An electroencephalogram signal processing method and device based on a non-invasive brain-computer interface

Abstract

This invention provides a method and apparatus for processing electroencephalogram (EEG) signals based on a non-invasive brain-computer interface, belonging to the field of medical device technology. Specifically, it includes: determining an identification and processing scheme for the EEG signal staging model of the target based on the variation of amplitude data of the electrodes of the target, combined with the risk of variation of the amplitude data of the target's electrodes and the similarity to other targets, as well as signal-stable electrodes; determining the interference risk type and location of the target based on the deviation data of the EEG signal staging results between different identification and processing schemes; and determining the processing method for the EEG signal at the location of the interfering electrode based on the similarity of the identification and processing schemes with identification deviations in each target, and the interference risk type of the target. This improves the accuracy of sleep staging results.

Description

Technical Field

[0001] This invention belongs to the field of medical device technology, and in particular relates to a method and device for processing electroencephalogram (EEG) signals based on a non-invasive brain-computer interface. Background Technology

[0002] The clinical gold standard for sleep staging is based on manual interpretation of polysomnography (PSG). This method requires subjects in a specialized laboratory to have dozens of sensors attached or connected to their head, face, chest, abdomen, and limbs, simultaneously collecting multiple physiological signals such as electroencephalogram (EEG), electrooculogram (EOG), electromyogram (EMG), electrocardiogram (ECG), and respiration. Subsequently, rigorously trained technicians, following the interpretation manual of the American Academy of Sleep Medicine (AASM), manually staging the sleep stages in 30-second increments. While comprehensive and accurate, this method has inherent drawbacks: it is highly invasive, costly, and inefficient. The complex wires and sensors not only cause significant discomfort and severely disrupt natural sleep patterns, triggering the "first night effect" and leading to data distortion, but the entire process also heavily relies on specialized facilities, expensive equipment, and manpower, making it impossible to achieve widespread and routine sleep monitoring.

[0003] To address the aforementioned technical issues, the invention patent application CN202511556423.6, "A Sleep Monitoring and Intervention Method and System Based on Electroencephalogram (EEG) Signals," integrates front-end adaptive processing, graph neural networks, and a deep learning model using Transformer for sleep staging. Based on this, it combines closed-loop phase-synchronous stimulation driven by a digital phase-locked loop (PLL) to achieve a seamless transition from sleep monitoring to closed-loop intervention, effectively improving the accuracy of sleep monitoring and the effectiveness of intervention. However, the following technical problems remain:

[0004] During sleep staging, due to the difference between the EEG measuring device and the user's head contour, poor contact at certain locations often occurs, leading to inaccurate measurement results. Therefore, how to assess the interference of electrode position on sleep staging results and generate corresponding EEG signal processing methods to minimize the interference of electrode position on sleep analysis results has become an urgent technical problem to be solved.

[0005] Therefore, there is an urgent need for a method and device for processing electroencephalogram (EEG) signals based on a non-invasive brain-computer interface. Summary of the Invention

[0006] To achieve the objectives of this invention, the following technical solution is adopted:

[0007] Specifically, this application provides a method for processing electroencephalogram (EEG) signals based on a non-invasive brain-computer interface, which includes:

[0008] S1 uses the parsing results of brain-computer interface data to determine the similarity of EEG signals between different electrodes, determines a reference sleep process during sleep based on the similarity, and determines the analysis and processing target of the EEG signals in the user based on the reference sleep process data and the variation of amplitude data of EEG signals in different reference sleep processes.

[0009] S2 determines the identification and processing scheme of the EEG signal staging model of the target based on the variation of the amplitude data of the target electrode and the similarity of the amplitude data of the target electrode with other targets and the signal stable electrodes.

[0010] S3 determines the interference risk type and interference electrode location of the analysis and processing target based on the deviation data of the EEG signal staging results between different recognition and processing schemes. Based on the similarity of the recognition and processing schemes with recognition deviations at the interference electrode locations in each analysis and processing target, and the interference risk type of the analysis and processing target, the processing method of the EEG signal at the interference electrode location is determined.

[0011] The beneficial effects of this invention are as follows:

[0012] By monitoring and analyzing the similarity of EEG signals among multiple electrodes distributed on the user's head, monitoring moments with stable electrode installation and minimal external interference are identified. Based on the proportion of such moments throughout the entire sleep process, the overall data quality and interference risk level of that sleep process are comprehensively assessed, enabling the identification of a reference sleep process. This assessment provides a reliable data foundation for subsequent analysis of brain-computer interface data.

[0013] Based on the reference sleep process data and the variation of EEG signal amplitude data in different reference sleep processes, the analysis and processing targets of the EEG signals in the user are determined. By using the distribution data of signal stable electrodes and signal fluctuation electrodes to be evaluated from the reliable reference benchmark in the user's reference sleep process, users with both types of electrodes can be screened. Combined with the reference sleep process data, users with less external interference, i.e., more reference sleep processes, can be screened. Thus, a reliable assessment of the interference risk of signal fluctuation electrodes can be achieved from users with less external interference risk.

[0014] In this application, the method for processing EEG signals to determine the location of interfering electrodes, for an identified potential interfering electrode location, comprehensively considers the scale of users with problems, the reliability of sleep analysis results of users with problems, and the dispersion of interfering electrode locations associated with interference risk. This achieves reliable identification of the interference risk and complexity of the interference pattern. Based on the evaluation results, it selects from different processing paths such as "directly performing feature inspection and possible exclusion", "conducting in-depth pattern analysis before making a decision", and "not processing for the time being", which further improves the reliability of the user's sleep staging results.

[0015] Furthermore, the analysis results of the brain-computer interface data are determined based on the monitoring data of the user's electroencephalogram (EEG) signals in different electrodes.

[0016] Furthermore, the similarity of the EEG signals between the electrodes is determined based on the similarity coefficient of the EEG signal characteristics between the electrodes at the same monitoring time.

[0017] Furthermore, the method for determining the reference sleep process during the sleep process is as follows:

[0018] Based on the similarity of EEG signals between electrodes at different monitoring times during the sleep process, the monitoring times at which the similarity coefficient of EEG signal features meets the requirements are determined and used as consistent monitoring times.

[0019] Based on consistent monitoring time data during the sleep process, it is determined whether the sleep process belongs to a reference sleep process.

[0020] Furthermore, the method for determining the type of interference risk of the target being analyzed and processed is as follows:

[0021] Each signal analysis combination is used as an identification and processing scheme. The deviation data of the EEG signal staging results between different identification and processing schemes are used to determine the reference staging result in the sleep staging process.

[0022] Based on the deviation between the sleep staging results of the aforementioned identification and processing scheme and the reference staging results, identify the identification and processing schemes that are inconsistent with the reference staging results, and use them as deviation identification schemes.

[0023] Using deviation identification scheme data in different sleep stages, the type of interference risk to the analysis and processing target is determined.

[0024] In a second aspect, the present invention provides a computer device comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the above-described method for processing electroencephalogram signals based on a non-invasive brain-computer interface when running the computer program.

[0025] Other features and advantages will be set forth in the following description, and the objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0026] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0027] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0028] Figure 1 This is a flowchart of a brainwave signal processing method based on a non-invasive brain-computer interface;

[0029] Figure 2 It is a flowchart of a method for determining a reference sleep process during sleep;

[0030] Figure 3 This is a flowchart illustrating the method for determining the identification and processing scheme of the target's EEG signal staging model.

[0031] Figure 4 This is a flowchart illustrating the method for determining the identification and processing scheme of the target's EEG signal staging model. Detailed Implementation

[0032] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0033] Example 1

[0034] like Figure 1 As shown, this application provides a method for processing electroencephalogram (EEG) signals based on a non-invasive brain-computer interface, specifically including:

[0035] S1 uses the parsing results of brain-computer interface data to determine the similarity of EEG signals between different electrodes, determines a reference sleep process during sleep based on the similarity, and determines the analysis and processing target of the EEG signals in the user based on the reference sleep process data and the variation of amplitude data of EEG signals in different reference sleep processes.

[0036] Furthermore, the analysis results of the brain-computer interface data are determined based on the monitoring data of the user's electroencephalogram (EEG) signals in different electrodes.

[0037] Furthermore, the similarity of the EEG signals between the electrodes is determined based on the similarity coefficient of the EEG signal characteristics between the electrodes at the same monitoring time.

[0038] It should be noted that for symmetrically distributed electrodes, if they are installed accurately and there is no external interference, the similarity coefficient of their EEG signal characteristics meets the requirements.

[0039] Specifically, such as Figure 2 As shown, the method for determining the reference sleep process during the sleep process is as follows:

[0040] In this embodiment, the user's sleep process is assessed to determine whether it falls within a reference sleep process—that is, a sleep process with a low risk of external interference—to ensure the accuracy and reliability of brain-computer interface (BCI) data parsing. The logic is to identify monitoring moments (called consistent monitoring moments) where the electrodes are stably installed and minimally affected by external interference by monitoring and analyzing the similarity of EEG signals among multiple electrodes distributed across the user's head. Based on the proportion of such moments throughout the entire sleep process, the overall data quality and interference risk level of that sleep process are comprehensively judged. This judgment provides a reliable data foundation for subsequent BCI data parsing.

[0041] S11 determines the monitoring time when the similarity coefficient of the EEG signal characteristics between electrodes meets the requirements based on the similarity of the EEG signals at different monitoring times during the sleep process, and uses it as the consistent monitoring time.

[0042] Similarity coefficient of EEG signal features: This refers to a numerical index calculated using mathematical methods (such as correlation coefficient, covariance, etc.) at the same monitoring time, used to quantify the degree of similarity between the features of EEG signals collected by two electrodes. The higher the coefficient, the more similar the signals from the two electrodes are in terms of waveform, frequency, or amplitude.

[0043] Consistent monitoring moment: refers to a specific monitoring point during sleep when the similarity coefficient of EEG signal characteristics between all or specific electrode pairs (especially symmetrically distributed electrodes) reaches or exceeds a preset requirement. This moment is identified as a consistent monitoring moment.

[0044] The purpose of this step is to filter out high-quality discrete points in the time dimension. Since EEG signals are easily affected by factors such as poor electrode contact, user movement, or environmental electromagnetic interference, calculating the similarity coefficient between the signals of the electrodes in real time allows for the objective identification of moments with less interference and stable signals. Its significance lies in providing a clean and reliable set of time-point samples for subsequent overall sleep process assessment, avoiding interference from low-quality data and ensuring the accuracy of the analytical basis.

[0045] Specific examples combining scenarios and keywords:

[0046] In a typical nighttime sleep monitoring scenario, the user wears an EEG cap equipped with symmetrical electrodes, such as the left frontal lobe electrode F3 and the right frontal lobe electrode F4, forming a symmetrical pair. At a given monitoring moment, the system calculates the similarity coefficient of the EEG signal characteristics acquired by electrodes F3 and F4 in real time. If the coefficient meets a preset requirement (e.g., exhibiting high similarity), it indicates that the electrodes are accurately positioned at that moment, and there is no significant head movement or external device interference; therefore, this moment is determined as a consistent monitoring moment. Conversely, if the similarity coefficient is too low, it suggests possible electrode loosening or transient interference.

[0047] S12 determines whether the sleep process belongs to a reference sleep process based on the consistent monitoring time data during the sleep process.

[0048] Consistent monitoring time percentage: refers to the percentage of total monitoring times during the entire sleep process that are identified as consistent monitoring times.

[0049] Preset consistent time percentage threshold: a pre-set standard value for judgment.

[0050] Reference sleep process: refers to the sleep process in which the proportion of consistently monitored moments is higher than the preset threshold throughout the entire sleep process. Such processes are considered to have a low risk of external interference, high data quality, and can be used as a reliable reference for subsequent analysis.

[0051] The purpose of this step is to make a holistic and comprehensive binary judgment on the data quality of the entire sleep process. Relying solely on the similarity of a single moment is insufficient to assess the entire sleep stage. By calculating the proportion of consistent monitoring moments and comparing it with a preset threshold for the proportion of consistent moments, the reliability and coverage of the data throughout the monitoring period can be quantitatively evaluated. Its significance lies in providing a clear decision rule: if the proportion is higher than the threshold, the entire sleep process is considered a reference sleep process, meaning that the data collected in this sleep session is generally reliable and suitable for high-precision applications such as model training, health assessment, or disease diagnosis; conversely, it suggests that the data collection may have been subject to significant interference, requiring caution or re-collection.

[0052] Specific examples combining scenarios and keywords:

[0053] Continuing with the sleep monitoring scenario described above, assume an 8-hour sleep monitoring session comprises 28,800 monitoring moments (one moment per second). Through the analysis in step S11, the system identifies 25,920 moments as consistent monitoring moments. The system then calculates the proportion of consistent monitoring moments (e.g., reaching 90%) and compares this proportion to a preset threshold for the proportion of consistent moments. If this threshold is set to 85%, then since 90% > 85%, the system ultimately determines that this sleep process belongs to the reference sleep process. This indicates that the overall quality of the monitoring data is high, and the risk of interference is low.

[0054] It is understood that if the proportion of consistent monitoring times in the monitoring time during the sleep process is greater than the preset consistent time proportion threshold, then the sleep process is determined to be a reference sleep process, that is, a sleep process with a low risk of external interference.

[0055] This method improves the objectivity and automation of brain-computer interface data quality assessment by introducing inter-electrode signal similarity analysis and time ratio judgment mechanisms, reduces reliance on human experience interpretation, and achieves quantitative measurement of the reliability of sleep process data. By consistently monitoring the clear indicator of time ratio, the interference risk is assessable and comparable, providing high-quality data screening assurance for downstream advanced analysis tasks such as EEG signal analysis, sleep staging, and disease biomarker identification. This ensures that these analyses are based on reference sleep process data with low interference risk, thereby improving the accuracy and credibility of research results or clinical diagnoses.

[0056] Furthermore, the reference sleep process data includes the proportion of the user's reference sleep process within their sleep process.

[0057] Furthermore, the variation of the amplitude data of the EEG signal during the reference sleep process is determined based on the rate of change of the amplitude characteristics of the EEG signal between adjacent consistent monitoring times during the reference sleep process.

[0058] Specifically, the method for determining the target of the analysis and processing of the user's electroencephalogram (EEG) signals is as follows:

[0059] In this embodiment, a screening mechanism is constructed to determine whether a specific user is suitable as a target for in-depth analysis of EEG signals (such as for high-precision sleep staging or pathological feature identification). This mechanism identifies user samples suitable for subsequent quantitative analysis of interference effects. The logic is that, to accurately assess the impact of a specific electrode as an "interference risk electrode" on the analysis results (such as sleep staging), user data with internal comparison conditions must first be obtained. That is, the user's data should simultaneously include signal-stable electrodes that provide reliable reference benchmarks and signal-fluctuating electrodes to be evaluated. This method uses progressive conditional judgments to screen users from the user group who meet these data characteristics and determines whether they belong to the target group for in-depth analysis of EEG signals.

[0060] S21 Based on the reference sleep process data, determine the proportion of the reference sleep process in the user's sleep process, and use the proportion of the reference sleep process in the user's sleep process as the reference process proportion;

[0061] Reference sleep process percentage: This refers to the percentage of all sleep processes in a user's history that are considered to have a low risk of external interference.

[0062] Reference process ratio: The calculation result of the above ratio is used as the direct output of this step.

[0063] This step serves as an initial screening process, designed to assess a user's ability and consistency in providing high-quality baseline data. Its significance lies in ensuring that candidate users possess a sufficient number of usable reference sleep process samples. This is a prerequisite for any meaningful historical data comparative analysis and statistical inference. If a user's reference process proportion is too low, it indicates that their data output quality is unstable or has been subject to long-term interference, lacking a reliable data foundation for in-depth comparative analysis; therefore, they should be excluded at this stage. This ensures that subsequent analyses are built upon a solid foundation of data.

[0064] S22 determines the variation of the amplitude range of the EEG signal between different reference sleep processes by measuring the variation of the amplitude data of the EEG signal in different reference sleep processes.

[0065] Variation in EEG signal amplitude data: This is an indicator describing the amplitude fluctuation characteristics of EEG signals within a single reference sleep process, which can be reflected by parameters such as the amplitude variation rate between adjacent consistent monitoring times.

[0066] Variation in amplitude range: This refers to the differences and fluctuation patterns observed when comparing the overall distribution range (i.e., amplitude range) of EEG signal amplitudes across different reference sleep processes for the same user.

[0067] This step aims to assess the long-term stability of the user's electrode signals across a cross-time dimension. Its significance lies in providing a quantitative basis for identifying electrodes with stable signals and those with fluctuating signals. For users with relatively stable physiological states, the amplitude range of resting-state EEG signals collected from different electrodes at different nights should maintain a certain consistency. By analyzing the variations in amplitude ranges, it is possible to objectively distinguish which electrodes exhibit consistent amplitude characteristics (stable) across different sleep stages, and which exhibit significant, unconventional shifts (fluctuations). The output of this step is the direct input for performing the electrode classification operation in step S23.

[0068] S23 determines whether the user is a target for EEG signal analysis and processing based on the reference process ratio and the variation in amplitude range of EEG signals between different reference sleep processes.

[0069] It should be noted that if the reference process ratio is less than the preset reference process ratio threshold, then the user is determined not to be a target for EEG signal analysis and processing.

[0070] Preset reference process ratio threshold: A pre-set minimum standard value used to determine whether the user's reference process ratio meets the standard.

[0071] This judgment is a fundamental threshold for data usability. If a user's reference process ratio is less than a preset threshold, it means that the user cannot consistently provide the amount of data required to meet the minimum quality standards. For studies aiming at rigorous comparative analysis, such users do not meet the basic requirements to be considered qualified samples, and therefore are directly deemed not to be part of the target data for EEG signal analysis. This improves the overall baseline level of the research cohort.

[0072] Additionally, it should be noted that if the reference process ratio is not less than a preset reference process ratio threshold, the following content is also included:

[0073] Based on the variation of the amplitude range of the EEG signal between different reference sleep processes, S231 determines the signal fluctuation electrode and the signal stability electrode among the electrodes, and determines whether the user has signal stability electrode and signal fluctuation electrode. If yes, proceed to step S232; otherwise, determine that the user is not the target of EEG signal analysis and processing.

[0074] It is understood that the method for determining the signal stabilizing electrode is as follows:

[0075] The deviation of the electrode at the endpoints of the amplitude range between the reference sleep process and other reference sleep processes is determined by the variation of the amplitude range of the EEG signal between the reference sleep process and other reference sleep processes. If the deviation at the endpoints of the amplitude range between different reference sleep processes is within a preset deviation range, then the electrode is determined to be a signal-stabilized electrode.

[0076] Furthermore, other reference sleep processes whose deviation from the endpoints of the amplitude interval between the reference sleep process and the reference sleep process is not within a preset deviation range are considered as deviation sleep processes. Reference sleep processes whose number of deviation sleep processes is greater than a preset deviation process number threshold are considered as suspected abnormal sleep processes. If the proportion of suspected abnormal sleep processes of the electrode in the reference sleep process is greater than a preset proportion threshold, then the electrode is determined to be a signal fluctuation electrode.

[0077] Signal-stabilized electrodes: These are electrodes whose amplitude ranges of EEG signals show minimal variation at the endpoints throughout multiple reference sleep periods, consistently remaining within a preset deviation range. They represent reliable and consistent signal acquisition points.

[0078] Signal fluctuation electrode: This refers to an electrode whose amplitude range frequently exceeds a preset deviation range at its endpoints across multiple reference sleep processes. Such electrodes are typically associated with a large number of deviant sleep processes. An electrode is identified as a signal fluctuation electrode when the proportion of suspected abnormal sleep processes across all reference processes exceeds a preset threshold.

[0079] Deviation sleep process: For a specific electrode, a single reference sleep process in which the endpoint deviation of its signal amplitude range exceeds a preset deviation range compared to the historical stable range of that electrode.

[0080] Suspected abnormal sleep processes: For a given electrode, those reference processes whose cumulative number of deviant sleep processes exceeds a preset threshold for the number of deviant sleep processes.

[0081] This step is central to building the internal comparative analysis model. Its significance lies in accurately identifying electrodes playing different roles within the user's electrode array. The signal-stable electrode group will serve as the internal "gold standard" or reliable reference in subsequent analyses; the signal-fluctuating electrode group is the target whose impact needs to be evaluated. This method achieves refined classification of electrode instability modes and resistance to random noise by setting multi-level thresholds (deviation, quantity, proportion). Its key significance lies in determining whether a user simultaneously possesses both types of electrodes, which is an absolute necessity for conducting internal influence comparative analysis. If a user has only stable electrodes or only fluctuating electrodes, an effective comparative study cannot be conducted within that user, and therefore they should be excluded.

[0082] S232 determines whether the proportion of the signal stabilizing electrode in all electrodes is greater than a preset electrode proportion threshold. If not, proceed to step S233. If yes, determine that the user belongs to the target of EEG signal analysis and processing.

[0083] Preset electrode percentage threshold: The minimum standard value used to determine whether a signal-stable electrode occupies a dominant proportion in the total number of electrodes.

[0084] This step assesses the adequacy and representativeness of the internal "gold standard" (stable electrodes). If the proportion of signal-stabilized electrodes exceeds a preset electrode proportion threshold, it indicates that signal acquisition over most of the user's head area is consistently stable and reliable. Analysis results derived from such a group of stable electrodes (e.g., sleep staging) have high confidence and are sufficient to serve as a robust internal benchmark for evaluating the impact of individual fluctuating electrodes. Therefore, users meeting this condition are ideal targets for analysis and can be directly identified.

[0085] S233 acquires the number of signal fluctuation electrodes and, in conjunction with the number of deviation sleep processes of the signal fluctuation electrodes, determines the interference risk value of the user, and determines whether the user belongs to the target of EEG signal analysis and processing based on the interference risk value.

[0086] It is understandable that if the user's interference risk value is greater than the preset risk threshold, then the user is determined to be the target of EEG signal analysis and processing, that is, the analysis determines whether sleep staging can still be accurately performed after the signals of certain electrodes are excluded.

[0087] Interference Risk Value: A comprehensive quantitative indicator that combines the number of signal fluctuation electrodes with the degree of their individual anomalies (typically characterized by the number of associated skewed sleep processes). A higher value indicates a greater number of unstable signal sources in the user data and a higher frequency of problems.

[0088] Preset risk threshold: A critical value used to determine the eligibility of end users based on the interference risk value.

[0089] This step handles situations where a stable electrode does not dominate but both types of electrodes coexist. Its purpose is to conduct an overall risk assessment to determine whether it's worthwhile to include this user in subsequent analyses. The interference risk value comprehensively reflects the breadth and depth of potential interference. If this value exceeds the preset risk threshold, it means the impact of the fluctuating electrode may be too widespread or severe, or the results of comparative analysis may be unreliable due to excessive background noise. Conversely, if the risk value is manageable, it indicates that despite local instability, the user's overall data environment still allows for effective internal comparisons to quantify the impact of specific fluctuating electrodes. This achieves refined management by maximizing the range of eligible samples while ensuring analytical quality.

[0090] In one possible specific embodiment, a method for screening target users for EEG signal interference impact analysis is implemented in a research project containing 300 candidate users, with the aim of providing a qualified sample for subsequent research on "evaluating the impact of specific fluctuating electrodes on the accuracy of sleep staging".

[0091] Data basis: For candidate user U_A, the system recorded 40 recent sleep monitoring sessions. Based on the prior judgment method, 32 of these were judged as reference sleep processes.

[0092] Execute step S21: Calculate the reference process ratio for user U_A: 32 / 40 = 80%.

[0093] In this embodiment, a preset reference process ratio threshold of 70% is set. If 80% ≥ 70%, the condition is met, and the process proceeds to step S22 and subsequent in-depth analysis.

[0094] Step S22: The system extracts 32 reference sleep process data of user U_A, and for each process and each electrode, calculates the 10th to 90th percentile of the EEG signal amplitude as the amplitude range of that process.

[0095] By comprehensively analyzing all 32 processes, the variation of the amplitude range for each electrode was determined.

[0096] Execute step S23 – sub-step S231: Based on the output of step S22, apply stability criteria to each of the 64 electrodes of U_A: the preset deviation range is set to ±12% of the endpoint of the common range of the electrode, the preset deviation process number threshold is set to 5 times, and the preset ratio threshold is set to 20%.

[0097] Judgment results: Of the 64 electrodes, 41 electrodes were classified as signal-stable electrodes, 16 electrodes were classified as signal-fluctuating electrodes, and the remaining 7 electrodes were considered neutral because they did not meet the clear classification criteria.

[0098] The system determines that user U_A has both a stable signal electrode and a fluctuating signal electrode, and therefore proceeds to sub-step S232.

[0099] Execute sub-step S232: Calculate the proportion of signal-stabilized electrodes: 41 / 64 ≈ 64.1%. In this embodiment, the preset electrode proportion threshold is set to 75%.

[0100] Judgment: 64.1% < 75%, the stable electrode has not reached the dominant proportion, so proceed to sub-step S233.

[0101] Execute sub-step S233: Obtain the number of signal fluctuation electrodes: 16, and calculate the interference risk value R. This embodiment uses the formula: R = (number of fluctuation electrodes) × (average number of deviation processes of the fluctuation electrodes). The calculated average number of deviation processes is 6.2, therefore R = 16 × 6.2 = 99.2.

[0102] In this embodiment, a preset risk threshold of 90 is set, and it is determined that the interference risk value R = 99.2 is greater than 90.

[0103] Final ruling: User U_A is a target for EEG signal analysis and processing. They meet all the progressive criteria: sufficient high-quality sleep (reference process ratio meets standard); they possess both a stable electrode group that can serve as an internal benchmark and a volatile electrode group to be evaluated; and the overall interference risk value is within an acceptable range, making them suitable for subsequent comparative analysis studies to quantify the impact of volatile electrodes.

[0104] This method employs a multi-level, progressively conditional automated screening process to precisely locate key research samples. It efficiently and accurately identifies individuals from a large pool of users whose data characteristics meet the requirements for quantitative research on the impact of interference. These users must simultaneously possess a reliable internal reference benchmark (stable electrode) and a clearly defined variable to be measured (fluctuating electrode). This is the cornerstone of effective internal comparative experiments. By setting preset reference process proportion thresholds, preset electrode proportion thresholds, and preset risk thresholds, this method ensures that the data from selected users meet the rigorous requirements of scientific analysis in terms of quantity, quality (benchmark reliability), and comparison environment (risk controllability). This fundamentally guarantees the feasibility of the subsequent analytical strategy of "assessing the impact of fluctuating electrodes using stable electrode results" and the reliability of the conclusions.

[0105] The output of this step (i.e., the users marked as the target of analysis and their corresponding stable / fluctuating electrodes) provides a validated, high-quality sample pool for statistical induction at a larger user group level. Based on this sample pool, the frequency and degree of influence of a specific electrode (such as Fp1) that commonly exhibits signal fluctuations among multiple users can be systematically analyzed, thereby achieving group identification and risk assessment of "interference risk electrodes".

[0106] S2 determines the identification and processing scheme of the EEG signal staging model of the target based on the variation of the amplitude data of the target electrode and the similarity of the amplitude data of the target electrode with other targets and the signal stable electrodes.

[0107] Specifically, such as Figure 3 As shown, the method for determining the identification and processing scheme of the EEG signal staging model of the analysis and processing target is as follows:

[0108] In this embodiment, a customized EEG signal staging strategy is dynamically generated for each identified target user. The core logic is to construct a systematic comparative analysis framework: using the user's own set of stable electrodes as a reliable benchmark, one or more stable electrodes are strategically combined to form various signal analysis combinations. By performing staging processing on these combinations and comparing the results, the aim is to accurately quantify the degree and pattern of influence of different signal fluctuation electrodes on sleep staging results. This method not only relies on the user's own electrode stability characteristics but also introduces the group indicator of fluctuation-related targets to distinguish between general and specific interference, thereby intelligently adjusting the scope and focus of the test and optimizing the efficiency of computational resource utilization while ensuring analytical depth.

[0109] S31 determines the signal fluctuation electrode of the target based on the variation of amplitude data of the electrode of the target being analyzed and processed;

[0110] Analysis and processing target: refers to qualified users identified after a rigorous pre-screening process, whose core characteristic is that they simultaneously possess reliable signal stabilization electrodes and signal fluctuation electrodes to be evaluated.

[0111] Signal fluctuation electrodes: Specifically refers to a set of electrodes that are judged to be inconsistent and unstable based on the historical changes in the amplitude data of the EEG signal during multiple reference sleep processes of the target user in the current analysis and processing.

[0112] This step is the starting point for solution development, and its significance lies in directly inheriting and applying the precise conclusions of the preliminary analysis stage. By explicitly listing all signal fluctuation electrodes for the current user, this step transforms the abstract "analysis and processing target" into a concrete and actionable list of analysis objects, providing clear inputs and objectives for subsequent combined construction and comparative experiments. It ensures that the entire processing solution is designed closely around the user's specific data quality issues.

[0113] In a sleep analysis project for user U_X, the system automatically outputs a list of electrodes showing signal fluctuations based on the amplitude variation history of multiple past sleep monitoring sessions. This list may include electrodes such as Fp1, F7, and T3. This list is the output of step S31, marking the formal entry of the personalized analysis for U_X into the solution construction phase.

[0114] S32 Based on the similarity between the signal fluctuation electrode of the analysis and processing target and other analysis and processing targets, the analysis and processing targets that belong to the signal fluctuation electrode in the signal fluctuation electrode are taken as fluctuation-related targets.

[0115] Fluctuation-related targets: For a specific signal fluctuation electrode, among all other analysis and processing targets besides the current target user, the set of users that are also identified as signal fluctuation electrodes. Their size reflects the prevalence of the electrode's instability.

[0116] Similarity: This is defined here as whether two different analysis target users have the same "fluctuation" attribute on a specified electrode. If they are the same, they are considered to have a high degree of similarity on that electrode.

[0117] This step elevates the analytical perspective from individual users to the user group level. Its significance lies in tracing and classifying the sources of interference. By calculating the number of targets associated with the fluctuations of each signal-fluctuating electrode, it can be determined whether the instability of that electrode is a problem specific to that individual user (such as personal wearing habits or local physiological structures) or an inherent, widespread interference characteristic across multiple users (such as its location in easily movable or sweaty areas). This distinction is crucial for the branching decisions in the decision tree in subsequent step S33 (especially in cases 3 and 4), enabling the processing solution to focus on interference patterns with greater general research value and practical guidance.

[0118] For the signal fluctuation electrode Fp1 of user U_X, the system queried the database of another 200 target users for analysis and processing, and found that 160 users also marked Fp1 as a signal fluctuation electrode. Therefore, these 160 users constitute the fluctuation correlation targets for electrode Fp1. This high correlation number suggests that Fp1 may be a generally susceptible electrode location.

[0119] S33 determines the identification and processing scheme of the EEG signal staging model of the target being analyzed and processed based on the signal fluctuation electrodes of the target being analyzed and processed, the fluctuation-related targets in different signal fluctuation electrodes, and the signal stabilizing electrodes.

[0120] The identification and processing scheme of the EEG signal staging model refers to a series of specific instructions generated for the target user of the current analysis and processing. The core is to define a set of signal analysis combinations to be tested and their generation rules.

[0121] Signal analysis combination: A subset of electrodes consisting of a basic combination and a wave electrode combination, which serves as the primary input to the phased model.

[0122] Basic assembly: A fixed set consisting of all signal stabilizing electrodes of the current user, serving as the reliability foundation for all comparative tests.

[0123] Wave electrode combination: A subset selected from the list of signal wave electrodes according to specific rules.

[0124] Focus on fluctuating electrodes: These are signal fluctuating electrodes whose number of fluctuating associated targets exceeds a certain threshold, representing unstable electrodes with a general population impact.

[0125] The model construction process using EEG signals as input and sleep stages as output is as follows: First, the raw multi-lead EEG signals are preprocessed, including filtering to remove power frequency and EMG artifacts, resampling to a uniform frequency, and applying segmentation and standardization operations. Then, time-frequency analysis methods (such as wavelet transform or short-time Fourier transform) are used to extract power spectrum features covering delta, theta, alpha, sigma, and beta bands from each channel signal. At the same time, time-domain nonlinear features (such as sample entropy and spectral entropy) and cross-channel synchronicity features (such as phase-locking values) are calculated to construct a high-dimensional feature vector. Then, this feature sequence is input into a hybrid neural network consisting of a one-dimensional convolutional layer and a bidirectional long short-term memory network layer. The convolutional layer is used to capture local spatiotemporal patterns, and the recurrent network layer is used to model the temporal dependence of sleep stages. Finally, a conditional random field layer is introduced at the output to perform temporal constraint optimization on the stage probability sequence output by the network, thereby generating a final sequence that conforms to the continuous law of sleep structure and is labeled with Wake, N1, N2, N3, and REM.

[0126] It should be noted that the identification and processing scheme for determining the EEG signal staging model of the analysis and processing target specifically includes:

[0127] Use signal stabilizing electrodes as the basic combination;

[0128] Case 1: Obtain the signal stabilizing electrode of the target for analysis and processing. If the proportion of the signal stabilizing electrode among all electrodes is greater than the preset electrode proportion threshold, freely combine different signal fluctuation electrodes to construct a fluctuation electrode combination. Combine the fluctuation electrode combination with the basic combination to form a signal analysis combination for the EEG signal staging model. Perform EEG staging processing based on the signal analysis combination.

[0129] Preset electrode proportion threshold: a threshold used to determine whether the signal-stable electrode dominates the total number of electrodes.

[0130] When the proportion of stable electrodes exceeds a preset electrode proportion threshold, it indicates that the user possesses a very sufficient and reliable signal baseline. At this point, the computational cost of comprehensively testing all possible combinations of fluctuating electrodes is relatively controllable, and the potential impact of each fluctuating electrode or any combination thereof can be explored without omission. The significance of this setup is to conduct the most comprehensive "stress test" on users with extremely high-quality data foundations to draw a complete impact spectrum, providing rich data for model robustness evaluation.

[0131] User U_A has a high percentage of stable electrodes. The system determines that condition 1 is met, and then performs a full combination enumeration on all signal fluctuation electrodes (e.g., 5) of U_A, generating multiple fluctuation electrode combinations such as "{fluctuating electrode A}", "{A, B}", up to including all 5. Each combination is then merged with the basic combination to form the final signal analysis combination for testing.

[0132] Case 2: If the proportion of the signal stabilizing electrode in all electrodes is not greater than the preset electrode proportion threshold, and if the number of signal fluctuation electrodes of the analysis and processing target is less than the preset fluctuation electrode number threshold, different signal fluctuation electrodes are freely combined to construct a fluctuation electrode combination. The fluctuation electrode combination is combined with the basic combination to form a signal analysis combination of the EEG signal staging model. EEG staging processing is performed based on the signal analysis combination.

[0133] Preset threshold for the number of oscillating electrodes: a threshold used to determine whether the total number of signal oscillating electrodes is too small.

[0134] When the total number of signal fluctuation electrodes is less than a preset threshold, even if the proportion of stable electrodes is not high, the computational burden of comprehensive testing remains light because the total number of combinations to be tested is limited (the number of combinations increases exponentially with the number of fluctuation electrodes). Simultaneously, the impact on the overall accuracy of the test results is minimal. The significance of this is that for users with only a few unstable electrodes, the exhaustive method can obtain a precise measure of the impact of these few "outliers" on the results at low cost, ensuring thorough analysis while avoiding the influence of too many signal fluctuation electrodes.

[0135] User U_B has only 3 signal fluctuation electrodes. The system determines that condition 2 is met, and then freely combines these 3 electrodes to generate 7 non-empty fluctuation electrode combinations, thereby forming 7 signal analysis combinations for phased processing.

[0136] Case 3: If the number of signal fluctuation electrodes of the analysis and processing target is not less than the preset threshold for the number of fluctuation electrodes, the average value of the fluctuation-related targets in different signal fluctuation electrodes is determined based on the fluctuation-related targets in different signal fluctuation electrodes. When the average value of the number of fluctuation-related targets in different signal fluctuation electrodes is greater than the preset threshold for the number of fluctuation-related targets, the different signal fluctuation electrodes are freely combined to construct a fluctuation electrode combination. The fluctuation electrode combination is combined with the basic combination to form a signal analysis combination of the EEG signal staging model. EEG staging processing is performed based on the signal analysis combination.

[0137] Average value of fluctuation-related targets: refers to the arithmetic mean of the number of fluctuation-related targets of all signal fluctuation electrodes of this user.

[0138] Preset threshold for the number of fluctuation-related targets: a threshold value used to determine whether the number of fluctuation-related targets is large enough.

[0139] When there are many user signal fluctuation electrodes and their average prevalence (the average value of fluctuation-related targets) is high, it means that these electrodes are likely to be sources of systemic interference. Although a comprehensive test involves many combinations and is computationally expensive, it is still necessary to conduct a comprehensive test to discover their common impact patterns because the interference patterns of these electrodes may have universal research value. The significance of this setup is that it strikes a balance between computational resources and potential major discoveries; when the prevalence is high enough, resources are inclined to be invested in in-depth exploration.

[0140] User U_C has 20 signal fluctuation electrodes, and the average value of their fluctuation-related targets is very high. The system determines that condition 3 is met. Despite the huge number of combinations, it still performs free combination of all 20 electrodes to generate a massive number of signal analysis combinations, aiming to explore the complex correlation and influence patterns among these highly prevalent fluctuation electrodes.

[0141] Case 4: If the average number of wave-related targets in different signal wave electrodes is not greater than a preset threshold for the number of wave-related targets, the signal wave electrodes with a number of wave-related targets greater than the preset threshold for the number of wave-related targets are taken as the wave electrodes of interest. Different wave electrodes of interest are freely combined to construct wave electrode combinations. The wave electrode combinations are combined with the basic combinations to form the signal analysis combination of the EEG signal staging model. EEG staging is performed based on the signal analysis combination.

[0142] Focus on volatile electrodes: In this case, it specifically refers to electrodes whose number of volatile associated targets exceeds a preset threshold for the number of volatile associated targets, that is, a universal subset selected from a large number of volatile electrodes.

[0143] When there are many user signal fluctuation electrodes but their overall prevalence is low, comprehensive testing may lead to insufficient accuracy of the overall test results. Since the reference staging result is determined based on the number of schemes with the highest staging consistency, exclusion can ensure the accuracy of the reference staging result. In this case, the strategy shifts to focusing on those fluctuation electrodes of general interest. This essentially performs a filtering process, concentrating analytical resources on electrodes most likely to generate transferable knowledge and have the greatest clinical or engineering guidance, while temporarily ignoring user-specific interference factors. The significance of this setup is that it significantly improves computational efficiency and analytical focus in large-scale analysis scenarios, allowing research to focus on solving common problems.

[0144] Specific examples:

[0145] User U_D has 18 signal fluctuation electrodes, but only 4 of them have a number of fluctuation-related targets exceeding the high threshold. The system determines that condition 4 is met, and then only performs free combination testing on these 4 fluctuation electrodes of interest, thereby reducing the number of signal analysis combinations that need to be tested from millions to 15, ensuring the accuracy of the reference staging results.

[0146] It should be noted that, in all cases, all electrodes are combined for signal analysis, and EEG staging is performed based on this combination of signals.

[0147] In one possible specific embodiment:

[0148] In a large, multicenter sleep EEG study, 300 participants have been identified as targets for analysis. This study focuses on one participant, U_E, and aims to determine the identification and processing strategy for their EEG signal staging model.

[0149] Given conditions: User U_E is using a 128-lead EEG cap. Preliminary analysis determined that they have 80 electrodes with stable signals, 35 electrodes with fluctuating signals, and the states of the remaining 13 electrodes were not included in this stage of analysis.

[0150] Execute step S31:

[0151] The system calls the electrode stability file of U_E and directly reads the list of identifiers of its 35 signal fluctuation electrodes, such as Fp1, Fp2, AF7, AF8, F7, F8, FT9, FT10, etc.

[0152] Execute step S32:

[0153] The system queries the database of the remaining 299 analysis targets to determine the number of associated targets for each signal fluctuation electrode of U_E. For example, electrode Fp1 has 250 associated targets, electrode F7 has 180, and electrode CP5 has 40.

[0154] Step S33: First, calculate the stable electrode ratio of U_E: 80 / 128 = 62.5%. This value is not greater than the preset electrode ratio threshold of 70%, so case 1 is not applicable.

[0155] Secondly, the number of signal fluctuation electrodes for U_E is 35, which is not less than the preset threshold of 25 for the number of fluctuation electrodes, so case 2 is not applicable.

[0156] Next, the average value of the fluctuation-related targets for all 35 signal fluctuation electrodes in U_E is calculated. Assume this average value is 90. Since an average value of 90 is not greater than the preset threshold of 200 for the number of fluctuation-related targets, case 3 is not applicable.

[0157] Therefore, the system automatically applies the processing logic of Case 4 to filter out electrodes with more than 200 wave-related targets from the 35 signal wave electrodes. Assuming that three electrodes, Fp1 (250), Fp2 (245), and AF8 (210), are selected, these three electrodes are defined as the wave-related electrodes of interest, and the basic combination of 80 signal-stabilizing electrodes for user U_E is determined.

[0158] By performing all possible non-empty free combinations on the three wave electrodes of interest, a total of 7 (i.e. 2^3 - 1) different wave electrode combinations are generated, for example: {Fp1}, {Fp2}, {AF8}, {Fp1, Fp2}, {Fp1, AF8}, {Fp2, AF8}, {Fp1, Fp2, AF8}.

[0159] Each wave electrode combination is combined with the basic combination, and then combined with the combinations constructed from all electrode positions to generate eight signal analysis combinations to be tested.

[0160] This method achieves dynamic optimization of computational and analytical resources by constructing a closed-loop decision-making framework that integrates individual electrode stability diagnosis, group interference pattern correlation, and adaptive scheme generation. By introducing a group-based indicator of fluctuation correlation targets and combining it with intelligent judgment of four scenarios, the method can automatically identify and focus on the interfering electrodes with the most universal research value. In real-world scenarios with a large number of electrodes and a vast combination space, this effectively avoids the combinatorial explosion problem, concentrating limited computational resources on analytical paths that generate high-value insights, and significantly improving the efficiency of large-scale data analysis.

[0161] This method significantly enhances the transparency and interpretability of EEG signal staging analysis. By systematically constructing comparative experiments based on signal-stabilizing electrodes, it establishes a clear causal relationship analysis chain. Researchers can clearly identify which electrodes contribute to each staging result and precisely quantify the changes in results resulting from adding or removing specific combinations of fluctuating electrodes. This provides a solid data foundation for evaluating the robustness of existing staging models to data quality and for the decision-making basis of localization models.

[0162] S3 determines the interference risk type and interference electrode location of the analysis and processing target based on the deviation data of the EEG signal staging results between different recognition and processing schemes, determines the overlap of interference electrode locations of the analysis and processing target under different interference risk types, and determines the processing method of the EEG signal at the interference electrode location based on the overlap.

[0163] Specifically, such as Figure 4 As shown, the method for determining the interference risk type of the analysis and processing target is as follows:

[0164] In this embodiment, the interference patterns exhibited by the target user after implementing a customized identification and processing scheme are systematically evaluated and classified to determine the type of interference risk. The core logic is as follows: by comparing and analyzing multiple sleep staging results obtained based on different signal analysis combinations, an internal consensus voting mechanism is established to determine the most reliable reference staging result, thereby identifying deviation identification schemes that significantly deviate from the reference result. Finally, by statistically analyzing the occurrence patterns, frequency, and proportion of deviation identification schemes in different sleep staging processes, the user's interference impact is summarized into different levels of interference risk types. This method achieves the quantification and sublimation from specific staging result differences to abstract risk level judgment, laying the foundation for determining the reliability of subsequent interference electrode locations.

[0165] S41 uses each signal analysis combination as an identification processing scheme, and determines the reference staging result in the sleep staging process by using the deviation data of the EEG signal staging results between different identification processing schemes.

[0166] Identification and processing scheme: refers to each specific analysis instruction generated by the preceding step (S33) for the current analysis and processing target. Its core is to run the phased model using a specific signal analysis combination.

[0167] Deviation data of EEG signal staging results: refers to the difference measurement data between the staging results output by different identification and processing schemes (i.e. different signal analysis combinations) in the same sleep staging process, such as the staging consistency coefficient between different staging results, the difference in duration of each sleep stage, etc.

[0168] Reference staging result: This refers to the most frequent staging result among multiple staging results obtained from all current identification and processing schemes within a single sleep staging process. It represents the consensus reached by the most common electrode combination schemes in this analysis.

[0169] The significance of this step lies in establishing a decision-making mechanism to determine the most reliable "gold standard" in a single analysis. Since the user's signal stabilizing electrodes are included in most or all signal analysis combinations, results agreed upon by the majority (especially those based on stabilizing electrodes) have higher confidence. By comparing the deviations between results and determining the reference staging result according to the "majority rule" principle, this method can resist abnormal interference introduced by a few signal fluctuation electrodes, thus anchoring a relatively reliable benchmark for subsequent deviation judgment.

[0170] In the analysis of a sleep staging process for user U_X, the system executed 10 different identification and processing schemes, resulting in 10 staging results. Comparison revealed that 7 schemes yielded completely consistent staging results (e.g., all dividing the night's sleep into the same N1, N2, N3, REM stage sequence), while the results from the other 3 schemes differed. Therefore, the staging result supported by these 7 schemes was established as the reference staging result for this process.

[0171] S42 determines the identification and processing scheme that is inconsistent with the reference staging result based on the deviation between the sleep staging result of the identification and processing scheme and the reference staging result, and uses it as the deviation identification scheme.

[0172] Deviation identification scheme: A scheme for identifying and handling discrepancies between the output sleep staging results and the reference staging results during the same sleep staging process.

[0173] The significance of this step lies in precisely locating the source of the discrepancy. Identifying the deviation identification scheme is equivalent to identifying which specific signal analysis combinations (i.e., which electrode combinations) produced results that differed from the mainstream consensus. This is directly related to the specific signal fluctuation electrodes included in these combinations, and is crucial for subsequent analysis of interference sources and patterns.

[0174] Specific examples:

[0175] Continuing with the previous example, the three identification and processing schemes that produced outputs different from the reference staging results are marked as deviation identification schemes in this sleep staging process. Further analysis reveals that these three schemes may all contain a specific signal fluctuation electrode Fp1, suggesting that Fp1 may be a potential source of interference causing discrepancies in the results of this analysis.

[0176] S43 uses deviation identification scheme data in different sleep stages to determine the type of interference risk to the analysis and processing target.

[0177] Sleep staging process: refers to the complete monitoring and analysis process of a user's single sleep episode. A user typically has historical data from multiple sleep staging processes.

[0178] Interference Risk Type: This is a risk level classification based on the overall performance pattern of the deviation identification scheme during multiple sleep phases. It comprehensively reflects the severity and persistence of the interference.

[0179] Type 1 risk: Highest risk level.

[0180] Category II risk type: Medium risk level.

[0181] Risk-free type: Lowest risk level.

[0182] It should be noted that the reference staging result is the sleep staging result with the highest number of consistent identification and processing schemes in the sleep staging process.

[0183] It is understandable that if a bias identification scheme exists in different sleep stages, then the interference risk type of the analysis and processing target is determined to be a type of risk.

[0184] If at least one biased identification scheme exists in every stage of a user's historical sleep stages, it indicates that the interference is persistent and systematic. This means that regardless of the electrode combination, some schemes will always deviate from the consensus due to interference, indicating a fundamental and long-term instability in the user's EEG signal quality. The significance of this setup lies in identifying the highest-risk users, and the reliability of their data needs to be given high priority.

[0185] In the analysis of user U_A's past 10 sleep staging processes, there was at least one deviation detection scheme in each instance. The system determined that the user's interference risk was a type I risk, indicating that the user may have a long-term, stable source of interference (such as a permanent poor contact of a certain electrode or the presence of a significant physiological artifact).

[0186] Furthermore, if there are uneven bias identification schemes in different sleep stages, the following should also be included:

[0187] S431 determines the number of deviation identification schemes in different sleep stages based on deviation identification scheme data in different sleep stages.

[0188] Number of deviation identification schemes: The number of deviation identification schemes in a single sleep staging process where a deviation exists.

[0189] The significance of this step lies in quantifying the "breadth" of interference in a single analysis. A large number of biased schemes means that multiple different electrode combinations have produced anomalous results, suggesting that the interference may not originate from a single electrode, but rather from a broader signal quality issue or model instability on the user's data.

[0190] S432 determines whether the average proportion of the deviation identification scheme in different sleep stages among all identification and processing schemes is greater than the preset deviation scheme proportion threshold. If so, the interference risk type of the analysis and processing target is determined to be a type of risk. If not, proceed to step S433.

[0191] The average proportion of deviation identification schemes among all identification and processing schemes: First, calculate the proportion of the number of deviation schemes to the total number of schemes in each process with deviation; then calculate the average of this proportion for all processes with deviation.

[0192] Preset deviation scheme percentage threshold: a standard used to determine whether the above average percentage is too high.

[0193] The significance of this step lies in assessing the "intensity" of interference. Even if deviations don't occur in every process, a high average percentage of deviations in those processes indicates a significant impact of interference, affecting a large number of analytical schemes. This is also a high-risk characteristic, and therefore it is classified as a risk type. This prevents a few minor deviations from lowering the overall risk assessment, ensuring sensitive identification of significant interference patterns.

[0194] Specific examples:

[0195] User U_B experienced deviations in 5 processes. On average, the percentage of deviations in these 5 processes was as high as 60%. This means that when deviations occurred, more than half of the analytical results were unreliable. Even if he had 5 other processes without any deviations, this highly disruptive intermittent interference could still classify his risk as the highest risk category.

[0196] S433 takes the sleep staging process without a bias identification scheme as a reliable identification process, and determines whether the proportion of the reliable identification process in the sleep staging process is greater than a preset process proportion threshold. If so, the interference risk type of the analysis and processing target is determined to be a risk-free type. If not, the interference risk type of the analysis and processing target is determined to be a second-class risk type.

[0197] Reliable identification process: refers to sleep staging processes that do not have any biased identification schemes, that is, a "pure" process in which all analysis schemes yield consistent results.

[0198] Preset process proportion threshold: A standard used to determine whether the reliable identification process is dominant.

[0199] The significance of this step lies in distinguishing between medium and low risk. When the disturbance is not systematic and its intensity is low, the risk level is determined by the "purity" of the data's reliability.

[0200] If the proportion of reliable identification processes is very high (exceeding the threshold), it indicates that the user's data quality is excellent most of the time, and occasional interference can be regarded as accidental events, so it is rated as a risk-free type.

[0201] If the proportion of reliable identification processes is low, it indicates that a significant portion of the user's processes are subject to interference (even if slight), resulting in some uncertainty in data quality; therefore, it is classified as a Category II risk. This setting achieves refined risk stratification.

[0202] In one possible specific embodiment, an interference risk type assessment is performed on the target user U_C for analysis and processing. U_C has analysis data from 8 historical sleep staging processes. For each process, the system previously generated 10 different identification and processing schemes for staging.

[0203] Execute step S41 (for each process):

[0204] For the first process, among the 10 options, 8 options give the same phased results (e.g., result R1). This result is determined as the reference phased result for this process. Similarly, reference phased results are determined for the remaining 7 processes.

[0205] Execute step S42 (for each process):

[0206] In the first process, there were two schemes that differed from the reference result R1. These two schemes were marked as deviation identification schemes for this process. After counting all 8 processes, it was found that deviation identification schemes existed in processes 1, 3, 4, 5 and 7 (a total of 5 processes had deviations). All schemes in processes 2, 6 and 8 had consistent results and no deviation schemes (i.e. these 3 processes were reliable identification processes).

[0207] Execute step S43:

[0208] First, we determine that since not all (8 times) of the process have deviations, we proceed to the analysis process of Case 2.

[0209] Execute sub-step S431: Count the number of deviation identification schemes in each of the 5 processes where deviations occurred. Assume the numbers are 2, 1, 3, 1, and 2 respectively.

[0210] Execute sub-step S432:

[0211] In each process where deviations exist, the percentage of the deviation scheme is calculated as follows: 2 / 10 = 20% for the first time, 1 / 10 = 10% for the third time, 3 / 10 = 30% for the fourth time, 1 / 10 = 10% for the fifth time, and 2 / 10 = 20% for the seventh time. The average of these five percentages is calculated as: (20% + 10% + 30% + 10% + 20%) / 5 = 18%. The preset deviation scheme percentage threshold in this embodiment is set to 25%.

[0212] Judgment: The average value of 18% is not greater than the threshold of 25%, so it does not meet the conditions for being classified into a risk type. Proceed to sub-step S433.

[0213] Execute sub-step S433: Determine that the reliable identification process is the 2nd, 6th, and 8th times, for a total of 3 times. Calculate the proportion of reliable identification processes: 3 / 8 = 37.5%. Set the preset process proportion threshold of this embodiment to 50%. Determine: 37.5% is not greater than 50%.

[0214] Final ruling: User U_C's interference risk type was determined to be Category II. This indicates that U_C experiences intermittent interference; approximately 37.5% of the data is of high quality, but slight to moderate discrepancies occur in 62.5% of the process, thus the reliability of his sleep stage identification results is relatively high.

[0215] This method achieves objective and quantitative rating of user data interference risks by constructing a complete evaluation chain, from comparing results from multiple schemes and establishing internal consensus benchmarks to statistical classification of risk patterns. This method eliminates subjective experience-based judgments and transforms the abstract issue of "data quality" into specific and comparable types of interference risks (Class I, Class II, and no risk) through clearly defined mathematical statistical rules (such as majority consensus, proportion mean, and percentage threshold). This lays the foundation for subsequently using risk levels to determine the reliability of the interference electrode locations obtained by the user.

[0216] Furthermore, the first type of risk is greater than the second type of risk, and the second type of risk is greater than the risk-free type.

[0217] Specifically, the method for determining the processing method of the EEG signal at the location of the interfering electrodes is as follows:

[0218] In this embodiment, for an identified potential interfering electrode location (e.g., electrode "Fp1"), the specific processing strategy for the signal at that location is automatically determined based on its performance data across multiple target user groups. The core logic involves a multi-level, group-based risk assessment: first, initial triage is performed based on the size of the affected users; second, for smaller user groups, the reliability of the signal (through interference risk type) and the complexity of the problem pattern (through similarity clustering of matching deviation schemes) are further examined; finally, based on the assessment results, a choice is made between different processing paths: "directly perform feature inspection and potential exclusion," "conduct in-depth pattern analysis before making a decision," and "temporarily leave it alone." This method aims to extract knowledge from group data to guide the standardized and intelligent management of specific electrode locations.

[0219] S51 uses the interference electrode location data in each analysis and processing target to determine the analysis and processing target whose electrode location belongs to the interference electrode location, and uses it as the interference analysis target to determine the interference risk type of the interference analysis target.

[0220] In the above steps, if the number of interference analysis targets at the interference electrode location is greater than the preset threshold for the number of interference targets, then the method for processing the EEG signal at the interference electrode location is to extract and process the EEG feature signals in different users, and determine whether there is an abnormality at the interference electrode location based on the EEG feature signals. When there is an abnormality, the interference electrode location does not need to be considered when performing sleep staging.

[0221] Additionally, it can be understood that if the number of interfering electrode locations is not greater than a preset threshold for the number of interfering targets, and if there is no risk type among the interfering risk types of the interfering analysis targets, it indicates that the probability of interfering electrode locations having an interfering risk is relatively high. Therefore, the processing method for determining the EEG signals of the interfering electrode locations is to extract and process EEG feature signals from different users, and determine whether there is an abnormality at the interfering electrode locations based on the EEG feature signals. When there is an abnormality, the interfering electrode locations do not need to be considered when performing sleep staging.

[0222] Furthermore, if there is a risk type among the interference risk types of the interference analysis target, then proceed to step S52.

[0223] Interference electrode location: refers to a specific electrode channel location that is identified as potentially problematic among multiple users (such as "C3", "O2").

[0224] Interference analysis target: refers to the target users of analysis and processing whose specific interference electrode locations are marked as signal fluctuation electrodes or related to deviation identification schemes in their data analysis.

[0225] Interference risk type: refers to the risk level assigned to the target user of the interference analysis according to the aforementioned method (such as step S43), including type I risk, type II risk, etc.

[0226] Preset threshold for the number of interference targets: a standard used to determine whether the size of the user group that is concerned about a certain electrode location is large enough.

[0227] The significance of this step lies in performing an efficient first-level triage, addressing the most common and well-defined situations. If an electrode location experiences problems across a large number of users (a large number of targets for interference analysis), this strongly suggests a widespread defect or inherent vulnerability to interference at that location. Regardless of individual risk levels, a standardized inspection process should be initiated first. This ensures that system resources address the most widespread potential problems first.

[0228] Conversely, if only a few users experience problems at that location, more careful identification is required. In this case, interference risk type is introduced as a second layer of filter: if none of these few users exhibit the most severe type of risk, it indicates that the probability of that location being incorrectly identified as an interference electrode location due to inaccurate sleep staging of the users is relatively low, and the standard inspection process is still used; only when the few problematic users include high-risk users does it mean that it may be necessary to investigate more complex interference patterns associated with severe risks, thereby triggering subsequent more refined analysis steps (S52).

[0229] The specific method for determining whether an abnormality exists, and the extraction and anomaly detection of EEG feature signals, is as follows: First, within the defined interference analysis target or all users, a robust feature set is extracted from the preprocessed EEG signals at specified electrode locations (or electrode combinations). This includes, but is not limited to, the absolute and relative power of specific frequency bands (such as Delta and Theta), the mean and variance of signal amplitude, sample entropy, and the coherence coefficient between the electrode and the signals of neighboring stable electrodes. Second, based on the above-mentioned feature set extracted from the same location (or combination) during the user's own historical reference sleep process... According to the data, the baseline distribution in its multidimensional feature space (such as the mean vector and covariance matrix) is calculated; then, for the current sleep process to be examined, the Mahalanobis distance between the extracted feature vector and the user's historical baseline distribution (or the weighted comprehensive score of the deviation of each dimension feature from the historical mean) is calculated as a quantitative indicator of the degree of deviation; finally, if the degree of deviation exceeds the dynamic threshold determined based on historical data statistics (such as the upper limit of the confidence interval of the historical baseline distribution), it is determined that the EEG feature signal of the electrode position (or electrode combination) in the current process is abnormal, that is, inconsistent with the normal state.

[0230] Specific examples:

[0231] Suppose we focus on electrode location "T7". Statistics show that out of 300 users analyzed, 240 identified "T7" as an interfering electrode location, meaning the number of interfering targets is 240. Due to this large number, far exceeding the preset threshold for the number of interfering targets, the system directly determines that the characteristic signal of the "T7" location should be extracted from all users for examination. If any abnormalities are found, the electrode will be ignored in the sleep stage analysis. This corresponds to the common situation where electrode "T7" may be easily compressed or subject to electromyographic interference due to its proximity to the ear.

[0232] S52 will identify the identification and processing scheme with identification bias in the target of analysis and processing, which is located at the position of the interfering electrode, as the matching bias scheme. Based on the similarity of the electrodes of the matching bias schemes of different targets of analysis and processing, the targets of analysis and processing with the same electrodes of the matching bias scheme will be classified into the same bias scheme combination.

[0233] In the above steps, the number of deviation scheme combinations is obtained, and it is determined whether the number of deviation scheme combinations is greater than the preset threshold for the number of deviation scheme combinations. If so, the deviation of other electrodes associated with the interference risk is relatively large. Therefore, when there are multiple other electrodes with deviations from the electrode position, the overall staging result may be deviated. Based on this, the EEG signal processing method for determining the position of the interfering electrode is to extract and process EEG feature signals in different users, and determine whether the position of the interfering electrode is abnormal based on the EEG feature signals. When there is an abnormality, the position of the interfering electrode does not need to be considered when performing sleep staging. If not, proceed to step S53.

[0234] Matching bias scheme: For a target user of interference analysis, the specific schemes that include the location of the interfering electrode and are marked as bias identification schemes in its historical analysis.

[0235] Deviation scheme combination: Users whose deviation schemes contain the exact same electrode combination are grouped together. This combination represents a fixed, associated electrode interference pattern when a specific electrode location causes a problem.

[0236] Preset deviation scheme combination quantity threshold: a quantity standard used to determine whether the deviation pattern caused by the position of the interfering electrode is concentrated or dispersed.

[0237] The significance of this step lies in further analyzing whether the interference is regular when a problem only affects a few high-risk users. By using cluster analysis to match the deviation scheme, we can observe "the degree of dispersion in the distribution of electrodes that cause problems together with fixed partner electrodes".

[0238] If the number of clustered deviation patterns is large (exceeding the threshold), it indicates that the problem pattern caused by the electrode location is highly discrete, inconsistent, and lacks a clear pattern. This "elusive" characteristic itself implies a higher risk, as it may reflect a complex or random interference mechanism. Therefore, the system will tend to adopt a conservative strategy, i.e., initiate standard inspection procedures.

[0239] If the number of clustered combinations is small, it indicates that the interference patterns are relatively concentrated and fixed, which makes it possible to conduct more accurate and costly analysis in the next step (such as extracting the characteristic signals of the entire electrode group), thus proceeding to S53.

[0240] Specific examples:

[0241] Continuing with the example of electrode "Fp1", let's assume it only causes problems among 50 high-risk users (one type of risk). We collect the matching deviation schemes from these 50 users and perform clustering. We find a wide variety of electrode combinations in these schemes, such as {[Fp1]}, {[Fp1, Fp2]}, {[Fp1, AF7]}, {[Fp1, F7, F3]}…, ultimately resulting in 40 different deviation scheme combinations. Since the number of combinations is extremely large (exceeding the preset threshold for the number of deviation scheme combinations), it indicates that when "Fp1" causes deviations, the "partner" electrodes involved are highly variable and the patterns are complex. Therefore, the system determines to apply the standard inspection procedure to "Fp1".

[0242] S53 determines the processing method for the EEG signal at the interference electrode location based on the interference analysis target and the deviation scheme combination data at the interference electrode location.

[0243] In the above steps, based on the deviation scheme combination data, the number of analysis and processing targets in the deviation scheme combination is determined, and it is determined whether there is a deviation scheme combination with a number of analysis and processing targets greater than a preset threshold for the number of analysis and processing targets. If so, proceed to the next step; otherwise, there is no need to extract EEG feature signals. Since the number of analysis and processing targets is small and the other electrodes associated with the identification deviation are relatively consistent, the risk of deviation is low, so there is no need to extract EEG feature signals.

[0244] A combination of bias schemes where the number of analysis and processing targets exceeds a preset threshold is designated as a focus scheme combination. EEG feature signals from all interfering electrode locations in the focus scheme combination are extracted and processed from all users. Only when the EEG feature signals from all interfering electrode locations are abnormal, i.e. inconsistent with the EEG feature signals during normal measurement, is it necessary to exclude all interfering electrode locations in the focus scheme combination before performing sleep staging processing.

[0245] Focus on scenario combinations: Among a limited number of deviation scenario combinations, those combinations that involve a number of users (the number of analysis and processing targets) exceeding a preset threshold for the number of analysis and processing targets represent a relatively dominant interference pattern involving a large number of users.

[0246] EEG characteristic signals: These are specific features extracted from raw EEG signals that characterize electrode status or signal quality, such as power in a specific frequency band, signal complexity, and coherence with neighboring electrodes.

[0247] The significance of this step lies in conducting a final cost-benefit analysis and decision-making for cases where the patterns are concentrated and fixed. When the interfering patterns are fixed (few combinations), but the number of users involved in the main patterns is also small, it means that the scope of the problem is very limited, and the cost-effectiveness of investing resources in in-depth feature inspection is low. Therefore, the decision is "no need to handle".

[0248] Only when there are a large number of concerned scenarios involving a significant number of users is it worthwhile to invest computational resources in in-depth verification. In this case, the processing strategy is upgraded: instead of checking the problematic electrode individually, the entire fixed electrode combination (all electrodes in the concerned scenario combination) is checked. Only when the characteristic signals of all electrodes in the combination are abnormal is the entire electrode group excluded in the phase. This setting is very precise and conservative: it improves the specificity of the judgment and avoids mistakenly excluding an entire functional electrode group due to a momentary abnormality in a single electrode.

[0249] It aligns with physiological and physical logic: when a group of spatially or functionally related electrodes simultaneously exhibit characteristic abnormalities, it is more likely to indicate a common source of interference (such as localized sweating or poor contact over a large area), in which case excluding the entire group is reasonable.

[0250] Decision-making based on "discrete distribution": The final solution also needs to consider the distribution of other electrode positions within the combination. If other electrodes are spatially discrete in the head, the probability of such a combination being abnormal is low, and the decision will be more conservative; if other electrodes are spatially concentrated, the probability of the decision supporting exclusion increases.

[0251] Specific examples:

[0252] For electrode "Pz", assuming that after S52, only two combinations of bias schemes are clustered: combination A {[Pz, POz]} (involving 15 users) and combination B {[Pz]} (involving 5 users).

[0253] In one possible specific embodiment

[0254] In a large EEG database, all the aforementioned analyses have been completed for 300 target individuals. The next step is to determine the EEG signal processing method for electrode location "AF3".

[0255] Execution step S51:

[0256] Statistics show that among the 300 targets analyzed, the "AF3" location of 240 users was identified as the location of the interfering electrode, meaning the number of interfering targets was 240.

[0257] The preset threshold for the number of interference targets is set to 200. If 240 is greater than 200, the process will proceed directly to the standard inspection procedure.

[0258] If the threshold is set to 250, the interference risk type of these 240 interference analysis targets is further examined. It is found that there are 12 users with a type 1 risk. Since there are users with a type 1 risk, the process proceeds to step S52.

[0259] Step S52: For the interference analysis targets, collect their respective matching deviation schemes (i.e., deviation identification schemes containing "AF3"), and cluster them according to the electrode combinations of these schemes to form deviation scheme combinations. For example, the matching schemes of users U1, U2, and U3 are all {[AF3, Fp1, F7]}, and they are grouped into the same group; the schemes of users U4 and U5 are {[AF3, F3]}, and they are grouped into another group.

[0260] Three different combinations of deviation schemes were obtained. The preset threshold for the number of deviation scheme combinations was set to 5. It was determined that 3 ≤ 5, so the condition for entering the standard check due to pattern dispersion was not met, and the process was transferred to step S53.

[0261] Execute step S53:

[0262] Analyze the number of users included in these three combinations of deviation schemes: Combination 1: {[AF3, Fp1, F7]}, containing 180 users; Combination 2: {[AF3, F3]}, containing 60 users; Combination 3: {[AF3]}, containing 3 users.

[0263] The preset threshold for the number of target users to be analyzed and processed is set to 50. It is determined that the number of users for combination 1 and combination 2 is greater than the threshold, so they are identified as the combination of the focus schemes.

[0264] The final processing method was determined as follows: For focus group combination 1 (electrode group [AF3, Fp1, F7]): EEG feature signals from three locations within this group were extracted from all users. The entire electrode group was excluded only when all three locations of the same user showed abnormalities during sleep staging.

[0265] For the second scheme (electrode group [AF3, F3]): the processing logic is the same as above, and both positions are excluded only when both signals are abnormal at the same time.

[0266] For combination 3 ([AF3] only): The automatic feature check process will not be started for the time being because the number of users involved has not reached the threshold.

[0267] The system integrates the distribution of other electrodes (Fp1, F7, F3) in the two focus scheme combinations and finds that they are all located in the frontal and frontal lobe regions, and are relatively concentrated in space. Therefore, it supports the decision logic of "if all electrodes in the same group are abnormal, the whole group will be excluded".

[0268] This method constructs a progressive decision tree, from group size and risk level to interference pattern similarity, to achieve intelligent and refined formulation of processing strategies for specific electrode locations, thus automating and standardizing the processing strategies. Based on historical group data, the system can automatically assign appropriate signal processing rules to each electrode location, reducing the subjectivity and workload of manual review and ensuring the consistency and efficiency of large-scale EEG data analysis, reflecting the hierarchical and economical nature of risk assessment. The method dynamically selects different strategies—from "rapid screening and exclusion" to "refined group verification" and then to "temporary no processing"—based on the breadth (number of users), depth (risk type), and pattern complexity of the problem's impact. This design ensures that computational resources are prioritized for "high-priority" problems with a wide impact, high risk, or unclear patterns, while low-cost or observational strategies are adopted for niche problems with limited impact and clear patterns, optimizing overall resource utilization.

[0269] Example 2

[0270] In a second aspect, the present invention provides a computer device comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the above-described method for processing electroencephalogram signals based on a non-invasive brain-computer interface when running the computer program.

[0271] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0272] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0273] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A non-invasive brain-computer interface based electroencephalogram signal processing method, characterized in that, Specifically, it includes: Based on the analysis results of brain-computer interface data, the similarity of EEG signals between different electrodes is determined. Based on the similarity, a reference sleep process is determined during sleep. Based on the reference sleep process data and the variation of amplitude data of EEG signals in different reference sleep processes, the analysis and processing target of EEG signals in the user is determined. Based on the variation of amplitude data of electrodes of the target being analyzed, and combined with the variation risk of amplitude data of electrodes of the target being analyzed, the similarity of the target being analyzed, and the signal-stable electrodes, the identification and processing scheme of the EEG signal staging model of the target being analyzed is determined. Based on the deviation data of EEG signal staging results between different recognition and processing schemes, the interference risk type and interference electrode location of the analysis and processing target are determined. According to the similarity of the recognition and processing schemes with recognition deviations at the interference electrode locations in each analysis and processing target, and the interference risk type of the analysis and processing target, the processing method of the EEG signal at the interference electrode location is determined.

2. The EEG signal processing method based on a non-invasive brain-computer interface as described in claim 1, characterized in that, The analysis results of the brain-computer interface data are determined based on the monitoring data of the user's electroencephalogram (EEG) signals in different electrodes.

3. The EEG signal processing method based on a non-invasive brain-computer interface as described in claim 1, characterized in that, The similarity of the EEG signals between the electrodes is determined based on the similarity coefficient of the EEG signal characteristics between the electrodes at the same monitoring time.

4. The EEG signal processing method based on a non-invasive brain-computer interface as described in claim 1, characterized in that, The method for determining the reference sleep process during the aforementioned sleep process is as follows: Based on the similarity of EEG signals between electrodes at different monitoring times during the sleep process, the monitoring times at which the similarity coefficient of EEG signal features meets the requirements are determined and used as consistent monitoring times. Based on consistent monitoring time data during the sleep process, it is determined whether the sleep process belongs to a reference sleep process.

5. The EEG signal processing method based on a non-invasive brain-computer interface as described in claim 4, characterized in that, If the proportion of consistent monitoring times in the monitoring times during the sleep process is greater than a preset threshold for the proportion of consistent times, then the sleep process is determined to be a reference sleep process, that is, a sleep process with a low risk of external interference.

6. The EEG signal processing method based on a non-invasive brain-computer interface as described in claim 1, characterized in that, The reference sleep process data includes the proportion of the user's reference sleep process during their sleep process.

7. The EEG signal processing method based on a non-invasive brain-computer interface as described in claim 1, characterized in that, The method for determining the target of analysis and processing of the user's electroencephalogram (EEG) signals is as follows: Based on the reference sleep process data, the proportion of the reference sleep process in the user's sleep process is determined, and the proportion of the reference sleep process in the user's sleep process is used as the reference process proportion. By analyzing the changes in the amplitude data of EEG signals during different reference sleep processes, the variation in the amplitude range of EEG signals between different reference sleep processes can be determined. Based on the reference process ratio and the variation in the amplitude range of EEG signals between different reference sleep processes, it is determined whether the user belongs to the target of EEG signal analysis and processing.

8. The EEG signal processing method based on a non-invasive brain-computer interface as described in claim 1, characterized in that, The method for determining the processing method of the EEG signal at the location of the interference electrodes is as follows: Using the interference electrode location data in each analysis and processing target, the analysis and processing target whose electrode location belongs to the interference electrode location is determined, and this target is used as the interference analysis target to determine the interference risk type of the interference analysis target. The identification and processing schemes with identification biases, which are located at the positions of interfering electrodes in the analysis and processing targets, are used as matching bias schemes. Based on the similarity of the electrodes of the matching bias schemes of different analysis and processing targets, analysis and processing targets with consistent electrodes in the matching bias schemes are grouped into the same bias scheme combination. Based on the interference analysis target and the combination data of the deviation scheme at the location of the interference electrode, a method for processing the EEG signal at the location of the interference electrode is determined.

9. The EEG signal processing method based on a non-invasive brain-computer interface as described in claim 8, characterized in that, If the number of interference analysis targets at the interference electrode location is greater than a preset threshold for the number of interference targets, then the processing method for the EEG signal at the interference electrode location is to extract and process EEG feature signals from different users, and determine whether there is an abnormality at the interference electrode location based on the EEG feature signals. When there is an abnormality, the interference electrode location does not need to be considered when performing sleep staging.

10. A computer device comprising: A memory and processor connected in communication, and a computer program stored in the memory and capable of running on the processor, characterized in that, when the processor runs the computer program, it executes a brain-computer interface-based electroencephalogram (EEG) signal processing method according to any one of claims 1-9.

Images