstract

The invention discloses an electroencephalogram data processing method for electroencephalogram glasses, which comprises the following steps: acquiring original electroencephalogram signals acquired by the electroencephalogram glasses, synchronously acquiring multi-mode auxiliary signals such as eye movement, acceleration and myoelectricity, and monitoring and acquiring quality parameters in real time; forming an initial data set containing the original electroencephalogram signal, the multi-mode auxiliary signal and a quality mark; taking the initial data set as input, and outputting a processed data set containing pure electroencephalogram signals and neural activity features through preprocessing, multi-modal artifact separation and personalized feature extraction in sequence; and packaging the processed data set and processing process meta-information thereof into a standard format file, executing local structured storage and cloud synchronous storage, recording a data operation log through a block chain technology, ensuring data security through an encryption technology, and completing whole-process data management.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent devices and relates to a data processing method for brain wave glasses. Background Art

[0002] As a wearable bioelectrical signal acquisition device, EEG glasses acquire EEG signals non-invasively and have broad application prospects in areas such as concentration training, sleep monitoring, and emotion recognition. However, existing technologies still face multiple challenges in the entire process of EEG data processing: During the data collection phase, traditional methods have weak adaptability to environmental interference and wearing conditions, and the original signal is easily affected by noise, resulting in unstable data quality. At the same time, the lack of synchronous recording of multi-dimensional auxiliary information brings difficulties to subsequent processing. In the data processing link, traditional algorithms have limited capabilities for signal denoising and feature extraction in complex scenarios, making it difficult to accurately separate interference components or capture individual differences in neural activity characteristics, affecting the reliability of analysis results. In terms of data storage and management, existing solutions have problems such as inconsistent formats, disordered storage, and insufficient security, resulting in low data reuse efficiency and the risk of sensitive information leakage.

[0003] In summary, existing technologies have shortcomings in many aspects of the full-process processing of EEG data. There is an urgent need for a systematic optimization method to improve data quality, processing accuracy, and management security, and to promote the in-depth application of EEG glasses in actual scenarios. Summary of the Invention

[0004] The present invention aims to provide a data processing method for brainwave glasses. Through a full-process design of multimodal data synchronous acquisition, intelligent hierarchical processing and standardized collaborative storage, it can improve the quality, processing accuracy and management security of brainwave data, and promote its application in scenarios such as concentration training.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: A method for processing EEG data using EEG glasses, comprising the following steps: S1. Acquire the original EEG signals collected by the EEG glasses, simultaneously acquire multimodal auxiliary signals such as eye movement, acceleration, and myoelectricity, and monitor the acquisition quality parameters in real time to form an initial dataset containing the original EEG signals, multimodal auxiliary signals, and quality markers; S2. Taking the initial dataset as input, the system sequentially performs preprocessing, multimodal artifact separation, and personalized feature extraction, outputting a processed dataset containing clean EEG signals and neural activity features. S3. Encapsulate the processed data set and its processing metadata into a standardized format file, perform local structured storage and synchronize cloud storage, and use blockchain technology to record data operation logs and encryption technology to ensure data security, completing the entire data management process.

[0006] Furthermore, the acquisition of the original EEG signal collected by the EEG glasses comprises the following steps: S11. Send microcurrent to the scalp through the dry electrodes of the EEG glasses, and measure the contact impedance of each electrode in real time; S12. If the contact impedance of an electrode exceeds a preset threshold, the user interface prompts the wearer to adjust the wearing position. If the impedance still does not meet the threshold after adjustment, signal interpolation compensation is enabled for adjacent electrodes, and the signal weight of the electrode exceeding the threshold is reduced. S13. The analog EEG signal after contact quality optimization is differentially amplified by the signal amplification module, and the interference is reduced by the common-mode interference suppression module. It is then converted into a digital signal by the analog-to-digital conversion module to obtain the original EEG signal.

[0007] Furthermore, the synchronous acquisition of multimodal auxiliary signals such as eye movement, acceleration, and electromyography includes the following steps: S14. Synchronously activate the eye movement detection module to collect eye movement signals and record eye movement events; Synchronously activate the acceleration detection module to collect head movement acceleration signals and record severe shaking events that exceed the preset acceleration threshold; Synchronously activate the electromyography detection module to collect electromyographic signals and record muscle contraction events; S15. Add an event marker to the time point of each multimodal auxiliary signal, with the marking accuracy being a preset accuracy.

[0008] Furthermore, the real-time monitoring and acquisition of quality parameters to form an initial data set including original EEG signals, multimodal auxiliary signals and quality markers includes the following steps: S16. Perform spectrum analysis on the original EEG signal to identify the ambient noise frequency band and mark periods of strong noise exceeding a preset noise amplitude threshold; S17. Recording the contact impedance value, signal amplitude and other quality parameters of each electrode; S18. Integrate the original EEG signals, multimodal auxiliary signals, and quality tags into an initial dataset, where each signal and quality tag contains timestamp information.

[0009] Furthermore, the method of taking the initial data set as input, sequentially performing preprocessing, multimodal artifact separation, and personalized feature extraction to output a processed data set containing pure EEG signals and neural activity features includes the following steps: S21. Apply bandpass filtering to the raw EEG signals in the initial dataset to remove low-frequency drift and high-frequency noise, and apply notch filtering to suppress ambient noise in specific frequency bands. If the scenario is event-related potential analysis, convert the reference electrode to a whole-brain average reference to obtain a preliminary denoised EEG signal. S22. Locate the artifact periods in the preliminary denoised EEG signals based on the multimodal auxiliary signals in the initial dataset; input the located EEG signals into a pre-trained deep learning model, and separate the pure EEG signals and artifact components through feature learning; verify the separation effect, and if it does not meet the standards, adjust the model parameters and re-separate, and finally output the pure EEG signals.

[0010] Furthermore, the method of taking the initial data set as input, sequentially performing preprocessing, multimodal artifact separation, and personalized feature extraction to output a processed data set containing pure EEG signals and neural activity features also includes the following steps: S23. Using the pre-stimulation preset duration data as the baseline, calculate the baseline mean and correct the post-stimulation data of the pure EEG signal to obtain a baseline-corrected EEG signal; S24. Detecting ERP components in the baseline-corrected EEG signal using a time series model. Dynamically adjusting model detection parameters based on user historical data to output ERP features. S25. Extract the power proportion of a preset frequency band from the baseline-corrected EEG signal through time-frequency analysis to reflect the user's neural activity state and output the frequency band power characteristics; S26. Calculate the phase-locking values ​​of multiple channels in the baseline-corrected EEG signal, evaluate the synchronization of neural activity in different brain regions, and output neural synchronization characteristics; S27. Integrate ERP features, frequency band power features, and neural synchronization features into neural activity features, and together with the pure EEG signals, form a processed data set.

[0011] Furthermore, encapsulating the processed data set and its processing metadata into a standardized format file includes the following steps: S31. Integrate the processed data set and the processing meta-information into data to be packaged; S32. Use the extended EDF+ format compatible with mainstream EEG analysis tools to define the file header information and data segment structure; S33. Encode the data to be encapsulated according to the specifications of the extended EDF+ format to generate a standardized format file containing a timestamp, event marker and complete meta-information.

[0012] Furthermore, the execution of local structured storage and cloud synchronization storage includes the following steps: S34. Create a local storage directory by the hierarchical structure of "user ID / collection date / task type"; S35. Store the generated standardized format file in the corresponding directory; S36. Generate an index file in the storage directory, recording the file name, file size, neural activity feature summary, and whether it contains artifacts.

[0013] Furthermore, the data operation log is recorded by blockchain technology, encryption technology is used to ensure data security, and the whole process of data management is completed, including the following steps: S37. Synchronize locally stored standardized format files to the cloud server via a secure transmission protocol; S38. In cloud servers, blockchain technology is used to record data operation events, including "upload," "download," and "modify" events. Each log entry contains the operation time, user ID, and file hash value, and the log content cannot be tampered with. S39. Desensitize user metadata in standardized format files and encrypt sensitive data using encryption algorithms.

[0014] The beneficial effects of the present invention are as follows: The present invention forms an initial data set through synchronous multimodal auxiliary signal acquisition and real-time quality monitoring, ensuring the reliability and traceability of the original data; based on the initial data set, preprocessing, multimodal artifact separation, and personalized feature extraction are sequentially performed to output high-quality pure EEG signals and neural activity features, improving signal processing accuracy; the processed data and metadata are encapsulated into standardized files, local structured storage is synchronized with the cloud, and data security is ensured through blockchain logs and encryption technology, optimizing the efficiency of the entire process management. The overall solution significantly improves the quality, analysis reliability, and management security of EEG glasses data through the collaborative design of acquisition, processing, and storage links, providing technical support for applications in scenarios such as concentration training. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the steps of the method of the present invention.

[0016] Figure 2 It is a detailed schematic diagram of step S1 of the present invention.

[0017] Figure 3 It is a detailed schematic diagram of step S2 of the present invention.

[0018] Figure 4 It is a detailed schematic diagram of step S3 of the present invention. DETAILED DESCRIPTION

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as commonly understood by those skilled in the art to which this invention belongs. The terms used in the specification of the application are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the accompanying drawings are intended to cover non-exclusive inclusions. The terms "first" and "second" in the specification and claims of the present invention and the accompanying drawings are used to distinguish different objects, not to describe a specific order.

[0020] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0021] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0022] The present invention provides Figures 1 to 4 In an embodiment of the present invention, a data processing method for brainwave glasses achieves the purpose of improving the quality, processing accuracy and management security of brainwave data through a full-process design of multimodal data synchronous acquisition, intelligent hierarchical processing and standardized collaborative storage, and promotes its application in scenarios such as concentration training. Specific embodiment 1 A method for processing EEG data using EEG glasses, comprising the following steps: S1. Acquire the original EEG signals collected by the EEG glasses, simultaneously acquire multimodal auxiliary signals such as eye movement, acceleration, and myoelectricity, and monitor the acquisition quality parameters in real time to form an initial dataset containing the original EEG signals, multimodal auxiliary signals, and quality markers; After the user puts on the EEG glasses, the data acquisition function is activated. The system obtains the original EEG signals through dry electrodes, and simultaneously activates eye movement, acceleration, electromyography and other sensors to collect multimodal auxiliary signals (such as eye movement events, head shaking events, muscle contraction events), and monitors the quality parameters of the acquisition process in real time (such as electrode contact status, environmental noise). Finally, the original EEG signals, multimodal auxiliary signals and quality marks are aligned and integrated according to timestamps to form the initial data set.

[0024] In this embodiment, through synchronous multimodal signal acquisition and quality monitoring, the initial data set completely retains the original EEG signals and interference source information (such as blinking and shaking events), providing accurate time correlation basis for subsequent artifact separation. At the same time, quality marking ensures the traceability of the data acquisition process and improves the reliability of the original data.

[0025] S2. Taking the initial dataset as input, the system sequentially performs preprocessing, multimodal artifact separation, and personalized feature extraction, outputting a processed dataset containing clean EEG signals and neural activity features. Taking the initial data set as input, the system sequentially performs preprocessing (filtering out irrelevant noise and adjusting the reference electrode), multimodal artifact separation (locating and separating interference artifacts based on auxiliary signals), and personalized feature extraction (extracting neural activity features by combining user historical data), and finally outputs a processed data set containing pure EEG signals and neural activity features.

[0026] In this embodiment, the preprocessing and artifact separation of the present invention effectively removes environmental noise and motion artifacts, thereby improving the signal-to-noise ratio of the EEG signal; personalized feature extraction is combined with user historical data to dynamically adjust parameters, so that the detection of neural activity characteristics (such as the P300 wave related to concentration) is more in line with individual differences of users, thereby improving the accuracy of feature analysis.

[0027] S3. Encapsulate the processed data set and its processing metadata into a standardized format file, perform local structured storage and synchronize cloud storage, and use blockchain technology to record data operation logs and encryption technology to ensure data security, completing the entire data management process.

[0028] The processed data set and its processing metadata (such as filtering parameters, artifact separation records, and feature extraction results) are encapsulated into standardized format files and stored locally according to the hierarchical structure of "user ID / acquisition date / task type", and an index file is generated to record key file information (such as file name and feature summary). At the same time, the file is synchronized to the cloud through a security protocol, and blockchain technology is used to record data operation logs to ensure traceability. User metadata is desensitized and sensitive data is encrypted to complete the entire data management process.

[0029] In this embodiment, standardized format files are compatible with mainstream analysis tools, lowering the threshold for data reuse; local hierarchical storage and index files significantly improve data retrieval efficiency; cloud synchronization combines blockchain and encryption technology to ensure the security of long-term data storage and ensure the traceability of the entire life cycle of data through operation logs, meeting the requirements for data rigor in medical, scientific research and other scenarios. Specific embodiment 2 In this implementation, the EEG glasses use flexible comb-shaped dry electrodes with a contact area of ​​15 mm², a built-in microcurrent generator with an output current that can be set to ≤10 μA, a signal amplification module, a 24-bit analog-to-digital converter, and a matching user interface that can be a mobile phone APP display.

[0031] The method of obtaining the original EEG signal collected by the EEG glasses comprises the following steps: S11. Send microcurrent to the scalp through the dry electrodes of the EEG glasses, and measure the contact impedance of each electrode in real time; Specifically, after the user wears the EEG glasses, the system sends a microcurrent to the scalp via dry electrodes. The microcurrent parameters can be set to 5μA intensity and 1kHz frequency. The system then monitors the contact impedance between each electrode and the scalp in real time, with a detection frequency of once per second. For example, if the impedance of the left temporal lobe electrode is detected to be 25kΩ and the impedance of the right parietal lobe electrode is 18kΩ, the preset threshold is 20kΩ.

[0032] S12. If the contact impedance of an electrode exceeds a preset threshold, the user interface prompts the wearer to adjust the wearing position. If the impedance still does not meet the threshold after adjustment, signal interpolation compensation is enabled for adjacent electrodes, and the signal weight of the electrode exceeding the threshold is reduced. Specifically, if the impedance of an electrode exceeds a preset threshold (e.g., 25kΩ > 20kΩ for the left temporal lobe electrode), the system prompts the user through the display: "Poor contact of the left temporal lobe electrode, please gently press to adjust"; After the user adjusts the impedance of the electrode, the impedance is still 23kΩ (still exceeds the threshold) when re-detected. The system uses the signal of the adjacent left parietal electrode (impedance 18kΩ) for linear interpolation compensation. The interpolation compensation formula is: ; and reduce the signal weight of the left temporal lobe electrode from 100% to 30%, so the final signal is .

[0033] S13. The analog EEG signal after contact quality optimization is differentially amplified by the signal amplification module, and the interference is reduced by the common-mode interference suppression module. It is then converted into a digital signal by the analog-to-digital conversion module to obtain the original EEG signal.

[0034] Specifically, the analog EEG signal (amplitude 10-100μV) after contact quality optimization is input into the signal amplification module, and the signal amplitude is increased to 80-800μV through differential amplification (gain × 8); in conjunction with the common-mode interference suppression module (suppressing 50Hz power frequency noise), the common-mode noise amplitude is reduced from 100μV to below 2μV; the amplified signal is converted into a digital signal (16-bit signed integer representation) at a sampling rate of 250SPS through a 24-bit analog-to-digital converter (resolution 0.1μV), and finally the original EEG signal with a timestamp is obtained (the format is binary stream + timestamp).

[0035] In this embodiment, the present application effectively solves the problem of signal distortion caused by loose wearing of traditional dry electrodes through real-time detection and dynamic adjustment of contact impedance. The interpolation compensation mechanism ensures that valid data can still be obtained when a single channel fails; differential amplification and common-mode suppression technology significantly reduces the interference of environmental noise on weak EEG signals, and the data resolution after analog-to-digital conversion meets the requirements of capturing μV-level neural activity, providing high-quality original input for subsequent processing. Specific embodiment three The synchronous acquisition of multimodal auxiliary signals such as eye movement, acceleration, and myoelectricity includes the following steps: S14. Synchronously activate the eye movement detection module to collect eye movement signals and record eye movement events; Specifically, a camera embedded in or mounted on the glasses tracks the user's pupil position (sampling rate 120Hz) and outputs electrooculogram (EOG) signals (amplitude 50-300μV). When rapid pupil closure (duration ≤ 100ms) is detected, it is identified as a "blink event" and recorded as the event type.

[0037] Synchronously activate the acceleration detection module to collect head movement acceleration signals and record severe shaking events that exceed the preset acceleration threshold; Specifically, the accelerometer collects X / Y / Z three-axis acceleration in real time (sampling rate 50Hz) and calculates the composite acceleration value. When the composite acceleration value is ≥0.1g (g is the acceleration due to gravity, 9.8m / s²), it is judged as a "severe shaking event" and the event type is recorded as "head shaking".

[0038] Synchronously activate the electromyography detection module to collect electromyographic signals and record muscle contraction events; Specifically, the electromyographic sensor collects bioelectrical signals (EMG, amplitude 100-1000μV, sampling rate 1000Hz) from facial muscles (such as the corrugator supercilii). When the signal amplitude exceeds 500μV and the duration is ≥50ms, it is identified as a "muscle contraction event" and recorded as the event type "frowning."

[0039] Furthermore, basic data collection is also included. The ambient light sensor is activated to collect light intensity (lux), the posture sensor detects sitting angle (°), the wearing timer records duration (minutes), the infrared sensor measures distance (cm), the motion accelerometer counts outdoor steps, and the microphone monitors ambient noise decibels (dB). This data is saved once a minute, and EEG signals are activated simultaneously. Four dry electrodes distributed across the frame (Fp1 / Fp2 channels on the forehead and T3 / T4 channels on the temporal lobe) collect raw EEG signals related to concentration, fatigue, and mood in real time at a certain sampling rate. Both types of data are synchronously transmitted to the cloud backend via wireless network.

[0040] S15. Add an event marker to the time point of each multimodal auxiliary signal, with the marking accuracy being a preset accuracy.

[0041] Specifically, the system adds an event marker to the signal time point of each sensor based on a unified timestamp (provided by the main controller clock with an accuracy of ±1ms): Electrooculogram (EOG): Add the marker "blink-10:00:05.123" to the time point where the "blink event" occurs (e.g., T=10:00:05.123s); Acceleration signal: Add the marker "head shaking - 10:00:07.456" at the time point where the "severe shaking event" occurs (e.g., T = 10:00:07.456s); EMG signal: Add the tag "Frown-10:00:09.876" to the time point where the "muscle contraction event" occurs (for example, T=10:00:09.876s).

[0042] In this embodiment, the synchronous acquisition of multimodal auxiliary signals provides a "time anchor point" for the location of artifacts in EEG signals (such as the temporal correlation between blink events and EEG artifacts), and the high precision of event marking (±1ms) ensures the accurate location of the artifact period; the event detection mechanism of electromyography and acceleration signals (such as amplitude threshold and duration) improves the reliability of interference source identification, provides rich auxiliary information for subsequent artifact separation, and significantly improves the efficiency and accuracy of artifact removal. Specific embodiment 4 The real-time monitoring and acquisition of quality parameters to form an initial data set including original EEG signals, multimodal auxiliary signals and quality markers includes the following steps: S16. Perform spectrum analysis on the original EEG signal to identify the ambient noise frequency band and mark periods of strong noise exceeding a preset noise amplitude threshold; Specifically, the system performs real-time spectrum analysis on the original EEG signal (a digital signal that has undergone contact optimization and analog-to-digital conversion through the second embodiment, with a sampling rate of 250 Hz): The analysis method uses fast Fourier transform (FFT) to calculate the power distribution in the 0-100 Hz frequency band, with a window length of 1 second (covering 250 sampling points), and a Hanning window to reduce spectral leakage; Noise identification uses a preset noise amplitude threshold of 5μV² (i.e., a frequency band's power density ≥ 5μV² is considered strong noise). For example, if the power density detected in the 50Hz band is 7μV² (exceeding the threshold) between 10:00:00 and 10:00:01, this period is marked as "50Hz strong noise period." If the power density detected in the 60Hz band is 6μV² (assuming a 60Hz power grid environment), this period is also marked as "60Hz strong noise period."

[0044] S17. Recording the contact impedance value, signal amplitude and other quality parameters of each electrode; The system records the following quality parameters in real time (summarized every 0.5 seconds and stored in a temporary cache): Contact impedance value: the test results of each electrode in claim 2 (e.g., 23 kΩ for the left temporal lobe electrode and 18 kΩ for the right parietal lobe electrode); Signal amplitude: the peak-to-peak value of the original EEG signal (e.g., the amplitude range at 10:00:00 is 15-90 μV); Sensor status: The acquisition status of multimodal auxiliary signals (e.g., eye tracker "normal operation" and accelerometer "stable data").

[0045] S18. Integrate the original EEG signals, multimodal auxiliary signals, and quality tags into an initial dataset, where each signal and quality tag contains timestamp information.

[0046] The system aligns the timestamps (with a timestamp accuracy of ±1ms) of the original EEG signals, the multimodal auxiliary signals (e.g., electrooculography, acceleration, and electromyography) collected in Specific Example 3, and the quality markers of steps S16-S17 (strong noise period, contact impedance exceeding standard marker, and event marker), and integrates them into a structured initial data set.

[0047] The data integration method is as follows: Original EEG signals: Arranged in chronological order, each sampling point contains a timestamp (e.g., T=10:00:00.004s, T=10:00:00.008s) and the corresponding amplitude (e.g., 15μV, 18μV); Multimodal auxiliary signals: The "blink event" marker of the electrooculogram signal (e.g., "blink - 10:00:05.123s"), the "head shake event" marker of the acceleration signal (e.g., "head shake - 10:00:07.456s"), and the "frown event" marker of the electromyography signal (e.g., "frown - 10:00:09.876s") are all associated with the signal amplitude at the corresponding time point; Quality marking: strong noise periods (such as "50 Hz strong noise - 10:00:00-10:00:01"), electrodes with excessive contact impedance (such as "left temporal lobe - 23 kΩ"), and signal amplitude ranges (such as "15-90 μV") are separately summarized in a quality information list.

[0048] Furthermore, the initial data set also includes minute-by-minute records of ambient light intensity (lux), sitting inclination angle (°), wearing time (minutes), infrared detection distance (cm), number of steps in outdoor activities, and ambient noise decibel value (dB).

[0049] In this embodiment, spectrum analysis and quality parameter recording realize "full-link monitoring" of the acquisition process. The marking of strong noise periods provides a direct basis for notch filtering in the preprocessing stage, avoiding the accidental deletion of valid signals by fixed parameter filtering; the recording of quality parameters (such as contact impedance and signal amplitude) provides quantitative indicators for data quality assessment (such as determining whether re-acquisition is necessary); the integration method of timestamp alignment ensures the temporal correlation between each signal and the marker, laying the foundation for subsequent multimodal collaborative processing (such as artifact separation based on event markers). Specific embodiment five The method takes the initial dataset as input, sequentially performs preprocessing, multimodal artifact separation, and personalized feature extraction, and outputs a processed dataset containing pure EEG signals and neural activity features, including the following steps: S21. Apply bandpass filtering to the raw EEG signals in the initial dataset to remove low-frequency drift and high-frequency noise, and apply notch filtering to suppress ambient noise in specific frequency bands. If the scenario is event-related potential analysis, convert the reference electrode to a whole-brain average reference to obtain a preliminary denoised EEG signal. Specifically, the initial data set (including the original EEG signal, the multimodal auxiliary signal and the quality mark) output by the specific embodiment 4 is used as input, and the following operations are performed: 1. Bandpass filtering removes low-frequency drift and high-frequency noise: A second-order Butterworth bandpass filter was applied to the raw EEG signal, with a low-cut frequency of 0.3 Hz (to filter out DC drift and very low-frequency noise) and a high-cut frequency of 30 Hz (to filter out myoelectric noise above 30 Hz). For example, drift components below 1 Hz (amplitude 20 μV) in the raw EEG signal were reduced to below 2 μV after filtering, and myoelectric noise at 35 Hz (amplitude 50 μV) was reduced to below 5 μV after filtering.

[0051] 2. Notch filtering suppresses ambient noise in specific frequency bands: Based on the "strong noise periods" marked in the initial dataset (e.g., periods with strong noise at 50 Hz), a 50 Hz notch filter (quality factor Q = 30) was applied to suppress the power-frequency noise. For example, the noise amplitude in the 50 Hz band was reduced from 10 μV to below 1 μV, while retaining valid EEG signals below 30 Hz (e.g., alpha waves between 8 and 13 Hz).

[0052] 3. Reference electrode conversion (event-related potential analysis scenario): If the current scenario is ERP analysis (such as a concentration test), convert the default reference electrode (FCz) to a whole-brain average reference (calculate the mean of all eight EEG channels as the new reference). For example, when the original signal is referenced to FCz, the Pz channel amplitude is 50μV. After converting to a whole-brain average reference (the whole-brain mean is 10μV), the Pz channel amplitude is adjusted to 40μV to eliminate individual differences in the reference electrode.

[0053] After the above processing, the EEG signal with preliminary denoising is obtained.

[0054] S22. Locate the artifact periods in the preliminary denoised EEG signals based on the multimodal auxiliary signals in the initial dataset; input the located EEG signals into a pre-trained deep learning model, and separate the pure EEG signals and artifact components through feature learning; verify the separation effect, and if it does not meet the standards, adjust the model parameters and re-separate, and finally output the pure EEG signals.

[0055] Specifically, based on the multimodal auxiliary signals (e.g., electrooculogram, acceleration event markers) in the initial dataset, artifact separation is performed on the preliminarily denoised EEG signal: 1. Artifact time period positioning: Based on the "blink event" mark of the EOG signal (such as "blink-10:00:05.123s"), the corresponding artifact period in the EEG signal (event time ±100ms, i.e., 10:00:05.023s to 10:00:05.223s) was located. Similarly, based on the "head shake event" mark of the acceleration signal (such as "head shake-10:00:07.456s"), the period from 10:00:07.356s to 10:00:07.556s in the EEG signal was located as the motion artifact area.

[0056] 2. Deep learning model separation artifacts: The localized EEG signal (with a 500ms window length, including artifact periods) is fed into a pre-trained Transformer model. The model uses a self-attention mechanism to learn artifact features (such as high-frequency components corresponding to blinks and low-frequency drift corresponding to head shaking) and outputs a clean EEG signal and artifact components. For example, given an EEG signal containing blink artifacts (amplitude 50μV, frequency 20Hz), the model separates the output into a clean signal (amplitude 30μV, frequency 10Hz, corresponding to alpha waves) and an artifact component (amplitude 20μV, frequency 20Hz, corresponding to blink interference).

[0057] 3. Separation effect verification and adjustment: Verify whether the alpha wave (8-13 Hz) power of the pure EEG signal is restored to the normal range (10-20 μV²). If the alpha wave power of a certain period is only 5 μV² (lower than the threshold), determine that the separation is not up to standard, adjust the number of attention heads of the model (such as from 4 to 8) or the learning rate (such as from 0.001 to 0.0005), re-separate the signal of the period, until the alpha wave power is restored to 12 μV² (up to standard).

[0058] In this embodiment, the band-pass filtering and notch filtering in the preprocessing stage effectively remove low-frequency drift, high-frequency electromyogram and power frequency noise, and the reference electrode conversion eliminates the interference of individual differences on ERP analysis, significantly improving the signal-to-noise ratio (SNR) of the signal; the artifact positioning of the multi-modal auxiliary signal combined with the feature learning of the deep learning model realizes the accurate separation of artifacts in a dynamic scene, and the verification mechanism ensures the reliability of the separation effect, providing high-purity EEG signals for subsequent feature extraction. Specific embodiment six The initial data set is input, and the pre-processing, multi-modal artifact separation, and personalized feature extraction are sequentially performed, and the processed data set containing pure EEG signals and neural activity features is output, which further includes the following steps: S23. Calculate the baseline mean based on the pre-stimulus baseline data for a predetermined time length, and correct the post-stimulus data of the pure EEG signal to obtain the baseline-corrected EEG signal; Specifically, the pure EEG signal (artifact separation completed) output by the specific embodiment five is input, and baseline correction is performed, wherein: Baseline selection: select the EEG data 100 ms before stimulation (no task state, such as user resting state) as the baseline, and calculate the amplitude mean (μ = 12 μV) of all sampling points in this period; Correction calculation: subtract the baseline mean from the amplitude of each sampling point in the EEG data 600 ms after stimulation (such as the period when the user sees the math problem, time range T = 10:00:10.000s to T = 10:00:10.600s) (corrected amplitude = original amplitude - μ). For example, a sampling point with an original amplitude of 25 μV is corrected to 13 μV (25 μV - 12 μV), eliminating the interference of resting state background activity.

[0060] S24. Detect the ERP component in the baseline-corrected EEG signal through a time series model, dynamically adjust the model detection parameters based on user historical data, and output the ERP feature; Specifically, the LSTM model is used to detect the ERP component in the baseline-corrected EEG signal (taking P300 wave as an example), wherein: Dynamic adjustment of model parameters: Based on the user's historical data (e.g., the average latency of the P300 wave in the user's past five tests was 320ms and the average amplitude was 15μV), the model detection time window was set to 250-400ms (covering the average latency ± 70ms) and the amplitude threshold was set to 10μV (lower than the average amplitude to improve detection sensitivity); Detection Process: The baseline-corrected signal (time window 250-400ms, corresponding to T = 10:00:10.250s to T = 10:00:10.400s) was fed into the LSTM model. The model output prediction result: a peak with an amplitude of 18μV was detected at T = 10:00:10.315s, identified as the P300 wave. Its latency (315ms) and amplitude (18μV) were recorded as ERP features.

[0061] S25. Extract the power proportion of a preset frequency band from the baseline-corrected EEG signal through time-frequency analysis to reflect the user's neural activity state and output the frequency band power characteristics; Specifically, the baseline-corrected EEG signals were analyzed by wavelet transform to extract the power proportions of the δ (0.5-4 Hz), α (8-13 Hz), and β (13-30 Hz) frequency bands, where: Wavelet transform parameters: Morlet wavelet (center frequency f0 = 5 Hz, bandwidth parameter σ = 2) is used to perform time-frequency decomposition of the signal and obtain the frequency band power at each time point; Power calculation: Calculate the total power of each frequency band within 600ms after stimulation (e.g., total power of delta band = 50μV², alpha band = 120μV², beta band = 80μV²). The sum of the total power is 250μV². Percentage output: δ = 50 / 250 = 20%, α = 120 / 250 = 48%, β = 80 / 250 = 32%. Based on the empirical threshold of α > 30%, the user is considered "relaxed" and recorded as the frequency band power feature.

[0062] S26. Calculate the phase-locking values ​​of multiple channels in the baseline-corrected EEG signal, evaluate the synchronization of neural activity in different brain regions, and output neural synchronization characteristics; Specifically, the phase locking values ​​of multiple channels (such as the frontal lobe Fz channel and the parietal lobe Pz channel) in the baseline-corrected signal are calculated to evaluate the synchronization of neural activity, where: Phase extraction: Perform Hilbert transform on the signals of the Fz and Pz channels respectively to extract the instantaneous phase (φ_Fz and φ_Pz); PLV calculation: Calculate the phase difference (Δφ = φ_Fz - φ_Pz) for all time points within 600ms after stimulation. PLV = |(1 / N) × Σe^(iΔφ)| (where N is the number of time points). For example, a PLV of 0.7 (a synchronization threshold of ≥ 0.6) indicates high synchronization between prefrontal and parietal neural activity, which is recorded as a neural synchronization feature.

[0063] S27. Integrate ERP features, frequency band power features, and neural synchronization features into neural activity features, and together with the pure EEG signals, form a processed data set.

[0064] Specifically, ERP features (P300 latency 315ms, amplitude 18μV), frequency band power features (α accounted for 48%, "relaxed state"), and neural synchronization features (PLV=0.7, "high synchronization") were integrated into neural activity features and together with pure EEG signals (including timestamps and amplitudes) formed the processed data set.

[0065] In this embodiment, baseline correction eliminates the interference of resting-state background activity on post-stimulation signals, thereby improving the detection sensitivity of ERP components; the timing model dynamically adjusts parameters (such as detection time window and amplitude threshold) based on user historical data, solving the problem of missed detection or misjudgment caused by individual differences in general models (the missed detection rate is reduced from 15% to 5%); time-frequency analysis and neural synchronization calculation quantify neural activity characteristics (such as concentration and relaxation state) from multiple dimensions. The integrated feature set provides comprehensive data support for applications such as emotion recognition and cognitive assessment. Specific embodiment seven The step of encapsulating the processed data set and its processing metadata into a standardized format file includes the following steps: S31. Integrate the processed data set and the processing meta-information into data to be packaged; Specifically, the processed data set (including clean EEG signals and neural activity features) and processing process metadata outputted in the sixth embodiment are used as input to perform data integration, wherein: Contents of the processed dataset: 1. Clean EEG signal: timestamp sequence (e.g., T=10:00:00.004s, T=10:00:00.008s) and corresponding amplitude (e.g., 13μV, 15μV); 2. Neural activity characteristics: ERP characteristics (P300 latency 315ms, amplitude 18μV), frequency band power characteristics (α accounts for 48%), and neural synchronization characteristics (PLV=0.7).

[0067] Processing process meta information content: 1. Preprocessing parameters: band-pass filter low cut 0.3 Hz, high cut 30 Hz, notch filter 50 Hz (Q=30); 2. Artifact separation records: use the Transformer model to separate the period 10:00:05.023s to 10:00:05.223s (eye blink artifact area), and the alpha wave power is 12μV² after separation (up to standard); 3. Feature extraction parameters: baseline length 100ms (mean 12μV), LSTM model detection time window 250-400ms, wavelet transform uses Morlet wavelet (f0=5Hz).

[0068] S32. Use the extended EDF+ format compatible with mainstream EEG analysis tools to define the file header information and data segment structure; Specifically, the extended EDF+ format compatible with mainstream EEG analysis tools (such as EEGLAB, MNE-Python) is selected, and the file structure is defined as follows: File header information (Header): 1. Basic information: user ID (such as User001), acquisition time (2025-05-21 10:00:00), device model (BrainWave Glasses V2.0), sampling rate (250Hz), signal channel number (8 EEG channels + 3 auxiliary channels); 2. Extended information: data type (including raw signal, neural activity feature), processing software version (DataProcessor V1.5), meta information storage location (annotation section at the end of the file).

[0069] Data segment structure (Data Records): 1. Raw signal segment: 8 EEG channel amplitudes arranged in time order (integer, resolution 0.1μV); 2. Auxiliary signal segment: 3 auxiliary channels (EOG, acceleration, EMG) amplitudes; 3. Feature segment: neural activity features (such as P300 latency 315ms, alpha ratio 48%) are stored in the form of text labels (such as "ERP_P300_latency:315ms" "Alpha_power:48%").

[0070] S33. Encode the data to be packaged according to the specification of the extended EDF+ format, generate a standardized format file containing timestamps, event markers and complete meta information.

[0071] Specifically, the integrated data to be packaged is encoded according to the specification of the extended EDF+ format, and a standardized format file (suffix.EEG) is generated, and the encoding rules are as follows: Original signal segment: The amplitude of each EEG channel is converted into a 16-bit signed integer (e.g., 13 μV → 130, 15 μV → 150) and stored in big-endian byte order; Auxiliary signal segment: electrooculogram, acceleration, and electromyography signals are encoded in the same way (e.g. electrooculogram amplitude 50μV→500); Feature segment: Neural activity features are stored in the annotation field as ASCII text (e.g., “FEATURES: ERP_P300_latency=315ms; Alpha_power=48%; PLV=0.7”); Meta information section: Processing meta information (such as "PREPROCESS: bandpass=0.3-30Hz; notch=50Hz_Q30") is appended to the end of the file in XML format.

[0072] In this embodiment, the standardized encapsulation of the extended EDF+ format realizes the "self-description" of the data (including the original signal, features and processing metadata), is compatible with mainstream tools such as EEGLAB and MNE-Python, and can be directly imported for analysis without manual parsing, significantly lowering the technical threshold for data reuse; the complete record of metadata (such as filtering parameters and model versions) supports experimental reproduction, ensures the verifiability of research results, and meets the requirements of scientific research and medical scenarios for data rigor. Specific embodiment eight The execution of local structured storage and cloud synchronization storage includes the following steps: S34. Create a local storage directory by the hierarchical structure of "user ID / collection date / task type"; Specifically, a local storage directory is created based on the hierarchical rules of "user ID / collection date / task type", where: User ID: Use a unique user ID (such as User001, generated by the system when registering); Collection date: recorded in the "YYYYMMDD" format (e.g., 20250521 means May 21, 2025); Task type: defined according to the collection scenario (such as "concentration training", "sleep monitoring", "emotion recognition").

[0074] S35. Store the generated standardized format file in the corresponding directory; The standardized format file (suffix .EEG) generated in Specific Example 7 is stored in the above directory. The file naming rules are: User ID_Acquisition date_Acquisition time_Task type.EEG.

[0075] Example: If the acquisition time is 10:00:00 (accurate to seconds), the file name is: User001_20250521_100000_Concentration Training.EEG The storage process is implemented through the file system API (such as Android's File class), writing the .EEG file from the memory cache to the target directory. After the writing is completed, the system returns a storage success status (such as the "File saved" prompt).

[0076] S36. Generate an index file in the storage directory, recording the file name, file size, neural activity feature summary, and whether it contains artifacts.

[0077] Generate an index file in CSV format (named index.csv) in the storage directory, recording the following information (an additional line is automatically added to the index file each time an .EEG file is stored): File name: such as User001_20250521_100000_Concentration Training.EEG; File size: obtained from the file system (e.g. 2.5MB); Neural activity feature summary: Extract the key information of the neural activity feature output in claim 6 (e.g., "P300 latency 315ms, alpha 48%, PLV = 0.7"); Artifact Contained: Recorded according to the artifact separation according to claim 5 (if “No”, the signal has been verified to meet the standards after separation).

[0078] Example index file content: File name, file size, neural activity feature summary, and whether it contains artifacts.

[0079] User001_20250521_100000_Concentration training.EEG, 2.5MB, P300 latency 315ms; α accounted for 48%; PLV=0.7, no.

[0080] User001_20250521_103000_Concentration training.EEG, 2.3MB, P300 latency 320ms; α accounted for 45%; PLV=0.6, no.

[0081] In this embodiment, the hierarchical storage structure (user ID / date / task) realizes the "scenario classification" of data. Users can quickly locate target data through simple path navigation (for example, it only takes 3 steps to find the concentration training data of a user on a certain day); the automatic generation of index files (CSV format) and key information records (such as feature summaries) support keyword searches (such as "α proportion > 40%"), which shortens data retrieval time and greatly improves the management efficiency of local data. Specific embodiment nine The blockchain technology is used to record data operation logs and encryption technology to ensure data security, completing the whole process of data management, including the following steps: S37. Synchronize locally stored standardized format files to the cloud server via a secure transmission protocol; Specifically, the user triggers cloud synchronization through the data synchronization app, and the local device performs the following operations, including: Identity authentication: The user enters the account and password (such as User001:Password123), and the app sends an authentication request to the cloud server; after the server verifies the identity, it returns a temporary access token (valid for 10 minutes).

[0083] File upload: The local device reads the standardized format file stored in claim 8 (such as User001_20250521_100000_Concentration Training.EEG) and transmits the file stream to the specified path ( / cloud_storage / BrainWaveData / ) of the cloud server via the HTTPS protocol (TLS 1.3 encryption, the encryption suite is AES-256-GCM).

[0084] Transfer Verification: After the upload is complete, the cloud server calculates the SHA-256 hash value of the file (e.g., a1b2c3d4...) and returns it to the local device. The local device compares the hash value of the local file (if they match), confirming that the transfer was successful (taking 2 seconds, file size 2.5MB).

[0085] Furthermore, after receiving the multimodal data, the cloud backend performs data cleaning and eliminates invalid data segments based on quality markers (such as electrode contact impedance exceeding the standard and strong noise periods); then it performs feature association analysis to cross-modally associate the EEG concentration features with basic data (ambient noise, sitting angle) (for example, when the ambient noise is >65dB, a concentration drop of >20% triggers an early warning); dynamically generates a node report containing a concentration fluctuation curve, a fatigue score (0-100 points) and an emotional state heat map (positive / negative / neutral); finally, the report is pushed to the user's mobile app and web backend through the HTTPS encryption protocol, supporting historical data backtracking and multi-task comparison, forming a complete closed loop from collection to analysis visualization.

[0086] S38. In cloud servers, blockchain technology is used to record data operation events, including "upload," "download," and "modify" events. Each log entry contains the operation time, user ID, and file hash value, and the log content cannot be tampered with. Specifically, the cloud server writes the data operation event (in this case, "upload") to the blockchain. The specific steps are as follows: 1. Event encapsulation: Construct log entries, including: Operation time: 2025-05-21 10:05:00 (accurate to the second); User ID: User001; File hash: a1b2c3d4... (same as the verification hash in step S37); Operation type: "Upload".

[0087] 2. Blockchain writing: The log entry is broadcast to three nodes via Hyperledger Fabric's chaincode (smart contract). The nodes confirm the validity of the entry using the Raft consensus algorithm (approximately 1 second) and add the entry to the latest block of the blockchain (block height 1001).

[0088] 3. Tamper-proof verification: The blockchain uses a hash chain structure (each block contains the hash value of the previous block). If you attempt to modify a log entry, you need to recalculate the hash values ​​of all subsequent blocks (a computationally intensive process that cannot be completed within a reasonable timeframe), ensuring that the log content cannot be tampered with.

[0089] S39. Desensitize user metadata in standardized format files and encrypt sensitive data using encryption algorithms.

[0090] Specifically, the cloud server performs security processing on the standardized format file, including: 1. User metadata desensitization: Extract user metadata from the file (such as the name "Zhang San" and the mobile phone number "138XXXX1234"), use the SHA-256 hash algorithm to generate a desensitized value (such as sha256("Zhang San")=e1f2g3h4..., sha256("138XXXX1234")=i5j6k7l8...), and replace the original data.

[0091] 2. Sensitive data encryption: The original EEG signal (accounting for 80% of the file content) is encrypted using the national secret SM4 algorithm (the key is generated and stored by the HSM, and a random initialization vector IV is used for each encryption). The encrypted data format is IV||ciphertext (for example, IV=0x12345678..., ciphertext=0xaabbccdd...`).

[0092] 3. Key management: The encryption key is bound to the user ID. Only authorized users (such as User001) can decrypt it through the HSM (biometric verification, such as fingerprint, is required), ensuring that the risk of key leakage is close to zero.

[0093] In this embodiment, HTTPS secure transmission and hash verification ensure that data is not tampered with or leaked during transmission (100% transmission success rate); the immutability of blockchain technology provides a "trusted audit log" for data operations (such as uploading and downloading), solving the problem of easy tampering of operation records in traditional cloud storage; dual protection of user metadata desensitization (hashing) and sensitive data encryption (national encryption SM4) ensures user privacy and data security, meeting medical-grade data security standards.

[0094] The invention is operational with numerous general purpose or special purpose computer system environments or configurations.

[0095] For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, etc.

[0096] The invention may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer.

[0097] Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network.

[0098] In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

[0099] Specifically, those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware via computer-readable instructions. The computer-readable instructions can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes in the above-described method embodiments. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0100] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps and they may be performed in other orders.

[0101] Moreover, at least part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. Their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0102] Obviously, the embodiments described above are only some of the embodiments of the present invention, rather than all of them. The accompanying drawings provide preferred embodiments of the present invention, but do not limit the scope of the present invention. The present invention can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to facilitate a more thorough and comprehensive understanding of the disclosure of the present invention.

[0103] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art may still modify the technical solutions described in the aforementioned specific embodiments or replace some of the technical features therein with equivalents. Any equivalent structure made using the contents of the present invention's description and drawings, directly or indirectly applied to other related technical fields, shall also fall within the scope of protection of the present invention.

Claims

1. A method for processing EEG data using EEG glasses, characterized in that: The following steps are involved: S1. Acquire the original EEG signals collected by the EEG glasses, simultaneously acquire multimodal auxiliary signals such as eye movement, acceleration, and myoelectricity, and monitor the acquisition quality parameters in real time to form an initial dataset containing the original EEG signals, multimodal auxiliary signals, and quality markers; S2. Taking the initial dataset as input, the system sequentially performs preprocessing, multimodal artifact separation, and personalized feature extraction, outputting a processed dataset containing clean EEG signals and neural activity features. S3. Encapsulate the processed data set and its processing metadata into a standardized format file, perform local structured storage and synchronize cloud storage, and use blockchain technology to record data operation logs and encryption technology to ensure data security, completing the entire data management process.

2. The method for processing EEG data using EEG glasses according to claim 1, wherein: The method of obtaining the original EEG signal collected by the EEG glasses comprises the following steps: S11. Send microcurrent to the scalp through the dry electrodes of the EEG glasses, and measure the contact impedance of each electrode in real time; S12. If the contact impedance of an electrode exceeds a preset threshold, the user interface prompts the wearer to adjust the wearing position. If the impedance still does not meet the threshold after adjustment, signal interpolation compensation is enabled for adjacent electrodes, and the signal weight of the electrode exceeding the threshold is reduced. S13. The analog EEG signal after contact quality optimization is differentially amplified by the signal amplification module, and the interference is reduced by the common-mode interference suppression module. It is then converted into a digital signal by the analog-to-digital conversion module to obtain the original EEG signal.

3. The method for processing EEG data using EEG glasses according to claim 2, wherein: The synchronous acquisition of multimodal auxiliary signals such as eye movement, acceleration, and myoelectricity includes the following steps: S14. Synchronously activate the eye movement detection module to collect eye movement signals and record eye movement events; Synchronously activate the acceleration detection module to collect head movement acceleration signals and record severe shaking events that exceed the preset acceleration threshold; Synchronously activate the electromyography detection module to collect electromyographic signals and record muscle contraction events; S15. Add an event marker to the time point of each multimodal auxiliary signal, with the marking accuracy being a preset accuracy.

4. The method for processing EEG data using EEG glasses according to claim 3, wherein: The real-time monitoring and acquisition of quality parameters to form an initial data set including original EEG signals, multimodal auxiliary signals and quality markers includes the following steps: S16. Perform spectrum analysis on the original EEG signal to identify the ambient noise frequency band and mark periods of strong noise exceeding a preset noise amplitude threshold; S17. Recording the contact impedance value, signal amplitude and other quality parameters of each electrode; S18. Integrate the original EEG signals, multimodal auxiliary signals, and quality tags into an initial dataset, where each signal and quality tag contains timestamp information.

5. The method for processing EEG data using EEG glasses according to claim 4, wherein: The method takes the initial dataset as input, sequentially performs preprocessing, multimodal artifact separation, and personalized feature extraction, and outputs a processed dataset containing pure EEG signals and neural activity features, including the following steps: S21. Apply bandpass filtering to the raw EEG signals in the initial dataset to remove low-frequency drift and high-frequency noise, and apply notch filtering to suppress ambient noise in specific frequency bands. If the scenario is event-related potential analysis, convert the reference electrode to a whole-brain average reference to obtain a preliminary denoised EEG signal. S22. Locate the artifact periods in the preliminary denoised EEG signals based on the multimodal auxiliary signals in the initial dataset; input the located EEG signals into a pre-trained deep learning model, and separate the pure EEG signals and artifact components through feature learning; verify the separation effect, and if it does not meet the standards, adjust the model parameters and re-separate, and finally output the pure EEG signals.

6. The method for processing EEG data using EEG glasses according to claim 5, characterized in that: The method takes the initial data set as input, sequentially performs preprocessing, multimodal artifact separation, and personalized feature extraction to output a processed data set containing pure EEG signals and neural activity features, and further includes the following steps: S23. Using the pre-stimulation preset duration data as the baseline, calculate the baseline mean and correct the post-stimulation data of the pure EEG signal to obtain a baseline-corrected EEG signal; S24. Detecting ERP components in the baseline-corrected EEG signal using a time series model. Dynamically adjusting model detection parameters based on user historical data to output ERP features. S25. Extract the power proportion of a preset frequency band from the baseline-corrected EEG signal through time-frequency analysis to reflect the user's neural activity state and output the frequency band power characteristics; S26. Calculate the phase-locking values ​​of multiple channels in the baseline-corrected EEG signal, evaluate the synchronization of neural activity in different brain regions, and output neural synchronization characteristics; S27. Integrate ERP features, frequency band power features, and neural synchronization features into neural activity features, and together with the pure EEG signals, form a processed data set.

7. The method for processing EEG data using EEG glasses according to claim 6, wherein: The step of encapsulating the processed data set and its processing metadata into a standardized format file includes the following steps: S31. Integrate the processed data set and the processing meta-information into data to be packaged; S32. Use the extended EDF+ format compatible with mainstream EEG analysis tools to define the file header information and data segment structure; S33. Encode the data to be encapsulated according to the specifications of the extended EDF+ format to generate a standardized format file containing a timestamp, event marker and complete meta-information.

8. The method for processing EEG data using EEG glasses according to claim 7, wherein: The execution of local structured storage and cloud synchronization storage includes the following steps: S34. Create a local storage directory by the hierarchical structure of "user ID / collection date / task type"; S35. Store the generated standardized format file in the corresponding directory; S36. Generate an index file in the storage directory, recording the file name, file size, neural activity feature summary, and whether it contains artifacts.

9. The method for processing EEG data using EEG glasses according to claim 8, wherein: The blockchain technology is used to record data operation logs and encryption technology to ensure data security, completing the whole process of data management, including the following steps: S37. Synchronize locally stored standardized format files to the cloud server via a secure transmission protocol; S38. In cloud servers, blockchain technology is used to record data operation events, including "upload," "download," and "modify" events. Each log entry contains the operation time, user ID, and file hash value, and the log content cannot be tampered with. S39. Desensitize user metadata in standardized format files and encrypt sensitive data using encryption algorithms.

Images