A brain wave based attention monitoring and alarm method, device and equipment

By constructing a clean EEG database, conducting chaotic dynamics analysis and personalized intervention, the problem of real-time and accurate monitoring of attention states in industrial work environments has been solved, enabling precise early warning and personalized alarms, and improving the foresight and effectiveness of attention risk management.

CN120713523BActive Publication Date: 2025-11-28BEIJING NAOZHI XINGYAO TECHNOLOGY CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202511135288.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-08-14
Publication Date
2025-11-28
Estimated Expiration
2045-08-14

AI Technical Summary

Technical Problem

Existing technologies struggle to achieve real-time, accurate monitoring and personalized early warning of workers' attention in industrial environments. Traditional methods lack in-depth analysis and personalized adaptation of brain cognitive activities, resulting in high false alarm rates and poor practicality.

Method used

By collecting EEG signals, constructing a purified EEG database, extracting alpha and beta wave power spectra, performing chaotic dynamics analysis, identifying attentional state domains, predicting critical decay moments, designing personalized intervention strategies, generating audio-visual and vibration alarm signals, and achieving multi-channel allocation and progressive stimulation.

Benefits of technology

It enables precise identification and intelligent early warning of attention states, improves the scientific nature and accuracy of monitoring, transforms into proactive prevention, and enhances the effectiveness and individual adaptability of alarms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120713523B_ABST
    Figure CN120713523B_ABST
Patent Text Reader

Abstract

The application discloses a brain wave-based attention monitoring and alarming method, device and equipment, constructs a purified brain wave library by collecting brain wave signals of operating personnel and carrying out artifact removal processing; extracts alpha wave and beta wave power spectrum, carries out chaos dynamics analysis on the alpha wave power spectrum to obtain a traction response degree, identifies attention anti-interference ability, and fuses to generate an attention state domain; carries out cluster analysis on the state domain to identify a high and low vigilance area, extracts peak brain wave features and a falling rate parameter, and predicts a critical attenuation moment; releases the peak brain wave features to generate a pre-charge activation sequence, carries out timing optimization and personalized intervention parameter matching; distributes multi-channel stimulation based on an early warning advance, decomposes the pre-charge activation sequence into a gradual stimulation string; extracts a stimulation parameter sequence to generate a brain wave synchronous stimulation signal through rhythm coding, converts the brain wave synchronous stimulation signal into an audible, visual and vibratory alarm signal, and realizes intelligent monitoring and timely alarming of the attention state of the operating personnel.
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biological signal processing, in particular to a brain electrical activity-based attention monitoring and alarm method, device and equipment. BACKGROUND

[0002] The cognitive state of personnel in an industrial operation environment directly affects operation safety and production efficiency, and fatigue and distraction are one of the main factors leading to operation accidents. Traditional personnel state monitoring mainly relies on external behavior observation and subjective evaluation, which has strong lag and insufficient accuracy, and it is difficult to realize real-time early warning and active intervention.

[0003] Existing biological signal monitoring technologies mostly use peripheral physiological indicators such as electrocardiogram and electromyogram, which indirectly reflect the brain cognitive activity and have limited sensitivity. Although some studies attempt to use brain electrical signals for state evaluation, they mostly stay at the level of simple frequency spectrum statistical analysis, lack of mining of deep dynamic characteristics of signals, and cannot accurately capture the complex change pattern of cognitive state. In addition, existing methods generally lack personalized adaptation mechanism, making it difficult to customize monitoring for physiological characteristics of different individuals, resulting in high false alarm rate and poor practicability. Therefore, there is an urgent need to develop a personnel monitoring technology that can deeply analyze the brain electrical activity rule, realize accurate state recognition and intelligent early warning. SUMMARY

[0004] The present application provides a brain electrical activity-based attention monitoring and alarm method, device and equipment, aiming to establish a purified brain electrical library through deep analysis and processing of the brain electrical signals of operating personnel, extract alpha wave and beta wave power spectrum and perform chaotic dynamics analysis, construct an attention state domain, identify high and low alert regions, predict critical attenuation time, design personalized intervention strategies, form progressive stimulation through multi-channel distribution, and finally generate sound and light vibration alarm signals, realizing intelligent monitoring, accurate early warning and timely alarm of the attention state of operating personnel.

[0005] The present application provides a brain electrical activity-based attention monitoring and alarm method, device and equipment, aiming to establish a purified brain electrical library through deep analysis and processing of the brain electrical signals of operating personnel, extract alpha wave and beta wave power spectrum and perform chaotic dynamics analysis, construct an attention state domain, identify high and low alert regions, predict critical attenuation time, design personalized intervention strategies, form progressive stimulation through multi-channel distribution, and finally generate sound and light vibration alarm signals, realizing intelligent monitoring, accurate early warning and timely alarm of the attention state of operating personnel.

[0006] The brain electrical signals of operating personnel are collected, and the brain electrical signals are subjected to artifact removal processing to construct a purified brain electrical library;

[0007] Based on the purified brain electrical library, alpha wave and beta wave power spectrum are extracted, chaotic dynamics analysis is performed on the alpha wave power spectrum to obtain a traction response degree, the attention disturbance resistance is identified based on the traction response degree, the beta wave power spectrum is fused with the attention disturbance resistance to generate an attention state domain;

[0008] performing cluster analysis on the attention state domain to identify a high vigilance zone and a low vigilance zone, extracting a peak brain electrical feature from the high vigilance zone through fluctuation analysis, extracting a decline rate parameter from the low vigilance zone, performing exhaustion prediction processing on the peak brain electrical feature and the decline rate parameter to generate a critical attenuation time point;

[0009] releasing the peak brain electrical feature before the critical attenuation time point to generate a pre-charge activation sequence, performing timing optimization processing on the pre-charge activation sequence to generate a ladder-type wake-up scheme, matching and adjusting the ladder-type wake-up scheme with the traction response to generate personalized intervention parameters, and determining an optimal stimulation time based on the personalized intervention parameters;

[0010] performing time difference analysis on the optimal stimulation time and the critical attenuation time point to generate a pre-warning lead time, performing multi-channel stimulation distribution based on the pre-warning lead time to generate a stimulation scheduling matrix, and decomposing the pre-charge activation sequence into an incremental stimulation string according to the stimulation scheduling matrix;

[0011] extracting a stimulation parameter sequence from the incremental stimulation string, performing rhythm coding on the stimulation parameter sequence to generate a brain electrical synchronization stimulation signal, and converting the brain electrical synchronization stimulation signal into an audible-light-vibration alarm signal.

[0012] The second aspect of the present application proposes an attention monitoring and alarm device based on brain electrical signals, comprising:

[0013] a signal acquisition module configured to acquire brain electrical signals of an operator, and perform artifact removal processing on the brain electrical signals to construct a purified brain electrical library;

[0014] an attention evaluation module configured to extract alpha wave and beta wave power spectra based on the purified brain electrical library, perform chaotic dynamics analysis on the alpha wave power spectrum to obtain a traction response, identify an attention disturbance resistance ability based on the traction response, and perform fusion processing on the beta wave power spectrum and the attention disturbance resistance ability to generate an attention state domain;

[0015] a mode storage module configured to perform cluster analysis on the attention state domain to identify a high vigilance zone and a low vigilance zone, extract a peak brain electrical feature from the high vigilance zone through fluctuation analysis, extract a decline rate parameter from the low vigilance zone, and perform exhaustion prediction processing on the peak brain electrical feature and the decline rate parameter to generate a critical attenuation time point;

[0016] an intervention design module configured to release the peak brain electrical feature before the critical attenuation time point to generate a pre-charge activation sequence, perform timing optimization processing on the pre-charge activation sequence to generate a ladder-type wake-up scheme, match and adjust the ladder-type wake-up scheme with the traction response to generate personalized intervention parameters, and determine an optimal stimulation time based on the personalized intervention parameters;

[0017] The scheduling control module is configured to perform time difference analysis on the optimal stimulation opportunity and the critical decay moment to generate an early warning lead time, perform multi-channel stimulation distribution based on the early warning lead time to generate a stimulation scheduling matrix, and decompose the pre-charge activation sequence into a progressive stimulation string according to the stimulation scheduling matrix.

[0018] The alarm execution module is configured to extract a stimulation parameter sequence from the progressive stimulation string, perform rhythm coding on the stimulation parameter sequence to generate an electroencephalogram synchronization stimulation signal, and convert the electroencephalogram synchronization stimulation signal into an audible, visual and vibratory alarm signal.

[0019] The third aspect of the present application provides a computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the electroencephalogram-based attention monitoring and alarm method disclosed in the first aspect when executing the program.

[0020] The beneficial effects of the present application are reflected in the following points: first, the nonlinear characteristics of the alpha wave power spectrum are analyzed by chaotic dynamics theory, the chaotic attractor is extracted and perturbation is injected to detect the critical point, the anti-disturbance characteristics of the attention system are quantified by obtaining the traction response, and a two-dimensional attention state domain is constructed combined with the beta wave power spectrum, which realizes the accurate quantification of the complex dynamic characteristics of the attention system that cannot be captured by traditional linear analysis, and improves the scientificity and accuracy of the attention state evaluation. Second, a complete technical link from high and low alertness area identification to critical decay prediction is established, the peak electroencephalogram features and the decline rate parameters are extracted by fluctuation analysis to predict the exhaustion, and the timing optimization processing of the pre-charge activation sequence and the stepwise wake-up scheme is combined to realize the early warning of attention decay and the accurate matching of individual intervention parameters, which changes passive monitoring into active prevention and improves the forward-looking and targetedness of attention risk control. Finally, the pre-charge activation sequence is decomposed into a progressive stimulation string of auditory, visual and tactile coordination by using multi-channel stimulation distribution and reticular activation path technology, a stimulation signal synchronized with individual electroencephalogram features is generated by rhythm coding, and an individualized audible, visual and vibratory alarm output is converted, which realizes the technical breakthrough from a single alarm mode to a multi-modal collaborative alarm, and improves the effectiveness and individual adaptability of the alarm.

[0021] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0022] The drawings herein show specific examples of the technical solutions described in the present application, and constitute part of the specification together with the specific embodiments, for explaining the technical solutions, principles and effects of the present application.

[0023] Unless otherwise specified or implied, the same reference signs in different drawings represent the same or similar technical features, and different reference signs can also be used to represent the same or similar technical features.

[0024] Figure 1 is a flowchart of a brain electrical activity-based attention monitoring and alarm method according to the present application.

[0025] Figure 2 is a structural block diagram of a brain electrical activity-based attention monitoring and alarm device according to the present application.

[0026] Figure 3 is a structural diagram of a computer device according to the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0028] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present application are only used to explain the relative positional relationship, movement condition, etc. between components in a certain posture (as shown in the drawings), and if the certain posture changes, the directional indications also change accordingly.

[0029] It should also be noted that when an element is referred to as being “fixed to” or “set on” another element, it can be directly on the other element or can have a middle element. When an element is referred to as being “connected” to another element, it can be directly connected to the other element or can have a middle element.

[0030] In addition, the description involving “first”, “second” and the like in the present application is only for the purpose of description, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the features defined with “first”, “second” can explicitly or implicitly include at least one of the features. In addition, the technical solutions of each embodiment can be combined with each other, but it must be based on the fact that a person skilled in the art can realize it, and when the combination of technical solutions contradicts each other or cannot be realized, it should be considered that the combination of technical solutions does not exist, and is not within the scope of protection claimed by the present application.

[0031] The technical solutions of the embodiments of the present application will be introduced below.

[0032] AsFigure 1 As shown, this embodiment of the invention provides a method for attention monitoring and alarm based on electroencephalography (EEG), including the following steps S110-S160:

[0033] Step S110: Collect the EEG signals of the workers, perform artifact removal processing on the EEG signals to construct a clean EEG database.

[0034] Specifically, the system collects electroencephalogram (EEG) signals from workers. A multi-channel EEG electrode array is deployed on the worker's head to record electrical activity signals from the cerebral cortex in real time. The EEG acquisition system employs a 64-channel high-density electrode configuration, with electrodes spatially arranged according to the international 10-20 standard system, covering major brain regions such as the frontal, parietal, temporal, and occipital lobes. The electrodes are made of Ag / AgCl material, with contact impedance controlled below 5kΩ to ensure the stability and reliability of signal acquisition. The EEG signal sampling frequency is set to 1000Hz, with a bandwidth range of 0.1-100Hz and a dynamic range greater than 110dB, capable of capturing the complete EEG spectrum from slow waves to fast waves. During acquisition, reference electrodes and ground electrodes are used to eliminate common-mode interference, and differential amplification technology is employed to improve the signal-to-noise ratio. The data acquisition system is equipped with a 24-bit analog-to-digital converter to ensure the accuracy and fidelity of signal digitization. The acquired EEG signals contain rich neural activity information, but also contain physiological artifacts such as electrooculography (EOG), electromyography (EMG), and electrocardiography (ECG), as well as technical artifacts such as power frequency interference and electrode movement. The raw EEG data is stored in standard EDF format, recording metadata information such as acquisition time, electrode location, and sampling parameters.

[0035] In some embodiments, the step of constructing a cleaned EEG database by performing artifact removal processing on the EEG signal includes: constructing an artifact feature map based on the EEG signal; performing phase inversion generation processing on the artifact feature map to obtain a cancellation waveform; performing superposition and neutralization processing on the cancellation waveform and the EEG signal; and establishing a cleaned EEG database based on the superposition and neutralization result.

[0036] Based on the electroencephalogram (EEG) signal, a feature map of artifacts is constructed. A multi-domain analysis method is used to extract features in three dimensions of time, frequency, and space. Time-domain feature analysis detects amplitude mutations, slope abnormalities, and duration characteristics of EEG signals through sliding window technology. The window length is set to 500 ms, and the overlap rate is 75%. The amplitude envelope A(t) = |H[x(t)]| of each channel signal is calculated, where A(t) is the amplitude envelope, H[x(t)] is the Hilbert transform, and x(t) is the original EEG signal. Frequency-domain feature analysis extracts the frequency spectrum characteristics of artifacts using short-time Fourier transform, focusing on characteristic frequency bands such as power frequency interference 50 Hz and its harmonics, electromyographic artifacts 20-200 Hz, and electrooculographic artifacts 0.5-4 Hz. Space-domain feature analysis identifies localized and diffuse artifacts by calculating the correlation and signal propagation pattern between electrodes. An artifact classification model is established to classify detected artifacts into categories such as electrooculographic artifacts, electromyographic artifacts, electrocardiographic artifacts, power frequency interference, and electrode artifacts. A pattern recognition algorithm is used to automatically label the spatiotemporal distribution characteristics of each type of artifact. An artifact intensity evaluation index is constructed to quantify the degree of pollution of EEG signals by different artifacts. A three-dimensional artifact feature map is generated, with the horizontal axis representing time, the vertical axis representing frequency, the depth axis representing spatial location, and the color coding representing artifact intensity. An artifact feature library is established to store the feature templates of typical artifact patterns, supporting fast artifact recognition and classification.

[0037] The artifact feature map is processed in reverse to obtain a cancellation waveform. Based on the constructed artifact feature map, a signal inversion technique is used to generate a cancellation waveform with the same amplitude and opposite phase as the artifact component. The generation process of the cancellation waveform needs to accurately maintain the time-frequency characteristics of the artifact signal, while achieving a 180-degree phase flip. For different types of artifacts, corresponding inversion generation algorithms are designed. For periodic artifacts such as power frequency interference, an adaptive notch filter is used to generate the inverse signal, and the center frequency and quality factor of the filter are dynamically adjusted according to the interference characteristics. For transient artifacts such as electrooculographic signals, wavelet decomposition is used to extract the artifact component, and then the wavelet coefficients are phase-flipped and reconstructed to generate the cancellation waveform. The generation function of the cancellation waveform is C(t) = -A(t) × cos(φ(t) + π), where C(t) is the cancellation waveform, A(t) is the artifact amplitude, φ(t) is the artifact phase, and π represents a 180-degree phase shift. A multi-channel cancellation strategy is designed to simultaneously process artifact signals from multiple electrode channels. A time synchronization mechanism for the cancellation waveform is established to ensure accurate time alignment between the cancellation signal and the artifact signal. The generated cancellation waveform has the same time-frequency characteristics as the artifact signal but opposite polarity, providing an accurate cancellation signal for subsequent superposition and neutralization processing.

[0038] The cancellation waveform is superimposed and neutralized with the brain electrical signal. An adaptive weighting strategy is adopted to dynamically adjust the cancellation strength according to the artifact intensity and distribution characteristics. The superposition operation model S_clean(t) = S_original(t) + α × C(t) is established, where S_clean(t) is the purified signal, S_original(t) is the original brain electrical signal, α is the adaptive weight coefficient, and C(t) is the cancellation waveform. The determination of the weight coefficient α is based on the least mean square error criterion, and the optimal weight value is solved by an iterative optimization algorithm. Real-time monitoring is implemented during the superposition process, and the signal-to-noise ratio change before and after superposition is calculated to ensure the artifact removal effect. A segmented processing strategy is adopted to divide long-term electroencephalogram recording into several time periods, and superposition and neutralization are independently performed on each time period. An evaluation index of neutralization effect is established, including artifact suppression ratio, signal distortion, spectral purity and other parameters. A multi-scale processing method is adopted to perform superposition and neutralization at different time scales, taking into account the removal of transient artifacts and persistent artifacts. The superposition and neutralization results include purified electroencephalogram and artifact removal report, providing a high-quality data source for the establishment of a purified electroencephalogram library.

[0039] Based on the superposition and neutralization results, a purified electroencephalogram library is established. A hierarchical storage architecture is adopted, and the data is classified and organized according to the data type, processing level, time sequence and other dimensions of the superposition and neutralization results. A data indexing system is established to record the acquisition time, subject information, processing parameters, quality rating and other metadata of each electroencephalogram recording. A data quality evaluation system is designed to quantitatively evaluate the purification effect by calculating the signal-to-noise ratio, spectral purity, artifact residual rate and other indicators. A data integrity checking mechanism is established to ensure the continuity and consistency of the data in the library. Data compression techniques are used to reduce storage space requirements while maintaining data losslessness. An efficient data retrieval interface is designed to support fast query of the required electroencephalogram data according to multiple conditions. A data backup and recovery mechanism is established to ensure the safety and reliability of the data in the library. A data export function is designed to support data output in multiple standard formats, meeting the needs of different analysis software. Statistical reports of the purified electroencephalogram library are generated, including data volume, coverage duration, quality distribution and other information.

[0040] In step S120, based on the purified electroencephalogram library, the alpha wave and beta wave power spectrum are extracted, the alpha wave power spectrum is subjected to chaotic dynamics analysis to obtain the traction response degree, the attention resistance ability is identified based on the traction response degree, and the beta wave power spectrum is fused with the attention resistance ability to generate an attention state domain.

[0041] Specifically, the alpha and beta power spectra are extracted based on the cleaned EEG database. The high-quality EEG signals in the cleaned EEG database are used as input, and the frequency domain analysis technique is used to extract the alpha and beta components closely related to the attention state. The frequency range of alpha wave is set to 8-13 Hz, and the frequency range of beta wave is set to 13-30 Hz. These two frequency bands are key indicators reflecting brain cognitive activity and attention level. Windowed Fourier transform is used for time-frequency analysis of EEG signals, with window length set to 2 seconds and overlap rate set to 50%, ensuring the balance of time resolution and frequency resolution of spectral analysis. The power spectral density PSD(f)=|X(f)| ² / T is calculated for each electrode channel, where PSD(f) is the power spectral density at frequency f, X(f) is the Fourier transform of the EEG signal, and T is the observation time window length. The alpha and beta components are extracted by a band-pass filter, and the filter is designed with a fourth-order Butterworth design, with a fluctuation of less than 0.5 dB in the passband and an attenuation of more than 40 dB in the stopband. The extracted alpha and beta signals are subjected to power calculation, and the Welch method is used to estimate the power spectrum to reduce the influence of random fluctuations on power estimation. The spatial distribution map of the power spectrum is established, and the power values at different electrode positions are mapped to the scalp surface to form the topological distribution of EEG power. The power spectrum is normalized by logarithmic transformation to eliminate the influence of individual differences on power level. The alpha and beta power spectrum database is generated, and the power spectrum features of each time period and each electrode position are recorded.

[0042] In some embodiments, the chaotic dynamics analysis of the alpha power spectrum obtains the traction response degree, including: extracting a chaotic attractor from the alpha power spectrum; injecting a perturbation into the chaotic attractor to detect a critical point; applying anti-chaos control at the critical point to obtain a stabilization difficulty; and backstepping the traction response degree based on the stabilization difficulty.

[0043] Chaotic attractor is extracted from the alpha wave power spectrum. The time series data of the alpha wave power spectrum is taken as the input, and the chaotic attractor structure hidden in it is extracted by using the phase space reconstruction technique. The chaotic attractor is a geometric structure that describes the long-term behavior of a chaotic system and reflects the internal dynamic law of the alpha wave power change. The phase space is reconstructed using the delay coordinate embedding method, the appropriate embedding dimension m and delay τ are selected, and the embedding vector X(t) = [x(t), x(t-τ), x(t-2τ),..., x(t-(m-1)τ)] is constructed, where x(t) is the alpha wave power time series and X(t) is the embedding vector. The minimum embedding dimension is determined by the false nearest neighbor method to ensure that the reconstructed phase space can fully unfold the geometric structure of the attractor. The mutual information function is used to determine the optimal delay parameter, so that the coordinate components in the phase space have the maximum information independence. The geometric characteristics of the attractor are identified in the reconstructed phase space, including the dimension, shape and topological structure of the attractor. The correlation dimension D_c of the attractor is calculated, which reflects the complexity and self-similarity of the attractor. The maximum Lyapunov exponent λ_max is used to judge the chaotic property of the system, and a positive value indicates the existence of chaotic behavior. The local stability of the attractor is analyzed, and the distribution of fixed points, periodic orbits and chaotic regions is identified.

[0044] Critical points are detected by injecting a small perturbation into the chaotic attractor. The injection position and injection method of the small perturbation are determined based on the phase space coordinates and trajectory density distribution of the chaotic attractor. The injection of the small perturbation is to identify the most sensitive region and critical transition point in different regions of the chaotic attractor. According to the signal amplitude characteristics and dynamic scale of the chaotic attractor, the form and strength of the perturbation signal are designed, and different types of perturbation signals such as white noise, sine wave and pulse are used for the frequency components of the chaotic attractor. Based on the original signal amplitude characteristics of the chaotic attractor, the perturbation strength is set to 1-5% of the amplitude of the chaotic attractor signal, ensuring that the perturbation is small enough not to destroy the original dynamic structure of the chaotic attractor. Along the trajectory path and high-density region of the chaotic attractor, the perturbation is injected at different positions in the phase space of the chaotic attractor, and the response characteristics of the entire chaotic attractor are systematically detected. The deviation degree and recovery time of the chaotic attractor trajectory after the injection of the perturbation are recorded, and the sensitivity of the chaotic attractor system to the perturbation is quantified. Based on the phase space structure of the chaotic attractor, the local Lyapunov exponent λ_local(x) is calculated, where x is the position in the phase space of the chaotic attractor, and the critical points in the chaotic attractor that are most sensitive to the perturbation are identified. For the spatial non-uniformity of the chaotic attractor, the spatial distribution characteristics of the critical points in the chaotic attractor are analyzed, and the regions with concentrated and dispersed critical points in the chaotic attractor are found.

[0045] The difficulty of stabilization is evaluated by applying anti-chaos control at the critical point. At the identified critical point, the OGY control method is used to apply small-amplitude control to the phase space coordinates of the critical point on the Poincare section, guiding the chaotic trajectory to the unstable periodic orbit. Based on the dynamic characteristics of the critical point and the requirements of the target trajectory, the control law u(t) = K[x(t) - x_target] is designed, where u(t) is the control input, K is the control gain matrix, and x_target is the target trajectory position near the critical point. According to the local stability characteristics of the critical point, the control gain is calculated through linearization analysis to ensure the stability and convergence of the control system implemented at the critical point. For the control response characteristics of different critical points, the control cost of the stabilization process is evaluated, including control energy, control time, and control accuracy at each critical point. Based on the control cost evaluation results of the critical points, the stabilization difficulty index SD = E_control / E_system is calculated, where SD is the stabilization difficulty of the critical point, E_control is the control energy at the critical point, and E_system is the natural energy of the system. By comparing the stabilization difficulty indexes of each critical point, the differences in stabilization difficulty of different critical points are analyzed, and the most difficult and easiest control regions are identified from all critical points.

[0046] For example, the response degree is determined based on the stabilization difficulty and the control cost index. The control cost index includes disturbance suppression time, energy consumption, and oscillation decay rate. The response level is determined based on the control cost index. The quantization parameter is set according to the response level. The stabilization difficulty is inversely mapped to generate the traction response degree using the quantization parameter.

[0047] The response level is determined based on the stabilization difficulty and the control cost index. The stabilization difficulty directly affects the resource consumption and time cost in the control process. The control cost index is evaluated using a multi-dimensional analysis method. Based on the numerical range of the stabilization difficulty, the consumption of time, energy, and dynamic characteristics is calculated. The basic cost coefficient is determined based on the size of the stabilization difficulty SD. The disturbance suppression time T_suppress = k1 × SD + T_base, where k1 is the influence coefficient of the stabilization difficulty on time, and T_base is the basic suppression time. The amplification factor of energy consumption is calculated based on the stabilization difficulty. The energy consumption E_cost = k2 × SD × ∫|u(t)|2dt, where k2 is the influence coefficient of the stabilization difficulty on energy consumption, and ∫|u(t)|2dt is the integral of the square of the control input over time. ²dt, where k2 is the influence coefficient of the stabilization difficulty on the energy consumption, and u(t) is the control input signal. Considering the influence of the stabilization difficulty on the oscillation characteristics, the oscillation decay rate D_rate=D_base / (1+k3xSD), where D_base is the basic decay rate, and k3 is the influence coefficient of the stabilization difficulty. The stabilization difficulty is taken as the weight adjustment factor, and the comprehensive evaluation function CI=(w1xT_suppress+w2xE_cost+w3x(1 / D_rate))x(1+αxSD) is established, where α is the comprehensive influence coefficient of the stabilization difficulty. Based on the joint distribution characteristics of the CI value and the stabilization difficulty, the response level is divided into three basic levels of high response, medium response, and low response according to the numerical range of the control cost index. The fuzzy set theory is used to process the fuzziness of the level boundary, and a continuous level evaluation system is established. Through statistical analysis of the stabilization difficulty, the statistical distribution characteristics of different levels are analyzed, and the typical numerical range and probability distribution of each level are determined.

[0048] Quantitative parameters are set according to the response level. Based on the three levels of high response, medium response, and low response determined, a dedicated quantitative parameter configuration scheme is established for each level. The setting of quantitative parameters needs to consider multiple factors such as level characteristics, individual differences, and time changes. For the high response level, the reference parameter values with high sensitivity are set, including small reference traction response, high response sensitivity coefficient, and short time constant; for the medium response level, the parameter values with medium sensitivity are set, and balanced response strength and moderate time constant are used; for the low response level, the reference traction response is set to be large, the sensitivity coefficient is low, and the time constant is long. According to the characteristics of each response level, a parameter adjustment mechanism is established, a fine adjustment strategy is used for the high response level, a coarse adjustment strategy is used for the low response level, and the baseline parameters are personalized adjusted according to the individual baseline characteristics and current state. Differentiated dynamic update algorithms are designed for different response levels, the high response level uses a fast response update mechanism, and the low response level uses a stability priority update strategy. Considering the mutual dependence between parameters in the same response level, an intra-level parameter coupling model is established to avoid parameter conflicts.

[0049] The stable difficulty is inversely mapped using a quantization parameter to generate a traction response. Inverse mapping is the conversion of the control domain's stable difficulty into the physiological domain's traction response characteristics. A mapping function TR = g(SD, P) is established, where TR is the traction response, SD is the stable difficulty, P is the quantization parameter vector, and g is a nonlinear mapping function. Multivariate nonlinear regression methods are used to determine the specific form of the mapping function, which includes polynomial functions, exponential functions, and neural network models. The least squares method or maximum likelihood estimation method is used to solve the parameters of the mapping function to ensure mapping accuracy and generalization ability. A quality evaluation system for mapping results is established to evaluate the mapping effect by calculating the mapping error and correlation coefficient. An inverse verification mechanism for mapping is designed to verify the consistency of the mapping by back-propagating the stable difficulty from the known traction response. A confidence interval calculation method for mapping is established to quantify the uncertainty and reliability of the mapping results. Cross-validation techniques are used to evaluate the generalization performance of the mapping model to ensure its effectiveness on new data. A real-time calculation algorithm for mapping is designed to support online fast calculation of traction response. A post-processing mechanism for mapping results is established, including smoothing filtering, outlier detection, and data fusion processing steps.

[0050] The attention disturbance resistance ability is identified based on the traction response. Attention disturbance resistance ability reflects the brain's ability to maintain attention in the face of external interference, and is a key indicator for evaluating the working state of operators. The mapping relationship between traction response and attention disturbance resistance ability is established, and high traction response usually corresponds to low attention disturbance resistance ability, indicating that the attention system is easily disturbed by external interference. A disturbance resistance grading system is designed to divide the attention disturbance resistance ability into high, medium, and low levels, each level corresponding to a different range of traction response. Fuzzy logic methods are used to handle the uncertainty of the grade boundaries to establish a continuous disturbance resistance evaluation function. The differences in traction response of different brain regions are analyzed, and the response of the frontal lobe region is usually related to executive control ability, while the response of the parietal lobe region is related to spatial attention. A multi-dimensional disturbance resistance evaluation model is established to consider the response characteristics of different brain regions and different frequency bands. The time stability of the disturbance resistance ability is calculated to analyze its trend in different time periods. An individual difference correction mechanism is established to adjust the evaluation standard of the disturbance resistance ability based on the individual's baseline level.

[0051] The beta power spectrum and the attention anti-disturbance ability are fused to generate an attention state domain. The beta wave generally reflects the excitability and alertness level of the brain, and complements the anti-disturbance characteristics of the alpha wave to form an attention evaluation dimension. A fusion weight distribution mechanism is established to determine the fusion coefficient according to the contribution of the beta wave and the anti-disturbance ability to the attention state. The fusion algorithm AS = w1 × P_beta + w2 × R_anti is designed, where AS is the attention state value, P_beta is the beta wave power feature, R_anti is the attention anti-disturbance ability, and w1 and w2 are dynamic weight coefficients. The principal component analysis method is used to determine the weight coefficients, and the main change mode of the beta wave power and the anti-disturbance ability is extracted. A coordinate system of the attention state domain is established, with the beta wave power as the horizontal axis and the attention anti-disturbance ability as the vertical axis, to construct a two-dimensional attention state space. Different attention state regions are defined in the state space, including highly concentrated, moderately concentrated, attention dispersed, and over-excited state regions. A clustering algorithm is used to analyze historical data to determine the boundaries and characteristics of each state region. A state transition rule analysis is established to study the transition mode and transition probability of the attention state between different regions. A real-time state positioning algorithm is designed to quickly determine the attention state position according to the current beta wave power and anti-disturbance ability.

[0052] In step S130, clustering analysis is performed on the attention state domain to identify a high alertness region and a low alertness region. Peak electroencephalogram features are extracted from the high alertness region through fluctuation analysis, and a decline rate parameter is extracted from the low alertness region. The peak electroencephalogram features and the decline rate parameter are used for exhaustion prediction to generate a critical decay time.

[0053] Specifically, clustering analysis is performed on the attention state domain to identify a high alertness region and a low alertness region. The K-means clustering algorithm is used to group the data points in the attention state domain, with the number of clusters set to 3-5, covering main state categories such as high alertness, moderate alertness, and low alertness. The cluster center C_i = (1 / n_i)∑x_j is calculated, where C_i is the center of the i-th cluster, n_i is the number of data points in the cluster, and x_j is a data point within the cluster. The silhouette coefficient is used to evaluate the clustering effect, ensuring high intra-class compactness and large inter-class separation. Density clustering method is used to identify abnormal points and noise points in the state domain, to exclude the interference of atypical states on the clustering results. A mathematical description of the clustering boundary is established, and a support vector machine method is used to construct the hyperplane separating different alertness regions. The statistical characteristics of each cluster are analyzed, including the center position, distribution range, data density, and other parameters. According to the numerical size of the beta wave power and the attention anti-disturbance ability, the high numerical region is marked as the high alertness region, and the low numerical region is marked as the low alertness region. A dynamic boundary adjustment mechanism for the alertness region is established to correct the region definition according to individual differences and time changes. An alertness region distribution map is generated to visually display the spatial distribution characteristics of different alertness states in the state domain.

[0054] In some embodiments, the extracting the peak electroencephalogram feature from the high vigilance zone by fluctuation analysis comprises: performing microscopic fluctuation analysis on the high vigilance zone to extract a fluctuation amplitude sequence; identifying minimum points from the fluctuation amplitude sequence; generating an enhanced signal based on the minimum points by noise amplification processing; and encoding the enhanced signal as the peak electroencephalogram feature.

[0055] Microscopic fluctuation analysis is performed on the high vigilance zone to extract a fluctuation amplitude sequence. The electroencephalogram data of the high vigilance zone is taken as input, and high-precision microscopic fluctuation analysis technology is used to extract the subtle fluctuation characteristics of the signal. Microscopic fluctuation analysis captures the fine structure and transient change process of the electroencephalogram activity in the high vigilance state. A high-resolution time window is established, with a window length of 100 milliseconds and a sliding step of 10 milliseconds, ensuring accurate capture of microscopic fluctuations. The power fluctuation amplitude A_fluct(i) = √[(P(i) - P_mean) ² ] is calculated for each time window, where A_fluct(i) is the fluctuation amplitude of the i-th window, P(i) is the average power in the window, and P_mean is the overall average power. A multi-scale analysis method is used to calculate the fluctuation amplitude at different time scales, revealing the hierarchical structure of the fluctuation. The Hilbert transform is used to extract the instantaneous amplitude and instantaneous frequency of the signal, and the modulation characteristics of the fluctuation are analyzed. A statistical distribution model of the fluctuation amplitude is established, and the shape parameters and eigenvalues of the distribution are calculated. Empirical mode decomposition method is used to decompose the fluctuation signal into intrinsic mode functions of different frequencies, and the fluctuation characteristics of each mode are extracted. The spatial correlation of the fluctuation amplitude is analyzed, and the fluctuation synchronicity between different electrode positions is calculated. Autocorrelation analysis of the fluctuation amplitude sequence is established to identify the memory effect and long-range correlation of the fluctuation.

[0056] The minimum value points are identified from the fluctuation amplitude sequence. A sliding window extreme value detection method is used to find the local minimum value points in each detection window of the fluctuation amplitude sequence. The significance index S_min of the minimum value points is calculated using the local statistical features of the fluctuation amplitude sequence, S_min=(A_neighbor-A_min) / σ_local, where S_min is the minimum value significance, A_neighbor is the neighborhood average value in the fluctuation amplitude sequence, A_min is the minimum value in the fluctuation amplitude sequence, and σ_local is the local standard deviation of the fluctuation amplitude sequence. Based on the signal-to-noise ratio features of the fluctuation amplitude sequence, the minimum value points with a significance index exceeding a set threshold are screened out to exclude the pseudo-minimum values caused by noise in the fluctuation amplitude sequence. The time interval and distribution law between the minimum value points in the fluctuation amplitude sequence are calculated by analyzing the time index information of the fluctuation amplitude sequence and analyzing the time distribution characteristics of the minimum value points in the fluctuation amplitude sequence. According to the amplitude distribution range of the fluctuation amplitude sequence, a depth classification system of the minimum value points is established, and the minimum values in the fluctuation amplitude sequence are classified into shallow minimum values and deep minimum values according to their relative depth. In view of the noise characteristics in the fluctuation amplitude sequence, a morphological processing method is used to optimize the detection results of the minimum value points in the fluctuation amplitude sequence, and the burrs and misjudgments in the detection of the fluctuation amplitude sequence are eliminated.

[0057] Noise amplification processing is performed based on the minimum value points to generate an enhanced signal. The identified minimum value points are used as key positions, and signal amplification technology is used to enhance the signal features of these positions to highlight their importance in the peak electroencephalogram features. Noise amplification processing is performed by amplifying the weak signals at specific positions to reveal hidden electroencephalogram activity patterns. An adaptive amplification function G(x)=α×exp(β×|x-x_min|) is designed, where G(x) is the amplification gain, x is the signal position, x_min is the minimum value point position, and α and β are amplification parameters. An amplification window is established around the minimum value point, and the window size is determined according to the influence range of the minimum value point. A nonlinear amplification strategy is used to amplify signals of different amplitudes by different amplification factors to avoid signal saturation. An adaptive adjustment mechanism for amplification strength is established to dynamically adjust the amplification parameters according to the dynamic range of the signal and the noise level. A band-pass filtering technique is used to preserve the signal components of specific frequency bands during amplification and filter out irrelevant frequency components. A multi-channel collaborative amplification method is used to consider the spatial correlation between different electrode channels.

[0058] The enhancement signal is encoded into peak electroencephalography features. Peak electroencephalography features are standardized templates of brain electrical activity of an individual in the optimal state of attention. By encoding, a feature vector containing multi-dimensional information in time, frequency, and space domains is formed, which is used to quantify the peak performance of an individual's attention and provide personalized reference standards for subsequent attenuation prediction and active intervention. A multi-dimensional feature vector extraction mechanism is established to extract key features in time, frequency, and spatial domains from the enhancement signal. The calculation of time domain features includes peak amplitude, peak width, rise time, fall time, and other parameters. Wavelet transform is used to extract frequency domain features to analyze the energy distribution of the enhancement signal at different frequency scales. Spatial domain features are described to record the distribution characteristics and propagation patterns of peak features at different electrode positions. Principal component analysis is used to reduce the dimensionality of the extracted multi-dimensional features, retaining the main variation characteristics. A standardized format for feature encoding is established, and vector quantization technology is used to map continuous features into discrete code words. The generated peak electroencephalography features provide standardized feature representations for attention state recognition and exhaustion prediction.

[0059] The decline rate parameter is extracted from the low alertness region. The decline rate parameter reflects the kinetic characteristics of the transition of attention from a higher level to a lower level, and is a key indicator for predicting attention exhaustion. The time derivative dA / dt of the attention indicator in the low alertness region is calculated, where A is the attention level and t is the time, and dA / dt represents the instantaneous change rate of attention. A sliding window method is used to calculate the local decline rate, with a window length of 30 seconds and an overlap rate of 50%, ensuring the time resolution of the rate calculation. An exponential fitting method is used to establish an attention decay model A(t)=A0×exp(-λt), where A0 is the initial attention level and λ is the decay constant, reflecting the size of the decline rate. The decay constant is estimated by least squares method, and the goodness of fit is used to evaluate the effectiveness of the model. The linear and nonlinear characteristics of the decline process are analyzed to identify different patterns of gradual decay and step decay. The variability of the decline rate is calculated to analyze the inter-individual and intra-individual variability characteristics. A multi-stage decline model is established to identify different stages and transition points in the attention decay process. State space method is used to analyze the kinetic characteristics of the decline process, and state transition rate and stability parameters are extracted.

[0060] The peak electroencephalogram feature and the decline rate parameter are processed to generate a critical attenuation time. The extracted peak electroencephalogram feature and decline rate parameter are used as input features to establish an attention depletion prediction model to determine the time point of critical attenuation. The critical attenuation time refers to the critical time point when the attention level drops to the level that cannot meet the requirements of normal work, which is an important early warning indicator for work safety management. A depletion prediction function T_critical=f(P_peak,R_decline) is established, where T_critical is the critical attenuation time, P_peak is the peak electroencephalogram feature, and R_decline is the decline rate parameter. Machine learning methods are used to train the prediction model, and support vector regression or neural network algorithms are used to establish nonlinear mapping relationships. The model parameters are determined through supervised learning of historical data, and the prediction accuracy is evaluated using cross-validation methods. A multi-time scale prediction mechanism is established to provide short-term, medium-term, and long-term attenuation time predictions. A method for calculating the prediction confidence interval is designed to quantify the uncertainty and reliability of the prediction results. A personalized threshold setting method is designed to consider the special requirements of different individuals and tasks.

[0061] In step S140, the peak electroencephalogram feature is released before the critical attenuation time to generate a pre-charge activation sequence. The pre-charge activation sequence is time-optimized to generate a stepwise wake-up scheme. The stepwise wake-up scheme is matched and adjusted with the traction response to generate personalized intervention parameters. The best stimulation opportunity is determined based on the personalized intervention parameters.

[0062] In some embodiments, the release of the peak electroencephalogram feature before the critical attenuation time to generate a pre-charge activation sequence includes: determining a release start time based on the critical attenuation time; retrieving the peak electroencephalogram feature at the release start time; performing time-domain stretching on the peak electroencephalogram feature to obtain a gradual waveform; and serializing the gradual waveform to generate a pre-charge activation sequence.

[0063] The release start time is determined according to the critical attenuation time. Based on the predicted critical attenuation time, a time backtracking algorithm is used to determine the optimal start time of pre-charge activation. A time budget allocation model is established, and the time period from the start time to the critical time is divided into three stages: activation preparation time, activation establishment time, and effect maintenance time. The release start time T_start is calculated as T_start = T_critical - T_activation - T_buffer, where T_start is the release start time, T_critical is the critical attenuation time, T_activation is the activation establishment time, and T_buffer is the safety buffer time. The typical range of activation establishment time is determined through statistical analysis, which is generally set to 5-15 minutes and adjusted according to individual differences. The safety buffer time is set to avoid time estimation errors, and the buffer time is usually 20-30% of the activation time. An individualized time parameter adjustment mechanism is established to adjust the time budget based on the individual's historical response data. A dynamic time adjustment strategy is used to adjust the start time based on real-time monitoring of attention decay speed. A time conflict processing method is designed to coordinate time when the start time conflicts with other tasks.

[0064] The peak electroencephalogram feature is retrieved at the release start time. At the determined release start time, the most matched pattern data for the current individual and state is retrieved from the peak electroencephalogram feature library. A similarity evaluation algorithm for pattern matching is established to determine the best matching pattern by calculating the similarity between the current electroencephalogram feature and the patterns in the library. A multi-dimensional feature matching method is used to consider the similarity of frequency domain features, time domain features, and spatial domain features. The matching degree score M_score is calculated as M_score = w1 x Sim_freq + w2 x Sim_time + w3 x Sim_spatial, where Sim_freq, Sim_time, and Sim_spatial are the similarity of frequency domain, time domain, and spatial domain, respectively, and w1, w2, and w3 are weight coefficients. A priority sorting mechanism for pattern retrieval is established to preferentially select patterns with high matching degree and good use effect. A pattern data integrity check algorithm is designed to ensure that the retrieved pattern data is not missing or damaged. A pattern retrieval cache mechanism is established to pre-load possible pattern data to reduce retrieval delay. When a single pattern cannot be completely matched, a pattern fusion technology is used to generate a new pattern by weighted fusion of multiple similar patterns.

[0065] Temporal stretching is performed on peak EEG features to obtain a gradual waveform. The goal of temporal stretching is to expand relatively short activation patterns in peak EEG features into longer gradual processes, avoiding overly abrupt activation. Based on the temporal parameters of the peak EEG features, especially the original peak width and rise / fall times, a linear stretching method is used to extend the time axis of the peak EEG features. The stretching factor is determined based on the ratio of the target activation duration to the original duration of the peak EEG features. Based on the distribution of key feature segments (such as peak regions and transition regions) in the peak EEG features, a nonlinear stretching function is designed, using exponential or logarithmic functions to achieve non-uniform temporal stretching of the peak EEG features, highlighting the key feature segments. Based on the individual response characteristics and sensitivity parameters reflected by the peak EEG features, an adaptive adjustment mechanism for the stretching parameters is established, adjusting the stretching intensity and shape of the stretching curve according to individual differences in the peak EEG features. To address the data density changes during the peak EEG feature stretching process, an interpolation algorithm is used to fill the data points during the stretching process, ensuring the smoothness and continuity of the stretched peak EEG feature waveform. Based on the frequency domain characteristic parameters of peak EEG features, filtering techniques are used to handle the spectral distortion introduced by peak EEG feature stretching, preserving the original frequency domain characteristics of the peak EEG features. According to the rise and fall time parameters of the peak EEG features, a method for controlling the gradient slope is designed, achieving a gradient effect by adjusting the rise and fall rates of the peak EEG feature waveform.

[0066] The gradient waveform is serialized to generate a pre-charge activation sequence. The time-stretched gradient waveform is used as the basic building block, and it is serialized based on its amplitude variation characteristics and temporal extension properties. Utilizing the multi-level structural features of the gradient waveform, a multi-channel sequence arrangement strategy is established, assigning different frequency components of the gradient waveform to corresponding brain regions. Based on the temporal extension characteristics of the gradient waveform, a temporal interleaving pattern is designed to stagger the activation timing of different channels based on the phase relationship of the gradient waveform, avoiding over-synchronous activation. Following the intensity increase pattern of the gradient waveform, a progressive sequence structure is adopted, starting from the initial low-intensity portion and gradually transitioning to the high-intensity activation segment. Based on the waveform interval characteristics of the gradient waveform, an interval control mechanism for sequence elements is established, using the natural interval of the gradient waveform as a control reference for activation density. Based on the periodicity of the gradient waveform, a repetition pattern is designed, determining the number of repetitions and the repetition intensity according to the repetition pattern of the gradient waveform. A sequence optimization algorithm is employed, using either a genetic algorithm or a particle swarm optimization algorithm to find the optimal sequence arrangement scheme. Based on the multi-channel synchronization characteristics of the gradient waveform, a sequence synchronization mechanism is established to ensure the time coordination and phase relationship of the multi-channel activated sequences.

[0067] The pre-charge activation sequence is processed in time optimization to generate a stepwise wake-up scheme. The design goal of the stepwise wake-up scheme is to avoid the impact of sudden strong stimulation on the brain according to the progressive characteristics of the pre-charge activation sequence, and to gradually improve the attention level in a mild incremental way. According to the three-section structure characteristics of the pre-charge activation sequence, a multi-stage wake-up model is established, and the whole wake-up process is divided into three main stages: start-up stage, acceleration stage and stable stage. According to the low-intensity activation section in the pre-charge activation sequence, the low-intensity stimulation corresponding to the first third of the pre-charge activation sequence is used in the start-up stage, and the duration is determined according to the time distribution of the sequence. Based on the medium-intensity activation section in the pre-charge activation sequence, the intensity of the stimulation is gradually increased according to the intensity increasing mode of the pre-charge activation sequence in the acceleration stage, and the intensity change curve follows the original change law of the sequence. According to the high-intensity activation section in the pre-charge activation sequence, the high-intensity activation mode of the last third of the pre-charge activation sequence is used in the stable stage, and the duration is determined according to the task requirements and sequence characteristics.

[0068] The stepwise wake-up scheme is matched and adjusted with the traction response to generate personalized intervention parameters. For the start-up stage parameters in the stepwise wake-up scheme, the intensity is adjusted according to the high and low of the traction response; for the increasing rate of the acceleration stage, the change gradient is adjusted based on the traction response; for the maintenance intensity of the stable stage, the final activation level is determined according to the traction response. The mapping relationship between response and intervention intensity is established, and the individual with high traction response corresponds to the lower intensity parameter configuration in the stepwise wake-up scheme, and the individual with low traction response corresponds to the higher intensity configuration. The matching function is designed, and the parameters AS_stage of each stage of the stepwise wake-up scheme are comprehensively calculated with the traction response TR: I_optimal=f(TR,AS_stage), wherein I_optimal is the optimal intervention intensity. The time sequence parameters of the stepwise wake-up scheme are used as constraint conditions, and the multi-element linear regression or neural network method is used to establish the matching model, and the model parameters are determined by training the historical data. Based on the multi-dimensional characteristics of the stepwise wake-up scheme and the individual differences of the traction response, the adjustment algorithm of personalized parameters is established, including the individual correction of multiple dimensions such as intensity adjustment, time sequence adjustment and frequency adjustment in the wake-up scheme.

[0069] The optimal stimulation timing is determined based on the personalized intervention parameters. According to the time window requirement in the personalized intervention parameters, a dynamic time window analysis method is used to find the optimal intervention timing within the time range defined by the personalized parameters. The configurations in the personalized intervention parameters are used as weight coefficients to calculate the timing suitability score S_timing = w1 x P_plasticity + w2 x A_network + w3 x (1 / D_interference), where the weight coefficients w1, w2, and w3 are determined by the personalized intervention parameters. Based on the effect expectation configuration in the personalized intervention parameters, a timing prediction model is established to predict the intervention effect at future time points based on historical intervention data calibrated by the personalized parameters. According to the response speed setting in the personalized intervention parameters, a real-time timing monitoring algorithm is designed to adjust the monitoring sensitivity according to the personalized parameters and capture the occurrence of the optimal intervention timing. Based on the trigger condition configuration in the personalized intervention parameters, a condition judgment mechanism for timing triggering is established, and the intervention program is automatically started when the conditions set by the personalized parameters are met.

[0070] In step S150, the optimal stimulation timing is analyzed with the critical decay time to generate a warning advance, and based on the warning advance, a multi-channel stimulation allocation is generated to generate a stimulation scheduling matrix, and according to the stimulation scheduling matrix, the pre-charge activation sequence is decomposed into a progressive stimulation string.

[0071] Specifically, the optimal stimulation timing is analyzed with the critical decay time to generate a warning advance. The warning advance reflects the time window from starting the intervention to preventing the complete decay of attention, which is a key parameter for intervention timing control. The basic time difference ΔT_basic = T_critical - T_optimal is calculated, where ΔT_basic is the basic time difference, T_critical is the critical decay time, and T_optimal is the optimal stimulation timing. A dynamic correction mechanism for the time difference is established, considering factors such as individual response delay, stimulation establishment time, and environmental interference factors. Time series analysis method is used to study the variation law of the time difference, and the periodicity and trend characteristics of the time difference are identified. A confidence interval calculation method for the time difference is designed to quantify the uncertainty and reliability of the time prediction. A multi-scale time analysis framework is established to analyze the time difference characteristics at different time scales such as seconds, minutes, and hours. The sliding window technique is used to analyze the local variation characteristics of the time difference, and the mutation points and stable intervals of the time difference are identified. The warning advance T_advance = ΔT_basic x α + T_safety is calculated, where α is the safety factor and T_safety is the safety buffer time. A personalized adjustment mechanism for the warning advance is established to adjust the advance parameter based on the individual's historical response data and current state.

[0072] In some embodiments, the generating a stimulation scheduling matrix based on the pre-warning advance includes: determining a stimulation window length by the pre-warning advance; dividing the stimulation window into auditory, visual and tactile sub-windows to obtain a window allocation table; configuring a phase offset in the window allocation table to form an alternating activation mode; and generating a stimulation scheduling matrix based on the alternating activation mode.

[0073] The stimulation window length is determined by the pre-warning advance. The determination of the stimulation window length needs to balance the activation effect and time efficiency, and ensure the best attention recovery effect in a limited pre-warning time. A calculation model of the window length L_window = T_advance x η - T_overhead is established, where L_window is the stimulation window length, T_advance is the pre-warning advance, η is the time utilization rate, and T_overhead is the time overhead of starting and ending. The influence of different window lengths on the activation effect is analyzed, and the optimal window length range is determined by historical data statistics. An adaptive adjustment mechanism of the window length is designed, and the window size is dynamically adjusted according to the response speed of the individual and the task urgency. A lower limit constraint of the window length is established to ensure that the window length is sufficient to support an effective activation process. A multi-objective optimization method is used to balance the relationship between the window length and the activation effect, and to find the optimal time configuration scheme. A hierarchical management system of the window length is established, and different levels of window length are set according to the urgency of attention decay.

[0074] The stimulation window is divided into auditory, visual and tactile sub-windows to obtain a window allocation table. Based on the determined stimulation window length L_window, time allocation is performed according to the characteristics and efficiency requirements of each sensory channel. The total duration of the stimulation window is taken as the allocation basis, and the goal of sub-window division is to make full use of the advantages of each sensory channel and realize the coordinated cooperation of multi-modal stimulation. According to the constraint condition of the stimulation window length, a channel time allocation algorithm is designed to ensure that the sum of the durations of the auditory, visual and tactile sub-windows is equal to the total length of the stimulation window: L_audio+L_visual+L_tactile=L_window. The channel allocation weight W_channel=f(E_efficiency,T_threshold,D_duration) is calculated, where E_efficiency is the channel efficiency, T_threshold is the activation threshold, and D_duration is the effect duration. Based on the front time allocation of the stimulation window, the parameter settings of the auditory sub-window are established, and 25-35% of the stimulation window length is allocated to the auditory channel to take advantage of the characteristics of fast conduction and immediate response of auditory stimulation. The middle section of the stimulation window is used to design the time arrangement of the visual sub-window, and 40-50% of the stimulation window length is allocated to the visual channel to take advantage of the high information capacity and strong attention grabbing ability of visual stimulation. In the later section of the stimulation window, the activation scheme of the tactile sub-window is configured, and 20-30% of the stimulation window length is allocated to the tactile channel to take advantage of the directness and persistence of tactile stimulation. At the boundary of the sub-windows in the stimulation window, a buffer time setting is established between the sub-windows to avoid activation interruption during channel switching. A window allocation table is generated to record the start time, duration, activation intensity and other parameter information of each sub-window in the stimulation window.

[0075] The phase offset configuration in the window allocation table forms an alternating activation mode. Based on the time parameters of each sub-window recorded in the window allocation table, including the start time and duration of the auditory sub-window, the visual sub-window, and the tactile sub-window, a phase offset strategy between channels is designed. Using the time allocation ratio in the window allocation table, a calculation model for phase offset is established: φ_offset = 2π × t_delay / T_period, where φ_offset is the phase offset angle, t_delay is the time delay in the window allocation table, and T_period is the activation period determined by the window allocation table. According to the time sequence arrangement of the three sub-windows in the window allocation table, a phase offset strategy for three channels is designed, and the activation times of the auditory, visual, and tactile three sub-windows in the window allocation table are re-adjusted with a 120-degree phase difference. Based on the activation intensity distribution in the window allocation table, an optimization algorithm for offset time is established to minimize the interference between channels and maximize the activation effect by adjusting the time parameters in the window allocation table. According to the channel sensitivity data of individuals in the window allocation table, a dynamic phase adjustment method is used to individually adjust the phase relationship in the window allocation table. Based on the buffer time setting in the window allocation table, a smooth transition mechanism for phase offset is designed to ensure the continuity and comfort of the channel switching process. Using the activation frequency information in the window allocation table, a rhythm control algorithm for alternating activation is established to adjust the activation frequency and interval time according to the rhythm requirements of the window allocation table. According to the multi-level time structure in the window allocation table, a multi-level alternating mode is designed to realize the alternating activation of channels at different time scales.

[0076] A stimulation scheduling matrix is generated based on the alternating activation mode. The information contained in the designed alternating activation mode, such as phase offset configuration, channel switching timing, and activation intensity variation, is converted into a standardized matrix representation. According to the time slicing structure and channel distribution characteristics in the alternating activation mode, a complete management framework for stimulation scheduling is constructed. According to the time division scheme in the alternating activation mode, the definition rules of matrix dimensions are established, the row dimension corresponds to the time slicing in the alternating activation mode, the column dimension corresponds to the stimulation channel in the alternating activation mode, and the matrix element represents the activation intensity of each channel at the corresponding time. According to the activation intensity classification in the alternating activation mode, the encoding method of matrix elements is designed, which maps the intensity variation in the alternating activation mode to a normalized numerical value, and the numerical value range 0-1 corresponds to from no activation to maximum activation. Based on the sparse activation characteristics in the alternating activation mode, a sparse representation method of the matrix is established, and the characteristics of a large number of zero activation periods in the alternating activation mode are used to reduce the memory occupation of the matrix through compression storage. According to the precise timing requirements in the alternating activation mode, a time synchronization mechanism of the matrix is designed to ensure that the matrix execution process strictly follows the time arrangement of the alternating activation mode.

[0077] In some embodiments, the step of decomposing the pre-charge activation sequence into a progressive stimulation string according to the stimulation schedule matrix comprises: extracting channel weight distribution from the stimulation schedule matrix; performing unequal segmentation of the pre-charge activation sequence based on the weight distribution to generate a segment set; cross-recombining the segment set to form a mesh activation path; and extracting node signals along the mesh path to form the progressive stimulation string.

[0078] The channel weight distribution is extracted from the stimulation schedule matrix. The generated stimulation schedule matrix is taken as input, and the activation intensity distribution characteristics of each channel in different time periods are extracted by a matrix analysis algorithm. The time sequence activation data of each channel is read from the column vector of the stimulation schedule matrix, and the row index of the matrix corresponds to the time slice, and the column index corresponds to the stimulation channel. The weight vector W_channel=[w1,w2,...,wn] of each channel is calculated, where wi represents the activation intensity value of the channel extracted from the ith row of the stimulation schedule matrix. Statistical analysis is performed on the data columns of each channel in the stimulation schedule matrix, and the characteristic parameters of the weight distribution are calculated, including mean, variance, skewness, kurtosis, and other statistical quantities. Based on the numerical range difference of different channels in the stimulation schedule matrix, a normalization processing mechanism for the weight distribution is established to ensure that the different channel weights extracted from the matrix are within the same dimension range. By analyzing the time dimension change of the stimulation schedule matrix, an analysis algorithm for the weight change trend is designed to identify the time evolution law of the weight distribution from the row sequence of the matrix. The weight time sequence data in the stimulation schedule matrix is subjected to frequency domain transformation, and the frequency domain characteristics of the weight distribution are studied by using the frequency spectrum analysis method to identify the periodic components of the weight change. Based on the similar activation patterns in the stimulation schedule matrix, a weight clustering analysis mechanism is established to group time periods with similar weight characteristics. A weight smoothing processing algorithm is designed to eliminate the mutation and noise interference in the weight distribution in view of the numerical mutation phenomenon in the stimulation schedule matrix.

[0079] The pre-charge activation sequence is unevenly cut based on the weight distribution to generate a segmented set. The extracted channel weight distribution is used as a cutting guide, and an uneven cutting strategy is designed in combination with the time structure characteristics of the pre-charge activation sequence. According to the intensity change pattern in the weight distribution and the activated nodes of the pre-charge activation sequence, a cutting point identification algorithm is established to determine the segmentation position of the sequence by gradient analysis of the weight distribution combined with the structure characteristics of the pre-charge activation sequence. The rate of change of the weight distribution is used as a cutting reference to calculate the weight change rate dW / dt, where W is the weight value in the weight distribution and t is the time. The cutting boundary is set at the position of the pre-charge activation sequence corresponding to the extreme value point of the weight distribution change rate. According to the complexity of the weight distribution and the density characteristics of the pre-charge activation sequence, an adaptive cutting strategy is designed to adjust the cutting granularity of the pre-charge activation sequence. Based on the statistical characteristics of the weight distribution and the time length constraint of the pre-charge activation sequence, a constraint condition for the cutting length is established to ensure that the segmentation length of the pre-charge activation sequence is within a reasonable range, avoiding excessive subdivision or excessive roughness.

[0080] The segmented set is cross-recombined to form a network activation path. The starting point and ending point information of each segment in the segmented set are used to establish a definition rule for path nodes, and the starting point and ending point of each segment in the segmented set are used as the nodes of the path network. According to the connection possibility and weight correlation of the segments in the segmented set, the topological structure of path connection is designed, and the directed graph model is used to describe the connection relationship between the segments in the segmented set. Based on the importance evaluation and mutual correlation strength of each segment in the segmented set, a path weight calculation method is established to determine the weight value of the path according to the importance and connection strength of the segments in the segmented set. The connectivity and shortest path characteristics of the network path are analyzed using graph theory algorithms for the constructed segmented set network graph. Based on the cross-recombination rule of the segments in the segmented set, a path intersection control mechanism is designed to control the interaction between paths by adjusting the intersection probability and intersection strength between segments in the segmented set. Considering the balance between the overall efficiency of the segmented set and the complexity of the path, a path redundancy management strategy is established to optimize the complexity of the path under the premise of ensuring the activation effect of the segmented set. For the multi-objective optimization requirements of the segmented set, a path optimization algorithm is used to find the optimal path combination of the segmented set through genetic algorithm or ant colony algorithm.

[0081] The node signals are extracted along the mesh path to form a progressive stimulation train. According to the connectivity analysis results and the shortest path characteristics in the mesh activation path, a path selection strategy for node traversal is established, and the traversal order is determined along the path with the optimal weight in the mesh activation path by using depth-first search or breadth-first search. Based on the weight values and connection strengths of the nodes in the mesh activation path, the extraction rules of the node signals are designed, and the amplitude and duration of the signals are determined according to the weight and connection characteristics of the nodes in the mesh activation path. The sequence of the node signals extracted along the mesh activation path is calculated, and the progressive index G_index = ∑|S(i+1)-S(i)| / n, where G_index is the progressive index, S(i) is the signal strength of the ith node in the mesh path, and n is the total number of path nodes. In view of the continuity requirement of the node signals in the mesh activation path, a signal smoothing processing mechanism is established to eliminate the mutations between the nodes by using interpolation and filtering techniques. When there are parallel branches in the mesh activation path, a multi-path signal fusion algorithm is designed to generate a comprehensive signal by weighted fusion of the parallel path signals in the mesh path. Based on the timing synchronization requirement of the mesh activation path, a signal timing synchronization control mechanism is established to ensure that the stimulation train extracted from the mesh path has accurate time characteristics. The generated progressive stimulation train fully embodies the structural characteristics and optimization effect of the mesh activation path.

[0082] In step S160, the stimulation parameter sequence is extracted from the progressive stimulation train, the rhythm coding of the stimulation parameter sequence is performed to generate an electroencephalogram synchronization stimulation signal, and the electroencephalogram synchronization stimulation signal is converted into an audible and visual vibration alarm signal.

[0083] The stimulation parameter sequence is extracted from the progressive stimulation train. The stimulation characteristic parameters are extracted from the time domain, frequency domain and amplitude domain of the progressive stimulation train. By analyzing the time structure of the progressive stimulation train, the time domain parameters including the stimulation duration, rise time, fall time, and intermittent time in the progressive stimulation train are extracted. Based on the amplitude variation characteristics of the progressive stimulation train, the stimulation amplitude parameter A_stim = max(S(t))-min(S(t)) is calculated, where A_stim is the stimulation amplitude, S(t) is the stimulation signal in the progressive stimulation train, and max and min represent the maximum and minimum value operations respectively. The frequency domain analysis is performed on the progressive stimulation train, and the dominant frequency, frequency spectrum width, and frequency modulation depth are extracted from the progressive stimulation train by using fast Fourier transform. In view of the multi-dimensional parameter characteristics of the progressive stimulation train, a parameter correlation analysis mechanism is established to identify the coupling relationship and constraint conditions between different parameters in the progressive stimulation train. Based on the parameter distribution characteristics of the progressive stimulation train, a parameter clustering method is used to group the parameters with similar characteristics in the progressive stimulation train to form a hierarchical structure of parameter clusters. In view of the dimension difference of the parameters in the progressive stimulation train, a parameter standardization processing algorithm is designed to map the parameters with different dimensions in the progressive stimulation train to a unified numerical range.

[0084] The rhythm coding of the stimulation parameter sequence generates the electroencephalogram synchronization stimulation signal. Based on the frequency characteristic parameters in the stimulation parameter sequence, a reference model of electroencephalogram rhythm is established, including the characteristic parameters of the main electroencephalogram frequency bands such as α rhythm, β rhythm, γ rhythm, etc. Using the dominant frequency and spectral width information in the stimulation parameter sequence, a rhythm matching algorithm is designed to align the frequency components of the stimulation parameter sequence with the target electroencephalogram rhythm. According to the frequency parameters in the stimulation parameter sequence and the target electroencephalogram frequency, the synchronization index Sync = |f_stim-f_brain| / f_brain is calculated, where Sync is the synchronization index, f_stim is the stimulation frequency in the stimulation parameter sequence, and f_brain is the target electroencephalogram frequency. Based on the phase information in the stimulation parameter sequence, a phase-locked technique is used to ensure that the stimulation signal generated based on the stimulation parameter sequence matches the phase of the electroencephalogram signal, establishing a stable synchronization relationship. Using the multi-frequency components in the stimulation parameter sequence, a multi-band coding strategy is designed to code the rhythm in multiple electroencephalogram frequency bands contained in the stimulation parameter sequence, realizing a multi-level synchronization mode. According to the individualized characteristic parameters in the stimulation parameter sequence, an adaptive adjustment mechanism of coding intensity is established to adjust the coding parameters based on the individual electroencephalogram characteristics and response sensitivity reflected by the stimulation parameter sequence. Using the modulation characteristics in the stimulation parameter sequence, a modulation coding technique is used to realize complex rhythm coding modes through amplitude modulation and frequency modulation parameters in the stimulation parameter sequence.

[0085] The electroencephalogram synchronization stimulation signal is converted into an acousto-optic vibration alarm signal. Alarm signal conversion is to map the individualized electroencephalogram synchronization features into effective external perception stimulation, realizing real-time monitoring and timely alarm of the attention state of the operating personnel. A multi-modal conversion matrix is established to map the frequency, amplitude, and phase characteristics of the electroencephalogram synchronization signal into physical parameters such as sound frequency, light intensity, and vibration mode. An algorithm for generating sound alarm signals is designed to convert the frequency characteristics of electroencephalogram α waves and β waves into audio signals in the audible range, with a frequency mapping relationship of f_audio=k×f_brain+f_base, where k is the frequency amplification coefficient and f_base is the base audio frequency. An intensity control mechanism for optical signal alarm is established to modulate the brightness and flicker frequency of the LED light according to the amplitude changes of the electroencephalogram signal. A mode coding of vibration alarm signal is designed to convert the rhythm characteristics of the electroencephalogram signal into the vibration mode and intensity changes of the vibration motor. A hierarchical alarm system is established to set different intensity levels of acousto-optic vibration alarm levels according to the degree of attention decay. An adaptive alarm strategy is adopted to dynamically adjust the intensity and mode of the alarm signal according to the environmental noise and individual sensitivity. A time sequence control algorithm for the alarm signal is designed to ensure the coordination and time synchronization of the three modalities of sound, light, and vibration. The generated acousto-optic vibration alarm signal can effectively remind the operating personnel of the changes in attention state, providing timely early warning protection for work safety.

[0086] In order to perform the above-mentioned method embodiment corresponding to a kind of EEG-based attention monitoring and alarm method, to realize corresponding function and technical effect. Figure 2 , Figure 2 The structure block diagram of the EEG-based attention monitoring and alarm device 200 provided by the embodiment of the application is shown.The part related to the embodiment is shown for convenience of description, and the EEG-based attention monitoring and alarm device 200 provided by the embodiment of the application includes:

[0087] The signal acquisition module 201 is configured to acquire the EEG signal of the operator, and to perform artifact removal processing on the EEG signal to construct a purified EEG library.

[0088] The attention evaluation module 202 is configured to extract the alpha wave and beta wave power spectrum based on the purified EEG library, to perform chaotic dynamics analysis on the alpha wave power spectrum to obtain a traction response degree, to identify an attention anti-disturbance capability based on the traction response degree, and to fuse the beta wave power spectrum with the attention anti-disturbance capability to generate an attention state domain.

[0089] The mode storage module 203 is configured to perform cluster analysis on the attention state domain to identify a high vigilance zone and a low vigilance zone, to extract a peak EEG feature from the high vigilance zone by fluctuation analysis, to extract a decline rate parameter from the low vigilance zone, and to perform exhaustion prediction processing on the peak EEG feature and the decline rate parameter to generate a critical attenuation time.

[0090] The intervention design module 204 is configured to release the peak EEG feature before the critical attenuation time to generate a pre-charge activation sequence, to perform timing optimization processing on the pre-charge activation sequence to generate a stepwise wake-up scheme, to match and adjust the stepwise wake-up scheme with the traction response degree to generate individualized intervention parameters, and to determine an optimal stimulation opportunity based on the individualized intervention parameters.

[0091] The scheduling control module 205 is configured to perform time difference analysis on the optimal stimulation opportunity and the critical attenuation time to generate a warning lead time, to perform multi-channel stimulation distribution based on the warning lead time to generate a stimulation scheduling matrix, and to decompose the pre-charge activation sequence into a gradual stimulation string according to the stimulation scheduling matrix.

[0092] The alarm execution module 206 is configured to extract a stimulation parameter sequence from the gradual stimulation string, to perform rhythm encoding on the stimulation parameter sequence to generate an EEG synchronous stimulation signal, and to convert the EEG synchronous stimulation signal into an audible, visual and vibratory alarm signal.

[0093] The above-mentioned EEG-based attention monitoring and alarm device 200 can implement an EEG-based attention monitoring and alarm method according to the above-mentioned method embodiment. The optional items in the above-mentioned method embodiment are also applicable to this embodiment, and will not be described in detail herein. The remaining content of the present embodiment can be referred to the content of the above-mentioned method embodiment, and will not be described in detail herein.

[0094] As shown in Figure 3 The third embodiment of the present application further provides a computer device, which comprises a memory 301, a processor 302, and a computer program stored in the memory 301 and capable of running on the processor 302, characterized in that the processor 302 implements the steps of the EEG-based attention monitoring and alarm method according to the first embodiment of the present application when executing the program.

[0095] The above-mentioned is only part or preferred embodiments of the present application, neither the text nor the drawings can limit the scope of protection of the present application, any equivalent structural transformation using the content of the present application specification and drawings, or direct / indirect application in other related technical fields are included in the scope of protection of the present application.

Claims

1. A method for monitoring and alerting attention based on electroencephalogram, characterized in that, The method comprises the following steps: Collecting the brain electrical signals of the working personnel, and performing artifact removal processing on the brain electrical signals to construct a purified brain electrical library; Based on the purified brain electrical library, extracting alpha wave and beta wave power spectrum, performing chaotic dynamics analysis on the alpha wave power spectrum to obtain a traction response degree, identifying attention noise resistance ability based on the traction response degree, and performing fusion processing on the beta wave power spectrum and the attention noise resistance ability to generate an attention state domain; Performing cluster analysis on the attention state domain to identify a high vigilance area and a low vigilance area, extracting peak brain electrical characteristics from the high vigilance area through fluctuation analysis, and extracting a decline rate parameter from the low vigilance area to generate a critical attenuation time through exhaustion prediction processing of the peak brain electrical characteristics and the decline rate parameter; Releasing the peak brain electrical characteristics before the critical attenuation time to generate a pre-charge activation sequence, performing time sequence optimization processing on the pre-charge activation sequence to generate a stepwise awakening scheme, matching and adjusting the stepwise awakening scheme with the traction response degree to generate individualized intervention parameters, and determining an optimal stimulation time based on the individualized intervention parameters; Performing time difference analysis on the optimal stimulation time and the critical attenuation time to generate a pre-warning lead time, performing multi-channel stimulation distribution based on the pre-warning lead time to generate a stimulation scheduling matrix, and decomposing the pre-charge activation sequence into an incremental stimulation string according to the stimulation scheduling matrix; Extracting a stimulation parameter sequence from the incremental stimulation string, performing rhythm coding on the stimulation parameter sequence to generate a brain electrical synchronization stimulation signal, and converting the brain electrical synchronization stimulation signal into an audible light vibration alarm signal.

2. The method of claim 1, wherein, The artifact removal processing on the brain electrical signals to construct a purified brain electrical library comprises the following steps: Constructing an artifact characteristic map based on the brain electrical signals; Performing inverse generation processing on the artifact characteristic map to obtain a cancellation waveform; Superimposing and neutralizing the cancellation waveform and the brain electrical signals; Based on the superimposition and neutralization result, a purified brain electrical library is established.

3. The method of claim 1, wherein, The chaotic dynamics analysis on the alpha wave power spectrum to obtain a traction response degree comprises the following steps: Extracting a chaotic attractor from the alpha wave power spectrum; Injecting a perturbation into the chaotic attractor to detect a critical point; Applying anti-chaos control at the critical point to obtain a stabilization difficulty; Based on the stabilization difficulty, the traction response degree is backstepped.

4. The method of claim 1, wherein, The peak brain electrical characteristic extraction from the high vigilance area through fluctuation analysis comprises the following steps: Performing microscopic fluctuation analysis on the high vigilance area to extract a fluctuation amplitude sequence; Identifying a minimum point from the fluctuation amplitude sequence; Performing noise amplification processing based on the minimum point to generate an enhanced signal; Encoding the enhanced signal into a peak brain electrical characteristic.

5. The method of claim 1, wherein, The peak brain electrical characteristic release before the critical attenuation time to generate a pre-charge activation sequence comprises the following steps: Determining a release starting time according to the critical attenuation time; Retrieving the peak brain electrical characteristic at the release starting time; Performing time domain stretching processing on the peak brain electrical characteristic to obtain a gradual waveform; Serializing the gradual waveform to generate a pre-charge activation sequence.

6. The method of claim 1, wherein, The multi-channel stimulation distribution based on the pre-warning lead time to generate a stimulation scheduling matrix comprises the following steps: Determining a stimulation window length through the pre-warning lead time; The stimulation window is divided into auditory, visual and tactile sub-windows to obtain a window allocation table; The window allocation table is configured with a phase offset to form an alternating activation mode; A stimulation scheduling matrix is generated based on the alternating activation mode.

7. The method of claim 1, wherein, The pre-charge activation sequence is decomposed into a progressive stimulation string according to the stimulation scheduling matrix, including: Extracting a channel weight distribution from the stimulation scheduling matrix; Unequally cutting the pre-charge activation sequence based on the weight distribution to generate a segmented set; Cross-recombining the segmented set to form a mesh activation path; Extracting node signals along the mesh path to form a progressive stimulation string.

8. The method of claim 3, wherein, The response degree is inversely deduced based on the stabilization difficulty, including: Based on the stabilization difficulty evaluation control cost index, the response level is determined, and the control cost index includes disturbance suppression time, energy consumption and oscillation decay rate; According to the response level, a quantization parameter is set; The stabilization difficulty is inversely mapped to generate a traction response degree using the quantization parameter.

9. An electroencephalogram-based attention monitoring and alarm device, characterized by comprising: It includes: The signal acquisition module is used for collecting the brain electrical signal of the operator, and the brain electrical signal is processed to remove artifacts to construct a purified brain electrical library; The attention evaluation module is used for extracting the alpha wave and beta wave power spectrum based on the purified brain electrical library, performing chaos dynamics analysis on the alpha wave power spectrum to obtain the traction response degree, identifying the attention disturbance resistance ability based on the traction response degree, and fusing the beta wave power spectrum with the attention disturbance resistance ability to generate an attention state domain; The mode storage module is used for clustering analysis of the attention state domain to identify a high vigilance area and a low vigilance area, extracting peak brain electrical features from the high vigilance area through fluctuation analysis, extracting a decline rate parameter from the low vigilance area, and performing exhaustion prediction processing on the peak brain electrical features and the decline rate parameter to generate a critical attenuation time; The intervention design module is used to release the peak brain electrical features before the critical attenuation time to generate a pre-charge activation sequence, perform timing optimization processing on the pre-charge activation sequence to generate a stepwise awakening scheme, match and adjust the stepwise awakening scheme with the traction response degree to generate individualized intervention parameters, and determine the best stimulation opportunity based on the individualized intervention parameters; The scheduling control module is used to perform time difference analysis on the best stimulation opportunity and the critical attenuation time to generate a pre-warning lead time, perform multi-channel stimulation distribution based on the pre-warning lead time to generate a stimulation scheduling matrix, and decompose the pre-charge activation sequence into a progressive stimulation string according to the stimulation scheduling matrix; The alarm execution module is used to extract a stimulation parameter sequence from the progressive stimulation string, perform rhythm coding on the stimulation parameter sequence to generate a brain electrical synchronous stimulation signal, and convert the brain electrical synchronous stimulation signal into an audible, visual and vibratory alarm signal.

10. A computer device, comprising: It includes a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and realize the method of any one of claims 1 to 8 when executing the computer program.

Citation Information

Patent Citations

  • Safety production monitoring method and device based on attention detection and electronic equipment

    CN115481650A

  • EEG concentration degree-distraction degree two-dimensional classification method based on nonlinear mathematics

    CN119202866A